Logarithms and Antilogarithms (Danuta Przeworska-Rolewicz (Auth.) )

Logarithms and Antilogarithms
Mathematics and Its Applications
Managing Editor:
M. HAZEWINKEL
Centre for Mathematics and Computer Science, Amsterdam, The Netherlands
Volume 437
Logarithms and
Antilogarithms
An Algebraic Analysis Approach
by
Danuta Przeworska-Rolewicz
Institute of Mathematics of the Polish Academy of Sciences,
Warsaw, Poland
SPRINGER SCIENCE+BUSINESS MEDIA, B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress.
ISBN 978-94-010-6194-0 ISBN 978-94-011-5212-9 (eBook)

DOI 10.1007978-94-011-5212-9
Printed on acid-free paper
AII Rights Reserved

© 1998 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1998
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner
TABLE OF CONTENTS
PREFACE. vii
1. Preliminaries. Introduction to Algebraic Analysis. 1
2. Basic equation. Logarithms and antilogarithms. 11
3. Logarithms and antilogarithms of higher order. 39
4. Logarithms and antilogarithms of operators having either finite
nullity or finite deficiency. 52
5. Reduction theorems. 63
6. Multiplicative case. 82
7. Leibniz algebras. 92
8. Linear equations in Leibniz algebras. 113
9. Trigonometric mappings and elements. 136
10. Semigroup properties of solutions to linear equations 149
11. Operator e hD . 156
12. Power mappings. Polylogarithmic functions. Nonlinear equations. 167
13. Smooth logarithms and antilogarithms. 187
14. Riemann-Hilbert type problems in Leibniz algebras. 201
15. Periodic problems. 216
16. Equations with multiplicative involutions of order N. 236
17. Remarks on the fractional calculus 256
APPENDIX. Functional shifts. By Z. Binderman. 259
AI. Functions of a right invertible operator. 259
A2. Functional shifts. 264
A3. Isomorphisms of spaces of functional shifts. 271
A4. Functional shifts induced by operators of complex
differentiation. 285
A5. Euler-Maclaurin type formulae. 296
A6. Differential and integral properties. 304
REFERENCES 324
SUBJECT INDEX 334
AUTHORS INDEX 341
LIST OF SYMBOLS 343
v
Leonhard Euler in his paper De 1a controverse en-
tre Mrs. Leibniz et Bernoulli sur 1es logarithmes des
nombres negatifs et imaginaires considered the rule
d(log x) = dx / x. He rejected an earlier suggestion of
Leibniz that this rule is only valid for positive real val-
ues of x with the following observation:
"( ... ) Car, comme ce calcul roule sur les quantites vari-
ables, c. a d. sur des quantites considerees en general,
s'il n'etait pas vrai generalement qu'il fOt d . lx = dx/x,
quelque quantite qu'on donne a x, soit possible positive
ou negative, ou meme imaginaire, on ne pourrait jamais se
servir de cette regie, la verite du calcul differentiel etant
fondeesur la generalite des regles qu'il renferme." *)
Memoirs de l'Academie des Sciences de Berlin 5(1749),
139-171, in OPERA, (1) 17,195-232; cf. Fraser F[I]
Beyond this (and extremely important), there is the

magic of "algorithms", or symbolism in mathematics.
Calculus itself shows the wonder of it. Various trans-
forms, generating functions, and the like perform almost
miraculously in mathematical applications.
S. M. Ulam. Adventures of a Mathematician.
Charles Scribner's Sons, New York, 1976
*) "( ... ) For, at this calculus con cernes variable quantities consid-
ered in general, if it were not generally true that d· lx = dx/x,
whatever value we give to X, either positive, negative, or even imag-
inary, we would never be able to make use of this rule, the truth of
differential calculus being founded on the generality of the rules it
contains."
vi
PREFACE
logarithm - the power to which a fixed number, called the base,

must be raised in order to produce a given number, the antilogaritm.
(The Reader's Digest Great Encyclopedic Dictionary)
Another old name of antilogaritm is numerus logarithmi. Logarithms and their

name have been invented by the Scottish mathematician JOHN NAPIER, Lord of
Merchiston (1550-1617) called also NEPER. Napier published the first tables of
logarithms in 1614 (cf. Np[l]).
The term logarithm is derived from Greek: >.0"(0<; = ratio, apdJp,o<; = number.
Nicolaus KAUFMANN, known better as MERCATOR, in his Logarithmo-technia
edited in 1668 presented an approach to the expansion of the function 10g(1 + x)
(cf. Mr[l]).
By more than three centuries properties of logarithms, their tables and slide rules
served as main tools for physycists and (later) engineers in order to calculate with a
requested exactness values of functions and other quantities under considerations,
in particular, values of trigonometric functions. For that reason, any syllabus in
Mathematics for primary and secondary schools even in the first half of 20th cen-
tury consists of several chapters devoted to properties of logarithms and knowledge
how to use them.
It was so when, as students of the first and second year of Mathematics at the
Warsaw University in 1950-52, we had Physical Laboratory. It was almost impos-
sible to finish and describe any experiment at time, since the obligatory calculus
of errors. Fortunately, Rolewicz and myself, we had slide rules of his father, who
was an engineer, and we knew how to use them. This permits us to be ready with
all calculations at the prescribed time.
Next twenty years brought an essential change. In early seventies any student
of engineering in Warsaw, hence also our son, would like to have an electronic
calculator. Even if it was not necessary, it was fashionable. Thus, after few years,
slide rules were no more used. Consequently, this part of syllabus in Mathematics
for schools, which was connected with logarithms, was quickly decreasing to zero.
At present, our grand daughter after 8 years of the primary school does not know
the word "logarithm", but she is familiar with a computer.
The academic and engineering career of logarithms gone with wind? At least,
this seems to be true for engineers and physicists.
Nevertheless, several problems in Mathematics without beautiful properties of log-
arithms and antilogaritms could be stated and solved in an artificial way, although
the logarithmic function has rather a strange nature.
vii
viii Preface
Recently, I had a feeling of a need of some" generalized" logarithmic and antilog-

arithmic functions. However, I did not find any definition, which satisfied my
requests. Finally, three years ago I have started with the following definition (cf.
PR[16J, PR[18j), which came useful much more than I expected.
Let X be a commutative algebra over a field IF with the characteristic zero. Let
D E L(X) be an operator with either finite nullity or finite deficiency and such
that the domain of D is a subalgebra of X. Let n : dom D --+ 2dom D be a
multifunction defined as follows:
nu = {x Edam D: Du = uDx} far u E dom D.
This multifunction is well-defined. Any invertible selector L of n is said to be a

logarithmic mapping and E = L -1 is said to be an antilogarithmic mapping.
Observe that the equation Du = uDx which appears in the definition of the
multifunction n can be written for invertible u's in the form: Dx = u- 1 Du. In
the classical case of the operator D = -itthis is the equation for the so-called
logarithmic derivative. Indeed, [lnu(t)l' = [U(t)]-IU'(t).
In PR[18] properties of logarithmic and antilogarithmic mappings in arbitrary
commutative algebras are studied (cf. also PR[16] for a particular class of algebras
with right invertible operators). In PR[18] there are given some remarks concerning
these mappings in a particular non-commutative case.
In PR[25] we have defined logarithmic and antilogarithmic mappings in non-
commutative algebras and we studied their properties. In general, we have omitted
there corollaries for commutative algebras, since that case was studied in details
in PR[18] (except, of course, some results unknown before).
The present book is an essentially extended (and corrected) version of PR[25]. It
should be mentioned that most of considerations for non-commutative algebras are
essentially different than those for commutative algebras. Also there are several
important results, which never have been published.
It should be pointed out that, due to the presented approach, there are derived
results unknown even in the classical case of the operator D = -it.
Logarithmic and antilogarithmic mappings can be used, for instance, in order to
solve linear and non-linear equations in algebras of square matrices. It seems that
some results might provide numerical algorithms for approximations of solutions.
This book is, in a sense, strictly connected with ALGEBRAIC ANALYSIS (cf.
PR[8j), although it is not a direct continuation of that book.
Appendix written by Zbigniew BINDERMAN gives another essential contribution
to Algebraic Analysis, which in some points is strictly connected with the main
stream of this book.
This book is not self-contained, as the previous books ofthe author. The material
is too reach to keep its volume in a reasonable size. However, I tried to make
Preface ix
it easy to understand for the reader acquainted with basic notions of Functional
Analysis and Linear Algebra. The only exception is few times used notion of locally
pseudo-convex and locally bounded spaces. This theory can be found in the book
of ROLEWICZ R[l). Another possibility is to restrict the corresponding results to
Banach spaces. Clearly, these results remain true, but much less interesting (cf.
the following diagram).
complete linear metric spaces

+
complete locally pseudo-convex spaces
./ '\t
complete locally bounded spaces complete locally convex spaces
'\t ./
Banach spaces
All proofs are closed with •. Definitions, examples, notes and remarks are closed
with D.
I hope that this book will be used not only by mathematicians, but also by these
non-mathematicians who are looking for direct methods in order to solve equations
and problems obtained in their works.
It would be nice to say about this book words, which were used by Mercator in
his Logarithmo-technia as a subtitle:
" Methodus construendi

LOGARITHMOS
Nova, accurata, & facilis ... "
Acknowledgment
A big part of this book has been written during my several visits in Institut fUr
Statistik und Mathematische Wirtschaftstheorie at Universitat Karlsruhe. I am
and I always will be very indebted to Professor Diethard Pallaschke and Authorities
of this University for their efforts in order to organize my life there in the possibly
best way. In particular, I appreciate very much that the preprint, a preliminary
version of this book, has been printed at Universitat Karlsruhe.
My gratitude should be extended to my husband, Professor Stefan Rolewicz, for
his permanent assistance.
Warszawa, May 25, 1997
Van uta Przeworska-Rolewicz

CHAPTER 1
PRELIMINARIES. INTRODUCTION TO ALGEBRAIC ANALYSIS
The term" Algebraic Analysis" was initially used by LAGRANGE L[1] in 1797 in the
title of his book
Theorie des Fonctions Analytiques contenant Les Principes du Calcul Differentiel,
degages de toute consideration d'infiniment petits, d'evanouissans de limites et de
fluxions, et reduit a I'analyse algebriques de quantites finies
in order to point out that most of results have been obtained by algebraic opera-
tions on analytic quantities. In that general and common sense this name, adopted
also by Cauchy, was used in 19th and in 20th century (d. FRASER F[IJ, JAHNKE
J[1J, also the author PR[26J, PR[28).
The main idea of Algebraic Analysis in its present, more strict, sense derives from
the fact that the differential operator D = ft
is right invertible in several functions
spaces.
Foundations of Algebraic Analysis are following (d. PR[8]).
As usually, denote by N, Z, Q, R, C the sets of all positive integers, integers, ratio-
nals, reals and complexes, respectively. Write No = NU{O}, Il4 = {r E R: r ~ O}.
Let X be a linear space (in general, without any topology) over a field IF of scalars
of the characteristic zero. In the sequel we shall write 1 for the unit in IF and 1.
for ~ . 1 (n EN). We also write n
• L(X) is the set of all linear operators with domains and ranges in X;
• dom A is the domain of an A E L(X);
• ker A = {x E dom A : Ax = O} is the kernel of an A E L(X);
• Lo(X) = {A E L(X) : dom A = X}.
An operator D E R(X) is said to be right invertible if there is an operator R E
Lo(X) such that RX C dom D and DR = I, where I denotes the identity operator.
The operator R is called a right inverse of D. By R(X) we denote the set of all
right invertible operators in L(X). Let D E R(X). Let RD C Lo(X) be the set of
all right inverses for D, i.e. DR = I whenever R E RD. We have
dom D = RX EB ker D, independently of the choice of an R E RD.
Elements of ker D are said to be constants, since by definition, Dz = 0 if and only
if z E ker D. The kernel of D is said to be the space of constants. We should point
out that, in general, constants are different than scalars, since they are elements
of the space X. For all R,R' E RD,X E X we have Rx - R'x E ker D, i.e. a
difference of two primitives of x is a constant. If two right inverses commute each
with another, then they are equal. Let
FD = {F E Lo(X) : F2 = F; FX = ker D and 3RE'RD FR = O}.
1
D. Przeworska-Rolewicz, Logarithms and Antilogarithms
© Kluwer Academic Publishers 1998
2 Chapter 1
Any F E F D is said to be an initial operator for D corresponding to R. One can

prove that any projection F' onto ker D is an initial operator for D corresponding
to a right inverse
R' = R - F'R independently of the choice of an R E 'RD.
If two initial operators commute each with another, then they are equal. Thus
this theory is essentially noncommutative. An operator F is initial for D if and
only if there is an R E 'RD such that
(1.1) F =I - RD on dom D.
Even more. Write'R D = {Ry}-yEf. Then, by (1.1), we conclude that 'RD induces
in a unique way the family F D = {F-y }-YEr of the corresponding initial operators
defined by means of the equality
(1.2) F-y = I - R-yD on dom D (-y E r).
Formula (1.2) yields (by a two-lines induction) the Taylor-Gontcharov Formula,

which playa fundamental role in our theory. Namely, let {'Yn} C r be an arbitrary
sequence. Then for all positive integers N the following identity holds:
N-l
(1.3) I = F-yo + L R-yo ... R-Yk_l F-Yk Dk + R-yo ... R-YN_l DN on dom DN.
k=O
Let F be an initial operator for D corresponding to an R E 'RD. Putting in (1.3)

R-Yn = Rand F-Yn = F (n EN), we obtain the Taylor Formula:
n
(1.4) 1= L RnFD n + RnD n on dom D n (n EN).
k=O
Let D E R(X). Then
(1.5) 'RD = {R + FA: A E Lo(X)}, FD = {F(I - AD): A E Lo(X)},
where F is an arbitrary (but fixed) initial operator for D corresponding to a right

inverse R of D. This means that it is enough to know one right inverse in order
to determine all right inverses and all initial operators.
Suppose that Db ... , Dm E R(X) and that the superposition D = D1 ... D m is
well-defined. Let Fj be an initial operator for D j corresponding to an Rj E 'RDj
(j = 1, ... , m). Write
Then D E R(X), R E 'RD and F is an initial operator for D corresponding to R.

Introduction to Algebraic Analysis 3
Some results can be proved also for left invertible operators. Write
• A(X) is the set of all left invertible operators belonging to Lo(X) (we assume
that dom T = X for T E A(X));
• LT = {S E L(X) ST = I} is the set of all left inverses to aTE A(X);
• gT = {G E L(X) G2 = G, kerG = TX and :3 SG = O} is the set of all
SECT
co-initial operators for aTE A(X);
• I(X) = R(X) n A(X) is the set of invertible operators.
Clearly, if ker D -I {O}, then the operator D is right invertible, but not invertible.
Here the invertibility of an operator A E L(X) means that the equation Ax = y
has a unique solution for every y E X. If D E I(X) then FD = gD = {O} and
RD = LD = {D- 1 }.
The above determined sets are, in a sense, dual. Namely, if D E R(X) and FE F D
corresponds to an R E RD, i.e. FR = 0 then R E A(X) and F E gD corresponds
to DE LD.
With these facts one can obtain Calculus and solutions to linear equations (under
appropriate assumptions on resolving equations). If the field IF is algebraically
closed then solutions of linear equations with scalar coefficients can be calculated
by a decomposition of a rational function into vulgar fractions (similarly, as in
Operational Calculus). If the space under consideration is an algebra (a linear
ring), one can obtain some more results. Also there is a reach theory of shifts and
periodic problems. Some of them will be discussed later.
Recall then that any solution of the equation
Dx = y, Y E X, D E R(X),
is of the form x = Ry + z, where R E RD and Z E ker D is arbitrary. This form is
independent of the choice of R. Indeed, if R' E RD and R' -I R then for all x E X
we have R'X - Rx E ker D. So that, a change of R implies only a change of the
constant z.
The equation
Tx = y, yET X, T E A(X)
has a unique solution x = Sy which is independent of the choice of S E LT.
Indeed, by definition, x = STx = Sy and kerT = {O}.
Write
N
Q(D) =L Qk Dk , where D E R(X), Qo, ... ,QN-l E L(X), QN = I.
k=O
If there is an R E RD such that the operator

N
Q(I,R) = L QkRN-k
k=O
4 Chapter 1
is invertible, then Q(D) E R(X) and R N [Q(I,R)]-1 E 'RQ(D). Thus an initial

value problem for Q(D), namely, the problem to find solutions of the equation
Q(D)x = y, Y E X, with the initial conditions
FDkx = Yk, where Yk E ker D are given (k = 0,1, ... , N -1),

has for every y, Yo, ... , Y N -1 a unique solution of the form
N-l m N-l
X = R N [Q(I,R)J-1[y - 2: (2: QkRm-k)YmJ + 2: RkYk'
m=O k=O k=O
i.e. this problem is well-posed.

An A E Lo(X) is said to be a Volterra operator if the operators I - AA are
invertible for every A E IF, i.e. if for every A E IF \ {O} the number 1/ A is a regular
value of A. The set of all Volterra operators belonging to L(X) will be denoted
by V(X).
Main advantages of Algebraic Analysis are: (i) simplifications of proofs due
to an algebraic description of problems under consideration; (ii) algorithms for
solving "similar" problems, although these similarities could be rather far each
from another and very formal, and (iii) several new results even in the classical
case of the operator ft
(what is, indeed, unexpected).
There are several applications to ordinary and partial differential equations with
scalar and variable coefficients, functional-differential equations and for discrete
analogues of these equations, for instance, for difference equations. There are also
some results for nonlinear equations.
It should be pointed out that in Algebraic Analysis a notion of convolution is not
necessary. Also there is no need to have a structure of the Mikusinski field (cf.
PRUDNIKOV Pru[I]). This, together with the noncommutativity of right inverses
and initial operators, shows the essential distinction of Algebraic Analysis from
Operational Calculus.
Example 1.1. Let X = e[O,IJ over IF, where either IF = IR or IF = C, D = ft.
Then ker D consists of all constant functions. The operators Ra = I:, a E [0,1]'
are Volterra right inverses of D. Observe that dim ker D = 1. The family of
initial operators induced by Ra is defined as follows: (Fax)(t) =
x(a) for x E X,
a E [0,1]. Consider operators of the form
_ ))
(Fax J.l"'---,1"""":'-'--'--
(t = a
m(s)x(s)ds
::c for X E X, a E [0,1]' m E X,
Ia m(s)ds
AFa + (1 - A)Fb, AFa + (1 - A)Fb, for bE [0,1], A E IF.
They are also initial operators, because they are projections onto the space of
constants. In several particular cases the corresponding right inverses have eigen-
values.
If we consider the space C[0,1] over C then the only continuous Volterra right
inverses are f:,a E [0,1]. However, if IF =
~, then one can find a Volterra right
inverse, which is not of that form (cf. the author and ROLEWICZ PRR[3], PRR[4]).
The Taylor formula applied to a function x E dom DN = GN[O, 1] gives a classical
Taylor formula with a remainder in an integral form. In order to obtain remainders
either in the Lagrange form or in the Cauchy form it is enough to apply the
Darboux property of continuous functions by two different estimations, without
any theorem about the intermediate value, like the Rolle theorem or Lagrange
theorem.
Observe that X = G[O, 1] is a commutative algebra (a linear ring) with respect to
the usual pointwise multiplication. The operator D satisfies the so-called Leibniz
condition:
D(xy) = xDy + yDx far X,y E dom D = G1 [O, 1]
(cf. Formula (1.6)).
We point out that the restriction to the interval [0,1] is not essential. The same
considerations are valid also for any interval [a, b] and for half-axis or the real line.
o
Example 1.2. Let X = G(O), where 0 = [0,1] x [0, 1], and (Dx)(t, s) = :tx(t, s).
of the equality: (Rax)(t, s) =

right inverses. Since
f:
Clearly, the operator D is right invertible and the operators Ra defined by means
x(r, s)dr, 0 ::; a :::; 1, x E X, are its Volterra
ker D = {x E dom D = G 1 [0, 1] : x(t, s) = <p(s) far all t, s E [0, In,
we conclude that dim ker D = +00. The induced family of initial operators is
defined as follows: (Fax)(t, s) = x(a, s) for all t, s E [0,1]. 0
Example 1.3. (Difference operators). Let X = (s)JF be the space of all sequences
{x n }, Xn E IF, n E N. Let D{x n } = {X n+l - x n } for all {x n } E X. Then the
operator Rm defined by means of the equality:
ifn:::; m;
(m E N)
if n ~ m + 1;
is a Volterra right inverse of D, i.e. I - ARm is invertible for every A E IF. Each of
the induced initial operators Fm maps any element {x n } into a constant sequence.
Namely, if we write en = 1 for all n E N, then we get Fm{x n } = xm{e n } (m EN).
The kernel of D consists of constant sequences. Thus dim ker D = 1. Using these
definitions, we can solve, for instance, difference equations of the form
N
L Pk(n)xn+k = Yn, where Pk(n) E IF (n EN),

k=O
6 Chapter 1
with either initial conditions: xm+i = y:" (m E N is fixed, j = 0, ... , N - 1) or

(by means of the Taylor-Gontcharov formula) with mixed boundary conditions:
xm+i = y:" (m = Po, ... ,PN-l EN are fixed, j = 0, ... , N - 1). 0
Note 1.1. Nguyen Van Mau (cf. N[l]) has shown that there is a right invertible
singular integral operator, which has no Volterra right inverses. 0
If X is an algebra over IF with a D E L(X) such that x, y E dom D implies
xy, yx E dom D, then we shall write D E A(X). If X is a commutative algebra
then we shall write DE A(X).
If D E A(X) then X is said to be a Leibniz algebra if D satisfies the Leibniz
condition:
(1.6) D(xy) = xDy + (Dx)y for x,y E dom D.
In general, write
(1.7) fD(X,y) = D(xy) - cD[xDy + (Dx)y] for X,y E dom D,
where CD is a scalar dependent on D only. Clearly, fD is a bilinear (i.e. linear

in each variable) form, which is symmetric when X is commutative, i.e. when
DE A(X). This form is called a non-Leibniz component. Non-Leibniz components
have been introduced for right invertible operators D E A(X) (cf. PR[8], Section
6.1).
By definition of the non-Leibniz component, if D E A(X) then the product rule
in X can be written as follows:
D(xy) = cD[xDy + (Dx)y] + fD(X,y) for X,y E dom D.
This representation is not uniquely determined, as it is shown by

Example 1.4. Suppose that D E A(X), X has unit e E dom D and there are
two different representations of the product rule for D. Namely
D(xy) = cv[xDy + (Dx)y] + fD(X, y), and D(xy) = fb(x, y)

for x, y E dom D, where fb is a bilinear form and cD = O. Putting y = e, we get
Dx = cD[xDe + (Dx)e] + fD(X, e), which implies (1 - cD)Dx = cDxDe + fD(X, e)
for the first representation. For the second one, if y = e, we obtain Dx = fb(x, e).
This implies
(1 - cD)fb(x, e) = (1 - cD)Dx = cDxDe + fD(X, e),

i.e. fD(X, e) = (1 - cD)fb(x, e) - cDxDe. We therefore conclude that fD = fb if
and only if CD = 0 = CD' 0
In order to determine uniquely the non-Leibniz component, here and in the sequel
we shall assume that fD does not contain any terms of the form QxDy,(J(Dx)y
(a., (3 E IF \ {O}). If it is the case, we say that fD is a proper non-Leibniz com-

ponent. In our subsequent considerations we shall use only proper non-Leibniz
components. Thus we shall omit the word proper, since this will not lead to any
misunderstanding.
Observe that, by (1.6), if D E A(X) then X is a Leibniz algebra if and only if
CD = 1 and fD = O.
Clearly, if D E A(X) and X is a Leibniz algebra with unit e, then e E ker D.
Indeed, De = De 2 = eDe + (De)e = 2De, which implies De = o.
Non-Leibniz components for powers of D E A(X) are determined by recurrence
(equivalent) formulae. Namely, we have
Proposition 1.1. Let DE A(X). Then for all kEN, X,y E dom Dk such that
fD(X, y) E dom Dk we have xy E dom Dk and
(1.8) Dk(xy) = ct[xDky + (Dkx)y] + fi;)(x,y),

where fii) = fD and for k = 2,3, ...
(1.9) fi;)(x,y) = ct[(Dx)Dk-ly + (Dk-1x)Dy]+
+ct-1[JD(X, Dk-1y) + fD(Dk-1x, y)] + Dfi;-l) (x, y)

or
(1.9' ) fi;)(x,y) = ct[(Dx)Dk-ly + (Dk-1x)Dy]+
+cD[J1k - 1) (x, Dy) + fi;-l) (Dx, y)] + D k- l fD(X, y),

i.e.
CDk = ct and fDk = fi;l for kENo
Moreover, fi;) are bilinear mappings of dom Dk xdom Dk into dom Dk (k EN).
Proof (by induction). For n = 1 Formula (1.8) is trivially satisfied (by definition).
Let then 2 S N E N be arbitrarily fixed. Suppose Formula (1.8) to be true for
k = 1,2, ... , N - 1. By (1.8), if x, y E dom DN then DN-1x, DN-I y E dom D.
Also xDN-I y, (DN-IX)Y E dom D and
D[DN-1(xy)] = D{C~-I[xDN-Iy + (DN-1x)y] + f1N - 1)(x,y)}

= c1j[XDN y + (Dx)DN-1y + (DN x)y + (DN-1x)Dy]
+ c~-1[J1N-I)(X, DN-1y) + ft-l) (DN-1x, y)] + D f1N- 1)(x, y)
= c1j[xDN Y + (DN x)y) + c1j[(Dx)D(N-1)y + (D(N-I)X)Dy)
+ c1j[f1N- 1)(x, DN-1y) + ft-I\DN-1x, y)) + D f1N- 1)(X, y)
= c1j[xDN y + (DNx)y) + fg")(x,y).
8 Chapter 1
A similar proof for Formula (1.9').

Observe that, by definition, f~) is a bilinear mapping of dom Dk xdom Dk into
dom D k (k = 2,3, ... ). '
•
Note 1.2. The proof of Proposition 1.1 is, as a matter of fact, the same as
that given in the book PR[8], Section 6.1, for commutative algebras with right
invertible operators, since the right invertibility of D was not used. We have only
to remember that, in general, yx "I xy. 0
It is easy to prove the following
Proposition 1.2. Suppose that D E A(X) and p "lOis an arbitrarily fixed scalar.
Then pD E A(X) and
(1.10)
In a similar manner as Proposition 1.1, we prove

Proposition 1.3. Suppose that D l , D2 E A(X), the superposition D = DID2
exists and DlD2 E A(X). Then
(1.11)
fVlV2(X,y)
= fVl (x, y) + DIfv. (x, y) + +cv l CV2 [(DIX)D2Y + (D2X)DIY]'
Corollary 1.1. Suppose that all assumptions of Proposition 1.3 are satisfied. If
DID2 = D2DI and D E A(X) then the non-Leibniz component for the superpo-
sition can be written in a symmetrized form
(1.12)
Proof. Use the equality yx = xy.

For higher powers of D in Leibniz algebras, by an easy induction from Formulae
•
(1.9) and the Leibniz condition, we obtain
Proposition 1.4. Suppose that DE A(X) and X is a Leibniz algebra. Then the
Leibniz formula holds:
(1.13) Dn(xy) = t
k=O
(~) (Dkx)Dn-ky far x, Y E dom D n (n EN),
i.e.
x, y E dom D n (n E N).
Recall that in the case of a right invertible D E A(X) the algebra X is said to be
aD-algebra.
Example 1.5. Suppose that X and D are defined as in Example 1.2. Then
X with the pointwise multiplication is a commutative Leibniz algebra, since the
operator D satisfies the Leibniz condition. 0
Example 1.6. Suppose that X and D are defined as in Example 1.3. Then X
with the coordinatewise multiplication is a commutative algebra. It is easy to
verify that the product rule for D is the following:
D(xy) = xDy + yDx + (Dx)(Dy) for all x = {xn},y = {Yn} EX.

Here CD = 1 and fD(x,y) = (Dx)(Dy) for X,y E X. Commmutative algebras
with the product rule
D(xy) = xDy + yDx + d(Dx)(Dy) for x, y E dom D, where dE IF \ {a},
are said to be quasi-Leibniz algebras (d. DUDEK D[l], who considered quasi-
Leibniz D-algebras, also PR[8]). 0
Example 1.7. Let X and D be defined as in Example 1.1. Then X with the
pointwise multiplication is a commutative algebra. Let n E N be arbitrarily fixed.
Consider the operator Dn. Then CDn = 1 and fDn are defined by Formulae (1.14).
Clearly, D does not satisfy the Leibniz condition for n f. 1, hence X is not a
Leibniz algebra with respect to Dn (n 2:: 2). 0
Example 1.8. Let X be defined as in Example 1.3 and let Dp{x n } = {xn+l -
px n }, where pElF is fixed. By similar arguments as those used in Example 1.3, we
prove that D E R(X). X with the coordinatewise multiplication is a commutative
algebra. It is not difficult to verify that the product rule for D is
Dp(xy) = p(xDpy + yDpx) + (Dpx)(DpY) + (p2 - p)xy
for x = {xn},y = {Yn} E X, i.e. here CDp = P and fDp(x,y) = (Dpx)(DpY) +

(P2 -p)xy. For p = 1, we get Example 1.6. For p = 0 we find CDo = 0, fDo(x,y) =
(Dpx)(DpY). 0
Several other examples will be considered in the next chapters.
Let D E A(X). The operator of multiplication by an arbitrarily fixed element a E
X will be denoted by the same letter. Thus we shall mean by Aa the superposition
of two operators:
Aax = A(ax) for A E L(X), a,x E X, ax E dom A.
Let D E A(X). We shall consider the following sets:

• the set of all multiplicative mappings (not necessarily linear) with domains
and ranges in X:
M(X) = {A: dom A c X and A(xy) = A(x)A(y) for X,y E dom A}j
10 Chapter 1
• the set of all multiplicative linear operators belonging to L(X):
M L(X) = M(X) n L(X);
• the set I(X) of all invertible elements belonging to X;

• the set of all elements from Y C X having n-th roots:
If x E In(Y) and yn =x then we write y = x 1 / n , (n EN).

Let X be an algebra. Define for an a E X an operator at- by the formula
(1.15) at-x = xa where x E X.
Then, by definition, for D E A(X)
(1.16) = {x E dom D : Dx - at-x = O}

= {x E dom D : Dx - xa = O}.
This denotation corresponds to the classical one:
ker(D - a) = {x E dom D : Dx - ax = O}.

Note that the operator at- is linear. The operator D - at-, as a sum of two
linear operators, is linear with respect to its argument. Clearly, if D E A(X) then
D - at- = D - a for Dx - xa = Dx - ax whenever a EX, x E dom D.
By F[t] we denote the set of all polynomials in the variable t with coefficients in
F.
We shall use several time the so-called Kronecker symbol, i.e.
I if j = kj
{
Ojk = 0 if j i- k.
CHAPTER 2
BASIC EQUATION. LOGARITHMS AND ANTILOGRARITHMS
We begin with
Definition 2.1. Suppose that D E A(X). Let Or> 0 1 : dom D --+ 2dom D be
multifunctions defined as follows:
(2.1) Oru = {x E dom D : Du = uDx} for u E dom D.
(2.2) OIU = {x E dom D : Du = (Dx)u} for u E dom D.

The equations
(2.3) Du = uDx for (u,x) E graph Or, Du = (Dx)u for (u,x) E graph 0 1
are said to be the right and left basic equations, respectively. Clearly,
(2.4) 0;1X = {u E dom D : Du = uDx} for x E dom D.
(2.5) 0/lx = {u E dom D : Du = (Dx)u} for x E dom D.
If D E A(X) then Or = 0 1 and we write Or = 0 1 = O. Clearly, in this case we

have dom Or = dom 0 1 = dom O. 0
The multifunctions Or, 01, 0 are well-defined. We shall start with right invertible
operators. Left invertible and invertible operators will be considered later. We
have
Proposition 2.1. Let D E R(X)nA(X). Then
dom Or n dom 01 =f:. 0.
Proof. Let z, z' E ker D. Then Dz = Dz' = 0, zDz' = 0, (Dz')z = 0. Thus z and
z' satisfy right and left basic equations. •
Proposition 2.2. Let D E R(X)nA(X). Then
domO;1 :J{x=R(u- 1 Du)+z: UEI(X)ndomD; zEkerD; RERD}=f:.0;
domO / 1 :J{x=R[(Du)u- 1]+z: uEI(X)ndomD; zEkerD; RERD}=f:.0.
Proof. Let u E I(X) n dom D, z E ker D, R E RD be arbitrarily fixed. If

Du = uDx then Dx = u- 1 Du. Hence x = R(u- 1 Du) + z and x E dom 0;1. A
similar proof for Oil. •
As a matter of fact, from the proof of Proposition 2.2 it follows that
I(X)ndom 0;1 :J {x = R(u- 1Du) +z: u E I(X)ndom D; z E ker D; R E RD};
I(X)ndom Oil :J {x = R[«Du)u- 1 ]+z : u E I(X)ndom Dj z E ker Dj R E RD}.
11
12 Chapter 2
Corollary 2.1. If DE R(X) n ML(X)nA(X) then
dom 0;1 J {x = (Ru-l)u + z: u E J(X) ndom D; z E ker D; R E RD} =f. 0;

dom 011 J {x = uRu- l + z: u E J(X) ndom D; z E ker D; R E RD} =f. 0.
If X has a unit e E dom D then e E dom Or n dom 0 1 • If D E A(X) then
e E dom O.
Proof. Let u E J(X) n dom D, z E ker D, R E RD. Consider the right basic equa-
tion. Since DE ML(X), we have Dx = u-1Du = (DRu-l)(Du) = D[(Ru-l)u].
Hence x = (Ru-l)u + z and x E dom 0;1. If X has a unit e E dom D then
Ore = {x E dom D : De = (Dx)e} = {x = e +z : z E ker D} = e + ker D =f. 0.
A similar proof for the left basic equation. If D E A(X) then e E dom O. •
Proposition 2.3. Let u E I(X) n dom D. If D E A(X) and either x E Oru,
y E nru or x E OIU, Y E OIU, then x - y E ker D. If D E A(X) and x E Ou,
Y E nu, then x - y E ker D.
Proof. If x, y satisfy the right basic equation then, by definition, Dx = u- l Du,
Dy = u- l Du. Hence D(x - y) = o. A similar proof if x satisfies the left basic
equation. If X is a commutative algebra then Or = 0 1• Again we conclude that
x - y E kerD. •
Proposition 2.4. For any Y C X we write
nr(y) = U nru, nl(y) = U O,U.

uEY uEY
Let D E R(X)n A(X). Then

(i) Or (ker D), 0 1 (ker D) C ker D;
(ii) O(ker D) C ker D, provided that DE A(X);
(iii) zu, uz E I(X) n dom Or n dom 0 1 whenever z E J(X) n ker D, U E I(X) n
dom Or n dom 0 1;
(iv) zu E J(X) n dom 0 whenever z E J(X) n ker D, u E J(X) n dom 0, provided
that D E A(X).
Proof. (i) Let z E ker D and Xr E dom Or, Xl E dom 0 1 be arbitrary fixed. Then
zDx r = Dz = 0, (Dx)z = Dz = 0, which implies that Orz, 0IZ C ker D.
(ii) follows immediately from (i), if we apply the commutativity of X.
(iii) Suppose that z, u satisfy the assumptions of (iii). By definitions, there is an
Xr E dom D such that DX r = u- l Du and Dz = o. Hence
D(zu) = CD [zDu + (Dz)u] + fD(Z, u) = cDzDu + fD(Z, u)
Basic equation 13
which implies
(ZU)-l D(zu) = U- 1 Z-l CDZ Du + U- 1 Z-l fD(Z, u)

=CDU- 1 Du + U- 1 Z-l fD(Z, u) = cDDx r + U- 1 Z-l fD(Z, u).
We therefore conclude that there is a Yr E dom D such that DYr = (ZU)-l D(zu).
Namely,
Yr = CDX r + R[u- 1 Z-l fD(Z, u)) for an R E RD.
Thus ZU E J(X) n dom Or. A similar proof shows that uz E J(X) n dom 0 / .
(iv) follows immediately from (iii) if we apply the commutativity of X. •
Suppose that (u r , x r ) E graph Or, (UI, XI) E graph OJ, Ln LI are selectors of On
0/, respectively, and E r , EI are selectors of 0;:-1,0,1, respectively. By definitions,
Lru r E dom 0;:-1, Erx r E dom On Llul E dom 0,1, E1xI E dom 0 1 and the
following equations are satisfied:
Let F be an initial operator for D corresponding to an R E RD and let Ur , UI E

J(X). Then these equations can be written in equivalent forms
Definition 2.2. Any invertible selector Lr of Or is said to be a right logarithmic

mapping and its inverse Er = L;:-l is said to be a right antilogarithmic mapping.
If (un x r ) E graph Or and Lr is an invertible selector of Or then the element Lru r
is said to be a right logarithm of Ur and Erx r is said to be a right antilogarithm
of Xr . By G[Or) we denote the set of all pairs (Lr' E r ), where Lr is an invertible
selector of Or and Er = L;:-l. Respectively, any invertible selector Ll of 0 1 is said
to be a left logarithmic mapping and its inverse EI = L,l is said to be a left
antilogarithmic mapping. If (UI, XI) E graph 0 1 and LI is an invertible selector of
01 then the element Llu is said to be a left logarithm of UI and Elxl is said to be
a left antilogarithm of XI. By G[Oz] we denote the set of all pairs (LI' E I ), where
LI is an invertible selector of 0 1 and EI = L,l.
H D E A(X) then 0 1 = Or = 0 and we write
Lr = LI = L, Er = EI = E, (L,E) E G[O).
Selectors L, E of 0 are said to be logarithmic and antilogarithmic mappings, re-

spectively. For any (u, x) E graph 0, (L, E) E G[O) elements Lu, Ex are said to
be logarithm of U and antilogarithm of x, respectively. 0
14 Chapter 2
Clearly, by definition, for all (Lr,Er) E G[Or], (ur,x r ) E graph Or, (Ll,EI ) E
G[Oz], (Ul, xz) E graph Oz we have
Moreover,
(2.8) uDLru - (DLlU)U = 0 for all U E dom Or n dom Ol.
Indeed, by Corollary 2.1 and our definitions, uDLru - (DLzu)u = Du - Du = O.

Similarly, if DE A(X), (L,E) E G[O] and (u,x) E graph 0, then we have
ELu = u, LEx =x and DEx = (Ex)Dx, Du = uDLu.

The equality (2.8) in the commutative case is trivially satisfied.
Proposition 2.5. Let DE R(X)nA(X). A right logarithm of zero is not defined.
If (Lr,E r ) E G[Or], (Lz,Ez) E G[Oz], then Lr(kerD \ {O}) C kerD, Er(kerD) C
kerD, Ll(kerD \ {O}) C kerD, El(kerD) C kerD. If DE A(X), (L,E) E G[O],
then a logarithm of zero is not defined and L(ker D \ {O}) C ker D, E(ker D) C
ker D whenever (L, E) E G[O]. In particular, Er(O), Ez(O), E(O) E ker D.
Proof. If x = Lr(O) then we have 0 = D(O) = ODx. Hence Lr(O) is not defined.
Suppose that Z E ker D is arbitrarily fixed and Z =I O. Then, by Proposition 2.4,
Lrz E ker D. Let x = Erz. Then DErz = (Erz)Dz = 0, which implies that
Erz E ker D. In particular, Er(O) E ker D. A similar prooffor left logarithms and
antilogarithms and, in the commutative case, for logarithms and antilogarithms .
Definition 2.3. A right logarithmic mapping Lr (left logarithmic mapping Ll,

•
logarithmic mapping L, respectively) is said to be of the exponential type if
Ll(UV) = Llu + Lzv for u, v E dom Ol; L(uv) = Lu + Lv for u, v E dom 0,

respectively. 0
Proposition 2.6. Let DE R(X).
(i) if DE A(X), (Lr,E r ) E G[Or], (Lz,Ez) E G[Oz] and Lr (Lz, respectively) is of
the exponential type then
whenever x, y E dom 0;1 (x, y E dom Oil, respectively);

Basic equation 15
(ii) if D E A(X), (L, E) E G[O] and L is of the exponential type then
E(x + y) = (Ex)(Ey) fOT x, y E dom 0- 1 ;
(iii) if Lr E M(X) then Er E M(X), if LI E M(X) then EI E M(X), if L E M(X)

then E E M(X);
(iv) if DE A(X) n ML(X) then L,E E M(X).
Proof. (i) Let (LnEr) E G[Or). Let u,v E dom Or and let x = Lru, y = Lrv.
Then x, y E dom 0;:-1 and u = Erx, v = ErY. If Lr is of the exponential type
then, by (2.6),
(ii) A similar proof in the commutative case. (cf. Definition 7.3, Theorem 7.1 and
Corollary 7.8).
(iii) If Lr E M(X) then
i.e. Er E M(X). Similar proofs for L1, EI and L, E.

(iv) Suppose now that that D E A(X) n M L(X). Let u, v E O. Then x = Lu,
y = Lv E dom 0- 1 and u = Ex, v = Ey. By our assumptions,
D[(Ex)(Ey))
= (DEx)(DEy) = [(Ex)Dx)[(Ey)Dy) =.
(Ex)(Ey)(Dx)(Dy) = (Ex)(Ey)D(xy)
Hence E(xy) = (Ex)(Ey). This implies that L(uv) = L[(Ex)(Ey)} = LE(xy) =

xy = (Lu)(Lv). •
Observe that, by Proposition 2.6(i),(ii), if either left (right) logarithmic mappings
or logarithmic mappings are of the exponential type, then the corresponding left
(right) antilogarithmic mappings and antilogarithmic mappings satisfy the same
functional equation as usual exponential functions cent, c,o: E lR (cf. Example
3.2).
Definition 2.4. Let D E A(X). By Lgr(D) (LgJ(D), respectively) we denote
the class of these D-algebras with unit e E dom Or (e E dom 0/, respectively)
for which there exist invertible selectors of Or (0/, respectively), i.e. there exist
(Lr,E r ) E G[Or] ((L/,Ed E G[Od, respectively). If D E A(X) then Lgr(D) =
LgJ(D). This class is denoted by Lg(D). 0
Proposition 2.7. Let X E Lgr(D) (X E LgJ(D), X E Lg(D), respectively). Let
(Lr,Er ) E G[Or], (L/,E1) E G[Od, (L,E) E G[O]. Ife E kerD, then Zr = Ere,
Z/ = Ele, Z = Ee E ker D and Lrzr = e, L/z/ = e, Lz = e.
16 Chapter 2
Proof By our assumptions, D E R(X) and De = O. By definition, DEre =

(Ere)De = O. Hence Zr = Ere E ker D and LZr = LrEre = e. Similar proofs for
L/,E/ and L,E. •
Theorem 2.1. Let either X E Lgr(D) or X E Lgl(D). Let R E RD and 9 = Re.
Then 9 E dam Or (g Edam 0 1, 9 Edam 0, respectively) if 9 E leX).
In other words: a right logarithm (left logarithm, logarithm) of 9 = Re exists if 9
is invertible.
Proof By our assumptions, De = 0, Dg = e, D2g = 0 and e E dam Or (e E
dam 0 1, e Edam 0, respectively). Let (Lr, Er) E G[Or], (LI,EI) E G[Od, (L, E) E
G[O]. If u E leX) n dam Or then the right basic equation Du = uDx may be
written as follows: Dx = u- 1Du, which implies that x Edam D and x = Lru. In
particular, if 9 E leX) then DLrg = g-l Dg = g-l, which implies Lrg = Rg- 1 +z,
where Z E ker D. A similar proof for left logarithms. •
Corollary 2.2. Suppose that X E Lg(D), R E RD and 9 = Re. Then
(i) 9 Edam 0 if and only if 9 E leX);
(ii) there is no 6 Edam 0 \ {O} such that 6g = 0, i.e. 9 is not a zero divisor.
Proof (i) Sufficiency is proved by Theorem 2.1.
Necessity. Let 9 Edam O. Then Lg is well-defined and, by the commutativity of
X, e = gDLg = (DLg)g. Hence 9 E leX).
(ii) If there is a 6 :j:. 0 such that 6g = 0 then 6 = g-l g6 = 0, a contradiction to our
assumptions. •
Corollary 2.3. Suppose that all assumptions of Theorem 2.1 (Corollary 2.2,
respectively) are satisfied. Let R1 :j:. R be a right inverse of D. Then Theorem 2.1
(Corollary 2.2, respectively) holds for gl = R 1e.
Proof Clearly, gle - 9 = R 1e - Re = Z E ker D (cf. PR[8], Section 2.1). This
implies that Dg1 = D(Re + z) = e, as before. •
Example 2.1. Suppose that X is a commutative D-algebra with unit e. Let
(L, E) E G[O] and let R E RD. Let 9 = Re Edam 0 and let v = Lg = LRe.
Then e = DRe = Dg = gDv, by our definitions. By Theorem 2.1, 9 E leX) and
DLg = g-l, i.e. DL(Re) = (Re)-l. Note that the element 9 = Re plays a role
of an argument. Indeed, in the classical case of IF = JR, X = C(ll4) with the
ft
pointwise multiplication, D = and R = J;
we have 9 = R1 = t, L = In and
d 1
DLg = -In t = _ = g-1 (far t :j:. 0).
dt t
o
Proposition 2.8. Let X E Lgr(D) (X E Lgl(D), X E Lg(D), respectively).
(i) H(Lr,Er), (L~,E~) E G[Or] and (Ur,X r ) E graph Or, Ur E leX), then
Basic equation 17
L;ul = Llul + Z, Ef(xi + z) = Elxl for a z E ker D.
(iii) If(L,E), (L',E') E G[O] and (u,x) E graph 0, U E leX), then
L'u = Lu + Z, EI(X + z) = Ex for a z E ker D.
Proof. (i) Let y = L~ur. Then y E Orur. By definition and Proposition 2.3, we
have Lru r - L~ur = Xr - Y = z E ker D. This implies that Erxr = Ur = E~L~ur =
E~y = E~(xr + z). The same proof for (ii) and (iii). •
Proposition 2.9. Let X E Lgr(D) (X E Lg1(D), X E Lg(D), respectively).
If (Lr,Er ) E G[Or], (unxr) E graph On L~ur = Lrur + z, E~(xr + z) = Erxr
for a z E kerD then (L~,E~) E G[Or]' If (L/,Ed E G[Od, (UI,X,) E graph 0 1,
L;ul = Llul + Z, EI(xl + z) = Elxl for a z E ker D then (L;,Ef) E G[Od. If
(L, E) E G[O], (u, x) E graph 0, L'u = Lu + z, E'(x + z) = Ex for a z E ker D
then (L',E') E G[O].
Proof. By our assumptions, Dz = 0, Xr = Lrur . Let y = Erx r . Hence
DE~(xr + z) = DErxr = (Erxr)Dx r = E~(xr + z)D(xr + z),

i.e. y = L~ur and E~(xr+z) satisfy the right basic equation. Thus y E dom Or. By
our assumption, Lr is invertible and L~ur = Lru r + z. This means that for every
fixed z E ker D and for every y E dom Or the equation Lru r = y - z has a unique
solution, which implies that L~ is invertible. Since y = L~ur = Lru r + z = Xr + z,
we have E~y = E~(xr + z) = Ur = Erxr and y = L~E~(xr + z) = L~E~y, Ur =
E~(xr + z) = E~y = E~L~ur' We therefore conclude that (L~,E~) E G[Or]' A
similar proof for left logarithmic and antilogarithmic mappings and for logarithmic
and antilogarithmic mappings. •
Proposition 2.10. Let either X E Lgr(D) or X E Lgl(D) or X E Lg(D). Let
(Ln Er) E G[Or], (LI, E I) E G[Od, (L, E) E G[O], respectively. Suppose that F
is an initial operator for D corresponding to an R E RD. Let L~ = Rm Dm L r ,
E~ = ErRmDm, L; = RmDmLI, E: = EIRmD m , L' + RmDmL, E' = ERmDm
for an arbitrarily fixed mEN, respectively. Let
respectively. Then FDiL~ = 0, FDiL; = 0, FDi L = 0 (j = 0,1, ... ,m - 1) and

(L~,E~) E G[Or]' (L;,E;) E G[Od, (L',E') E G[O] and
L
m-l m-l
L~ = (I - RjFDj)L r , E~(I - L Rj FDj) = Er;
i=O j=O
18 Chapter 2
m-l m-l
L: = (I - L RjFDj)L/, EJ(1 - L Rj FDj) = E/;

j=O j=O
m-l m-l
L' = (I - L RjFDj)L, E'(1 - L RjFDj) = E;
j=O j=O
respectively.
Proof. Let j = 0,1, ... ,m - 1. Since FR = 0, by definition, we get FDjL~ =
F Dj Rm Dm Lr = F Rm-j Lr = O. By the Taylor Formula,
m-l
RmD m = 1 - L RjFDj on dom Dm.
j=O
Hence for Ur E dom Or, Xr = LU r E dom Dm we have
L~E~Rm Dmx r = L~Erxr = R m D m LrErx r = R m Dmx r .

Similar proofs for left logarithmic and antilogarithmic mappings and for logarith-
mic and antilogarithmic mappings. •
Propositions 2.8, 2.9, 2.10 together imply the following
Corollary 2.4. Let X E Lgr(D) (X E Lg1(D), X E Lg(D), respectively). Then
right (left) logarithms and antilogarithms (logarithms and antilogarithms, respec-
tively) are uniquely determined up to a constant. These constants are additive for
right (left) logarithms and logarithms and multiplicative for right (left) antiloga-
rithms and antilogaritms.
Definition 2.5. Let F be an initial operator for D corresponding to an R E RD.
Let mEN be arbitrarily fixed. Let X E Lgr(D). Let (Lr,Er ) E G[Or]. If
FDj Lr = 0 for j = 0,1, ... , m - 1, then (Lr' Er) is said to be m-normalized by
R and we write (Lr,E r ) E GR,m[Or]. Let X E Lgl(D). Let (L/,E!) E G[Od. If
FDjLt = 0 for j = O,I, ... ,m -1, then (L/,Et) is said to be m-normalized by
R and we write (Lt, Et) E GR,m[Od. Let X E Lg(D). Let (L, E) E G[O]. If
F Dj L = 0 for j = 0,1, ... , m - 1, then (L, E) is said to be m-normalized by Rand
we write (L,E) E GR,m[O]. 0
Corollary 2.5. Let X E Lgr(D) (X E Lg1(D), X E Lg(D), respectively). Let
F be an initial operator for D corresponding to an R E RD. Then always exist
(Lr,Er ) E GR,l[Or], (Lt,E/) E GR,d0tl, (L,E) E GR,dO].
Proof Let (Lr,E r ) E G[Or] be given. If (Lr,Er) f/. GR,l[Or] then set L~ =
(1 - F)Lr, Er = E~(I - F). Proposition 2.10 for m = 1 implies that (L~, E~) E
GR,dOr]. The same prooffor (L/,E1) and for (L,E). •
Theorem 2.2 . If X E Lgr(D) then there are R E RD and (Lr,Er ) E GR,dOr]
such that for u, v E 1(X) n dom Or and x = Lru, y = Lrv we have
(2.9) Lr(uv) = cDLrv + R[CDV-l(DLru)v + V-lU- l fD(U, v)],

Basic equation 19
(Erx) (ErY)
(2.10) = Er{CDY + R[CD(Ery)-l(Dx)(Ery)
+ (Ery)-l(Erx)-l fD(Erx, Ery))}.
If X E Lgl(D) then there are R E 'RD and (LI,EI) E GR,dnd such that for
u, v E leX) n dom nl and x = Lu, y = Lv we have
(EIX) (ElY)
(2.12) = EI{CDX + R[CD(EIX)-l(Dy)(EIX)
+ fD(EIX, ElY)(Ely)-l(EIX)-l]}.
If X E Lg(D) then there are R E 'RD and (L, E) E G R,l [n] such that for u, v E
leX) n dom n and x = Lu, y = Lv we have
(2.9') L(uv) = cDLu + R[CDDLv + fD(U,V)V-1U- 1)],
(Ex)(Ey)
(2.10') = E{CDX + R[CDDy
+ fD(Ex, Ey)(Ey)-l(Ex)-l]}.
Proof. Let u, v E leX) n dom nr . Then Du = uDLru, Dv = vDLrv and, by

Corollary 2.5,
DLr(uv) = (UV)-l D(uv) = v-1u-1{[CD[uDv + (Du)v] + fD(U,V)}

= CD[(V- 1Dv + v-1U-1(Du)v] + v-1u- 1fD(U, v)
= CD [DLrv + v-1(DLru)v] + V-1u- 1fD(U, v)
= D{CDLrv + R[CDV-1(DLru)v + V-1U- 1fD(U,V))}
for an R E 'RD. Moreover, if x = Lru, y = Lrv, then
= uv = ErLr(uv)
= Er{cDLru + R[CDV-1(DLru)v + V-1U- 1fD(U, v))}
= Er{CDX + R[CDY + (Ery)-l(Dx)(Ery)
+ (Ery)-l (ErX)-l fD(Erx, Ery))}.
Similar proofs for left logarithms and antilogarithms. Let X E Lg(D). Then, by
the commutativity and (2.9) (or (2.11)), v-1(DLu)v = v-1vDLu = DLu. This
implies (2.9') and (2.10'). •
20 Chapter 2
Corollary 2.6. Suppose that X E Lgr(D) (X E Lgl(D), X E Lg(D), respec-

tively). Let (Lr,E r ) E G[Or], (L/,Et) E G[Od, (L,E) E G[O). Then
(I - 2cD)DL r e,
(2.13) fD(e, e) = { (1 - 2cD)DL1e,
(1 - 2cD)DLe,
respectively. Moreover,
(2.14) DLre = DLle = DLe = De, provided that CD i- ~.
Proof. Since De = De 2 = cv[eDe + (De)e + fD(e, e)) = 2c DDe + fD(e, e), we find
(2.15) (1- 2cD)De = fD(e,e).
Now, applying the proof of Theorem 2.2 in the case u = v = e, we get Formulae
(2.13) and (2.14). This, and (2.15) together imply for CD i- ~ that DLre =
DLle = DLe = I_leD fD(e, e) = De. •
Corollary 2.7. If X E Lgr(D) then there are R E RD and (Lr,Er ) E GR,dOr)
such that
(2.16) Lru = cDLre + R[cDDLru + u- I fD(U, e)) for u E J(X) n dom Or.
If X E Lg1(D) then there are R E RD and (Ll,Et) E GR,dOd such that
(2.17) L1u = cDL1e + R[CDDLIU + fD(e,u)u- I ) for u E J(X) n dom 0 1,

If X E Lg(D) then there are R E RD and (L, E) E G R,I [OJ such that
(2.16') Lu = cDLe + R[CDDLu + u- I fD(U, e)J for u E I(X) n dom O.
Proof. Putting v = e in (2.9), we get (2.16). Putting v = e in (2.9'), we get (2.16').

In order to obtain (2.17), firstly we put u = e in (2.11). Putting in the obtained
formula v = u, we get (2.17). •
Proposition 2.11. If X E Lgr(D) then there are R E RD and (Lr,E r ) E
GR,I[OrJ such that
(2.18) (1 - cD)Lru = cDLre + R[u- I fD(U, e)) for u E I(X) n dom Or.
If X E Lg 1(D) then there are R E R D and (LI' EI) E G R,dOd such that
(2.19) (1 - cD)LIU = cDLle + R[fD(e, u)u- I ) for u E J(X) n dom 01.

If X E Lg(D) then there are R E RD and (L, E) E G R,I [OJ such that
(2.18') (1- cD)Lu = cDLe + R[u- I fD(u,e)) for u E I(X) n dom O.

Basic equation 21
°
Proof. Let F be an initial operator for D corresponding to R. Since (Ln Er) E
GR,l[Or), we have FLru = for U E I(X)ndom Or. Since F = I -RD on dom D,
by Formula (2.18) we get for U E I(X) n dom Or
cDLre + R[u- 1 fD(U, e)] = (I - cDRD)Lru = (I - cDI + cDF)Lru = (1- cD)Lru,
i.e. Formula (2.18). Similar proofs for Formulae (2.19), (2.18').

Proposition 2.12. Let F be an initial operator for D corresponding to an
•
R E RD. Let either X E Lrg(D) and (Lr, Er) E GR,dOr] or X E Lg1(D)
and (L/,Et) E GR,dOd or X E Lg(D) and (L,E) E GR,dO]. Then
(i) (1- 2cD)L r e = RfD(e, e), (1- 2cD)L/e = RfD(e, e), (1- 2cD)Le = RfD(e, e),
respectively;
(ii) CD = 1/2 implies fD(e, e) = 0;

(iii) CD =I- 1/2 implies Lre = (1 - 2CD)-lRfD(e,e) and FLre = 0, L/e =
(1- 2CD)-lRiD(e,e) and FL/e = 0, Le = (1- 2CD)-lRiD(e,e) and FLe = 0,
respectively;
(iv) CD =I- 1/2 and IDee, e) = ° implies Lre = 0, L/e = 0, Le = 0, respectively;

(v) CD =I- 1/2 and e E ker D implies Lre = 0, L/ = 0, Le = 0, respectively;
(vi) Lre = e + z, L/e = e + z, Le = e + z, respectively, where z E ker D;
(vii) if X is a Leibniz D-algebra then Lre = 0, L/e = 0, Le = 0, respectively;
(viii) (1 - 2cD)De = IDee, e);

(ix) CD =I- 1/2 implies De = (1 - 2CD)-1 IDee, e);
(x) CD =I- 1/2 implies Lre = e - Fe, L/e = e - Fe, Le = e - Fe, respectively.
Proof. Suppose that X E Lgr(D). Put u = e in (2.18). Then (1 - cD)Lre =
cDLre + RiD(e, e), which implies (i). Acting on both sides of (i) by the opera-
tor F we get (1 - 2cD)FLre = FRiD(e,e) = 0. If CD = 1/2 then iD(e,e) =
DRiD(e, e) = D(I- 2cD)L r e = 0. Let CD =1-1/2. Since (Lr,E r ) is I-normalized,
°
we have F Lre = and
1 1
Lre = 1
-
2
CD
RiD(e, e) + F Lre = 1
-
2
CD
RiD(e, e).
If, in addition, fD(e,e) = °

then Lre = 0. If CD =I- 1/2 and e E kerD, then
Lre E ker D, which implies Lre = FLre = 0. By definition, De = eDLre, which
implies D(Lre - e) = 0. Hence Lre = e + z, where z E ker D. If X is a Leibniz
D-algebra then CD = 1 and e E ker D. Hence, by (v), Lre = 0. Since De = DLre,
(i) implies IDee, e) = DRfD(e, e) = (1 - 2cD)DL r e = (1 - 2cD)De. Thus for
CD f. 1/2 we get (ix). Moreover, Le = l-~CD RfD(e, e) = RDe = e - Fe. Similar
proofs if X belongs either to Lgl(D) or to Lg(D). •
22 Chapter 2
Theorem 2.3. Let X E Lgr(D) and let (Lr,Er ) E GR,dOr] for an R E RD.
Then the mapping
(2.20) gD,r(U) = u- 1 fD(U, e) for U E leX) n dom Or
is constant if and only if CD = 1. Let X E Lg1(D) and let (LI, E I ) E GR,dOd for
an R E RD. Then the mapping
(2.21) gD,/(U) = fde, u)u- 1 for U E leX) n dom 0 1

is constant if and only if CD = 1. Let X E Lg(D) and let (L, E) E G R,l [0] for an
R E RD. Then the mapping
(2.20') gD(U) = u- 1 fD(U, e) = fD(e, u)u- 1 for u E leX) n dom 0
is constant if and only if CD = 1.

Proof. Suppose that X E Lgr(D), CD :j:. 1 and gD,r is constant, i.e. there is
an element a E X such that gD,r(U) = a for all U E leX) n dom Or. Then, by
Proposition 2.11,
1
Lru = --(cDLre + Ra) for all U E leX) n dom Or,
I-CD
i.e. Lr is constant. This contradicts our assumption that Lr is invertible.

Conversely, put CD = 1 in Formula (2.18). Then we get RgD,r(U) = -Lre, which
implies gD,r(U) = DRgD,r(U) = -DLre = -De for all U E leX) n dom Or, which
means that gD,r is a constant mapping. Similar proofs for gD,1 and for gD. •
Theorem 2.3 gives us a good criterion for checking which selectors of Or, 0 1 and
o are invertible.
Lemma 2.1. Let D E A(X). Let X have unit e E dom D. Then CD = 0 if and
only if fD(x,e) = Dx for all x E dom D.
Proof. By our assumptions,
D(xy) = cD[xDy + (Dx)y] + fD(X,y) for x,y E dom D.
Recall that, by definition of the (proper) non-Leibniz component, this representa-

tion is uniquely determined. Putting y = e, we get
Dx = D(xe) = cD[xDe + (Dx)e] + fD(X, e) = CD (xDe + Dx) + fdx, e).

Then
(1- cD)Dx = cDxDe + fD(x,e) for x E dom D.
If CD = 0 then we obtain fD(X, e) = Dx.
Basic equation 23
Conversely, suppose that fD(X, e) = Dx for x E dom D. Then
(1- cD)Dx = cDxDe + Dx, i.e. - cDDx = cDxDe.
If CD -=I- 0 then Dx = xDe for all x Edam D. On the other hand, if x = e, then we
=
have -cDDe cDeDe = =
cDDe, i.e. 2cDDe O. Since CD -=I- 0, we find De O.=
This implies Dx = xDe = 0 for all x E dom D, which means that D is a trivial
operator. This contradicts our assumption. Hence CD O. = •
Theorem 2.4. Suppose that either X E Lgr(D) or X E Lg1(D) or Lg(D) and
CD = O. Let a E IF\ {O}. If a -=I- 1 then au f/. dom Or whenever u E dom Or (au f/.
dom Ot whenever u E dom Ol, au f/. dom 0 whenever u E dom 0, respectively).
In other words: If CD = 0 then intersections of the domains of the multifunctions
Or, 0 1, 0 with lines passing by zero consist of one point only.
Proof. By Corollary 2.5, there are (Lr' Er) E GR,t[Or] for an R E RD. Suppose
that a E IF \ {O} and u E dom Or. By our assumption, au E dom Or. Since
CD = 0, Formula (2.18) implies that Lru = RgD,r(u). Hence
Lr(au)
= RgD,r(au) = R[(au)-l fD(au, e)]
= R[a-1au-1fD(u,e)] = R[u-1fD(u,e)] = RgDAu) = Lru.
Hence u = ErLru = ErLr(au) = au which implies a = 1. Similar proofs for 01
andbO. •
Corollary 2.8. Suppose that all assumptions of Theorem 2.4 are satisfied. Then
-e f/. dom On -e f/. dom 0 1, -e f/. dom O.
Proposition 2.13. Let CD -=I- O. Let X E Lgr(D) and let (Lr, Er) E GR,l[O] for
an R E RD. Then u 2 E leX) n dom Or if and only if u E leX) n dom Or. If it is
the case then
(2.22)
Conversely, the equation
(2.23) wu+uw =v
with v = ci/u[DLru2 - u- 2fD(U, u)] has a solution w = Lru. Let X E Lg1(D)
and let (LI,Et) E GR,l[O] for an R E RD. Then u 2 E J(X) n dom Ol if and only
if u E leX) n dom Ol. If it is the case then
(2.24)
Conversely, the equation (2.23) with v = cZ;1[DL/u2 - R[fD(U,U)U- 2]u has a

solution w = LIU.
24 Chapter 2
Proof Let n = 2. Sufficiency is already proved by Theorem 2.2 (without the

assumption that CD =f. 0).
Necessity. Suppose that u 2 E dom nr. It means that there is an x E dom D such
that Du 2 = u 2 Dx. Let w = DLru. Then
WU + uw= (DLru)u + uDLru

= u[DLru + u-1(DLru)u]
= c1:/u[Dx - u- 2 fD(U, u)] = v.
A similar proof for left logarithms.
In the commutative case we obtain
•
Corollary 2.9. Suppose that CD =f. 0, X E Lg(D) and (L, E) E GR,dn] for an
R E RD. Let n 2: 2. Then un E I(X) n dom n if and only if u E I(X) n dom n.
In that case,
n-2 n-2
Lu n = (CD L ct + 2cv)Lu + R[u- n L uk fD(U, u n- 1- k )].
k=O k=O
Proof Let n = 2. Sufficiency is already proved by Theorem 2.2 (without the

assumption that CD =f. 0).
Necessity. Suppose that u 2 E dom n. This means that there is an x E dom D
such that Du 2 = u 2 Dx. Let y = c~ [x - Ru- 2 fD(U, u)], where R E RD. Clearly,
y E dom D and
= u- 2 Du 2 = u- 2 [2cDUDu + fD(U, u)]

= 2CDU- 1Du + u- 2 fD(U, u),
which implies Dy = u- 1 Du. Hence u E dom n.

Using the same arguments, by an easy induction we get the required formula for
all n 2: 2 •
Note 2.1. Corollary 2.9 permits us to calculate nth roots of elements u E
In(dom n), i.e. elements u 1 / n , by means of their logarithms in the same man-
ner as in the classical case. Similarly, Proposition 2.13 permits us to calculate
square roots of elements u E 12 (dom nr), 12 (dom n,), i.e. elements U 1 / 2 by their
logarithms. It is also possible to calculate nth roots of those elements by means
of Theorem 2.2, however the complexity of calculation is growing together with
n. It should be also pointed out an essential role of Equation (2.23). In the
commutative case, this equation is reduced to the equation w = ~U-lv, which
leads to the formula: Lu = 2;D R[DLu 2 - u- 2 fD(e, e)]. 0
Basic equation 25
Theorem 2.5. Let a E IF\ {O}. Let CD =1= O. If X E Lgr(D), (Lr,Er ) E GR,dOr]
for an R E RD and u E J(X) n dom Or then au E J(X) n dom Or and
If X E Lgl(D), (L/,El) E GR,dOd for an REnD and u E J(X) n dom Ol then

au E J(X) n dom Ol and
Ll(au) = Llu + LI(ae) - Lie = Llu + z, where z E ker D.
If X E Lg(D), (L,E) E GR,l[O] for an R E RD and u E J(X) n dom 0 then

au E J(X) n dom 0 and
L(au) = Lu + L(ae) - Le = Lu + z, where z E ker D.
Proof. Suppose that a E IF \ {O} and u E J(X) n dom Or. Since e E dom Or, we
conclude that ae E dom Or. Indeed, by definition,
which implies Lr(ae) = e + z', where z' E ker D. Again by definition, DLre =
eD Lre = De, which implies Lre = e + zit, where zit E ker D. This, and Formulae
(2.9), (2.16) together imply that
Lr(au) = Lr(aeu) = cD[Lru + Lr(ae)] + R[(ae)-lu- 1 !D(ae, u)] =
= cD[Lru + Lr(ae)] + R[u- 1 !D(U, e)] = cDLru + cDLr(ae) + RgD,r(U) =

= Lru - Lre + Lr(ae) = Lru - e - zit + e + z' = Lru + z,
where z = z' - zit E ker D. Similar proofs for left logarithms and logarithms. •
Corollary 2.10. Let CD =1= o. If X E Lgr(D), (Lr, Er) E GR,l[Or] for an R E RD
and u E J(X) n dom Or, then -u E J(X) n dom Or and
Lr( -u) = cD(Lru + Lre) + RgD,r(U).

In particular, -e E J(X) n dom Or and Lr( -e) = Lre. If X E Lg1(D), (Ll' Ez) E
GR,d0zj for an REnD and u E J(X) n dom 0 1, then -u E J(X) n dom 01 and
LI(-u) = cD(Llu + Lie) + RgD,I(U).

In particular, -e E J(X) n dom 0 1 and L/(-e) = Lie. If X E Lg(D), (L,E) E
G R,l [OJ for an R E RD and u E J(X) n dom 0, then -u E J(X) n dom 0 and
L(-u) = cD(Lu + Le) + RgD(U).

In particular, -e E J(X) n dom 0 and L( -e) = Le.
26 Chapter 2
Proof. By Theorem 2.5, -e E leX) n dom Or. By our assumption, if F is an

initial operator for D corresponding to R, then RDLr = (I - F)L r = L r . Hence
Formulae (2.9) and (2.13) together imply that
= Lr[(-e)(-e)]
= cDL r ( -e) + R[CD( _e)-l DLr( -e)( -e) + (_e)-2 fD( -e, -e)]
= cDL r ( -e) + cDRDL r ( -e) + RfD(e, e) = 2CDL r (-e) + RfD(e, e)
= 2c DL r (-e) + (1- 2cD)L re,
which implies 2CDL r ( -e) = 2cDL re. Since CD =I 0, we get Lr( -e) = Lre. Suppose
that U E leX) n dom Or. By Formula (2.9),
Lr( -U)
= Lr[u(-e)]
=cDL r ( -e) + R[CD( -e)-l(DLru)( -e) + (_e)-lu- 1fD(U, -e)]
= cDLr(-e)+ cDRDLru + R[u- 1 fD(u,e)]
=cD[Lru + Lre] + RgD,r(U).
Similar proofs for left logarithms and logarithms.
Corollary 2.11. Suppose that IF = <C and CD =I O. If X E Lgr(D), (Lr' Er) E
•
GR,dOr] for an R E RD and u E leX) n dom Or, then iu E leX) n dom Or and
Lr(iu) = ~(LrU + Lre) + 2~D R[gD,r(U) - u- 2 fD(U, u)].
If X E Lg1(D), (LI,EI) E GR,d0tl for an R E RD and U E leX) ndom 0 1, then

iu E leX) n dom 0 1 and
Ll(iu) = ~(L1U + Lie) + 2~D R[gD,I(U) - fD(U,U)U- 2].
If X E Lg(D), (L,E) E GR,l[O] for an R E RD and U E I(X) n dom 0, then

iu E leX) n dom 0 and
L(iu) = ~(Lu + Le) + 2~D R[gD(U) - fD(U, u)u- 2].
Proof. Corollaries 2.7 and 2.10 together imply that
cD(Lru + Lre) + RgD,r(U)

= Lr( -u) = Lr[(iu)(iu)]
=2CDLr(iu) + R[(iu)-2 fD(iu, iu)]
=2CDLr(iu) + R[u- 2fD(U, u)].
Basic equation 27
Since CD =I- 0, we get the required formula for Lr(iu). Similar proofs for left
logarithms and logarithms. •
Write for A E L(X)
(2.25) VA"(X) = {a EX: 1 - Aa is invertible},
(2.26) vl(x) = {a EX: 1 - aA is invertible},
(2.27) WA(X) = {a EX: 1 - Aa+- is invertible},
where Aa+-w = A(wa) for all a E X (d. Formula (1.15).

Clearly, if the algebra X under consideration is commutative, then WA(X)
vr(x). Moreover, Ae E Vl(X) n Vl(X) for all A E VIFA, where
(2.28) VIFA = {O =I- A ElF: 1 - AA is invertible}.
By definition, A E vlFA whenever 1/ A is a regular value of the operator A.

Proposition 2.14. Suppose that D E R(X)n A(X), Xl E dom nil, x; = DXI E
VR(X) for an R E RD, F is an initial operator for D corresponding to Rand El
is a selector of of nl . Then
Moreover, EIX =I- 0 whenever x E dom nl ·

Proof The operator (I - RxD- l F is well-defined by our assumptions. Since
0= DFEI = D(1 - RxDEI = DEI - DRx;EI = DEI - X;EI' for Xl E dom nil
we have DE/Xl = (EXI)Dxl = XjEIXI. Hence the left basic equation is satisfied.
Moreover, for every Xl E dom ni there is a unique Ul such that LIUI = Xl, namely,
U/ = E/Xl. Suppose that E/x = 0 for an X E dom nl. Then X = Ll(E1X) = L10, a
contradiction with Proposition 2.5. •
Proposition 2.15. Suppose that D E R(X)n A(X), Xl E dom nil, x; = Dx/ E
VA(X) for an R E RD, F is an initial operator for D corresponding to R and El
is a selector of of nl . Then
Moreover, Elx =I- 0 whenever X E dom nl .

Proof By our assumptions, we have DEI = (1 - x;R)-1(x;FE1) + DFEI =
(I - X;R)-1(x;FE1). This implies x;FEI = (I - x;R)DEI = DEI - x;RDEI =
DE1-X;(1-F)EI =DEI-X;EI+x;FEI. Hence DEI =X;El. Further arguments
are the same as in the proof of Proposition 2.14. •
28 Chapter 2
Proposition 2.16. Suppose that D E R(X)nA(X), Xr E dom 0;1, X~ = DX r E

WR(X) for an R E RD, F is an initial operator for D corresponding to Rand Er
is a selector of Or. Let the operator Rx~+- be defined as in Formula (2.27), i.e.
Rx~+-w = R(wx~) for wE X. Then
and (Lr,Er)EG[Or]' Er=L;l.

Moreover, Erx "I 0 whenever x E dom 0;1.
Proof. By our assumptions, for all x E dom Or we have DErx = (Erx)Dx =
(Erx)x~ = (FErx + RDErx)x~. Thus
which implies DErx = (I - R xJ- 1[(FErx)x~. Hence Erx is of the required form.
Further arguments are the same as in the proof of Proposition 2.14. •
As it was mentioned, the set WA(X) defined by Formula (2.27) has not been
considered in the commutative case. The following fact (used in the proof of
Proposition 2.16) has not been stated before:
Proposition 2.17. Suppose that D E R(X)nA(X), F is an initial operator for
D corresponding to an R E RD and a E WR(X), Then Dv = va if and only if
v = R(I - Ra+-)-l[(Fv)a] + Fv.
Proof. Necessity. Dv = va = (Fv+RDv)a, which implies (Fv)a = Dv-(RDv)a =
(I - Ra+-)Dv. By our assumption, the operator J - Ra+- is invertible. Hence
v = RDv + Fv = R(J - Ra+-)-l[(Fv)a] + Fv.
Sufficiency is proved by checking.
Proposition 2.18. Suppose that X is a D-algebra with CD "11. Let the mappings
•
gD,T) gD,I, gD defined by Formulae (2.20), (2.21), (2.20'), respectively, be not
constant. Let the mappings A r , AI, A be defined by the following formulae:
1
Aru = -l--[cDe + RgD,r(u)]
-CD
for u E J(X) n dom Or, where R E RDi
1
AIU = -1--[cDe
-CD
+ RgD,I(U)] for U E J(X) n dom 0 1, where R E RDi
1
Au = -1--[cDe
-CD
+ RgD(U)] for U E J(X) n dom 0, where R E RD.
If Ar is invertible then there are (Lr' Er) E GR,dOr] such that Lr = A r , Er = A;l,
i.e. X E Lgr(D). If Al is invertible then there are (Lt,EI) E GR,d0tl such
that Ll = AI, El = A,l, i.e. X E LgJ(D). If A is invertible then there are
(L,E) E GR,l[Od such that L = A, E = A-1, i.e. X E Lg(D).
Basic equation 29
Proof By definition, we have
1
DAru = --[cDDe + gD,r(U))
1-CD
far u E J(X) n dom nr .
On the other hand, Du = D(ue) = cD[uDe + (Du)e) + fD(u,e), which implies

(1 - cD)Du = cDuDe + fD(U, e). Since CD =I- 1 we get
u- 1 Du = _I_[cDe + u- 1 fD(u,e) = _I_[cDDe + gD,r(U)) = DAru.

1- CD 1- CD
Hence x = Aru E dom n;1 and Du = uDx. This implies that there are (LT' Er) E
G[nr) such that x = Lru, u = Erx, i.e. Lr = Ar, Er = A;I. By definition,
(Lr. Er) is I-normalized. Similar proofs for Al and A. •
Example 2.2. Let X be a commutative D-algebra with the product rule
(2.29) D(xy) = (Fx)Dy + (Fy)Dx far x,y E dom D,
where F is an initial operator for D corresponding to an R E RD. In the classical

case of the operator D = ft
in the space X = era,
b) over IR with the pointwise
multiplication and with R = J:, i.e. with (Fx)(t) = x(a) for x EX, this is the
so-called point derivation. Clearly, here CD = 1 and fD(X, y) = (Fx - x)Dy +
(Fy - y)Dx. We shall consider two cases:
(i) e E ker D. Then Fe = e and fD(u,e) = 0 for u E dom D. Hence the mapping
gD is constant.
(ii) By (2.29), for x = y = e we get De = De 2 = 2(Fe)De, which implies
(e - 2Fe)De = O. Suppose that 2Fe = e. Then RDe = e - Fe = 2Fe - Fe =
Fe = e - RDe. Hence RDe = ~e, which implies De = ~De. Thus De = 0 and
e E kerD as in (i).
If neither e E ker D nor e - 2Fe = 0, however (e - 2Fe)De = 0, we do not
know whether the mapping gD is constant. On the other hand, it is well-known
that dom D = RX E9 ker D. Suppose that x,y E RX. Then Fx = Fy = 0 and
D(xy) = O. Hence RX f/. Lg(D). If x, y E ker D then Fx = x and Fy = y. Hence
ker D is a Leibniz D-algebra. 0
Example 2.3. Suppose that D E A(X) has unit e. Algebraists consider usually
instead of the non-Leibniz component fD used here (cf. PR[8) and other author's
papers) the following bilinear form for D:
HD(x,y) = D(xy) - xDy - (Dx)y far x, y E dom D.
Note that in their considerations dom D = X. We shall compare properties of HD

and fD in order to show why for our purposes it is better to apply fD.
(i) Clearly, D satisfies the Leibniz condition if and only if HD = O.
30 Chapter 2
(ii) HD(x,e) = -xDe; HD(e,X) = -(De)x for all x E dom D. In particular,

HD(e, e) = -De.
Indeed, HD(X, e) = D(xe) - xDe - (Dx)e = Dx - xDe - Dx = -xDe. A similar
prooffor HD(e,x). Putting x = e, we get HD(e,e) = -eDe = -De.
(iii) Let X E Lgr(D) n LgI(D). Let
GD,r(X) = x-1HD(x,e); GD,I(X) = HD(e,x)x- 1
Jor x E J(X) n dom Or n dom 0l.

Then GD,r(x) = xx- l HD(e,e) = -De, GD,I(X) = HD(e,e)x-lx = -De, i.e.
both mappings G D,r and G D,l are always constant. Theorem 2.3 showed us that
the corresponding mappings 9D,r, 9D,1 induced by JD are constant if and only if
CD = l.
(iv) CD = 1 if and only if HD = JD.
Indeed, for all x, y E dom D we have
HD(XY) = D(xy) - xDy - (Dx)y = CD [xDy + (Dx)y]- xDy - (Dx)y + JD(X, y) =
= (CD -l)[xDy + (Dx)y] + JD(X,y).

This implies (iv).
(v) Suppose that DE A(X) and consider the Duhamel condition: D(xy) = xDy =
yDx for x, y E dom D. Clearly, we have now CD = ~ and JD = 0 (cf. PR[8]).
By definition, for all x, y E dom D we have H(x, y) = D(xy) - xDy - yDx =
xDy - yDx - yDx = -yDx. Let y = e. Then Dx = -H(x,e) = -xDe.
Suppose now that x E J(X) n dom 0 and L is a selector of O. Then, by definition,
DLx = X-I Dx = x-1xDe = De, which implies Lx = RDe + z, where z = Fx E
ker D. Suppose that L is an invertible selector of O. This contradicts Theorem
2.3, since CD :f:. 1 and fD = O. Hence logarithmic mappings do not exist.
(vi) Let D E M(X), i.e. D(xy) = (Dx)(Dy) for x, y E dom O. Then CD = 0
and fD(X,y) = (Dx)(Dy). Moreover, De = e (cf. Chapter 6). Hence 9D,r{X) =
X-I Dx = DLrx, 9D,I(X) = (Dx)x- l = DL1X, where Lr, Ll are selectors of Or, Oz,
respectively. On the other hand, G D,r{X) = X-I HD(x, e) = -xx-IDe = -e and
GD,I(X) = HD(x,e)x- 1 = -(De)xx- l = -e.
Points (iii)-(vi) show us that the mapping HD does not distinguish between alge-
bras in the same manner as it is done by means of JD. 0
Example 2.4. Let X be a non-commutative algebra over F. Let a, bE X do not
vanish simultaneously. Write Dx = xa + bx for x E X. Clearly, dom D = X,
hence D E A(X). We find for all x, y E X
fD(X, y) = D(xy) - xDy - (Dx)y = xya + bxy - x(ya + by) - (xa + bx)y =
Basic equation 31
= xya + bxy - xya - xby - xay - bxy = -x(a + b)y.

Thus here CD = 1 and fD(X,y) = -x(a+b)y.
Let x, y E X be arbitrary. Denote by [x, y] and lx, y[ the commutator and anti-
commutator, respectively, i.e. [x, y] = xy - yx and lx, y[= xy + yx.
Consider some particular cases.
(i) b = -a, hence Dx = xa - ax = [x, a]. Since fD(X, y) == 0, X is a Leibniz
algebra (d. also Theorem 4.3). Suppose that X has unit e and a is an involution,
i.e. a2 = e. Then x E kerD, i.e. [x,a] = 0, if and only if x =]y,a[, y E X (d.
PR[5]). This means that an element x commutes with a if and only if x is an
anticommutator with a.
This result can be generalized. Namely, consider any proper two-sided ideal J in
i.e. such a that a2 = e + g, g E J. Then in the
X and an almost involution a,
quotient algebra XI J the coset
only if x =]y, a[+h, y E X, hE
°
[a] = a + J is an involution. Thus [x, a] = if and
J (d. PR[G]).
(ii) b = a, hence Dx = xa + ax =lx, a[ and fD(X, y) = -2xay (d. Equation
°
(2.3». Similarly, as in (i), suppose that X has unit e and a is an involution. Then
lx, a[= if and only if x = [y, al, y EX. This result is dual to the previous one in
(i) and also can be generalized for almost involutions (d. PR[G]).
(iii) Let X have unit e. Let Dx = xa + bx, where a, b are involutions. Then
x E ker D, i.e. xa + xb = 0, if and only if x = ya - by, y E X. A necessary and
sufficient condition for the equation xa + bx = y to have a solution for ayE X
is that ya - by = 0. If it is the case, then x = !ya + za + bz, where z E X is
arbitrary.
Indeed, suppose that x E ker D, i.e. xa = -bx. Write y = !xa. Since a2 = e = b2 ,
we get ya = !xa 2 = !x and by = !b(xa) = -!b2 x = -~x. Thus x = ya - by.
Sufficiency is proved by checking. Suppose that xa + bx = Y EX. By our
assumptions,
ya - by = (xa + bx)a - b(xa + bx) = xa 2 - bxa + bxa - b2 x =X- X = 0.
On the other hand, let ya - by = 0. Put x = ha. Then
°
Thus the condition ya - by = is sufficient for the equation xa + bx = y to
have solutions. Moreover, !ya is a particular solution of this equation. Clearly
za + bz E ker D for any z EX, which implies the required form of the general
solution to the equation Dx = y.
A duality of the above statements is easy to see if we let: D+x = xa - bx for
x EX. Then they may be rewritten as follows:
32 Chapter 2
Observe that this implies DD+ = 0 and D+D = O.

Selectors of Dr and Dl (up to a constant) are of the form
Lru = e + u-1bua if and only if [u, b] = 0, u E J(X),
LlU = e + buau- 1 if and only if [u, a] = 0, u E J(X).
Indeed, by definition, for u E J(X) we have DLrx = u-1Du = u-1(ua + bu) =

a + u-1bu. This implies that (up to a constant, i.e. element of the form za + bz,
z E X) we have
provided that the necessary and sufficient condition u-1(Du)a - bu- 1Du = 0 is
satisfied. Thus
=u-1(ua + bu)a - bu-1(ua + bu)

=ua 2 + u-1bua - bu-1ua - b2 u
= e - e + u-1bua - bu-1ua
= (u-1b - bu-1)ua.
But u E J(X). Also a E J(X), since a2 = e. Hence we get u-1b - bu- 1 = 0,

i.e. ub - bu = O. We therefore conclude that the condition, we are looking for, is
[u, b] = o. A similar proof for left logarithms.
Suppose now that J is a proper two-sided ideal in X and a is an almost involution.
Similarly, as before, we prove that x E ker D if and only if x = ya - by + h, y EX,
hE J. Clearly, if b = -a, D is a commutator, if b = a, D is an anticommutator.
Observe that in such cases operators D are not left invertible, since ker D f:. {O},
hence also not invertible.
(iv) Let X have unit e and let either a = 8e, b = 0 or a = 0, b = 8e, where
8 E IF \ {O}. Then Dx = 8x and fD(x, y) = -28xy. If D E A(X) then X is the
so-called quasi Leibniz algebra (cf. DUDEK D[l], also PR[8]).
(v) In general, we have
D 2x = D(xa + bx) = (xa + bx)a + b(xa + bx) = xa 2 + b2 x + 2bxa.

Hence D2 is not of the form of D, since the component 2bxa.
A fundamental paper of LUMER and ROSENBLUM (cf. LR[l]) is concerned with
the solvability of the equation 2::;=0 ajxbj = y. Solutions of that equation in the
case of algebraic aj, bj with single characteristic roots have been given in PR[6]
(Chapter 17, Section 1; cf. also NGUYEN VAN MAU N[l]). Expressions of the
forms 'LJ=l (-l)jc!ta j xbJ and 'L J=l (-l)jc n a j xb n - j are considered, for instance,
Basic equation 33
by TONG (cf. To[l]) in the algebra of all bounded linear operators over a Hilbert
space. ApOSTOL and FIALKOV (cf. AF[l]) have defined in a Banach algebra A
operators of multiplication MA,B mapping A into itself as follows: MA,B(X) =
AX B for X E A, where A, B E A are given. Every finite sum of such operators:
S = L~l MAi,Bi is said to be an elementary operator. In their cited paper there
are studied structural properties of elementary operators. 0
The operator D considered in Example 2.4 may be invertible. Namely, we have

the following
Theorem 2.6. Let X be a non-commutative algebra (over IR or C) with unit e

and a complete linear metric space. Write for a EX, b E I(X)
L
00
(2.30) X(a, b) = {y EX: (_l)nb-(n+1)ya n is convergent}.

n=O
Let Dx = xa + bx for x EX. Then
L
00
(2.31) Dx =y for Y E X(a,b) if and only if x= (_l)nb-(n+1)ya n ,

n=O
i.e. the operator D is invertible on the subspace X (a, b). Moreover, selectors of
nr and nl are of the form
L
00
(2.32) Lru = (-1) n b-(n+1)(a + u-1bu)a n for u E dom nr,

n=O
L (_1)nb-(n+1)(uau- 1 + b)an
00
Llu = for u E dom nl ,

n=O
provided that a + u-1bu E X(a, b), a + u-1bu E X(a, b), respectively.
Proof. The statement (2.31) is well-known and can be easily checked. If u E

I(X) ndom nr and Dx = u- 1Du then xa+bx = Dx = u- 1Du = u-1(ua+ bu) =
a+u-1bu. This, and (2.31) together imply that Lru is of the required form (2.32).
Similarly, we find the form of Llu, since (Du)u- 1 = (ua + bu)u- 1 = uau- 1 + b. •
Example 2.5. Suppose that X is a non-commutative algebra (over IF) with unit
e. Let a E X be nilpotent of order N i.e. aN = 0 and a k ¥- 0 for k = 1, ... , N - 1.
Let Dx = xa + bx, where bE I(X). Then Dx = y, Y E X, if and only if
N-l
X =L (_l)nb-(n+1)ya n .
n=O
34 Chapter 2
Indeed, suppose that x is of this form. By our assumptions,
Dx = xa + bx
N-I
= L (_l)n[b-(n+I)ya n+I + b-nya n ]
n=O
N N-I
= - L (_l)nb- n ya n + L
(_l)nb- n ya n
n=l n=O
N-I
- (_l)Nb- N ya N + L
(_l)n[b- n ya n - b-nya n ] + y
n=l
=y.
Conversely, suppose that x satisfies the equation Dx = y. This implies that

x + b-1xa = b-1y. We multiple this equation by _b- l from left and by a from
right k times (k = 1, ... , N - 1). Then we get N equations:
x + b-1xa = b-1y, .......................................................................... .
+ (_l)kb-(k+I) xa k+l = (_1)kb-(k+ 1 )yak , ................. .

(_l)kb- k xa k
(_1)N- 1b-(N-l)xa N - 1 + (_1)N- 1b- N xa N = (_1)N- 1b- N ya N - 1,
i.e. (_1)N- 1b-(N-I)xa N - 1 = (_1)N- 1 b- N ya N - I .
If we add these N equations, then we get the required formula.
Similarly, if a is invertible and b is nilpotent of order N, we conclude that Dx =y
if and only if
N-I
X =L (_l)nb n ya-(n+I)n.
n=O
Observe that, due to the assumption that a is nilpotent, we obtain a result similar
to that given in Theorem 2.6 without any topological assumption about the space
X. It is so, because the series in (2.30) has only a finite number of terms. We
conclude that the operator D is invertible on the whole space X. In order to find
formulae for selectors of Or and OJ it is enough to take in Formulae (2.32) finite
sums with the first N terms. 0
Example 2.6. Denote by X = (s)JR the space of all real sequences {Xn}nEN with
the coordinatewise addition and multiplication by scalars. Let the multiplication
in X be defined as a convolution, i.e. let the product rule be of the form
n
X * Y = {Xn} * {Yn} = {L XkYn+l-d for x, Y E X.
k=l
Then X is a commutative algebra with unit e = {8In } and without zero divisors
(cf. for instance PR[8], Example 6.1.3a). Consider the operator D defined by the
Basic equation 35
equality: D{x n } = {nx n }. Clearly, D E I(X) and D-I{x n } = {~Xn}. Moreover,

for all x, y E X we have
x * Dy +y * Dx
n n
= {L:: xk(n + 1- k)YnH-d + {L kXkYnH-k}
k=l k=l
n
= {L [(n + 1 - k) + klxkYnH-d
k=l
n
= {(n + 1) L XkYn+l-d = D(x * y) + x * y.
k=l
Hence CD = 1 and the non-Leibniz component is fD(x, y) = X * y. The basic
equation Du = u * Dx has a unique solution x provided UI ::j; 0 and this solution
determines the logarithmic mapping, i.e. we have
1
L
n 1
Lu =x = {-
n
kUkVnH-k}, where VI = - and
k=l
Xl
Consider now the space the same space X with the coordinatewise multiplication.
Then X is a commutative algebra with unit e = {en}, where en = 1 for all
n E N and without zero divisors. Let D'{x n } = {Xn+l - x n } for x E X. Clearly,
D' E R(X) and the operator R' defined as follows: R'x = Y, where Yl = 0,
Yn = I:;';;;~ Xk for n 2: 2, is a right inverse of D'. Moreover,
ker D' = {z = {zn} EX: V zn
nElIi
= C E lR}.
The product rule is of the form
D'(xy) = xD'y + yD'x + (D'x)(D'y) for x,y E X,

i.e. X is a quasi Leibniz algebra. Since u = {un} E J(X) if and only if Un ::j; 0 for
all n E N, from the basic equation we get x = R'(u- l D'u) + z, where u E J(X),
Z E ker D' and the logarithmic mapping is Lu = x, where
n
Xl = C E lR, xn=L
k=O
o
Example 2.7. Suppose that X E Lg(D) and (L,E) E G[Ol. Consider the
Cartesian product
X = 'X- -x..... .x. -X,' where x = (Xl, ... ,xn ) E X for Xl, ... ,xn E X,
n-times
36 Chapter 2
with the coordinatewise addition, multiplication and multiplication by scalars.

Clearly, X is again a commutative algebra. Write
Ax = (AXl' ... , Ax n ) for any A E L(X), Xl, ... , Xn E dom A.
By definition, A E L(X). Moreover, U = (Ul, ... ,Un ) E J(X) if and only if
----
Ul, "',U n E J(X). Write also
n=ox ... xO.

n-times
It is easy to verify that (i, E) E G[n). Indeed, if U E J(X) then
Di = D(Lul, ... ,Lun)
which proves that i is a selector of n.

The invertibility of L immediately implies
the invertibility of L. Thus E = i-I. 0
Theorem 2.7. Suppose that X E Lgl(D), (Lt, Ed E G[Oz) and
(2.33) A E ML(X), dom A ~ dom D, DA = dAD on dom D, dE J(X).
H(u,x) E graph 0 1 then (Au, Ax) E graph 01 and LIAu = ALIU, EIAx = AElx.
Proof. Let (u, x) E graph 0 1. Then, by definition, Du = (Dx)u, x = Llu, U = Elx.
Let x = Ax. Then, by (2.33),
DAu = dADu = dA[(Dx)u) = d(ADx)Au = (DAx)Au = (Dx)Au,
i.e. LIAu = x = Ax = ALIU. Hence AEIX = Au = EILIAu = EIX = EIAx. •

An immediate consequence of Theorem 2.7 is
Corollary 2.12. Suppose that all assumptions of Theorem 2.7 are satisfied and A
is invertible. H(u,x) E graph 0 1 then (A-lu,A-lx) E graph 0 1 and ALIA-lu =
LIU, AEIA-lx = E/x.
Corollary 2.13. Suppose that X E Lg(D), (L, E) E G[O) and Conditions (2.33)
hold. H(u,x) E graph 0 then (Au, Ax) E graph 0 and LAu = ALu, EAx = AEx.
Proof. Since X is commutative, we have Du = (Dx)u = uDx (u, x) E graph O. •
Another situation is for right logarithms and antilogarithms.
Theorem 2.8. Suppose that X E Lgr(D), (Lr, Er) E G[Or) and Conditions
(2.33) hold. Define a multifunction 0' : dom D ----t 2dom D as follows:
O~u = {x E dom D': D'u = uD'x} for u E dom D', where D' = d- l D.
Basic equation 37
Then (Au, Ax) E graph D~ whenever (u,x) E graph D~ and there are (L~,E~) E
G[D~]such that L~Au = ALru, E~Ax = AErx.
Proof. By definition, dom D' = dom D. Let (u,x) E graph Dr. Then Du = uDx,
x = Lru, u = Erx. Let x' = Ax = ALru. Then, by (2.33),
D' Au = d- l DAu
= d-ldADu = ADu = A(uDx) = (Au)ADx
= (Au)d-ldADx = (Au)d- l DAx = (Au)D'Ax = (Au)D'x',
which implies L~Au = x' = Ax = ALru. Hence E~Ax = E~x' = E~L~Au = Au =

AErx. •
Corollary 2.14. Suppose that all assumptions of Theorem 2.8 are satisfied and
A is invertible. If (u, x) E graph Dr then (A-lu, A-IX) E graph Dr and L~u =
ALrA-lu, E~ = AErA-lx.
Example 2.8. Suppose that 0 c IR. Then X = C(O) is a commutative algebra
with the pointwise multiplication. Let D = ft.
Let a function hE Cl(O) such that
h' > 0 on 0 be given. By our assumption, there exists the inverse function h- l E
Cl(O) and h' E leX). Define an operator A by means of the equality: (Ax)(t) =
x(h(t)) for x E C(O), tEO. The operator A is an operator of transformation
of argument. By definition, A E ML(X), A is invertible and DA = h'AD on
dom D = Cl(O). Indeed, for all X,y E X and tEO we have A(xy)(t) =
(xy)(h(t)) = x(h(t)y(h(t)) = [(Ax)(Ay)](t), (A-lx)(t) = x(h- 1(t)) and
(DAx)(t) = :t[(Ax)(t)] = h'(t)x'(h(t)) = h'(t)(ADx)(t).
This implies that Alnx = In Ax for x E dom D, i.e. lnx(h(t)) = Inx(s)Js=h(t).

o
Example 2.9. If X is a commutative Leibniz algebra for aD E L(X) then X is a
Leibniz algebra for aD' = dD, where d EX \ {OJ. Indeed, for all x, y E dom D' =
dom D we have
D'(xy) = dD(xy) = d(xDy + yDx) = xdDy + ydDx = xD'y + yD'x.

In particular, if D E R(X), R E RD and 9 = Re E leX), then X is a Leibniz
algebra for Dn = gnD (n EN). Let X E Lg(D) and let (L, E) E G[D]. If n = 1
then antilogarithms induced by D1 = dD are E 1(),.g) = gA = E()"Lg) , so that
L1 = L. Observe that in the classical case of the operator D = ft
and R = J;
in the space X = C(R) we have Dn = tnft. Let again n = 1. The corresponding
logarithms are (up to an additive constant) Lx = lnx, as in the case of the operator
ft
D = and antilogarithms are of the form E(),.t) = etA (c E R). 0
Properties of logarithmic and antilogarithmic mappings induced by an invertible
operator D can be obtained if we put ker D = {OJ. Clearly, in that case, FD = {OJ,
38 Chapter 2
i.e. there is no non-trivial initial operators. Left invertible operators can be studied
in a similar manner. In both cases of left invertible and invertible operators the
multifunctions Dr, Dl and D are single-valued. Moreover, if D is left invertible
then dom D = X. Hence the assumption that the domain of D is a sub algebra of
X is automatically satisfied. Nevertheless, we shall consider later some properties
of logarithmic and antilogarithmic mappings induced by left invertible operators.
CHAPTER 3
LOGARITHMS AND ANTILOGARITHMS OF IDGHER ORDER
We shall now consider logarithmic and antilogarithmic mappings induced by pow-

ers Dn (n E N) of the operator D under consideration. We begin with
Definition 3.1. Let n E N. Let D E A(X). Let Or,n, Ol,n : dom Dn -+ 2dom Dn
be multifunctions defined as follows:
The equations
(3.3)
Dnu = (Dnx)u for (u,x) E graph 0 1
are said to be the right and left basic equations, respectively. Clearly,
(3.4) O;::;x = {u E dom D n : Dnu = uDnx} for x E dom D n ,
(3.5) O~~x = {u E dom D n : Dnu = (Dnx)u} for x E dom Dn.
If D E A(X) then Or,n = Ol,n for all n E N and we write Or,n = Ol,n = On. If it
is the case, then right and left basic equations do coincide. If it does not lead to
any misunderstanding then we shall omit the index n = 1, i.e. we shall write (as
before), Or,1 = Or, !i 1,1 = 0 1, 0 = 0 1 . 0
Clearly, the multifunctions now defined have similar properties as these introduced
before and we shall not repeat all considerations of Section 2.1. We have only to put
D n instead of D. Recall that a power of a right invertible (left invertible, invertible)
operator again is right invertible (left invertible, invertible, respectively).
By definitions,
Let n E N. Suppose that (un x r ) E graph Or,n, (UI, xL) E graph Ol,n, Lr,n, LI,n
are selectors of Or,n, Ol,n, respectively, and Er,n, EI,n are selectors of O;'~, O~~,
respectively. By definitions, Lr,nur E dom O;,~, Er,nxr E dom Or,n, LI,nUI E
dom O~~, E1,nXI E dom !il,n and the following equations are satisfied:
39
40 Chapter 3
Definition 3.2. Let n E N. Any invertible selector Lr,n of Or,n is said to be a

right logarithmic mapping of order n and its inverse Er,n = L;'; is said to be a
right antilogarithmic mapping of order n. IT (Ur,X r ) E graph Or,n and Lr,n is an
invertible selector of Or,n then the element Lr,nur is said to be a right logarithm
of order n of Ur and Er,nxr is said to be a right antilogarithm of order n of Xr .
By G[Or,nl we denote the set of all pairs (Lr,n, Er,n), where Lr,n is an invertible
selector of Or,n and Er,n = L;::;. Respectively, any invertible selector Ll,n of Ol,n
is said to be a left logarithmic mapping of order n and its inverse El,n = L~~ is
said to be a left antilogarithmic mapping of order n. IT (Ul, Xl) E graph Ol,n and
Ll,n is an invertible selector of Ol,n then the element LI,nU is said to be a left
logarithm of order n of Ul and EI,nXI is said to be a left antilogarithm of order n of
Xl. By G[OI,nl we denote the set of all pairs (Ll,n, El,n), where Ll,n is an invertible
selector of Ol,n and El,n = L~~.
IT D E A(X) then Ol,n = Or,n = On for all n E N and we write
Selectors Ln, En are said to be logarithmic and antilogarithmic mappings of order

n, respectively. For any (u, x) E G[OnJ elements Ln U, Enx are said to be logarithm
of order n of U and antilogarithm of order n of x, respectively. For the sake of
brevity, we say sometimes about these mappings of order n that there are nth
right (left) logarithms and antilogarithms and nth logarithms and antilogarithms,
respectively. 0
Let n E N. Clearly, by definition, for all (Lr,n, Er,n) E G[Or,n], (u r , x r ) E
graph Or,n, (Ll,n, El,n) E G[OI,n], (Ul, xz) E graph Ol,n, (Ln, En) E G[On],
(U, X) E graph On we have
(3.7)
D n EI,nXI = (Dnxz)EI,nxz, DnUI = (Dn LI,nUz)UI;

DnEnx = (Enx)Dx = (Dx)Enx, Dnu = uDnL n = (DnLnu)u.
Recall that for a DE R(X) we have

n-l
(3.8) ker Dn = lin (L:: Rj Zj : Zo, ... , Zn-l E ker D ; R E RD} (n EN),
j=O
Higher order 41
i.e. ker Dn consists of all D-polynomiaJs of degree n - 1. Let F be an initial

operator for D corresponding to an R E RD. Let n E N. Then R n E RDn. By
the Taylor Formula
n-l
I = L Rk F Dk + R nD n on dom D n (n E N),
k=O
we conclude that the operator L:~':;~ Rk F Dk is an initial operator for Dn corre-

sponding to Rn (cf. PR[8]).
Theorem 3.1. Let n E N. If X E Lgr(D) then X E Lgr(Dn) and there are
R E RD and (Lr,n, Er,n) E G R,n[Or,n] such that for u E I(X) n dom Or,n
(3.9) cDLr,le + R[cDDLr,lU + u- 1 !n(u,e)] ifn = 1;

= { cDLr,n-lU + Rn[CD(DLr,lU)D n- 1Lr,n-IU
+U-1!D(U,Dn-1Lr,n_lU)] ifn 2: 2.
If X E Lg,(D) then X E Lg,(Dn) and there are R E RD and (Ll,E1) E GR,l[0tl
such that for U E I(X) n dom Ol,n
(3.10) cDL1,le + R[CDDL1,lU + !D(u,e)u- 1] ifn = 1;

={ cDL1,n-l U + Rn[CD(Dn-l Ll,n-l u)DL1,1 u
+!D(Dn-lL1,n_lU,U)U-1] ifn 2: 2.
If X E Lg(D) then X E Lg(Dn) and there are R E RD and (Ln,En) E GR,n[On]
such that for u E I(X) n dom On
cDLle + R[CDDL1u + u- 1 !D(U, e)] if n = 1;

= { cDLn-1u + Rn[CD(DL1u)D n- 1Ln-1u
+u- 1 !D(U, Dn-l Ln-1u)] if n 2: 2.
Proof. For all n E N we have Dn E R(X) and Rn E RDn whenever D E R(X)
and R E RD. Let X E Lgr(D). We shall prove by induction that there is
(Lr,n,Er,n) E GR,n[Or] such that (3.9) holds. For n = 1 this is proved by Formula
(2.9) with v = e.
Suppose this formula to be true for an arbitrarily fixed n 2: 2. By our assumptions
we have F Dj Lr,nu = 0 for j = 0,1, ... , n - 1. Let u E I(X) n dom Or,n+l and
let x = Lr,nu. Then Dnu = uDnx . By definition, DLr,lU = u- 1 Du. Let y =
Rn+l(u- 1Dn+lu). Then
Dn+ly
= u- 1 Dn+lu = u- 1 D(Dnu) = u- 1 D(uDnx)
= U-1{CD[uDn+1x + (Du)Dnx] + !D(u,Dnx])
=cD[Dn+1x + + u- 1 !D(U, Dnx)
(u- 1 Du)Dnx]
= Dn+l{CDLr,nu + cDRn+l[(DLr,lU)DnLr,nu
+ u- 1 !D(U, D nLr,nu )]} ,
42 Chapter 3
which implies that y E dom Or,n and
where W E ker Dn. If w = 0 then y is of the form (3.9). Suppose that w f:. O. By our
assumption, F Diy = cDF Di Lr,nu = 0 for j = 0, ... , n - 1. Put y' = Rn+1 Dn+1y
(cf. Proposition 2.10). Then FDiy' = FRn+1-iD n+1 y = 0 for j = O,l, ... ,n, i.e.
y' is of the form (3.9). We still have to prove that y = Lr,n+l U, where L r,n+1 is an
invertible selector of Or,n+1. The basic equation Dn+1u = uDn+1y together with
the conditions F Diy = 0 for j = 0,1, ... , n can be written as follows:
+L
n
y = Rn+1(u- 1 Dn+1u) RiFDiy = R n+1(u- 1 D n+1 u ).
i=O
Hence for every u E J(X) n dom Or,n+l there is a unique y E dom 0;';+1. This
implies that there exists an invertible selector E r,n+1 of 0;';+1 and L r,n+1 =
-1
E r,n+l·
In order to complete our inductive proof, we have to show Formula (3.9) for n = 2.
This is proved in the same manner, as the implication n --t n + 1. We have only
to apply (3.9) for n = 1.
Similar proofs for nth left logarithms and antilogarithms and for nth logarithms
and antilogarithms. •
In a similar manner as Theorem 2.2 we prove
Theorem 3.2. Let 2 ~ n E N. If X E Lgr(D) then X E Lgr(Dn) and there are
R E RD and (Lr,n, Er,n) E G R,n[Or,n) such that for u, v E J(X) n dom Or,n and
x = Lr,nu, y = Lr,nv
(3.11)
= Er,n{ci)y + R n [ci) (Er,ny)-l (Dnx)(Er,nY)

+ (Er,ny)-l (Er,nx)-l fj;')(Er,nx, Er,nY))}·
If X E Lgl(D) then X E Lgl(Dn) and there are R E RD and (L/,Et) E GR,dOd
such that for u, v E J(X) n dom O/,n and x = Lu, y = Lv
(3.12)
= Et,n{ci)x + Rn[ci)(Et,nx)-l(Dny)(Et,nx)
+ fj;') (EI,nX, EI,ny)(EI,ny)-l(EI,nX)-l)}.
Higher order 43
If X E Lg(D) then X E Lg(Dn) and there are R E RD and (Ln, En) E GR,n[On)
such that for u, v E leX) n dom On and X = Lnu, y = Lnv
(3.11')
Corollary 3.1. Right (left) logarithms and antilogarithms of order n (if they
exist) are uniquely determined up to a D-polynomial of degree n -1, i.e. elements
ofker Dn (cf. Corollary 2.4).
Corollary 3.2. Let n E N. If X E Lgr(D) then there are R E RD and
(Lr,n, Er,n) E GRn,dOr,n] such that
(3.13)
Lr,nu = cDLr,ne + Rn[CDDnLr,nu + u- 1 f1n )(u, e)] for U E leX) n dom Or,n,
i.e.
(3.14)
i.e.
(3.16)
If X E Lg(D) then there are R E RD and (Ln, En) E G Rn ,1 [On] such that
i.e.
(3.14')
Proof. Putting v = e in (3.11), (3.12), (3.11') we obtain (3.13), (3.15), (3.13'),

respectively. These equalities imply immediately equalities (3.14), (3.16), (3.14') .
Corollary 3.3. Let n E N. If X E Lgr(D) then there are R E RD and

•
(Lr,n, Er,n) E GRn,l[Or,n] such that
(3.17)
(3.18)
44 Chapter 3
If X E Lg(D) then there are R E RD and (Ln' En) E G Rn,1 [Or,n] such that
(3.17')
Proof Putting in (3.13), (3.15), (3.13') u = e we obtain (3.17), (3.18), (3.17'),

respectively. •
Corollary 3.4. Let F be an initial operator for D corresponding to an R E
RD. Let either X E Lgr(D) and (L r,I,Er,t) E GR,I[Od or X E Lgl(D) and
(L t ,I,Et ,t) E GR,I[OI] or X E Lg(D) and (LI,Et) E GR,dOl]. Then
(i) (1 - 2cD)FLr,ne = 0, (1- 2cD)FLt,ne = 0, (1- 2c D )FL n e = 0, respectively,
for all n E N;
(ii) (1 - 2cD)Dne = f1n)(e, e) for all n E N;
(iii) CD = 1/2 for an n E N implies ft) (e, e) = 0;
(iv) CD i- 1/2 for all n E N implies FLr,ne = 0, FLI,ne = 0, FLne = 0, respec-
tively, for all n E N.
Proof Let X E Lgr(D). Since FR = 0, acting on both sides of (3.14) by F we
get (i). By definitions, Dne = eDn Lr,ne, and (3.14) implies that
(1- 2cD)D n e = (1- 2CD)DnLr,ne = DnRnefj;)(e,e) = f1n )(e, e)
for n E N. If CD = 1/2 for an n E N then, by (ii), f1n )(e,e) = 0. If CD i- 1/2 for

all n E N then (i) and (ii) together imply (iv). Similar proofs if X belongs either
to Lgl(D) or to Lg(D). •
Note 3.1. A Leibniz D-algebra is not a Leibniz Dn-algebra for n ~ 2 even in the
commutative case. Indeed, by the Leibniz formula, if n ~ 2 then, in general,
fDn (x, y) = fj;)(x, y) = I:: (~)

k=O
(Dkx)Dn-ky i- 0 (x, y E dom Dn).
o
Open Question 3.1. Does there exist a D-algebra X with unit which is a Leibniz
Dn-algebra for an n ~ 2 ? 0
Theorem 3.3. Let n E N. Let x E leX). If X E Lgr(Dn), (Lr,n, Er,n) E G[Or,n]'
a E X and Rna E dom O;:;~ for an R E RD, then
(3.19) x E ker(D n - at-) if and only if x = Er,n(Rna + z),

where z E ker Dn is arbitrary. If X E Lgl(Dn), (Lt,n, Et,n) E G[Ot,n] a EX, and
Rna E dom 0E:'~ for an R E RD, then
(3.20) x E ker(D n - a) if and only if x = Et,n(Rna + z),

Higher order 45
where Z E ker D n is arbitrary. If X E Lg(Dn), (Ln, En) E G[OnJ, a E X and

Rna Edam 0;;-1 for an R E RD, then
(3.19') x E ker(D n - a) = ker(D n - a+-) if and only if x = En (Rna + z),

where Z E ker Dn is arbitrary.
Proof Sufficiency is proved by checking.
Necessity. Suppose that x E J(X) n ker(D n - a), i.e. Dnx = ax. Then Dn LI,nX =
(Dn x )x- 1 = a, which implies LI,nX = Rna + z, where z E ker Dn is arbitrary. We
therefore conclude that x = E1,n(Rn + z), where z E ker Dn is arbitrary. A similar
proof the operator D - a+-. In the commutative case we have Dn - a+- = Dn - a,
which implies (3.19'). •
Proposition 3.1. Let n E N. Suppose that X E Lgr(D) n Lg1(D), (Lr,n, Er,n) E
G[Or,nJ, (LI,n, EI,n) E G[OI,n] and e E ker Dn. Write
(3.21)
(3.22) A1,nU = (Dn LI,nU)U for U E domOI,n'
Then
(3.23) Ar,nu + AI,n(e - u) =0 for u E dom Or,n n dam Ol,n,
(3.25) Ar,nu + Ar,n(e - u) =0 for u E dom Or,n,
A1,nU + AI,n(e - u) =0 for U Edam Ol,n,
(3.26)
Proof. We shall prove only Formulae (3.23) and (3.24). The proofs of Formulae
(3.25), {3.26} are similar. Let then n E N. By our assumptions, for U Edam Or,nn
dom Ol,n we find
0= Dne = Dnu + Dn(e - u)

=uDnLr,nu + [(Dn LI,n(e - u)]{e - u)
= Ar,nu + AI,n(e - u).
46 Chapter 3
Let u E I(X) n dom Or,n n dom Ol,n. Then

Ar,nu + A l,n u - 1
= uD n Lr,nu + (Dn LI,nu - 1)u- 1
= =
D n u+D n u- 1 Dn(u+u- 1).
Clearly, if in Proposition 3.1 the algebra X is commutative, then Ar,nu = A1,nU.

•
So that we have
Corollary 3.5. Let n E N. Let X E Lg(D), (Ln' En) E G[OnJ and let e E ker Dn.
Write
(3.27)
Then
(3.28) Anu + An(e - u) = 0 for u E dom On,
(3.29) Anu + Anu-1 = Dn(u + u- 1) for u E I(X) n dom On.

Theorem 3.4. If X E Lg(D) and D E ML(X) then X E Lg(Dn) for all
n E N and there are (Ln' En) E G[OnJ n M(X) such that for some R E RD, all
(u, x) E graph On and all n E N
n-1
(3.30) Lnu = II Rj L 1u,
j=O
n-1 n-1
(3.31) II En(Rjx) = En(II Rjx) = E 1x.
j=O j=O
Proof Since D E ML(X) n R(X), also Dn E ML(X) n R(X) and Rn E RDn

whenever R E RD and n E No Proposition 2.6 implies that L1 = Land E1 =
E E M L(X). Theorem 3.1 implies that X E Lg(D). If (Ln, En) E G[OnJ then
Ln,En E ML(X), since Dn E ML(X).
We shall prove by induction that there exist (Ln' En) E G[OnJ such that (3.30)
and (3.31) hold. For n = 1 (3.30) is trivial. Suppose (3.30) to be true for an
arbitrarily fixed n E N. Let u E dom On+l and let Yn = Lnu. Then, by definition,
Du = UDY1, Dnu = uDn Yn and
Dn+1u = D(Dnu)
= D(uDnYn) = (Du)Dn+lYn
= (uDyt}Dn+lYn = u(Dn+l R nY1)D n+lYn
n-1
=uDn+l [(Rnyt}YnJ = uDn+l [(Rnyt) II Rj L1 uJ
j=O
n
= uD n+1 II Rj L1 U = uDn+lYn+l.
j=O
Higher order 47
Hence Yn+1 = Ln+1 u is of the form (3.30). Clearly, for all n E N
n-1 n-1 n-1

E 1x = U = EnLnu = En(IT Rj L1U) = En(IT Rjx) = En(Rjx). IT
j=O j=O j=O
Lemma 3.1. Let X be a D-algebra with CD ¥- O. Then for all u E I(X) n dom D
•
r
= c /u- 1[De - cD(Du)u- 1 - fD(U,U- 1)]
-l[D e - CDU -lDU - f D(-1
= CD U ,U)]-1
U .
Proof. Let U E J(X) n dom D. Then
De = D(uu- 1) = CD [uDu- 1 + (Du)u- 1] + fD(U,U- 1)],
De = D(u- 1u) = cD[(Du- 1)u + u- 1Du] + fD(U- 1,U),

which implies the required formulae.
Theorem 3.5. Let F = C and let CD ¥- O. If X E Lgr(D), (Lr,n, Er,n) E G[Or,n]'
•
(n = 1,2), and u E I(X) n dom Or,2, then for an R E RD
(3.32)
L r,l(iu) = cDLr,2U + R2{[De - cDDLr,lU - iD(U- 1,u)]DLr,lU + fD(u- 1,Dun.
If X E Lgl(D), (L1,n, El,n) E G[OI,n], (n = 1,2), and U E J(X) n dom 01,2, then
for an R E RD
(3.33)
L 1,1(iu) = cDL1,2U + R2{(DL1,lu)[De - cDDL1,lU - iD(U,U- 1)] + fD(Du,u- 1n·
If X E Lg(D), (Ln' En) E G[On], (n = 1,2), and u E J(X) n dom O2 , then for an
RERD
Proof. Let u E I(X) n dom Or,2. Then, by definitions and Lemma 3.1,
D 2 L r ,1(iu)
= D[DL r,l (iu)D(iu)] = D[-iu- 1iDu] = D(u- 1Du)
=CD[U- 1D 2u + (Du- 1)Du] + fD(U-l, u)
=cDLr,2U + CDCv1u- 1[De - CDU- 1Du - fD(U- 1,u)]u- 1Du
+ fD(U- 1, Du)
=cb[Lr,2U + [De - cDDLr,lU - fD(U- 1,u)]DLr,lU + fD(U- 1, Du).
48 Chapter 3
By Corollary 3.1, we get (3.32) for an R E RD. Similar proofs for (3.33) and
(3.32'). •
Corollary 3.6. Suppose that all assumptions of Theorem 3.5 are satisfied and X
is a Leibniz algebra. Then
Lr,l(iu) = Lr,2U - R2(DLr,lU)2, Ll,l(iu) = LI,2U - R 2(DLI,lU)2,
Ll(iu) = L 2u - R2(DLIU)2,
respectively.
Example 3.1. Let X E Lg(D) be a Leibniz D-algebra. By (3.1), if (Ln, En) E
GR,n[On] for an R E RD and u E dom On (n E N) then
Lnu = Ln-1u + Rn[(DLIU)Dn-l Ln-lU].
o
Example 3.2. Let IF = ~, X = C(~) and D = ft.
Then X with pointwise
multiplication is a commutative Leibniz D-algebra. By Theorem 3.1, X E Lg(Dn)
for all n E N. By Corollary 3.1, logarithms and antilogarithms of order n are
uniquely determined up to a polynomial of degree n -1. Let u(t) = eAt. Note that
here and in the sequel we shall denote by "e" the basis of an exponential function,
i.e. the number
00 1
e=2:-
n=O n!
in order to distinguish between this number and the unit e of the algebra under
consideration. Fix n E N. Let (Ln' En) E G[f!n]. Since Dnu = ).nu, by definition,
we have DnLnu = u-1Dnu = ).n. This implies that
where a E ~ is arbitrary, i.e. LneAt = ~()'t

n.
+ a)n.
Let (Fx)(t) = x(O) for x E X. Clearly, since a = 0, we get FLne>..t = ~an
n. = o.
Moreover, again by definition, we have
o
Example 3.3. Denote by C(Oj IF) the space of all functions continuous on a set
o and with values in IF (considered as a linear space over IF). In this example 0
will be a closed interval in R
Fix c: > O. Let X = C([c:,7r - c:];~) and let D = ft.
Then X with pointwise
multiplication is a commutative Leibniz D-algebra belonging to Lg(D). By similar
arguments to those used in Example 3.2, for u(t) = sin t, we find
L1u = lnu, E1lnu = u,

Higher order 49
1 . tj
Lju = -~(arcsin u)1 = ~ (j = 2k, kEN),
J. J.
Lju = J E:
t (t - s)j
.,
J.
In sin sds (j = 2k + 1, kEN).
We also have L 2 cott = 2lnu = 2lnsint.

Similarly, in the space Y = C([-7r/2 + €,7r/2 - €]; JR) for v(t) = cost we find that
L 1v = lnv, E 1 lnv = v,
1 . tj
L·v = --(arccosv)1 = -
J j! j!
(j = 2k, kEN),
Ljv = I t
-7r/2+E:
(t
.,
J.
s)j
In cos sds (j=2k+l, kEN).
Also L 2 tant = 2lnv = 2lncost.

Let now
x' = C([€, 7r - €j;C), Y' = C([-7r/2 + €,7r/2 - €]; C).
By definitions, X c X' and Y C Y'. Clearly, by the de Moivre formula, we get
iarcsinu = Ll(iu+~) EX',
i arccos v = Ll (v + i ~) E Y'.
o
Example 3.4. Let X be a noncommutative algebra (over JR or C) with unit e
and a complete linear metric space. Similarly as in Theorem 2.6, Formula (2.30),
write for a E X, bE J(X):
= {y EX: L
00
(3.34) X 2 (a, b) (n + 1)( _1)nb-(n+2)ya n is convergent}.

n=O
Let Dx = xa + bx for x E X. Then D 2 x = xa 2 + b2 x + 2bxa for x E X (cf.

Example 2.4(iv)). Similarly, as in Theorem 2.6, we prove that
= y for Y E X 2 (a, b) if and only if x = L

00
(3.35) D 2 x (n+l)(_I)nb-(n+2)ya n ,
n=O
i.e. the operator D2 is invertible on the subspace X 2(a, b). Moreover, selectors of
Or,2 and 01,2 are of the form
=L
00
(3.36) Lr,2U (n+l)( -1)nb-(n+2l(a2+u-lb2u+2u-lbua)an for U E dom 0,

n=O
50 Chapter 3
=L
00
LI,2U (n + 1)( -ltb-(n+2) (ua 2u- 1 + b2 + 2buau- 1)an for u E dom 0 / ,

n=O
provided that a 2 +u- 1b2u+2u- 1bua E X 2(a, b), a 2+u- 1b2u+2buau- 1 E X 2(a, b),
respectively.
Indeed, for x E X we have
= xa 2 + b2x + 2bxa
=L
00
(n + 1)(_1)n[b-(n+2)yan a2 + b2b-(n+2)yan + 2bb-(n+2)yan a]

n=O
L
00
= (n + 1)( _1)n[b-(n+2 ya n+2 + b-nyan + 2b-(n+1)yan+1]

n=O
L
00
= (-l)n[(n - 1) + (n + 1) - 2n]b- n ya n - 2b- 1ya + 2b- 1ya = y.

n=2
Let x = LT,2U for such u that a2 + u- 1b2u + 2u- 1bua E X 2(a, b). Then
D 2x = u- 1 D 2u = u- 1(ua 2 + b2u + 2bua) = a2 + u- 1b2u + 2u- 1bua.

This, and (3.35) immediately imply (3.36). A similar proof for left logarithms. 0
Example 3.5. Let X be a noncommutative algebra over IF. Let a, b E X do not
vanish simultaneously. Write Dx = xa + bx for x EX. Clearly, dom D = X,
hence X E A(D). Here CD = 1 and fD(x, y) = -x(a + b)y. Let a 2 = b2 = e (cf.
Example 2.4(iii». Then D 2x = xa 2 + b2x + 2bxa = 2(x + bxa) for x E X. By an
easy induction we prove that
We therefore conclude that the equation Dnx = y is equivalent to the equation

Dx = Yn, where
This, and Example 2.4(iii) together imply that for all n E N
L T,2nU = e + u -lbYna = e +2- n u -1 yu if and only if [u, b] = 0, u E I(X),
L T,2n-lU = e+u -lbYna = e+ 2- n +1bya if and only if [u, b] = 0, u E I(X),

Ll,2nU = e + bUYnau-1 = e + rnbubyau- 1 if and only if [u, a] = 0, u E I(X),
L1,2n-1U = e+buYnau-1 = e+rn+1buyau- 1 if and only if [u,a] = 0, u E I(X).
o
Higher order 51
Example 3.6. Let X and D be defined as in Example 3.5 and let a, b be idem-
potents, i.e. a2 = a, b2 = b. By an easy induction, we get
Dnx = Dx + (2n - 2)bxa = xa + bx + (2n - 2)bxa for n E N.
Just by definition, for n E Nand u E J(X) we have
Indeed,
A similar proof for left logarithms. In particular, if b = e then, again by an easy

induction, we prove that the operator D is invertible and D-n = y - 2~;;-1 ya for
y E X (n EN). Let u E J(X). Since u-1bua = u-1ua = a, we find
L -L 2n-l(L) (2n-2)(2n-2n+l)
T,n U - T,lU - ~ T,lU a+ 2n a.
A formula for left logarithms is much more complicated. A similar result for left
logarithms can be obtained if a = e. D
CHAPTER 4
LOGARITHMS AND ANTILOGARITHMS OF OPERATORS

HAVING EITHER FINITE NULLITY OR FINITE DEFICIENCY
Let A E L(X). Recall that the nullity and the deficiency of A are
aA = dim ker A, (3A = codim A(dom A) = dim XjA(dom A),
respectively. The index KA of an operator A E L(X) having either finite nullity or

finite deficiency is defined as follows:
{3A - aA if aA < +00, {3A < +00,

{
KA = +00 if aA < +00,
-00 if (3A < +00.

Denote by JC(X) the ideal of all finite dimensional operators belonging to Lo(X).
It is well-known that an operator A E L(X) has the finite nullity (respectively,
deficiency) if and only if there is a K E JC(X) such that
A = ~ + K, where ~ E A(X), A = D + K, where D E R(X),
respectively (cf. the author and ROLEWICZ PRR[I]). Clearly, the index KA is
finite if and only if there is a K E JC(X) such that A = T + K, where T E I(X).
Theorem 4.1. Suppose that D E R(X), K E ,qX) and D' = D + K. Let
X E Lgr(D), (Lr,Er) E G[OT]' Write
O~ : dom D' -+ 2dom D' ,
= {x E dom D' : D'u = uD'x},

O~u u E dom D' = dom D.
Let u E J(X) n dom Or and let
(4.1) L~u = Lru + R[u- 1 Ku - KLru] for an REnD.
Then the mapping L~ is a selector ofO~. If L~ is invertible and E~ = (L~)-l then

(L~, E~) E G[O~] and
(4.2) Erx = E~{x + R[(Erx)-l KErx - Kx]}, where x = Lru.

Let X E Lg1(D), (Ll,Et) E G[Oz]. Write
52
Operators with finite nullity and/or deficiency 53
0; : dom D' ---+ 2dom D' ,
O;u = {x E dom D' : D'u = (D'x)u}, u E dom D' = dom D.

Let u E leX) n dom 0/ and let
(4.3) L;u = L/u + R[(Ku)u- 1 - KL/u] far an R E 'RD.
Then the mapping L; is a selector of 0;. If L; is invertible and E; = (LD- 1 then

(L;, En E G[OD and
(4.4) E/x = E;{x + R[(KElx)(Erx)-l - Kx]), where x = Ltu.
Let X E Lg(D), (L, E) E G[O]. Write
0' : dom D' ---+ 2dom D' ,
O'u = {x E dom D': D'u = uD'x}, u E dom D' = dom D.

Let u E J(X) n dom 0 and let
(4.1') L'u = Lu + R[u- 1Ku - KLu] far an R E 'RD.
Then the mapping L' is a selector of 0'. If L' is invertible and E' = (L,)-l then
(L',E') E G[O'] and
(4.2') Ex = E'{x + R[(Ex)-l KEx - Kx]), where x = Lu.
Proof Let X E Lg1(D), (Ll, El) E G[Od and let D' = D + K. Let u E J(X) n
dom 0 1 and let x = L/u. Then E/x E leX). Let x' = x+R[(Ku)u-1-Kx]. Then
Dx' = Dx + DR[(Ku)u- 1 - Kx] = Dx + (Ku)u- 1 - Kx
and
D'u = Du + Ku = (Dx)u + Ku = [(D'x) - Kx]u + Ku =
= D'{x + R[(Ku)u- 1 - Kx]}u = (D'x')u.
Hence Dx' = DL;u which proves that Li is a selector of 0;. If L; is invertible,
E; = (LD- 1 and x = LIU then E1x = u = E;Liu, which implies (4.4). Similar
proofs for selectors of Or and O. _
Definition 4.1. By Lgr#(D) (LgI#(D), respectively) we denote the class of
algebras with the operator DE A(X) such that D E A(X) and Or = Lr (0 1 = L l ,
respectively) is invertible. If D E A(X) then Lgr#(D) = LgI#(D). This class is
denoted by Lg#(D). D
In the same manner as Theorem 4.1 we obtain
54 Chapter 4
Theorem 4.2. Suppose that D E A(X), K E IC(X) and D' = D + K. Let

n~, n; be defined as in Theorem 4.1. Let X E Lgr#(D), (Lr,E r ) E G[nr]. Let
u E leX) n dom nr and let
(4.5) L~u = Lru + S[u- l Ku - KLru] for an S E CD.
Then the mapping L~ is a selector ofn~. If L~ is invertible and E~ = (L~)-l then

(L~, E~) E G[n~] and
(4.6) Erx = E~{x + S[(Erx)-l KErx - Kx]}, where x = Lru.

Let X E Lgl#(D), (Ll, El ) E G[nr]. Let u E leX) n dom nl and let
(4.7) L;u = LlU + S[(Ku)u- l - KLlu] for an S E CD.
Then the mapping L; is a selector ofn;. If L; is invertible and E; = (L;)-l then

(L;, En E G[n;] and
(4.8) Elx = E{{x + S[(KElX)(Erx)-l - Kx]), where x = Llu.

Let X E Lg#(D), (L, E) E G[n]. Let u E leX) n dom n and let
(4.5') L'u = Lu + S[u- l Ku - KLu] for an S E CD.
Then the mapping L' is a selector ofn'. If L' is invertible and E' = (L,)-l then
(L', E') E G[n'] and
(4.6') Ex = E' {x + S[(EX)-l KEx - Kx]}, where x = Lu.

Lemma 4.1. Suppose that all assumptions of Theorem 4.1 (respectively, of The-
orem 4.2) are satisfied. Let
n
Kx = l: <Pk(X)Xk for x E X,
k=l
where <Pl, ... , <Pn belong to a conjugate space X' (i.e. a total space of linear func-
tionals over X what means that <p(x) = 0 for all <P E X' and an x E X implies
x = 0) and Xl, ... ,xn E X are linearly independent. Then the operator 1- RK
(respectively, 1- SKY is invertible if and only if
(respectively, ~(S) =I 0). If it is the case, then

n
V = (I - RK)-lW = W + l: VkRxk,
k=l
where (VI, ... , v n ) is the unique solution of the Cramer system:

n
L [b"jk - CPj(RXk)]Vk = cPj(w) (j = 1, ... ,n),
k=l
and Vk = CPk(V), i.e. Vk = [~(R)]-IVk(W), Vk(W) is the determinant obtained

from ~(R) if we put instead of the k-th column the column (CPI (w), ... , CPn(w)) (a
formula for (J - SK)-l is obtained if we put S instead of R).
Proof. We obtain the Cramer system of algebraic equations by standard techniques
for finite dimensional operators. This systems has a unique solution if and only
if its determinant ~(R) is different than zero. If it is the case, then the operator
J - RK is invertible. Respectively, if ~(S) =f:. 0 then J - SK is invertible. •
An immediate consequence of Theorem 4.1 (respectively, Theorem 4.2) and Lemma
4.1 is
Corollary 4.1. Suppose that all assumptions of Theorem 4.1 (respectively, of
Theorem 4.2) are satisfied. Let
n
Kx = L CPk(X)Xk for x E X,
k=l
where CPI, ... ,CPn belong to a conjugate space X' (cf. Lemma 4.1), Xl, ... ,Xn E X
are linearly independent. If there is an R E RD (respectively, an S E .cD, a
T E I(X)) such that
~(R) = det (b"jk - CPj(RXk))j,k=I, ... ,n =f:. 0,
(respectively, ~(S) =f:. 0, ~(T) =f:. 0) then there exist invertible selectors ofO~, OJ
and 0', which are of the form (4.1) (respectively, of the form (4.1'), (4.3), (4.5),
(4.5'), (4.7)).
Proof. Let (u, x) E graph Or. By our assumptions, we have X = Lru, i.e. x' = L~u
and
x' = L~u = Lru + R[u- l Ku - KLru] = x + R(u- l Ku - Kx).
This implies (J - RK)x = x' - R(u- l Ku) = w. Putting V = x in Lemma 4.1 we
conclude that for every x' there exists a unique u such that L~u = x' (x is already
fixed). This means that L~ is invertible. The same arguments are used in the case
of left logarithms, logarithms and left invertible operators. •
As in Chapter 2, we get
Proposition 4.1. Suppose that D E A(X) n A(X). Then
dom Or:::> {x = S(u-IDu): u E J(X) ndom D;S E .cD} =f:. 0;
dom 01 :::> {x = S[(Du)u- l ] : u E J(X) n dom D; S E .cD} =f:. 0
56 Chapter 4
(cf. Proposition 2.2).

Note 4.1. By definition, if DE A(X) then X is a commutative algebra and the
domain of D is a subalgebra of X. On the other hand, since D is left invertible,
we have dom D = X. Hence the assumption that the domain of D is a subalgebra
of X is trivially satisfied. 0
Proposition 4.2. Let U E J(X). If D E A(X) n A(X) and either x E Oru,
y E Oru or x E OIU and y E OIU then x = y. If D E A(X) n A(X) and x, y E Ou
then x = y. In particular, Or(O) = 0 1(0) = 0(0) = O. (cf. Propositions 2.3).
In other words: If either D E A(X) n A(X) or D E A(X) n A(X), then the
multifunctions Or, 0t, 0 are single-valued.
Proof Let x, y E Or. By our assumptions, we have Dx = u- 1 Du = Dy. Since
kerD = {O}, we get x = y. •
In view of Proposition 4.2, if ~ E A(X) then we may write
(4.9)
(4.10)
(4.9') L = 0, E = 0- 1 for (L, E) E G[O], i.e. L E I(X).
In the case of left invertible and invertible operators some pathologies may appear,
as it is shown by
Theorem 4.3. Let A E A(X). Consider the following conditions:
(i) X is a Leibniz algebra, i.e. CA = 1 and fA = 0;
(ii) X has a unit e;
(iii) ker A = {O}, i.e. A is left invertible.
Then these 3 conditions cannot hold simUltaneously, i.e. if ker A = {O} then either
X is not a Leibniz algebra or X has not a unit.
Proof. Suppose that X is a Leibniz algebra with unit e and ker A = {O}. Then
=
Ae Ae2 = = = =
2eAe 2Ae which implies Ae O. Thus e 0, a contradiction. •
Example 4.1. Suppose that X = G[O, T] and
D = :t with dom D == {x E G'[O, T] : x(O) = O}.
It is easy to verify that X (with respect to the usual addition, multiplication by
scalars and multiplication of elements) is a Leibniz algebra without unit and that
the operator D is invertible and its inverse is D-1 = J;. 0
This example can be generalized in the following manner:
Example 4.2. Suppose that X is a Leibniz D-algebra with unit e and F is an

initial operator for D corresponding to an R E RD. Let D' = D with
dom D' = {x E dom D: Fx = O}.
By our assumption, e E ker D. Hence Fe = e :f:. O. We therefore conclude that X

is a Leibniz algebra without unit, D' is invertible and (D,)-l = R. D
An example of an algebra with a multiplicative invertible operator will be shown
by Theorem 6.l.
Note 4.2. Recall that for D E I(X) we have .cD = RD = {D- 1 }. D
We shall examine now logarithmic and antilogarithmic mappings in commutative
algebras for the sake of brevity. Clearly, all following results might be formu-
lated and proved also in the non-commutative case. An immediate consequence
of Propositions 4.1 and 4.2 is
Proposition 4.3. Suppose that D E A(X) n A(X) and n = L E I(X). Then
Lu = S(u- 1 Du), [I - S(Dx)]Ex = 0,
where S E CD, (u, x) E graph L, u E I(X).

An immediate consequence of Proposition 4.3 is
Corollary 4.2. Suppose that D E A(X) n A(X) and n = L E I(X). Write
E=L-l and
(4.11) 8(S) = {x E dom E: ker[I - S(Dx)] :f:. O} for S E CD.
If8(S) = 0 then Ex = O. Hence Lu = 0 for all (u,x) E graph L, i.e. logarithmic

and antilogarithmic mappings do not exist.
Definition 4.2. We denote by Lg~(D) the class of commutative algebras with
unit e belonging to Lg#(D) such that e E dom L. D
Clearly, by this definition and Corollary 4.2, if X E Lg~(D) then there is an

S E .cD such that the set 8(S) defined by Formula 4.11 is not empty. Then we
get
Lu = S(u- 1 Du), Ex = u for u E I(X) n dom L.
In this case the operator I - S(Dx) is not left invertible (invertible). However,
if I - S(Dx) is right invertible then Ex = Rz, where R E R1-S(Dx) and z E
ker[I - S(Dx)] (d. Propositions 2.15 and 2.16). D
An immediate consequence ofthe fact that n is single-valued (d. Proposition 4.2)
is
Corollary 4.3. Suppose that X E Lg#(D). Then logarithms and antilogarithms
are uniquely determined.
(d. Propositions 2.8, 2.9 and Corollary 2.4).
58 Chapter 4
Proposition 4.4. Suppose that X E Lg#(D). Then the logarithm and antiloga-
rithm of zero are not defined.
Proof. Suppose that L E I(X). If x = L(O) then x should satisfied the equality
o= DO = ODx. Thus L(O) is not well determined. By definition, we have
DE(O) = E(O)D(O) = O. Since ker D = {O}, we conclude that E(O) = O. This
implies that L(O) = LE(O) = 0, a contradiction (cf. Corollary 4.3). •
Proposition 4.5. Suppose that X E Lg~(D), S E LD, e E DX and 9 = Se.
Then 9 E dom L if and only if 9 E J(X). In that case, 9 is not a zero divisor.
Proof. Since e E DX and Se = 9 = SDg, we get Dg = e. The basic equation
implies that e = Dg = gDLg = (DLg)g. Clearly, 9 is not a zero divisor (cf.
Corollary 2.2). •
Theorem 4.4. If X E Lg# (D) and (L, E) E G[L] then for u, v E J(X) n dom L
and x = Lu, y = Lv we have, independently of the choice of an S E LD,
(4.12) L(uv) = cD(Lu + Lv) + S[U-1V- 1fD(U, v)],
(4.13) (Ex)(Ey) = E{CD(X + y) + S[(EX)-l(Ey)-l fD(Ex,Ey)]}.

Proof. Let u, v E J(X) n dom L. If S E LD then
L(uv) = SDL(uv)
= SD{CD(Lu + Lv) + S[U-1V- 1fD(U,v)]}
= cD(Lu + Lv) + S[u-1v- 1fD(U, v)]
(cf. the proof of Theorem 2.2). By Proposition 4.4, Formula (4.12) is independent
of the choice of a left inverse S, and Formula (4.13) is an immediate consequence
of the definition and (4.12). •
Corollary 4.4. Suppose that CD :/: 0, X E Lg#(D) and (L,E) E G[L]. Then
n dom L if and only if u E J(X) n dom L. In that case
u 2 E J(X)
Lu 2 = 2CDLu + S[u- 2 fD(U, u)] (S E LD)

Corollary 4.5. Suppose that X E Lg#(D) and (L,E) E G[L]. If D E ML(X)
then Land E are multiplicative (cf. Proposition 2.6).
Theorem 4.5. Suppose that X E Lg~(D) and (L,E) E G[L]. Let gD be defined
by Formula (2.20'). Then
(4.14) (1 - cD)Lu = cDLe + SgD(U) far u E J(X) n dam L.
Moreover, if CD :/: 1 then the mapping gD is not constant.

Proof. Formula (4.14) follows immediately from our assumptions and Theorem
4.4 (cf. Proposition 2.11). The remaining part of the proof is similar to that of
Theorem 2.3. (cf. also Theorem 4.1). •
Theorems 4.1, 4.3 and Proposition 4.4 together imply (cf. Proposition 2.12) the
following
Proposition 4.6. Suppose that X E Lg~(D) and (L, E) E G[L]. Then
(i) (1- 2cD)Le = SfD(e,e) , where S E CD;
(ii) (1 - 2cD)DLe = (1 - 2cD)De = fD(e, e);
(iii) CD = 1/2 implies fD(e, e) = 0;
(iv) CD ::j: 1/2 implies Le = e .
Similarly to Theorems 2.4, 2.5 and Corollaries 2.8-2.11 we obtain
Proposition 4.7. Suppose that X E Lg#(D) and CD = O. Let a E IF \ {OJ.
Let u E dom L. If a ::j: 1 then au ¢ dom L. In particular, if X E Lg~(D then
-e ¢ dom L.
Proposition 4.8. Suppose that X E Lg~(D) and CD ::j: O. Let a E IF \ {OJ. Let
u E dom L. Then au E dom L and
L(au) = cnlLu + L(ae)] + SgD(U) = Lu + cnlL(ae) - Le] (S E CD).
Proof. Let a E IF\ {OJ. Let u E dom L. Theorems 4.4 and 4.5 together imply that
L(au)
= L(u· ae) = cD[Lu + L(ae)] + Sa-1u- 1fD(U, ae)
= cD(Lu + L(ae)] + a-1au-1SfD(U, e) = cD[Lu + L(ae)] + SgD(U)
= Lu - cDLe - SgD(U) + cDL(ae) + S9D(U) = Lu + cD[L(ae) - Le] .
Proposition 4.9. Suppose that D E A(X) n A(X), X has unit e and CD ::j: 1. Let
•
the mapping A with dom A :J dom n be defined by the following formula:
1
Au = -l--[cDe
- CD
+ SgD(U)] for u E J(X) n dom A, where S E CD,
and the mapping gD(U) = u- 1 fD(U, e) defined by Formula (2.20') is, by Theorem
4.3, not constant. If A is invertible then (A, A- 1) E G[n], i.e. X E Lg(D).
Proof By definitions, for all u E J(X) n dom A we have
1 1
SDA = Au = -l--[cDSDe + SgD(U)] = S--[cDe + gD(U)],
- CD 1- CD
60 Chapter 4
which implies
1
DAu = -l--[De + gD(U)]
-CD
(Up to an s E ker S which can be assumed to be zero). Further arguments are the
same as in the proof of Proposition 2.18. •
In the same way, as in Chapter 3, we introduce right and left logarithms and
antilogarithms of the higher order. Instead of Theorem 3.1 and Corollaries 3.1,
3.3 we have the following
Theorem 4.6. Let the IDultifunctions Or,n, Ol,n, On (n E N) be defined as
in Definition 3.1. Let (Lr,n, Er,n) E G[Lr,nj, (Ll,n, EI,n) E G[LI,nj, (Ln, En) E
G[Ln]. Then dom Lr,n+1 C dom Lr,n, dom L1,n+1 C dom L1,n, dom Ln+1 c
dom L n , respectively (n EN). If X E Lgr#(D) then X E Lgr#(Dn), if X E
Lgl#(D) then X E Lg1#(Dn), if X E Lg#(D) then X E Lg#(Dn) for all n E N.
Moreover, if (Lr,l, Er,l) E G[Lr,lj, (L1,1> El,d E G[LI,lj, (L1' Ed E G[Llj, then
(Lr,n, Er,n) E G[Lr,nj, (Ll,n, El,n) E G[L1,nj, (Ln, En) E G[Ln] for every n E N
and for (ur,xr), (vr,Yr) E graph Lr,n, (Ul,Xc), (Vt,Yl) E graph Ll,n, (u,x), (v,y) E
graph Ln we have
(4.15) =cDLr,n-1U + sn[cD(DLr,1U)Dn-1 L r,n-1U

+ u- l fD(U, D n- l Lr,n-l u)] (n ~ 2),
while for all n E N
(4.16) Lr,n(uv) = c'D(Lr,nu + Lr,nv) + Rn[u-Iv-ll~;')(u,v)j,
(Er,nx)(Er,nY)
(4.17) = Er,n{CD(X + y)
+ sn[(Er,nx)-l(Er,ny)-l fi;')(Er,n x , Er,nY)]} ,
L1,nU
(4.18) = cDL1,n-lU + sn[cD(DL1,lU)D n- 1LI,n-lU
+ u- l fD(U, D n- 1L1,n-lU)] (n ~ 2),
while for all n E N
(El,nx)(E1,nY)
(4.20) = El,n{CD(X + y)
+ sn[(E1,nX)-1(El,ny)-1 f1n)(El,nx ,E1,nY)]},
Lnu
(4.15') = cDLn-lu + sn[cD(DL1u)D n- 1Ln-1U
+U-l!D(U,Dn-1Ln_lU)] (n ~ 2),
while for all n E N
(4.16')
(Enx)(EnY)
(4.17)
= En{CD(X + y) + sn[(Enx)-l(Eny)-l !1n)(Enx, EnY)]},
independently of the choice of S E CD, i.e. Lr,n, Er,n, LI,n, EI,n, L n, En (n E
N are uniquely determined. If X E Lgr#(D) (X E Lg1#(D), X E Lg#(D),
respectively), then
(4.21)
(4.22)
(4.21')
respectively.
Lemma 4.2. Suppose that D E A(X) n A(X). If DE ML(X) and S E CD then
S E ML(X).
Proof. Let D E ML(X). Let x,y E X, S E CD and let u = Dx, v = Dy. Then
x = Su, y = Sv and S(uv) = S[(Dx)(Dy)] = SD(xy) = xy = (Su)(Sv). •
By similar arguments to those used in the proof of Theorem 3.4, applying Lemma
4.2, we obtain
Theorem 4.7. Suppose that X E Lg#(D) and D E ML(X). Then X E
Lg#(Dn) for all n E N, (Ln,En) E G[Ln] n ML(X) and for (u,x) E graph L
andnEN
n-l
(4.23) Lnu = II sj L1u,

j=O
n-l n-l
(4.24) II En(Sjx) = En(II sjx) = E1x
j=O j=O
independently of the choice of S E CD.

62 Chapter 4
Theorem 4.8. Let n E N and let a E X. Suppose that X E Lgr#(Dn),

(Lr,n, Er,n) E G[Lr,nJ (X E Lg1#(Dn), (L"n, E"n) E G[L"n]) and sna E dom Er,n
(sna E dom E"n, respectively) for an S E CD. Then
(i) Xo E ker(Dn - at-) if and only ifxo = Er,n(Rna);
(ii) Xo E ker(Dn - a) if and only if Xo = E"n(Rna).
Proof. (i), (ii) Sufficiency is proved by checking.
Necessity. (ii) Suppose that Xo E ker(Dn - a). Then Dnxo = axo. Hence
Dn L"nXO = a. If S E CD then Lnxo = sn Dn Lnxo = sna implies that
Xo = E"n(sna). A similar proof for the necessity of (i). •
CHAPTERS
REDUCTION THEOREMS
We shall consider now a class of commutative D-algebras which are important in

some applications and have non-Leibniz components of a particular form. Non-
commutative algebras can be treated in a similar manner, but for them all results
are much more complicated in the formulation (cf. Example 5.8). To begin with,
recall the following
Theorem 5.0. (cf. TARGONSKI T[l]) Let X be a commutative algebra of real
(complex) valued functions of a real (complex) variable. Let A E L(X) satisfy the
so-called Bourlet condition:
A(xy) = f(x,y,Ax,Ay) for X,y E dom A

(cf. C. BOURLET, C. R. Acad. Sci. Paris, 124(1827)). If f is an entire function of
its variables, then f is necessarily of the form
f(x, y, Ax, Ay) = o:(Ax)(Ay) + {3(xAy + yAx) + 'YXY,

where 0:, {3, 'Y are given scalars.
If X has a unit e and there is no zero divisors in X, then A is either a scalar
multiple of the identity or belongs to one of the following three types:
(i) Ao(xy) = o:(Aox)(Aoy), where Ao =A- pI (p is a scalar dependent on 0:,
{3, 'Y);
(ii) A(xy) = ~(xAy + yAx);
(iii) Al (xy) = xAIy + yAIx, where Al = A + 'YI.
Note that in the case (i) Al = o:Ao = o:(A - fL). In the case (ii) A is a multiplication
by a fixed element. Namely, Ax = xAe for all x E dom A. In the case (iii) we
have e E ker AI.
In the sequel we shall not need always the assumption that algebras under consid-
eration are without zero divisors and coefficients 0:, {3, 'Yare scalars. Nevertheless,
the Targonski theorem motivates in a sense the following:
Definition 5.1. Denote by AD(a, c, d) the set of these commutative D-algebras
X with unit e E dom D and with ker D =J {O} without zero divisors for which the
following condition is satisfied:
D(xy) = c(xDy + yDx) + d(Dx)(Dy) + axy for X,y E dom D,
where a, c, d E X depend on D only and do not vanish simultaneously. (cf. TAR-
GONSKI T[l] for scalar a, c, d). 0
Clearly, for X E AD (a, c, d) the non-Leibniz component is of the form
63
64 Chapter 5
fD(x, y) = (c - e)(xDy + yDx) + d(Dx)(Dy) + axy

and CD = 1.
Let r E IF. In the sequel we shall write r instead of re whenever it does not lead
to any misunderstanding.
Proposition 5.1. Let X E AD(a, c, d). Suppose that X E Lg(D), u, v E leX) n
dom D and (L, E) E G[DJ. If Hu = dDLu + c then
(5.1) H(uv) = (Hu)Hv + h, where h = ad - c2 + C.
Proof. Let u, v E leX) n dom D. Then
DL(uv)
= (uv)-l D(uv)
= u-1v-1[C(uDv + vDu) + d(Du)Dv + auvJ
=c(v- 1Dv + u- 1 Du) + d(u- 1Du)(v-1Dv) + au-1uv-1v
=c(DLu + DLv) + d(DLu)DLv + a
and
H(uv) = dDL(uv) +c
= d[c(DLu + DLv) + d(DLu)(DLv) + a] + c
= d 2 (DLu)DLv + cd(DLu + DLv) + ade + c
= (dDLu + c) (dDLu + c) + (ad - c2 + c)
= (Hu)(Hv) + h.

•
Example 5.1. Let X E AD(a, c, d) and let u, v E leX) n dom D.
(i) If a = d = 0, c = e, then X is a Leibniz D-algebra, e E ker D and H = I.
(ii) If a = 0, c = e, d = doe and do E IF \ {O}, then X is a quasi-Leibniz algebra,
i.e.
D(xy) = xDy + yDx + do(Dx)Dy for x, y E dom D,
H = doDL + e, H(uv) = do(Hu)(Hv).
(d. Examples 1.6 and 2.4(iv)).
(iii) If a = d = 0, c = ~e, then X is a simple Duhamel D-algebra: D(xy) = xDy =
yDx, i.e.
1
D(xy) = 2(xDy + yDx) for x,y E dom D and H = I.
Reduction theorems 65
(cf. Example 2.3(v)).

(iv) If a = c = 0, d = e, then D E M L(X) (cf. Example 1.8 for p = 0). 0
Consider now examples of D-algebras with non-scalar coefficients a, c, d.
Example 5.2. Let X = (s)JF be the space of all sequences x = {x n }, Xn E IF
(n EN), with the coordinatewise convergence. Clearly, X is a commutative algebra
over IF with the usual coordinatewise multiplication of elements. Write 8mn = 1 for
m = nand 8mn = 0 otherwise (m, n E Z), 8 = {8 mn }. Let D{x n } = {(n+ l)x n+l}
and R{x n } = {~xn-d, where we set Xn-k = 0 whenever n ::::; k. It is easy to
verify that D E R(X) and R E 'RD. An initial operator for D corresponding to R
is defined by F {xn} = Xl {8ln }. Moreover, we have
1
D(xy) = {--I}(Dx)Dy, where x = {x n }, y = {Yn} E X.
n+
Thus here we have a = c = 0, d = {n~l}. Hence h = o. o

Example 5.3. Let X = (s)JF be defined as in Example 5.2. Consider the operator
of forward shift: D{x n } = {xn+d. The operator D is mUltiplicative: D(xy) =
(Dx)(Dy) for X,y E dom D = X. The operator D is also right invertible and
ker D = {c8 : c E IF}, so that dim ker D = 1. The operator of backward shift:
R{x n } = {xn-d (where we set as in Example 5.2, Xn-k = 0 whenever n ::::; k)
is a right inverse of D. If F{x n } = x 1 8 then F is an initial operator for D
corresponding to the right inverse R. If Fl {xn} = (Xl + x2)8 then Fl is also an
initial operator for D corresponding to the right inverse Rl {xn} = {xn-d - x 1 8.
It is easy to verify that R is a Volterra right inverse, i.e. every.A E IF \ {a} is
a regular value of R. However, if Xl i- 0 then -1 is an eigenvalue of Rl and
ker(I + Rt} = xt{( _1)n}.
Clearly, the basic equation Du = uDx can be written as {un+l} = {unxn+d.
Then dom 0 = {u : Un i- 0 for n 2: I}. Let (u, X) E graph o. Then logarithms of
u and antilogaritms of X are of the form
for n = 1;
where Xn = { 0 Un for n 2: 2;
Un-l
1II
1 for n = 1;
X2 for n = 2;
where u~ = n
Xj for n 2: 3;
j=2
respectively. Clearly, L,E E M(X) (cf. Proposition 2.6).
Let P = {Pn} E X and Po = 1. Let Dpx = Dx - px = {xn+l - Pnxn} for

x = {Xn} E X (cf. Example 1.8). It is easy to verify that Dp E R(X) and that
66 Chapter 5
Rp E RD, where
if n = 1;
if n = 2;
otherwise.
Moreover,
ker Dp = {z = r{po ...Pn-d : r E IF},
e = {en}, where en = 1 for n E N, and
Here we have e = p, d = e, a = p2 - p, hence h = ad - e2 + e = 0. o

Corollary 5.1. Let X E AD(a, c, d) and (L, E) E G[n). If Le = °then ad = 0.
c
°
Proof. If Le = then He = dDLe + c = c. This, and (5.1) together imply that
= He = He 2 = (He)(He) + h = c2 + (ad - c2 + c) = ad + c. Hence ad = 0. •
Proposition 5.2. Suppose that X E AD(a,c,d). If DO = dD + cI then X E
ADo(h,O,e), where h = ad - c2 + c.
Proof. By definition, dD = DO - cI and for x, y E domD. Hence
= dD(xy) + cxy
= d[c(xDy + yDx) + d(Dx)Dy + axy) + cxy
= c(xdDy + ydDx) + (dDx)dDy + (ad + c)xy
= c[x(DOy - cy) + y(DOx - ex)) + (DOx - cx)(DOy - cy) + (ad + c)xy
= c(xDOy + yDOx - xDOy - yDOx) + (DOx)DOy
+ (ad + c - 2c2 + e2 )xy
= (DOx)(DOy) + hxy.
(cf. also LAUSCH and the author LPR[2)). •

Corollary 5.2. Suppose that X E AD(a, c, d) and d E I(X). If DO = dD + cI
and a = cd-1(c - e) then DO E ML(X), i.e. X E ADo(O,O,e).
Proof. Indeed, if a = cd-1(c - e) then h = ad - e2 + e = c(c - e) - e2 + c = O. •
Proposition 5.3. Suppose that X E AD(a,c,d) and a = cd-1(c - e), dE I(X).
If the operator dI + Re is invertible for an R E RD then DO = dD + cI E
R(X) n ML(X), RO = d(dI + Rc)-lRd- 1 E RDo, i.e. X E ADo(O,O,e). In
particular, if
(5.2) c = Qe, d = {3e, where Q,{3 E IF\ {O}, 73 E VF-R for an R E RD

-Q
then DO E R(X) n ML(X).

Proof. By our assumptions, the operator (dl + RC)-l is well-defined and
= (dD + cl)d(dl + RC)-lRd- 1

=dD(dl + Rc)d-1d(dl + RC)-l Rd- 1
=dDRd- 1 = dld- 1 = I.
By Corollary 5.2, DO E ML(X), hence X E ADo(O,O,e). Clearly, if Condition
(5.2) is satisfied then the operator dl + cR = ~(I + ~R) is invertible. •
Note 5.1. If c = 0, then the assumption that the operator dl + Rc is right in-
vertible, is automatically satisfied. Similar results can be obtained by substitution
Do = Dd+cl. 0
Proposition 5.4. Let X E AD(O, e, <5e) with <5 E IF \ {O}. If ~ E 1JFR for an
R E 'RD then DO = <5D + I E ML(X) n R(X) and W = l-R E 'RDo, hence
X E ADo(O,O,<5e).
Proof. Here we have a = 0, c = e, d = <5e. Hence h = ad - c2 + c = O. Further
arguments are the same as these used in the proof of Proposition 5.3. •
Proposition 5.5 Suppose that X E AD(a,c,d), dE leX), there is an R E 'RD
such that the operator dl +Rc is invertible and a = d-1c(c-e). Let DO = dD+cI,
0° : dom DO = dom D --+ 2dom DO,

OOu = {y E dom DO: DOu = uDOy} for u E dom DO.
If (LO, EO) E G[OO] then X E Lg(D) and (L, E) E G R,l [0], where
Proof. By our assumptions and Proposition 5.3, DO E M L(X) n R(X). By

Proposition 2.6, £0, EO E ML(X). Let (u,x) E graph 0, (u,y) E graph 0°. Then
x = Lu, u = Ex, y = LOu, u = EOy and
duDx = dDu = (dDO - cI)u = dDou - cu = duDy - cu.
Hence for u E leX)
which implies Lu = R(DO LOu - d-1c) + z, where z E ker D. Arguing as before,

without loss of generality we may put z = 0 (cf. Corollaries 2.4, 2.5). Then
Ex = u = ELu = ER(DO LOu - d-1c) = EO LOu = EOy. It is easy to verify that
= =
for every x there is a unique u such that Lu x, namely, u EO(Dx + Dd-1c).
Hence L is an invertible selector of O. •
68 Chapter 5
Theorem 5.1. Suppose that X E AD(O, c, 0) and e - c is not a zero divisor. If

either c = ~e or e E ker D or e - 2c is not a zero divisor then X (j. Lg(D).
Proof. By our assumption, D(xy) = c(xDy + yDx) for x,y E dom D. Then
De = De 2 = 2ceDe = 2cDe, i.e.
(e - 2c)De = O.
Suppose that e E ker D and (L, E) E G[n]. Then for every x E [dom n \ ker D] n
J(X) we have Dx = D(xe) = c(xDe + eDx) = cxDe + cDx, i.e. (e - c)Dx = 0
and DLx = (e - c)x-1xDx = 0, which implies DLx = O. Hence Lx = z E ker D
and x = Ez E ker D, which contradicts our assumption that x (j. ker D.
If e - 2c is not a zero divisor then, by (5.2), we get De = 0, i.e. e E ker D. Again
we conclude that invertible selectors of n do not exist.
Suppose then that c = ~. Let (L, E) E G[n]. Let u E dom n and let x = Lu.
Then u = Ex. Since here CD = ~ and fD = 0, we get
1 1 1 1
Lu = L(ue) = L[(Ex)(Ee)] = LE['2(x + Le)] = '2 (x + Le) = '2 Lu + '2Le.
Hence Lu = Le for all u, i.e. L is a constant mapping. This contradicts our
assumption that L is invertible. •
Example 5.4. Suppose that X = C([O, T]; q. Consider the so-called Pommiez
operator:
x(t) - x(O)
if t ~ 0;
(Dx)(t) = { t
x'(O) if t = O.
(cf. LINCHUK Li[l]).
Note that dom D = {x E C[O, T] : 3x'(0)}. The operator D is right invertible and
has a right inverse defined by (Rx)(t) = tx(t) for x E X. An initial operator F
corresponding to R is (Fx)(t) = x(O). Clearly, X is a D-algebra with respect to the
pointwise multiplication. This algebra has a unit e E ker D, namely the function
e(t) == 1. We shall show that in the case of the Pommiez operator invertible
selectors of n do not exist.
Indeed, let x, y E dom D. It is easy to verify that
1
D(xy) = '2[(x + Fx)Dy + (y + Fy)Dx] for x, y E dom D.
Suppose that (L, E) E G[n]. Let x, y E RX. Since F R = 0, we have Fx = Fy = 0

and D(xy) = ~(xDy + yDx) for x,y E RX. By the same arguments, as these
used in the proof of Theorem 5.1 for the case c = ~e, we conclude that invertible
selectors of n do not exist on RX c dom D. •
ExalTIple 5.5. Suppose that X E L(D) is a Leibniz D-algebra with unit e,

(L, E) E G[Ol, 9 = Re for an R E RD and the operator R' = 1- tgR is invertible.
Write D' x = gx - Dx for x Edam D = dam D'. The product rule for D' is
D'(xy) = gxy - D(xy) = gxy - xDy - yDx for x,y Edam D'.
Thus CD' = -1, fD' = gxy and X E AD,(g,-e,O), Le. here C = -e =fi e and
2c - e = -3e E I(X). We therefore conclude that X ft Lg(D'). In the classical
case, when X = C[O, T], D = ft,
R = J~, g(t) = (Re)(t) == t, we find D'x = tx-x'
for x E C I [0, T]. Clearly, the operator R' = I - ~t J~ is invertible. The operator
D' is the so-called the Hermite derivative (cf. ANTOSIK and MIKUSINSKI AM[I]).
o
Proposition 5.6. Suppose that X E AD(a, c, 0) and 2c - e E I(X). Let b =
a(2c - e)-I. Suppose that the operator I + Rb is invertible for an R E RD. Then
(i) DO = D + bE R(X), RO = (I + Rb)-I R E RDo and X E AD(O, c, 0);
(ii) X E Lg(D) only if c = e.
Proof. If the operator 1+ Rb is invertible for an R E RD then DO = D + b =
D(I + Rb)-I E R(X) and RO = (I + Rb)-I R E RDo. Moreover, for x, y E
dom DO = dom D we have
DO(xy) = D(xy) + bxy

= c(xDy + yDx) + axy + bxy
= c[x(DO - b)y + y(DO - b)x] + (a + b)xy
= c(xDOy + yDOx) + (a + b - 2bc)xy
= c(xDOy + yDOx)
for a = b(2c - e) = 2bc - b. Hence X E ADo (0, c, 0). By Theorem 5.1, invertible
selectors of 0 exist only if c = e. •
Propositions 5.5, 5.6 and Theorem 5.1 together imply that any algebra X E
AD(a, c, d) belongs to one of three types: either X is a Leibniz algebra or D
may be reduced by a substitution to a multiplicative operator or logarithmic and
antilogarithmic mappings do not exist.
We shall now construct logarithmic mappings induced by polynomials in a a right
invertible operator D. Recall that, by Theorem 3.1, if X E Lg(D) then X E
Lg(Dn) for all n E N.
TheorelTI 5.2. Suppose that X E Lg(D), (Ln' En) E G[On] and
N
P(D) =L Pk Dk , Po,···,PN EX, PN = e.
k=O
Set
70 Chapter 5
OOu = {y E dom DO : DOu = uDOy} far u E dom DO.

Let Ok be defined as in Theorem 3.1 and let (Lk' Ek) E O[Ok] for k = 1,2, ... , N.
Suppose that the operator
N
P(I,R) =L Pk RN - k
k=O
is invertible for an R E 'RD and
N
(5.3) LOu = RN[P(I, R)t 1(pou- 1 + L PkDLku) far u E dom ON n leX).
k=1
If the mapping £0 is invertible then X E Lg(DO) and (£0, EO) E 0 R,N[O°], where
EO = (£0)-1.
Proof. By our assumptions, DO = P(D) E R(X) and RO = RN[p(I, R)]-1 E 'RDo
(cf. PR[8], also Chapter 1). Let (u, x) E graph 0°. Then we have
N
= u- 1P(D)u = u- 1 L Pk Dku
k=O
N N
= L Pk U - 1 Dku = Pou- 1 + L Pk Dk Lku,
k=O k=1
which implies that x is of the form (5.3), i.e. x = £Ou. If the mapping LO is
invertible and EO = (LO)-1 then (£0, EO) E 0[0°]. Let F be an initial operator
for D corresponding to R. Clearly, since F R = 0, we have F Di LOu = 0 (j =
0,1, ... ,N -1). •
Note 5.2. If PN =j:. e, however PN E leX), then by substitution: p~ = pi/Pk we
obtain a polynomial
N
PI(D) = L p~Dk
k=O
satisfying the conditions of Theorem 5.2. o
Theorem 5.3. Suppose that X E Lg(D), (Ln, En) E O[nn] and
N
(5.4.) Q(D)=L Qk Dk , QkEL(X) (k=O,l, ... ,N-l), QN=I.
k=O
Set
DO = Q(D), no: dom DO = dom DN ---t 2dom DO,
nou = {y E dom DO: DOu = uDOy} far u E dom DO.

Let Ok be defined as in Theorem 3.1 and let (Lk, Ek) E G[OkJ [or k = 1,2, ... , N.
=L
N
(5.5) Q(I, R) QkRN-k
k=O
is invertible [or an R E RD and
lEthe mapping LO is invertible then X E Lg(DO) and (LO, EO) E GR,N[ooJ, where
EO = (Lo)-l.
Proof By our assumptions, we conclude that DO = Q(D) E R(X) and RO =
RN[Q(I, R)J-1. Let LO be defined by Formula (5.6). Since, by our definitions,
Q(D)R N = Q(I, R), we find, for all u E I(X) n dom 0° and for x = Uu that
= Q(D)LOu = Q(D)RN[Q(I, R)]-1[u- 1Q(D)u]

=Q(I, R)[Q(I, R)t 1[u- 1Q(D)u]
= u- 1Q(D)u = u- 1DOu.
This proves that LO is a selector of 0°. Further arguments are the same as in the
proof of Theorem 5.2. •
Theorem 5.4. Suppose that X E Lg(D), (Ln' En) E G[OnJ and
N
(5.7) Q(D) = L DkQk'
k=O
Qk E L(X), QkX E dom DN (k = O,l, ... ,N -1), QN = I.

Set
DO = Q(D), 0°: dom DO = dom DN --+ 2dom DO,
OOu = {y E dom DO : DOu = uDOy} for u E dom DO.

Let Ok be defined as in Theorem 3.1 and let (Lk, E k ) E G[Ok] [or k = 1,2, """' N.
N
(5.8) Q(I, R) =L RN-kQk
k=O
is invertible for an R E RD, (LN' EN) E G R,N[ON]. Write for u E dom 0° n I(X)
72 Chapter 5
If the mapping £0 is invertible then X E Lg(DO) and (LO, EO) E C R,N[nO], where
EO = (£0)-1.
Proof. By our assumptions, we have DO = Q(D) E R(X) and W = [Q(I, R)]-l RN
(cf. PR[8], Section 3). The similarity property:
(cf. PR[8], Corollary 3.2.1) implies that for x = LOu, where £0 is defined by
Formula (5.9), we get
u- 1 Q(D)u
= u- 1 DN Q(I, R)u
= u- 1 (Q(I, R)u)DN LN(Q(I, R)u).
Further arguments are the same as in the proofs of Theorem 5.2 and 5.3. •
Note 5.3. We can obtain similar results (under appropriate assumptions) for
operators of the form
All these operators are right invertible (cf. PR[8], Section 3). NGUYEN VAN MAU
(cf. N[2)) proved that the resolving operators Q(I, R) and Q(I, R) are simultane-
ously invertible (left invertible, right invertible) or not. In the same paper he has
considered operators of a more general form
M N
Q[D] = L L D m AmnD n , with Amn E L(X)
m=O n=O
which are again right invertible (under appropriate assumptions). Thus, apply-
ing his results, one can construct logarithmic mappings for the operator Q[D]
(provided that they exist) in a similar manner as in Theorems 5.3 and 5.4. 0
Proposition 5.7. Suppose that X E Lg(D') n Lg(D"), both superpositions
D' D", D" D' exist and D'D" = D" D' = D on dom D = dom D' n dom D". Let
(L', E') E C[n'], (L", E") E C[n"], where
n': dom D' ~ 2dom D', n'u = {x E dom D': D'u = uD'x}, u E dom D',
n": dom D" ~ 2dom D",

n"u = {x E dom D" : D"u = uD"x}, u E dom D".
Write
n : dom D ~ 2dom D, nu = {x E dom D : Du = uDx}, u E dom D.

Let u E leX) n dom n. Let R' E 'RD" R" E 'RD'" Then

(i) there is a selector L of n of the form
Lu = CD' L"u + R[cD' (D' L'u)D" L"u + u- 1 fD' (u, D" L"u)],
where R = R'R" E'RD;

(ii) if CD' i- 0 then L is invertible;
(iii) if CD' = 0, however the mapping f defined by feu) = fD,(u,D"L"u) for
u E dom n is invertible, then L is invertible;
(iv) if L is invertible and E = L- 1 then (L,E) E G[n];
(v) if e E ker D" and L"e = 0 then Le = O.
Proof. (i) Let (u, x) E graph nand u E leX). Then, by definition,
DLu= Dx
=u- 1 Du = u- 1 Du = u- 1 D' D"u
= u- 1 D'(uu- 1 D"u) = u- 1 D'(uD" L"u)
= u- 1 {CD' [(D'u)D" L"U + uD' D" L"u] + fD' (u, D" L"u)}
= CD,[(u- 1 D'u)D"L"u+ DL"u] + u- 1 fD'(U,
D"L"u)
= CD' [(D' L'u)D" L"u + DL"u] + u- fD' (u, D" L"u),
1
which implies that L is of the required form (cf. Formula (1.4)).

(ii) Let CD' i-
O. For every x E dom n- 1 and for every fixed x" E dom n"
such that D" x" E leX) there is a unique x' E dom (n,)-l satisfying the equation
x = CD'X" +R[CD' (D'x')D"x"] such that F'x' = 0, where F' E FD' corresponds to
R'. Namely, x' = R'[(D"x")-l D(cD~x - x"]. Having already uniquely determined
x', we conclude that u = E' x' is a unique solution of the equation Lu = x.
(iii) and (iv) are obvious.
(v) Suppose that e E ker D. Then L'e, L"e E ker D which implies D'L'e =
D" L" e = O. Hence Le = CD' L" e = O. •
Since, by our assumptions, D'D" = D" D' = D on dom D, interchanging the roles
of D' , D", we conclude that there is a selector L of n of the form
Lu = cD"L'u + R"R'[CD,,(D'L'u)D"L"u + fD,,(u,D'L'u)]

such that Lu = 0, provided that e E ker D and L'e = O. If we apply Formula
(1.12) instead of Formula (1.11) then we obtain a symmetric form of selectors L
(but more complicated).
Corollary 5.3. Suppose that all assumptions of Proposition 5.7 are satisfied.
Suppose, moreover, that X is a Leibniz D'-algebra and D" -algebra. Then X E
Lg(D), (L, E) E G[f!J, where E = L -1,
L(uv) = Lu + Lv + Rd(u, v),
74 Chapter 5
d(u, v) = (D' L'u)D" L"V + (D" L"u)D' L'v far u, v E dom 0

and
(Ex)(Ey) = E[x + y + Rd(Ex, Ey)], far x, y E dom 0- 1 .
Proof. By our assumption, CD' = 1. This, and Proposition 5.7(ii), (iv) now implies
that X E Lg(D) and (L, E) E C[O]. For all x, y E dom D we have
D(xy) = D'D"(xy)
= D'(xD"y + yD"X)
= (D'x)D"y + xD'D"y + (D'y)D"x + yD'D"x
= xDy + yDx + (D'x)D"y + (D"x)D'y.
Hence for u, v E J(X) n dom 0 we get
DL(uv)
= (uv)-1 D(uv)
= (uv)-I[uDv + vDu + (D'u)D"v + (D"u)D'v]
= V-I Dv + u- 1Du + (u- 1D'u)(v- 1D"V) + (u- 1D"u)(V- 1D'v))
= DLu + DLv + (D' L'u)D" L"V + (D" L"u)D' L'v
= DLu + DLv + d(u, v).
Further arguments are the same as in the proof of Theorem 2.2 (if we apply the
commutativity of the algebra X). •
Example 5.6. Suppose that E = [0, a] x [O,b] E IR!, X = C(E;IR), D' = and
D" = %s' Then X is a Leibniz D'-algebra and D"-algebra with respect to the
:t
pointwise multiplication, i.e. CD' = CD" = 1 and fD' = fD" = 0. This algebra
has a unit e. Namely, e(t, s) == 1. Moreover, e E ker D' n ker D". Note that both,
ker D' and ker D" are infinite dimensional. The operator
EP
D = atas'
as a superposition of right invertible operators, is again right invertible. It is
well-known that
a2 x a2 x
atas = asat'
provided that both these derivatives are continuous. Several examples of right
inverses to the operator D can be found in the book PR[8]. The operator D
satisfies all conditions of Corollary 5.3, 0
Proposition 5.8. Suppose that X E Lg(D(j)) (j = 1, .. " n) and the operator
n
n n
D =L D(j) with dom D = dom D(j)
j=1 j=l
is right invertible. Let (L(j), E(iJ) E G[O(il], where
O(j)U = {x E dom D(j) : D(j)u = uD(j)x}, U E dom D.

Write
0: dom D ----+ 2dom D, Ou = {x E dom D: Du = uDx}, u E dom D.
Then
(i) there is a selector L of 0 of the form
n
Lu = L L(j)u far u E J(X) n dam 0;
j=1
(ii) if L is invertible and E = L -1 then (L, E) E G[O];

(iii) if L(j)e = 0 for j = 1, ... , n then Le = O.
Proof. Let u E J(X) n dam O. By definition, we have
DLu = u- 1 Du
n n
= u- 1 L D(j)u = L u- 1 D(j)u
j=1 j=1
n n
=L D(j)L(j)u = D L L(j)u,
j=1 j=1
which implies (i). Points (ii) and (iii) are obvious. •

Note 5.4. Proposition 5.8 can be applied in order to construct logarithmic map-
pings for partial differential operators of the first order with variable coefficients:
To do this, one can apply the construction of right inverses for D given by VIR-
SIK V[I] (in a more general case, over JR). This construction, however, is too
complicated to be given here as an example. D
Proposition 5.9. Suppose that X E Lg(D(j)) (j = 1, ... , n) and the operator
n
n n
D= L Drj) with dom D = dom D(j)
j=1 j=l
76 Chapter 5
is right invertible. Let (L(j),E(j)) E G[O(j)], where
O(j)U = {x E dom D(j) : D(j)u = uD(j)x}, U E dom D.

Write
0: dom D --+ 2dom D, Ou = {x E dom D: Du = uDx}, for u E dom D.
Then
(i) there is a selector L of 0 of the form
n
Lu = L {CD(j) [L(j)u + Rfj) (D(j)L(j)u)2] + u- 1 fD(j) (u, D(j)L(j)u)}
j=l
for u E J(X) n dom 0 and an R E 'RDi

(ii) if L is invertible and E = L -1 then (L, E) E G[O]i
(iii) if L(j)e = 0 for j = 1, ... , n then Le = O.
Proof. Let u E J(X) n dom O. By Theorem 3.1, we have
DLu = u- 1 Du
n n
-_ u- 1 'L..- 2 U
(j) = 'L..-
" D2 U " u- 1 D (j)
j=l j=l
n
= D L {cD(j)[L(j)u
j=l
which implies (i). Points (ii) and (iii) are obvious.

Corollary 5.4. Suppose that all assumptions of Proposition 5.9 are satisfied and
•
X is a Leibniz D(j)-algebra (j = 1, ... , n). Then there is a selector L of 0 such
that
L(uv) = Lu + Lv + Rc5(u, v),
where
n
c5(u,v) = 2L (D(j)L(j)u)D(j)L(j)v for u,v E dom O.
j=l
If L is invertible then X E Lg(D), (L, E) E G[O], where E = L -1 and
(Ex)(Ey) = E[x + y + Rc5(Ex, Ey)], for x, Y E dom 0- 1 .

Proof. It is easy to verify that
n
D(xy) = xDy + yDx + L (D(j)x)D(j)Y for x, y E dom D.
j=l
Let u, v E J(X) n dom n. Then

DL(uv) = (UV)-I D(uv)
n
= (uv)-l[uDv + vDu + 2 L (D(j)u)D(j)v]
j=l
n
= u- 1 Du + V-I Dv + 2 L(u- 1 D(j)u)v- 1 D(j)v
j=l
n
= DLu + DLv + 2 L (D(j)L(j)u)D(j)L(j)v
j=l
= DLu + DLv + 6(u,v).
Further arguments are the same as in the proof of Theorem 1.2 (if we apply the
commutativity of the algebra X). •
Example 5.7. Let 0 E JRn be a domain with the boundary ao of the Liapunov
type. Let X = C(OjJR) and let D(j) = 8~. (j = 1, ... ,n), t = (tI, ... ,tn ) EO.
)
Clearly, X is a Leibniz D(j)-algebra with unit e(t) == 1 with respect to the pointwise
multiplication. The Laplace operator
is right invertible in X (cf. PR[8]) and satisfies all condition of Corollary 5.4. 0
Example 5.8. Let D E A(X). Denote by ND(a,b,c,d) the class of algebras X
with the following product rule for D:
(5.10) D(xy) = cxDy + b(Dx)y + d(Dx)Dy + axy for x,y E dom D,
where a, b, c, d EX. Non-Leibniz components of algebras belonging to the class

ND(a, b, c, d) have been considered by KORNACKI Kor[1]. He has shown that for
X END (a, b, c, d) the following assertions hold:
(i) (e - c - b)De = d(De)2 + aj
(ii) if e E ker D then a = OJ
(iii) if e E ker D then (c - e)Dx = (b - e)Dx = OJ
(iv) if d = a = 0 and e - c - b is not a zero divisor then e E ker Dj
78 Chapter 5
(v) if b - e and e - e are not zero divisors then dom D = ker D.

Write now
(5.11) ND(a,e,d) = {XE ND(a,b,e,d) : b = e}.
It means that the product rule for an X E ND(a, e, d) is of the form
(5.12) D(xy) = e[xDy + (Dx)y] + d(Dx)Dy + axy whenever x, y E dom D.
Denote by yc the eommutant of a subalgebra Y C X E ND(a, e, d), Le.
y c = {y E Y: V xy = yx}.
xEY
Suppose that X E NCa,e,d) and e,d E Xc. If DO = dD + cl then

DO(xy)=(DOx)DOy+hxy for x,yEdomD, where h=ad+e-e2 •
If h = ad+ e - e2 = °
then DO = dD + eI E ML(X), i.e. X E NDo(O,O,e). 0
Consider now left invertible operators.
Definition 5.2. Denote by BD(a,e,d) the class of algebras X with D E A(X)
such that the product rule is given by the same formula as in Definition 5.1:
D(xy) = e(xDy + yDx) + d(Dx)Dy + axy for x, y E dom D = X,

where a, e, d E X depends on D only and do not vanish simultaneously. We denote
by B'b(a, e, d) the class of algebras with unit e belonging to BD(a, e, d). 0
Taking into account Theorem 5.1 and Proposition 5.6, we conclude that Theorem
4.3 implies the following
Theorem 5.5. IE X E BD(O, e, 0) then X (j Lg#(D).

Proposition 5.10. Let X E BD(a, e, d). Suppose that X E Lg#(D) and (L, E) E
G[L]. If Hu = dDLu + e then
H(uv) = (Hu)Hv + h, for u, v E I(X) n dom L where h = ad - e2 +e

(d. the proof of Proposition 5.1).
° °
Proposition 5.11. Let X E B'b(a, e, d). Suppose that X E Lg#(D) and (L, E) E
G[L]. IE Le = then ad = (ef. Corollary 5.1).
Proposition 5.12. Suppose that X E B'b(a, e, d), dE I(X), DO = dD + eI and
a = ed-1(c - e). Then
(i) D E ML(X), i.e. X E B~o(O,O,e);
(ii) if the operator dI + Se is invertible for an S E CD then DO E A(X) and
SO = d(dI + Sc)-lSd- 1 E CDO;
(iii) if e = o:e, d = (3e, where 0:,(3 E IF \ {O} and -fjo E vIFS, then DO E A(X) n
ML(X) and SO = (0:1 + (3S)-lS E CDo.
Proof (cf. the proof of Proposition 5.2, Corollary 5.2 and Proposition 5.3). We
only have to check that SO E £ DO. Indeed,
SO DO = d(dI + Se)-lSd-l(dD + eI) = d(dI + Se)-l(Sd-ldD + Sd-leI) =
= d(dI + Se)-l(I + Sd-le) = d(dI + Se)-l(dI + Se)d- l = I.
Proposition 5.13. Suppose that all assumptions of Proposition 5.12 are satisfied.
•
Let DO = dD + eI,
nOu = {y E dom DO: DOu = uDOy} for u E dom DO.

If (LO, EO) E G[LO] then X E Lg# (D) and (L, E) E G[L], where

Theorem 5.6. Suppose that X E Lg#(D), (Ln' En) E G[Ln] and
N
P(D) =L Pk Dk , Po, ···,PN E X, PN = e.
k=O
Let
DO = P(D), nO: dom DO = dom DN --t 2dom DO,
nOu = {y E dom DO: DOu = uDOy} for u E dom DO.
Let nk be defined as in Definition 3.1 and let (Lk, Ek) E G[Lk] for k = 1,2, ... , N.
N
P(I, S) = L Pk RN - k
k=O
is invertible for an S E CD and
N
LOu = SN[P(I, S)]-l(pou- l + L PkDLku) for u E dom LN n I(X).
k=l
If the mapping L O is invertible then X E Lg# (DO) and (LO, EO) E G[LO], where
EO = (LO)-l.
80 Chapter 5
Proof. By our assumptions, DO = P(D) E A(X) and SO = SN[p(I, S)l-1 E LDO.

Indeed, we have
N N
P(D) =L PkSN-kDN-kDk =L PkSN-kDN = P(I,S)D N .
k=O k=O
Further considerations are similar to those in the proof of Theorem 5.2, if we take
into account the fact that ker D = {o}. •
Theorem 5.7. Suppose that X E Lg#(D), (Ln, En) E G[Lnl and
N
Q(D) =L QkD\Qk E L(X) (k = 0,1, ... , N - 1), QN = I.
k=O
Set
DO = Q(D), nO: dom DO = dom DN ~ 2dom DO,
nOu = {y E dom DO: DOu = uDOy} for U E dom DO.
Let nk be defined as in Definition 3.1 and let (Lk' E k ) E G[Lkl for k = 1,2, ... , N.
N
Q(I, S) = L QkSN-k
k=O
is invertible for an S E SD, (LN' EN) E G[LNl and
If the mapping £0 is invertible then X E Lg#(DO) and (LO,EO) E G[LO], where

EO = (LO)-1 (cf. Theorem 5.3).
Similarly, we have
Theorem 5.B. Suppose that X E Lg#(D), (Ln, En) E G[Lnl and
N
(5.13) Q(D) =L DkQk,
k=O
Qk E L(X), QkX E dom DN (k = 0,1, ... ,N -1), QN = I.

Set
DO = Q(D), nO: dom DO = dom DN ~ 2dom DO,
nOU = {y E dom DO: DOu = uDOy} for u E dom DO.

Let Dk be defined as in Definition 3.1 and let (Lk, E k ) E G[Lk) for k = 1,2, ... , N.
N
Q(I,S) = L SN-kQk
k=O
is invertible for an S E CD, (LN' EN) E G[LN) and
for U E dom LO n J(X). If the mapping LO is invertible then X E Lg#(DO) and

(LO, EO) E G[LO), where EO = (LO)-l (cf. Theorem 5.4).
Note 5.5. Similar results can be obtained for operators of the form
since all these operators are left invertible under assumptions of Theorems 5.7 and
5.8, respectively (cf. Note 5.3). 0
To conclude Sections 2-5, note that the existence (or non-existence) of right and
left logarithmic and antilogarithmic mappings and logarithmic and antilogarithmic
mappings is a property of an algebraic structure together with an operator and
not a property of the operator only.
In the next sections we shall consider particular properties in algebras with mul-
tiplicative operator D and in Leibniz algebras and formulae of solutions to linear
equations.
CHAPTER 6
MULTIPLICATIVE CASE
We shall see that in the case of algebras with multiplicative operators they may
appear some pathologies. One of them has been shown by Theorem 2.4. Now we
shall study these operators in commutative algebras.
To begin with, we shall prove the following
Lemma 6.1. Suppose that X is a commutative algebra over IF with unit e,
A E M(X), A"f 0, dom A has no zero divisors and e E dom A. Then
(i) Ae = e;
(ii) Ax E leX) and Ax- 1 = (AX)-l for x E leX) n dom A;
(iii) Axn = (Ax)n for all x E dom A, n E N;
(iv) Ax E lz(dom A) and Ax 1 / Z = (Ax?/z for x E 12(dom A).
Proof. (i) Since Ae = Ae z = (Ae)(Ae), we find (Ae)(Ae - e) = O. Since dom A
has no zero divisors, we have either Ae = 0 or Ae = e. Suppose then that Ae = o.
Then for all x E dom A we have Ax = A(xe) = (Ax)(Ae) = 0, which contradicts
our assumption that A "f o. Thus Ae = e.
(ii) Let x E I(X)ndom A. Then e = Ae = A(xx- I ), which implies that Ax E leX)
and (AX)-I = AX-I.
(iii) is proved by an easy induction.
(iv) Let x E 12(dom A). Then there is a y such that y2 = x, i.e. y = XI/2. Hence
(Ay)2 = Ay2 = Ax, which implies that Ax E 12(dom A) and AX I / 2 = (AX)I/Z . •
Proposition 6.1. Suppose that X is a D-algebra with unit e E dom D, D E
M L(X) and dom D has no zero divisors. Then
(i) De = e;
(ii) Dx E leX) and Dx- I = (DX)-I for x E leX) n dom D;
(iii) Dxn = (Dx)-n for xE dom D, n E N;
(iv) Dx E 12(dom D) and DX I / 2 = (DX)I/2 for x E 12(dom D);
(v) ker D n leX) = 0 and (dom D)(ker D) C ker D.
Proof. Assertions (i)-(iv) are immediate consequences of Lemma 6.!.
(v) Suppose that z E ker D n leX). By (i), De = e. Since Dz = 0, we get
e = De = D(zz-l) = (Dz)Dz- 1 = 0, a contradiction. Let x E dom D, z E ker D
82
Multiplicative case 83
be arbitrary. Then D(xz) = (Dx)(Dz) = 0, which implies xz E ker D. The

arbitrariness of x and z imply (v). •
In this section we shall admit the following condition:
[M] X E Lg(D), D E M L(X) and dom D has no zero divisors.
By Condition [M], we have X E AD(O, 0, e) n Lg(D). Since ker Dc dom D, the
property that ker D has no zero divisors assumed in the definition of the class
AD (a, c, d) is automatically satisfied.
Proposition 6.2 Suppose that Condition [M] holds, (L, E) E G[O] and e E
domO- I . Then
= Le = e;
(i) e E dom 0 n dom 0- 1 and Ee
1
(ii) Ex E I(X) and Ex- = (EX)-1 for all x E I(X) n dom 0- 1 ;
(iii) Exn = (Ex)n for all x E dom 0- 1 , n E N;
(iv) Ex E 12(dom 0- 1 ) and EX 1/ 2 = (EX)I/2 for x E 12(dom 0- 1);
(v) if aJ3 E IF and ae, /3e E dom 0- 1 then E(a/3e) = E(ae)E(/3e)j
(vi) 0 ¢ dom 0- 1 .
Proof. By Proposition 2.6, Land E are multiplicative. Points (i)-(iii) are proved
by Lemma 6.1(i)-(iii). Moreover, if Ee = e then Le = e.
(iv) Let x E 12(dom 0- 1 ). By Proposition 6.1, x 1/ 2 E dom D and Dx 1/ 2 =
(DX)I/2. Hence Ex = E(X 1/ 2x 1/ 2) = (EX 1/ 2)2, i.e. Ex E 12(dom 0- 1 ) and
Ex 1/ 2 = (EX)I/2.
(v) Let a, /3 E IF and ae, /3e E dom 0- 1 . Then E(a/3e) = E(ae/3e) = E(ae)E(/3e).
(vi) Suppose that 0 E dom 0- 1 . By Proposition 2.5, z = E(O) E ker D. By
definition, Dz = DE(O) = E(O)De = ze = z, which implies z = 0, i.e. E(O) = O.
Hence LO = LEO = 0, i.e. 0 E dom O. This contradicts Proposition 2.5. •
Definition 6.1. Suppose that either IF = C or IF = R, Condition [M] holds and
(L, E) E G[O]. We say that E is multiplicative with exponent). if there is a
). E R+ \ {O} such that
E(ae) = a"e for a E R+ \ {O}.
o
Proposition 6.3. Suppose that all assumptions of Definition 6.1 are satisfied,
E is multiplicative with exponent). > 0, x E dom 0- 1 and a E R+ \ {O}. If
ax E dom 0- 1 then a = 1.
Proof. Let a E R+ \ {O}. Let x, ax E dom 0- 1 . By Proposition 6.2(i),
e E dom 0- 1 . Suppose that ae E dom 0- 1 . By Proposition 6.2(v), E(ax) =
E(ae)Ex. By Definition 6.1, ae = LE(ae) = L(a"e). By Theorem 2.5, a A = 1,
i.e. ). = O.This contradicts our assumptions. Hence a = 1. •
84 Chapter 6
Note 6.1. Theorem 2.4 and Proposition 6.3 together show that few results of the
author's papers PR[16] and PR[18] are trivial. 0
Theorem 6.1. Suppose that Condition [M) holds and (L, E) E G[O). Let
y = {x E dom 0- 1 : u = Ex E J(X)}.
Then
(6.1) Y n ker D = {O}, i.e. Diy is invertible.
Proof By our assumption, D E M L(X). Hence CD = 0 and !D(U, v) = (Du)(Dv)

for u,v E dom D. Let (L,E) E GR,dO). Let U E J(X)ndom 0 be arbitrarily fixed
and let x = Lu. Then u = Ex E J(X). By definition, DEx = (Ex)Dx. Since X
is a commutative algebra, either by Formula (2.10) or by Formula (2.12),
Ex = E(ex)
= (Ee)(Ex) = E{R[e- 1(Ex)-1(De)(DEx)]}
= ER[e(Ex)-1e(Ex)Dx) = E(RDx) = E(x - Fx),
where F is an initial operator for D corresponding to R. Hence x = LEx =

LE(x - Fx) = x - Fx. Thus Fx = 0 and x E RX. Since dom D = RX EB ker D,
the arbitrariness of x E Y implies (6.1). •
Proposition 6.4. Suppose that Condition [M] holds and (L, E) E G[O]. Then
elements E( -e), L( -e) do not exist, in particular, -e ~ dom 0 n dom 0- 1 .
Proof By Theorem 2.5, -e ~ dom O. By our assumptions, if we put r = E( -e)
then we get
r2 = E( -e)E( -e) = E[( -e)( -e)] = E(e 2) = Ee = e.

Hence either r = e or r = -e.
Suppose that r = e. Then we have E( -e) = r = eEe and e = Le = Lr
LE( -e) = -e, which leads to a contradiction.
Suppose that r = -e. Then E( -e) = r = -e = -Ee, -e = -LE( -e) =
L(-Ee) = L(-e), which implies -e E dom 0, a contradiction. •
Corollary 6.1. Suppose that IF = C, Condition [M) holds and (L, E) E G[O].
Then elements E(ie), L(ie) do not exist, in particular, ie ~ dom 0 n dom 0- 1 .
Proof. Suppose that the element E(ie) exists. By Proposition 6.2(iii), we get
E(ie)E(ie) = E(i 2e) = E( -e). This contradicts Proposition 6.4. A similar proof
for L(ie). •
Lemma 6.2. Suppose that D E M L(X) and F is an initial operator for D
corresponding to an R E RD. Then
(i) R(xy) = (Rx)(Ry) - F[(Rx)(Ry)] for x, y E X;
(ii) if FE M L(X) then REM L(X);

(iii) if F E M L(X) then F[(Rx)(Ry)) = 0 for x, y EX;
(iv) if REM L(X) and F is almost averaging, i.e. F(zx) = zFx for z E ker D,
x E X, then FE ML(X).
Proof Recall that FR = O. Let x, y E X. We have DR(xy) = xy =
(DRx)(DRy) = D[(Rx)(Ry). Hence D[R(xy) - (Rx)(Ry)) = 0, i.e. R(xy) =
(Rx)(Ry) + z, where z = -F[(Rx)(Ry)) E ker D.
(ii) If FE M L(X) then
R(xy) = (Rx)(Ry) - F[(Rx)(Ry)) = (Rx)(Ry) - (FRx)(FRy) = (Rx)(Ry).
(iii) If REM L(X) then R(xy) = (Rx)(Ry). Hence
F[(Rx)(Ry)) = (FRx)(FRy) = O.
(iv) Let X,y Edam D. Then x = Ru + Z1, y = Rv + Z2, where u, v E X, Z1 = Fx,
Z2 = Fy E ker D and
F(xy) = F[(Ry + z1)(Rv + Z2))

= F[(Ru)(Rv)) + F(Z1Ru + z2Rv) + F(Z1Z2) =
= FR(uv) + z1FRv + z2FRu + z1Fz2
=Z1Z2 = (Fx)(Fy).
This proves that F is multiplicative.
Proposition 6.5. Let Condition [M) hold. Suppose that (L, E) E G[O), F E
•
M L(X) is an initial operator for D corresponding to an R E RD and x E VA(X) n
dom O. Then Ex = (I - XR)-1 z, where z E ker D is arbitrary.
Proof Since F E M L(X), Lemma 6.2(ii) implies that REM L(X). Let u = Ex.
Then Du = uDx = (DRu)(Dx) = D(xRu) and D(I - xR)u = 0, which implies
u = xRu+z, where z E ker D is arbitrary. Since x E VA(X)ndom D, the operator
1- xR is invertible and u = Ex is of the required form (d. Formula (2.26) and
Proposition 2.15). •
Proposition 6.6. Suppose that Condition [M] holds, a E X and there is an
R E RD such that Ra, 9 = Re E dom 0- 1 . Let (L, E) E G[O). Then
(6.2) x = E(Ra)E(g + z) E ker(D - a) for all z E ker D.
Proof By our assumption, D E M L(X). Since
D(g + z) = DRe + tJz = e,

86 Chapter 6
we get
Dx = D[E(Ra)E(g + z)]
= [DE(Ra)][DE(g + z)] = E(Ra)aE(g + z)e
= aE(Ra)E(g + z) + ax.
Proposition 6.7. Suppose that all assumptions of Proposition 6.6 are satisfied
•
and a, Ra E I(X). If x E ker(D - a) then x is of the form (6.2).
Proof. Proposition 6.2 and our assumptions together imply that E(Ra) E I(X)
and [E(Ra)]-1 = E[(Ra)-1]. Write u = x[E(Ra)]-1. Then x = E(Ra)u and
aE(Ra)(Du - u)
=aE(Ra)Du - aE(Ra)u = [DE(Ra)]Du - aE(Ra)u
= D[E(Ra)u] - aE(Ra)u = Dx - ax = O.
Since a, Ra E I(X), we conclude that Du = u = ue = uD(g + z), where z E ker D

is arbitrarily fixed. Hence u = E(g + z) and x is of the form (6.2). •
An immediate consequence of Propositions 6.6 and 6.7 is
Corollary 6.2. Suppose that Condition [M) holds, (L, E) E G[O], a E I(X) and
there is an R E RD such that 9 = Re E dom 0- 1 and Ra E I(X)ndom 0- 1 . Then
x E ker(D - a) if and only if x = E(Ra)E(g + z), where z E ker D is arbitrary.
Proposition 6.S. Suppose that either IF = C or IF = JR, Condition [M) holds,
(L, E) E G[O], E is multiplicative with an exponent A > 0, R E RD, 9 = Re E
I(X) n dom 0- 1 , a E :F and a = ae. If ag = aRe E dom 0- 1 then
(i) a = 1, i.e. Ra = g;
(ii) x E ker(D - I) if and only if
(6.3) x = E[g(g + z)], where z E ker D is arbitrary.
Proof. (i) is an immediate consequence of our assumptions and Proposition 6.3.

(ii) Put a = e in Corollary 6.2. Since E is multiplicative, by (6.2), we get
x = E(Re)E(g+z) = E(g)E(g+z) = E[g(g+z»), where z E ker D is arbitrary .
Theorem 6.2. Suppose that all assumptions of Corollary 6.2 are satisfied and
•
1 E vFR, i.e. the operator I - R is invertible (ef. (2.28)). Then all solutions of the
equation
(6.4) Dx = ax + y, yEX
are of the form
(6.5) x = Xo + Xl, where Xo E ker D and
(6.6)
Proof. By our assumptions, the operator I - R is invertible. Write
Then
Let Xl = E(Ra)u. We get
= D[E(Ra)u] - aE(Ra)u = [DE(Ra)]Du - aE(Ra)u

=aE(Ra)Du - aE(Ra)u = aE(Ra)(Du - u) = aE(Ra)Ya
= aE(Ra)a- l E(Ra)-ly = y.
Hence Xl is a particular solution of Equation (6.4). This, and Corollary 6.2 to-
gether imply that all solutions of that equation are of the form (6.5)-(6.6). •
An immediate consequence of this theorem and Proposition 6.8 is
Corollary 6.3. Let either IF = C or IF = JR. Suppose that Condition [M] holds,
(L, E) E G[n]; E is multiplicative with an exponent A > 0, R E RD and 9 = Re E
leX) n dom n- l . If 1 E vlF'R (i.e. if 1 is a regular value of R; cf. (2.28)), then all
solutions of the equation
(6.7) (D - l)x = y, YEX
are of the form X = Xo + Xl, where
(6.8) Xo = E[g(g + z)] E ker(D - AI), z E ker D is arbitrary,
(6.9) Xl = E(g)(l - R)-l R[(Eg)-ly].
Theorems 2.2 and 6.2 together imply

Corollary 6.4. Suppose that Condition [M] holds, (Ln' En) E G[nn] for all
n E N, a E leX) and there is an R E RD such that gn =
Rne E dom n;:;-l,
Rna E leX) n dom n;:;-l and 1 E vlF'R n (i.e. 1 is a regular value of Rn). Then all
solutions of the equation
(6.10) Dnx = ax + y, y E X, (n E N)
88 Chapter 6
are of the form x = Xo + Xl, where
Le.
n-l
Z= L Rk Zk, ZO, ... , Zn-l are arbitrary,
k=O
Note 6.2. If F is a multiplicative initial operator for D corresponding to R then,

by Lemma 6.2, R is multiplicative and Rna = (Ra)n, gn = Rne = (Re)n = gn.
By Theorem 3.4, also En E ML(X). Thus En(Rna) =
En[(Ra)n] =
[En(Ra)]n
and En(Rne + z) = En(gn + z). 0
We have proved that, in general, logarithms and antilogarithms are uniquely de-
termined up to a constant (cf. Corollary 2.4). In the multiplicative case the
dependence of a constant can be given in a different way. We will show this for
commutative algebras.
Proposition 6.9. Let X E Lg(D) and let (L,E) E OR.dO]. If DE ML(X) then
(V, E') E 0[0], where
L'u = L(zu), E'x = zEx for (u,x) E graph 0, z E ker D.
Proof. Let z E ker D. Let z' = Lz, i.e. z = Ez'. By Proposition 2.5, z' E ker D.
Since D E M L(X), by Proposition 2.6, L, E E M(X) and
= (Ez')Ex = zEx, L(zu) = (Lz)Lu.

E(z'x)
Hence DE(z'x) = D[(Ez')Ex] = D(zEx) = (Dz)DEx = 0, E(z'x)D(z'x)
E(z'x)(Dz')Dx = 0 and DL(zu) = D[(Lz)Lu] = (DLz)DLu = 0 = (zu)D(zu).
We therefore conclude that (L',E') E 0[0]. •
Proposition 6.10. Suppose that X E Lg#(D), (L, E) E alL], a E leX), D E
ML(X), Sa E dom E for an S E CD, Xo = E(Sa) E leX) and 1 E VFS then
x = xo(I - S)-1[(Sa-1)(Sxi)1)(Sy)]
is a unique solution of the equation
(6.12) Dx = ax + y, y E dom L
independently of the choice of S.
Proof. Let x be a solution of Equation (6.12) and let v = xi)lx. Recall that
Xo = E(Sa) E ker(D - a). Since D E ML(X), by our assumptions, we find
axov + yax + y = Dx = D(xov) = (Dxo)Dv = aXoDv,

which implies Dv - v = a-1xo1y. By Lemma 4.2, S E ML(X). Thus v - Sv =

S(Dv - v) = S(a- 1x o 1y) = (Sa- 1)(Sxo 1)Sy. By our assumption, the operator
I - S is invertible. We therefore conclude that x is of the required form. The
uniqueness of the solution follows from the fact that S is left invertible. •
Proposition 6.11. Suppose that D E ML(X)n A(X) n A(X), X has unit e E
dom D and dom D has no zero divisors. Then
(i) De = ei
(ii) Dx E I(X) and Dx- 1 = (Dx)-1 whenever x E I(X) n dom Di
(iii) Dx- n = (Dx)-n whenever xE dom D, n E Ni
(iv) Dx E 12(dom D) and Dx 1/ 2 = (DX)1/2 whenever x E h(dom D) (d. Propo-
sition 6.1.).
We will make use of the following condition:
[Ml# X E Lg~(D), DE ML(X) and dom D has no zero divisors.
If Condition [Ml# is satisfied, then X E B~(O,O,e)nLg#(D) (d. Definition 5.2).
Proposition 6.12. Suppose that Condition [MJ# holds, (L, E) E G[LJ and
e E dom L-1. Then
(i) e E dom L n dom L -1 and Ee = Le = ei

(ii) Ex E I(X) and Ex- 1 = (EX)-1 for all x E I(X) n dom L-1 i
(iii) Exn = (Ex)n for all x E dom L-1, n E N.
Definition 6.2. Suppose that either IF = C or IF = JR., Condition [Ml# holds
and (L, E) E G[L]. We say that E is multiplicative with exponent A if there is a
A E JR.+ \ {O} such that
E(ae) = aAe for a E JR.+ \ {O}
(d. Definition 6.1). o

E is multiplicative with exponent A > 0, x Edam L- 1 and a E ~ \ {O}. If
ax E dom L- 1 then a = 1 (d. Proposition 6.3).
Proposition 6.14. Suppose that Condition [Ml# holds and (L, E) E G[L). Then
elements E( -e), L( -e) do not exist, in particular, -e ¢ dom L n dam L -1 (d.
Proposition 6.4).
Proposition 6.15. Suppose that IF = C, Condition [Ml# holds and (L, E) E
G[L). Then elements E(ie), L(ie) do not exist, in particular, ie ¢ dam Lndom L- 1
(d. Corollary 6.1).
90 Chapter 6
Theorem 6.3. Suppose that Condition [Ml# holds, (Ln, En) E G[Lnl for n E N,
a E J(X) and there is an S E CD such that sne E dom L;;l, sna E J(X)ndom L;;l
and 1 E V]Fsn. Then the equation
Dnx = ax + y, y E X, (n E N)
has a unique solution x = Xo + Xl, where
Xo = En(sna)En(sne) E ker(D n - a),
Xl = En(sna)(J - sn)-lsny~n' where y~n = a-l[En(sna)]-ly

(cf. Corollary 6.4).
Properties of multiplicative operators in commutative algebras have been studied,
for instance, by DUDEK D[2], also by the author and LAUSCH LPR[2].
Suppose now that D E A(X) n M(X) and X has unit e. Clearly, De = e. We
have shown that in the commutative case several operators may be reduced (by a
substitution) to a multiplicative operator. Also, a multiplicative D induces multi-
plicative logarithmic and antilogarithmic mappings. For D E A(X) the situation
is, in general, different.
Proposition 6.16. Let D E M(X)n A(X). Suppose that X E Lgr(D) has the
unit e and (Lr,Er ) E G[Orl. Then
(6.13) DLr(uv) = V-I (DLru)v for u, v E J(X) n dom Or.
Suppose that X E Lg1(D) and (Ll,El) E G[Ozl. Then
(6.14) DL1(UV) = (Du)(DL1V)U-l for U, v E J(X) n dom Ol.
Proof Since DE M(X), we have
i.e. Formulae (6.13). A similar proof for (6.14).

•
and F E ML(X) is an initial operator for D corresponding to an R E RD' Then
(6.15) DLr(uv) = D[(Rv-l)(Lru)v] for u,v E J(X) n dom Or,
DLl(UV) = D[u(Llv)Ru- 1 ] for u, v E J(X) n dom 0"

respectively.
Proof By our assumptions, we have
which implies Lru = (Ru-1)u + Zr, where Zr E ker D. Since F is multiplicative

and FR = 0, we conclude that FLru = FZr = Zr. Moreover, by (6.13), we have
i.e. the first of Formulae (6.15). A similar proof for left logarithms.
Theorem 6.4. Suppose that D E M(X)n A(X), X E Lgr(D) n Lgl(D) has the
•
unit e, (LnEr) E G[Or], (Ll,El ) E G[Od and U,V E leX) n dom Dr n dom Ol.
Then
(6.16)
In particular,
(6.17)
Proof Formulae (6.13) and (6.14) together imply that
D[LI(V-1u)Lr(uv) - V-I (LIU) (Lru)v]
= D[Ll(V-1u)]D[Lr(uv)]
- D[v-1(LlU)]D[(Lru)v]
= D[v-1(LIU)]VV- 1D[(Lru)v]
- D[v-1(Llu)(Lru)v]
= D[v-1(L1U)]D[(Lru)v]
- D[V-l(LIU)(Lru)v] =
= D[v-1(L1u)(Lru)v]
- V-I (L1U)(Lru)v] = 0,
which proves (6.16). In particular, if we put u = e, then we get (6.16). Indeed, by

definitions, DLle = DLre = De = e, which implies
i.e. Formula (6.17) holds.

•
CHAPTER 7
LEIBNIZ ALGEBRAS
We begin with
Definition 7.1. Suppose that DE A(X). Then X is said to be a Leibniz algebra
if the Leibniz condition
(7.1) D(uv) = uDv + (Du)v whenever u,v E dom D.

is satisfied, i.e. if CD = 1, fD = O. If D E R(X)nA(X) then X is said to be a
Leibniz D-algebra (d. PR[8], also Chapter 1). 0
In the case of left invertible and invertible operators the situation is shown by
Theorem 4.3.
The proof of Theorem 4.3(a) immediately implies
Corollary 7.1. Suppose that D E A(X} and X is a Leibniz algebra with unit e.
Then e E ker D.
Definition 7.2. Let D E A(X}. Denote by Lr(D) and LI(D) the classes of these
Leibniz algebras for which there exist invertible selectors of Or, 01, respectively.
If X has unit e then we assume that e E dom Or (e E dom 0 1, respectively). If
D E A(X} then we write Lr(D) = Lr(D) = L(D). 0
Corollary 7.2. If X E Lr(D) (X E LI(D), X E L(D), respectively) has unit e,
then D is not left invertible, hence not invertible. If D E R(X} then
(7.2) e E ker D C dom Or n dom 01.
Proof. If either D E A(X) or D E I(X) c A(X), i.e. D is either left invertible or

invertible, then ker D = {O}. Hence, by Theorem 4.3, X has no unit. If DE R(X)
then, by Corollary 7.1, e E ker D. Thus D is not left invertible. By Proposition
2.1 and Corollary 2.1, we get (7.2). •
Corollary 7.3. Suppose that either X E Lr(D) or X E LI(D) and X has unit e.
Let (Lr,Er ) E G[Or]' (LI,Et) E G[Od. If DE R(X) then DLre = 0, DLle = 0,
i.e. Lre, Lie E ker D.
Proof. By Corollary 7.1, De = O. Since CD = 1, fD = 0, Corollary 2.6 implies
DLre = DLle = De = O. •
Corollary 7.4. Let D E R(X). Let X has unit e. If X E Lr(D) then there are
R E RD and (Lr,Er) E GR,l[Or] such that Lre = O. If X E LI(D) then there are
R E RD and (LI' EI) E GR,dOd such that Lie = O. If X E L(D) then there are
R E RD and (L, E) E G R,d!}] such that Le = O.
Proof. By our assumption, CD = 1, fD = O. Thus, by Formula (2.16),
92
Leibniz algebras 93
where F is an initial operator for D corresponding to R. This implies Lre = O.

=
Similarly, Formula (2.17) implies Lie 0 and Formula (2.16') implies Le O. • =
Corollary 7.5. Let either X E Lr(D) and (Lr,Er) E GR,l[Or] or X E L,(D) and
(L" EI) E GR,l[Od or X E L(D) and (L, E) E GR,I[0r] for an R E RD. If X has
unit e then for all n E N we have Lr,ne = 0, LI,ne = 0, Lne = 0, respectively.
Proof. Since X is a Leibniz algebra, e E ker D. By the Leibniz formula,
Since CD = 1, Corollary 3.3 now implies Lr,ne = 0, L"ne = 0, Lne = 0, respectively

(n EN). •
Corollary 7.6. Suppose that D E R(X) and X has unit e. If X E Lr(D) and
(LnEr) E G R,l[Or] for an R E RD then
Lr( -u) = Lru for u E leX) n dom Or.
If X E LI(D) and (LI,EI) E GR,I[0d for an R E RD then
Ll(-U) = Llu for u E leX) ndom 0,.
If X E L(D) and (L, E) E GR,I[O] for an R E RD then
L( -u) = Lu for u E leX) n dom O.
Proof. Since CD = 1 and fD = 0, by Corollary 7.4, Lre = 0, Lie = 0 and Le = O.

Since gD,r = gD,1 = gD = 0, Corollary 2.10 implies the assertion. •
Corollary 7.7. Let IF = C. Let X be an algebra with a unit e. If X E Lr(D)
and (Lr,Er ) E GR,dOr] for an R E RD then
Lr(iu) = ~Lru = ~Lr( -u) for all u E I(X) n dom Or
In particular, Lr(ie) = O. If X E LI(D) and (L"Et) E GR,I[O,] for an R E RD

then
LI(iu) = ~LIU = ~LI( -u) for all u E leX) n dom 0 1•
In particular, L/(ie) = O. If X E L(D) and (L,E) E GR,dOI] for an R E RD then
L(iu) = ~Lu
2
= ~L(-u)
2
for all u E leX) ndom O.
94 Chapter 7
In particular, L(ie) = o.
Proof. By our assumptions, we have CD = 1, fD = 0 and Lre = 0, L/e = 0,
Le = O. This, and Corollaries 2.10,2.11, 7.5 together imply our statement (cf. the
proof of Corollary 2.11). •
Proposition 7.1. Let X be an algebra with unit e. If X E Lr(D) then there are
(Lr,E r ) E G[Or) such that
(7.3) D[Lr(uv) - Lrv) = V-I (DLru)v far u, v E I(X) n dom Or.
If X E L/(D) then there are (L/, El) E G[Od such that
(7.4) D[Ll(UV) - L/u) = u(DL/v)u- 1 far u, v E I(X) n dom 0/.
If X E L(D) then there are (L, E) E G[O) such that
(7.3') D[L(uv) - Lu) = u(DLv)u- 1 far u, v E I(X) n dom o.
Proof. (cf. the proof of Theorem 2.2). Let u, v E I(X) n dom Or. Then Du =
uDLru, Dv = vDLrv and
= (UV)-I D(uv) = v-IU-I[uDv + (Du)v)

= [v- I Dv + V-IU-I(Du)vj = DLrv + v-I(DLru)v,
which implies (7.3). A similar prooffor left logarithms and logarithms. •
Proposition 7.1 can be formulated in a more symmetric way. Namely, we have
Proposition 7.1'. If X E Lr(D)n L/ (D) has unit e, then there are (Lr, Er) E
G[Or], (Ll, Ed E G[Od such that for all u, v E I(X) n dom Or n dom 0/ we have
(7.5) DLr(uv) = V-I D(Lru + Ltv)v, DL/(uv) = uD(Lru + L/v)u- I ,

i.e.
(7.6)
Proof. Let u, v E I(X) n dom Or n dom 0 1. Then, by definition,
= DLru + DL/v = u- I Du + (Dv)v- I

=u-I[(Du)v + uDv)v- 1 = u-ID(uv)v- 1
= {VV-IU- I D(uv)v- I = V(UV)-I D(uv)v- I = vDLr(uv)v-Ij
u- I D(uv)V-IU-IU = u- I D(uV)(UV)-IU = u- I DL1(UV)U.

Leibniz algebras 95
This immediately implies Formulae (7.5). If we add both sides of these formulae
and divide the sum by 2, then we get Formula (7.6). •
Proposition 7.2. Suppose that X is an algebra with unit e and D E R(X). If
X E Lr(D) then there are (Ln Er) E G[Or], Z E ker D, REnD such that
where x = Lru, y = Lrv, U, v E leX) n dom Or.

If X E LI(D) then there are (Ll, EI) E G[Od, Z E ker D, REnD such that
(7.8)
where x = Lju, y = Ljv, u, v E leX) n dom 0 1•

If X E L(D) then there are (L, E) E G[O], Z E ker D, REnD such that
(7.7') (Ex)(Ey) = E(x + y + z),

where x = Lu, y = Lv, u, v E leX) n dom O.
Proof. Suppose that D E R(X) and (Ln Er) E G[Or]. Let u, v E leX) n dom Or.
Then x = Lru, y = Lrv E dom 0;:-1. Formula (7.5) implies that
= ErLr(uv)
= Er{Lrv + R[v- 1 (DL ru)v + z]}
= Er{y + R[(Ery)-l (Dx)(Ery)] + Z},
for some Z E ker D and REnD.
Proposition 7.3. Suppose that D E R(X). If X E Lr(D)n Ll(D) has unit e,
•
then there are (Lr, Er) E G[Or], (LI' Ed E G[Ot) such that for an REnD and for
all u, v E leX) n dom Or n dom 0 1
or, in a symmetric form,
(7.10)
Proof. Let D E R(X). Our conclusion follows immediately from Proposition 7.1',
if we take into account that (Lr, E r ), (LI' EI) are I-normalized by R. •
Definition 7.3. Suppose that X E Lr(D) has unit e and there are (Lr,Er ) E
G[Or] such that Equation (7.3) is satisfied. Then Lr is said to be of the right
almost exponential type (shortly: RAE-type). Suppose that X E LI(D) has unit
96 Chapter 7
e and there is (LI, El) E G[Ot] such that Equation (7.4) is satisfied. Then LI is
said to be of the left almost exponential type (shortly: LAE-type) (cf. Definition
2.3). 0
Theorem 7.1. Suppose that D E A(X), X has unit e and (Lr,Er ) E G[Or]
((L1,E1) E G[Od, respectively). Then Lr is of the RAE-type (LAE-type, respec-
tively) if and only if X is a Leibniz algebra.
Proof. Sufficiency has been proved by Proposition 7.1. Suppose now that Lr is of
the RAE-type, i.e. Equation (7.3) holds. Then for all u, v E I(X) n dom Or we
have
(uv)-l[D(uv) - uDv - (Du)v]

= (uv)-l[D(uv) - uvv- 1Dv - uVV-1U-1(Du)v]
= (UV)-l D(uv) - V-I Dv - V-1U-1(Du)v
= DLr(uv - DLrv - V-I (DLru)v
= D[Lr(uv) - Lrv)- v-1(DLru)v = O.
This implies D(uv) = uDv + (Du)v, i.e. the Leibniz condition (7.1). A similar
proof for left logarithms. •
In the commutative case we have
Corollary 7.8. Suppose that D E A(X), X E Lg(D) has unit e and (L,E) E
G[O]. Then L is of the exponential type if and only if X is a Leibniz algebra, i.e.
if X E L(D).
In other words: In commutative algebras with unit the Leibniz condition is a nec-
essary and sufficient condition for logarithms to be of the exponential type. Thus,
by Proposition 2.6, the Leibniz condition is a necessary and sufficient condition
for the corresponding antilogarithms to satisfy the classical functional equation for
exponential functions: E(x + y) = (Ex)(Ey) whenever x,y E dom 0- 1 .
Proof. By our assumptions, Or = Ol = 0, Lr = Ll = L. Let u, v E I(X) n dom O.
Then Equation (7.3) is reduced to the equation D[L(uv) - Lv] = V-I (DLu)v =
v-1vDLu = DLu. Hence L(uv) = Lu + Lv + z, where z E ker D. By Corollary
2.5, we can admit z = O. We therefore conclude that L is of the exponential type
(cf. Definitions 2.3, 7.3 and Proposition 2.6). •
Corollary 7.8 well motivates the use of name antilogarithmic mapping for a map-
ping inverse to a logarithmic mapping, since, in general, antilogarithmic mappings
are not exponentials.
Theorem 7.2. Suppose that D E R(X), X E Lr(D)nLI(D) has unit e and
(Lr,Er ) E GR,dOrj, (Ll,El) E GR,dOt] for an REnD. Ifu E I(X) n dom D
then u E dam Or n dam Ol and
(7.11)
Leibniz algebras 97
Proof. By Corollary 2.1, dom Or n dom 0 1 =1= 0. Let U E leX) n dom D. Let
Aru = R[u- 1Du], A1u = R[(Du)u- 1], where REnD is arbitrarily fixed. Clearly,
Ar and Al are selectors of Or, 0 1, respectively. Indeed, DAru = u- 1 Du, DAlu =
(Du)u- l . We therefore conclude that u E dom Or n dom 01.
Suppose then that u E I (X) ndom Or ndom 01. Let v = u -1. Since X is a Leibniz
algebra, we have D(uv) = =
De = O. By Corollary 7.4, Lre O. By Theorem 7.1,
Lr is of the RAE-type, Ll is of the LAE-type. By Equation (7.3),
0= DLre
= DLr(uv) = DLrv + v-1(DLru)v = DLru- 1 + u(DLru)u- 1
= DLru- 1 + (Du)u- 1 = DLru- 1 + DLlu = D(Lru- l + L1u),
which implies Lru- 1 + L1u = z E ker D. Let F be an initial operator for D

corresponding to R. Since L r , Ll are I-normalized by R, we get z = Fz =
FL ru- 1 + FLlu = 0, which implies (7.11). The same result can be obtained by
means of Equation (7.4). •
Clearly, we also have a symmetric property:
(7.11')
provided that all assumptions of Theorem 7.2 are satisfied.

Theorem 7.2 and Corollaries 7.4, 7.8 immediately imply
Corollary 7.9. Suppose that DE R(X), X E L(D) has unit e and (L,E) E G[O).
Then Le = 0 and Lu- l = -Lu for u E leX) n dom O.
Theorem 7.3. Suppose that D E R(X), X E Lr(D)nLI(D) has unit e and
(LnEr) E GR,l[Or], (LI,Et) E GR,l[Oz). Then
(7.12) (Erx)EI( -x) = El( -x)Erx = e, i.e. El( -x) = (ErX)-l

whenever x E dom 0;:-1, -x E dom 0,1.
Proof. Let x E dom 0;:-1, -x E dom 0,1 and let u = Erx, v = El( -x). Hence
Lru = x, LIV = -x. This implies Lru + LIV = x - x = 0, i.e. LIV = -Lru. By
Theorem 7.2, we find L/v = -Lru = LIU- 1. Hence u E leX) n dom Or n dom 01
and
Theorem 7.3 and Corollary 7.9 immediately imply

•
Corollary 7.10. Suppose that D E R(X), X E L(D) has unit e and (L, E) E
G[O). Then E(O) = e, -x
E dom 0- 1 and E( -x) = (EX)-1 for x E dom O.
Theorem 7.4. Suppose that D E R(X) and X E Lg(D) is a Leibniz algebra

with unit e. Let u E dom D. Then u E leX) if and only if u E dom O.
98 Chapter 7
In other words: An essential property of Leibniz commutative algebras with right

invertible operators is that their elements have logarithms if and only if they are
invertible.
Proof. Suppose that D E A(X). Let u E J(X) n dom D. Theorem 7.2 implies
that u E dom 0 (cf. also Corollary 7.9).
Conversely, suppose that u E dom O. Let (L, E) E C[O]. Let x = Lu. Then
u = Ex. Since D satisfies the Leibniz condition, Corollary 7.10 implies that
-x E dom 0- 1 and E( -x) = (EX)-l = u- 1 • Hence u E J(X). •
Note 7.1. If we restrict ourselves to studies of commutative Leibniz algebras then
Theorem 7.4 is a generalization of Corollary 2.2(i) concerning only elements of the
form g = Re, where R E RD. 0
In the non-commutative case the condition u E J(X) in Theorem 7.2 is, in general,
sufficient but not necessary for u in order to belong to dom Or n dom 0/, as it is
in Theorem 7.4 in the commutative case (cf. Theorem 7.10 of DI BUCCHIANICO,
DB[l]). Similarly, in the non-commutative case the condition x E dom 0;:-1 does
not imply that -x E dom 011. Thus the existence of Erx does not imply the
existence of E/ ( - x). 0
If an algebra under consideration has no unit, Theorems 7.2 and 7.3 can be mod-
ified in the following manner:
Proposition 7.4. Suppose that X E Lr(D)nL/(D), D E R(X), ker D =I- {O} and
(Lr,Er ) E C[Or], (L/,E,) E C[O,]. Then
(uv)DLrv + (DL/u)(uv) = 0 for u, v E dom Or n dom 0" uv E ker Dj
Proof. Let u, v E dom Or n dom 0/. Let x = L/u, Y = Lrv. Then, by definitions,
uvDLrv + (DL/u)(uv) = uDv + (Du)v = D(uv) = OJ
(E/x)(Ery)Dy + (Dx)(E/x)(Ery) = (uv)DLrv + (DL/u)(uv) = O.
Proposition 7.5. Suppose that D E R(X).

•
If X E Lr(D) then there are
(Lr,Er ) E C[Orl such that
(7.13)
for x = Lru, z' = Lrz, u E J(X) n dom Or, Z E ker D.
If X E L/(D) then there are (L/,E/) E C[O,] such that
(7.14)
for x = L/u, Z' = L/z, U E J(X) n dom 0 " z E ker D.

Leibniz algebras 99
Proof. Suppose that D E R(X) and (Lr,E r ) E G[Or]. Let U E J(X) n dom Or
and let z E ker D. Then x = Lru, z' = Lrz E dom 0;1 and z' E ker D. Formula
(7.7) implies that
Lr(zu) = Lr(zErx) = LrEr(x + z') = x + z' = Lru + LIz.
A similar proof for left antilogarithms.

•
In order to calculate logarithms of powers of elements we need
Proposition 7.6. Suppose that D E A(X) and X is a Leibniz algebra with unit
e. Write
n n
(7.15) In(x, y) = L xk(Dy)x- k , Tn(X, y) = L x-k(Dy)x k ,
k=O k=O
In(x) = In(x, x), Tn(X) = Tn(X, x) for x, Y E J(X)n Edam D, n E No,
where XO = e. Then
(7.16) Dxn = In_dx)x n - 1 = Xn - 1Tn _l(X) for x E J(X) n dam D, n E N.
Proof, by induction. For n = 1 (7.16) is trivially satisfied. Suppose that Dxn =

In_l(X)X n - 1 for an arbitrarily fixed n E No. Then
= (Dnx)x + xn Dx = In_l(X)X n- l x + xn Dx
=In-l (x)x n + xn Dx
n-l
=L xk(Dx)x-kxn + xn(Dx)x- n+n
k=O
n
= L xk(Dx)x-kx n = In(x)x n .
k=O
The equality Dxn = Xn-1Tn(x) is proved in the same manner, if we calculate

Dxn = D(x . xn). •
Immediate consequences of Formulae (7.15) and (7.16) are

Then
100 Chapter 7
Corollary 7.12. Suppose that D E A(X) and X is a Leibniz algebra with unit
e. Then for all n E N and x E J(X) ndom D such that xDx - (Dx)x = 0 we have
(7.17) In(x) = rn(x) = (n + l)Dx, i.e. Dxn = nxn- 1Dx.

In particular, if DE A(X) then Formulae (7.17) hold.
Theorem 7.5. Suppose that X E Lr(D)nL/(D) has unit e E dom 0;:-1 ndom Oj1
and (Lr, Er) E G[Or), (L/, Ed E G[Od. Then for all n E Nand u E J(X) n
dom Or n dom Ol
where rn, In are defined by Formulae (7.15).

Proof. Let n E N. Let u E J(X) n dom Or n dom Ol. By Formulae (7.16) and
(7.17), we find
DLru n = u- n Dun
=u- nu n- 1r n_1(u) = u- 1r n_1(u)
n-1
= u- 1 L u-k(Du)u k
k=O
n-1
= L U- kU- 1(Du)u k
k=O
n
=L u-k(DLru)u k = r n-1(u,DL ru).
k=O
A similar proof for left logarithms.

Corollary 7.13. Suppose that X E L(D) has unit e E dom 0- 1 and (L, E) E
•
G[O]. Then
(7.18) DLu n = nDLu, for all n E Nand u E J(X) n dom O.
In particular, if DE R(X) and (L,E) E GR,dO] for an R E RD then
(7.19) Lu n = nLu, E(nx) = (Ex)n for all n ~ 2, u E J(X) n dom, x = Lu

Note 7.2. Suppose that D E A(X), X is a Leibniz algebra with unit e and a
complete linear metric space and either IF = IR or IF = Co Let {an} C IF. Write
x O = e and
=L
00
s(x) anx n for x E x.

n=O
Leibniz algebras 101
Suppose that D is closed. Proposition 7.6 implies a necessary condition for the
series
=L
00
(Ds)(x) an+lx n
n=O
to be convergent. Namely, if
then the series Ds(x) is not convergent. Clearly, this condition is reduced to the
condition
lim nan::J 0,
n-too
if X is a commutative algebra. o
Definition 7.4. Suppose that either IF = R or IF = <C and D E A(X) with unit e
is a complete linear metric space. Write x O = e and
n
= L;'
00
(7.20) e'" for x E X

n.
n=O
whenever this series is convergent. The function e'" is said to be an exponential

function. By definition, eO = e. 0
Definition 7.5. Let either IF = R or IF = <C. Suppose that X E Lgr(D) n Lgl(D)
with unit e Edam 0;:-1 n dom 0,1 is a complete linear metric space. Write
n
L;'
00
(7.21) ED(X) = {x Edam 0;1 n dam 0,1 :

n.
is convergent and
n=O
xy = yx implies e"'+Y = e"'e Y= eYe'" whenever these series are convergent}.

The same definition can be used for X E Lg~(D). o
Clearly, if X E Lg(D) then xy = yx, which implies e"'+Y = e"'eY = eYe'" whenever
the corresponding series are convergent.
Theorem 7.6. Suppose that either IF = R or IF = <C, X E Lr(D)nLl(D) with unit
e E dom 0;:-1 n dam 0,1 is a complete linear metric space and (Lr,Er ) E G[Or],
(LI' E 1) E G[Od. Write
(7.22) X D = {x E ED(X) : xDx = (Dx)x}.

In particular, if X E L(D) then XD = dom D and Or = Ol = 0, (L, E) E G[O]. If
D is closed then
(7.23) De'" = e'" Dx = (Dx)e'" for x E X D ,

102 Chapter 7
i.e. x E dom nr (x E dom nl , x E dom n, respectively) and

(7.24) e'" = Erx = Elx = Ex and Lre'" = LIe'" = Le'" = x far x E XD,
respectively.
Proof. Let x E X D • Our assumptions and Corollary 7.12 together imply that
Dxn = nx n- l Dx. Hence
n n-l n-l xn
L D; =De+ L ~Dx= L L
00 00 00 00
(x_ 1),Dx= -Dx=e"'Dx,

n=O n. n=l n. n=l n . n=O n!
which implies (7.23). By definitions and (7.23), we get (7.24).

Note 7.3. Let 9 = Re E J(X) for an R E 'RD. It is easy to verify that 9 E X D .
•
Moreover, In(g) = rn(g) = (n + 1)e for n E No. 0
Corollary 7.14. Suppose that all assumptions of Theorem 7.6 are satisfied, 9 =
Re and P(g) E dom n;:l n dom nil for an R E 'RD and a P(t) E IF[t]. Then
(i) Dgn = ngn-l for all n E Ni
(ii) P(g) E XDi

(iii) if D is closed and
n
P(t) = L Pk tk E IF[t]
k=O
then
n-l
(7.25) DeP(g) =eP(g)P'(g), where P'(t) =L (k+1)pk+1tk.
k=O
Proof. (i) By our assumptions, Dg = DRe = e. Hence In(g) and rn(g) defined by
Formulae (7.15) have the form
n n
In(g) =L gk(Dg)g-k =L gkg-k = (n + 1)e; rn(g) = (n + 1)e (n EN).
k=O k=O
This, and Formulae (7.16) together imply that Dgn = In_l(g)gn-l = gn-lrn_l(g)
= ngn-l (n EN).
(ii) By (i), we get
n n n
DP(g) =D L Pkl = poDe + L PkDl =L Pk(k + 1)gk-l = P'(g).
k=O k=l k=l
Hence P(g) and DP(g) commute each with another, since both are polynomials
in 9 with scalar coefficients. This implies that P(g) E XD. Now (ii) and (7.23)
together imply (7.25). •
Now we shall consider particular properties of commutative Leibniz algebras with

right invertible operators, i.e. Leibniz D-algebras. In order to do this (for the sake
of brevity), we shall admit the following condition:
[L] X E Lg(D) is a Leibniz D-algebra with unit e
(i.e. a commutative Leibniz algebra with unit and with D E R(X)).
If the condition [L] is satisfied then, by definition, X E AD(O, e, 0) and e E ker DC
dom n. By Corollary 7.8, L is of the exponential type whenever (L, E) E G[n].
By Corollaries 7.6 and 7.7, -u,iu E leX) n dom n whenever u E leX) n dom n
and L( -u) = Lu, L(iu) = ~L( -u) = ~Lu. Recall that in Corollary 7.7 we have
IF = C. By Theorem 7.4, u E dom n
if and only if u E leX). By Corollary
3.3, if (Ll,El) E GR,l[n] for an R E RD, then Lne = 0 for all n E N. The
list of particular important properties of logarithms and antilogarithms in Leibniz
D-algebras can be easily prolongated.
Proposition 7.7. Suppose that Condition [L] holds and (L,E) E G[n]. Then
(i) Ex, E( -x) E leX) and (EX)-l = E( -x) whenever x E dom n- 1 ;
(ii) E(nx) = (Ex)n whenever x E dom n- 1 , n E N;
(iii) Ex E In(dom n- 1 ) and (Ex)1/n = E(~x) whenever x E In (dom n- 1 ), n EN;

(iv) (Ex)q = E(qx) whenever x E dom n- 1 , q E Q;
(v) Lu q = qLu whenever u E dom n, q E Q;
(vi) L(uv- 1 ) = Lu - Lv whenever u, v E dom n;
(vii) E(O) = e;
(viii) if IF = C then Ll(iu) = L 2u - R2(DLIU)2;
(ix) (L', E') E C[n], where
L'u = L(zu), E'x = zEx far (u, x) E graph n, z E ker D.
Proof. (i) Let x E dom n-l. By Theorem 7.3 and Proposition 2.5, we have
e = E(O) = E(x - x) = (Ex)E( -x). Hence both, Ex and E( -x) belong to leX)
and (EX)-l = E( -x).
Assertion (ii) is proved by an easy induction.
(iii) Let n E N. Suppose that x E dom n-l. Then we have Ex = E(n~x) =

[E(~x)]n. Hence Ex E In(dom n- 1 ) and (Ex)l/n = E(~x).
Assertions (ii) and (iii) together imply (iv).
(v) Let u E dom n, q E Q. Then x = Lu and u = Ex. Hence, by (iv), L(u q) =

L(Ex)q = LE(qx) = qx = qLu.
104 Chapter 7
(vi) Let u, v E dom 0 and let x = Lu, y = Lv. Then u, v E leX) and u = Ex,
v = Ey. Since L is of the exponential type, by (i),
L(uv- 1 ) = Lu + L(v- 1 )
=Lu + L(Ey)-l = Lu + LE( -y)
= Lu - y = Lu - Lv.
(vii) By Proposition 2.12(vii), Le = O. Thus, by definition, E(O) = ELe = e.

(viii) follows immediately from Theorem 3.1 if we put there CD = 1, fD = 0, since
X is a commutative algebra.
(ix) Let z E ker D and let z' = Lz, i.e. z = Ez'. By Proposition 2.5, z' E ker D.
Since L is of the exponential type, we get L(zu) = Lu + Lz for u E dom 0 and
E(x + z') = (Ez')Ex = zEx for x E dom O. Proposition 2.9 now implies that
(L', E') E G[O] (cf. Proposition 6.9 in the multiplicative case). •
Theorem 7.7. Suppose that Condition [L] holds and X is a complete linear
metric space over JR, (L, E) E G[OJ, L is continuous on dom 0, E is continuous on
dom 0- 1 . Then
(Ext = E(rx), Lu r = rLu for (u,x) E graph 0, r E JR \ {OJ.
Proof. Fix r E JR \ {O}. Then there is a sequence {qn} C Q, qn > 0, such

that limn --+ oo Iqn - rl = O. This, our assumptions and assertions (i),(iv),(v) of
Proposition 7.7 together imply our conclusion. •
Definition 7.6. Suppose that [L] holds and (L, E) E G[O]. Then L is said to be
natural if L(Pne) = elnpn, where Pn is the n-th prime (n EN). 0
Proposition 7.8. Suppose that Condition [L] holds, (L, E) E G[O] and L is
natural. Then
L(qe) = elnq, E(elnq) = qe for 0 < q E Q.
Proof. By Proposition 7.7,
L(!!.e)
q
= L[(pe)(qe)-l] = L(pe) - L(qe), L(pqe) = L(pe)q = qL(pe)
for positive p, q E Q. There are positive integers M and N such that q = ~ and
= II
nN mM
N prj
nj' M = II pSk
mk'
j=l k=l
where nj, mk, rj, Sk E N, nj -:j; mk (j = 1, ... , nNj k = 1, ... , mM). This implies
that
nN mM
= L(Me) - L(Ne) = L rjL(Pnje) - L SkL(Pmk e)

j=1 k=1
nN 1nM M
= elL rj In(Pnj) - L Sk In(p1nk)] = eln N = elnq.
j=1 k=1
Moreover, E(elnq) = EL(qe) = qe.

Corollary 7.15. Suppose that Condition [L] holds, X is a complete linear metric
•
space over JR, (L, E) E G[O], L is natural and continuous on dom 0 and E is
continuous on dom 0- 1 . Then
(7.26) L(ru) = Lu + elnr, rEx = E(xlnr)

Jor (u,x) E graph 0, r E Rr \ {OJ.
Proof. Let 0 < q E Q, (u, x) E graph O. Then
L(qu) = L(qeu) = L(qe) + Lu = Lu + elnq,
= qeEx = E(elnq)Ex = E(xelnq) = E(xlnq).
qEx
Having thus proved (7.26) for 0 < q E Q, we prove it for r E Rr \ {OJ by similar
arguments to those used in the proof of Theorem 7.7. •
Theorem 7.S. Suppose that Condition [L] holds, (L, E) E G[OJ, F is a multi-
plicative initial operator for D corresponding to an R E 1?.D and 9 = Re E I(X).
Then
(i) all monomials, i.e. elements of the form Rnz, where z E ker D, n E N, belong
to dom 0 and
1
L(Rnz) = nLg + Lz + L( ,e) Jor z E ker D, n E Nj
n.
(ii) all monomials are invertible;

(iii) if L is natural, i.e. L(ne) = elnn for n E N, then
L(Rnz)=nLg+Lz-elnn! JorzEkerD, nEN.
Proof. Since X is a Leibniz D-algebra and F E ML(X), we have
(7.27)
106 Chapter 7
(d. the author and VON TROTHA PRT[l], also PR[B], Section 6.2). By Corollary
7.B, L is of the exponential type. Since 9 E J(X), we have 9 E dom n. Thus we
get
gn 1 1
L(Rnz) = L(z,) = Lg n + Lz + L(,) = nLg + Lz + L(,e).
n. n. n.
(ii) is an immediate consequence of (i) and Theorem 3.4.

(iii) Suppose that L is natural. Then, by Proposition 7.8, L(~e) = eln~ =
•
n.
-elnn!.
all monomials are not zero divisors.
Proof It follows immediately from Theorem 7.8, since invertible elements cannot
be zero divisors (d. also Corollary 2.3). •
Proposition 7.9. Suppose that Condition [L] holds, (L, E) E c[n], F is an
initial operator for D corresponding to an R E 'RD and 9 = Re E J(X). Then
(i) Dn g -l = (_l)nn!g-n+l for n E N;
(ii) DnLg = (_l)n-l(n _l)!g-n for n E N;
(iii) if FE ML(X) then Fg-l does not exist.
Proof By definition, Dg = DRe = e E ker D and Dgn = ngn-l Dg = ngn-l for
all n E N. Since 9 E J(X), we have 0 = De = D(gg-l) = gDg- l + g-1 Dg =
gDg- l + g-l, which implies that Dg-l = _g-2. We shall prove by induction
that
(7.2B) Dg- n = _ng-(n+l) for n E N.
Indeed, for n = 1 (7.2B) is already proved. Suppose (7.28) to be true for an

arbitrarily fixed n E N. Then
=g-1 Dg-n + g-n Dg-l

= g-l( _ng-(n+l») _ g-n g -2 = -(n + 1)g-(n+2).
We prove (i) also by induction. For n = 1 (i) is already proved. Suppose now (i)
to be true for an arbitrarily fixed n E N. Then
Now we prove (ii) by induction. Let n = 1. We have shown in Example 2.1 that
DLg- l = g-l. Suppose (ii) to be true for an arbitrarily fixed n E N. Then
(iii) Since e E ker D, 9 E leX), FR = 0 and F E M L(X), we get e = Fe =

F(gg-l) = (Fg)(Fg- 1 ) = (F Re)(Fg-l) = 0, which contradicts the assumption
that e is the unit of X. •
Note that in the classical case Fg- 1 = tlt=o, hence it does not exist (cf. Example
2.1).
For the sake of brevity, we admit here the following condition:
[LJ# X E Lg#(D) is a Leibniz algebra.
Similarly, as Proposition 7.7, we prove
Proposition 7.10 Suppose that Condition [LJ# holds and (L, E) E G[L]. Then
(i) E(nx) = (Ex)n whenever x E dom L -1, n E N'j
(ii) Ex E In(dom L- 1 ) and (Ex)l/n = E(~x) whenever x E In(dom L- 1 ), n E N'j
(iii) (Ex)q = E(qx) whenever x E dom L-1, q E Qj
(iv) Lu q = qLu whenever u E dom L, q E Q.
Theorem 7.9. Suppose that Condition [LJ# holds and X is a complete linear
metric space over JR., (L, E) E G[L], L is continuous on dom L, E is continuous
on dom L -1. Then
(Ext = E(rx), Lu r = rLu lor (u,x) E graph L, r E JR. \ {o}.

We shall give examples of results obtained using Banach algebras techniques (with-
out proofs).
Suppose then that X is a Banach algebra with unit e (either IF = JR. or F = q.
Following DI BUCCHIANICO (cf. DB[IJ, Theorem 1.3; DB[2], Chapter 3), write
(7.29) exp(X) = {x EX: 3 x = eY },

yEX
where eY is defined by the series (7.16). Here this series is convergent for all y E X.
A component of a set n c X is a connected subset of n which is not contained
in a larger connected subset of n. Denote by 91 this component of leX) which
contains the unit e (equipped with the norm topology induced by the topology of
X). It is well-known that in commutative Banach algebras with unit exp(X) = 91.
In particular, exp(X) is closed in leX). A path from a to b in X is a continuous
function I : [0,1] ---+ X such that 1(0) = a and 1(1) = b. Then the following
theorem holds:
Theorem 7.10. (cf. DI BUCCHIANICO DB[I], DB[2]) Let X be a commutative
Banach algebra (over JR. orC) with unit e. Then x E exp(X) ifand only if x E leX)
and there is a path I in leX) from ae to x for an a E C \ {o}.
Unfortunately, this theorem is false when X is not commutative An example is the
Banach algebra X = B(H) of all bounded linear operators over a Hilbert space H.
108 Chapter 7
(cf. DB[l], Remark 1.4, also DB[2], Chapter 3). Then J(X) is pathwise connected
but need not to be equal to exp(X).
Another result concerning a connection between the invertibility of an element
and existence of its logarithmic derivative can be found in in a paper of BART,
ERHARDT and SILBERMANN BES[l].
Observe that these results are, in a sense, related to Theorems 2.1 and 7.4.
We shall see now that some algebras "close" in a sense to Leibniz algebras must
be Leibniz. We shall give here without the proof the following
Theorem 7.11. (cf. SMERL S[l]). Let X be an infinite dimensional Banach
space and let X = B(X) be the algebra of all bounded linear operators defined
on E. Let y be a standard subalgebra of X, i.e. Y contains the ideal of all finite
dimensional operators. Let a function 'P : Il4 -t Il4 be such that
lim 'P(t) = o.
t-++oo t
Let V E L(X) with dom V = Y have the property:
(7.30) IIV(ST) - SVT - (VS)TII < 'P(IISII . IITII) far all S, T E y.
Then there is an A E X such that
(7.31) VS = SA - AS far all S E y.
Theorem 7.12. (cf. Smerl S[2]). Let X, Y be Banach spaces and let dim X =
00. Let X, Y be standard operator subalgebras of B(X), B(Y), respectively (cf.
Theorem 7.11). Let € > O. Suppose that 'P : X -t Y is a bijective mapping such
that
II'P(ST) - 'P(S)'P(T) II ~ € far all S,T EX.
Then 'P is of the form 'P(T) = ST S-1 for T EX, where S : X -t Y is either a
bounded linear bijective or a conjugate bounded linear bijective operator.
Corollary 7.17. Suppose that all assumptions of Theorem 7.11 are satisfied. Ten
X = B(X) is a Leibniz algebra.
Proof. If 'P = 0 then the Leibniz condition is automatically satisfied. Suppose
then that Condition (7.30) is satisfied with 'P t= O. By (7.31), we have
V(ST)
= ST A - AST = ST A - SAT + SAT - AST
=S(T A - AT) + (SA - AS)T
=S(VT) + (VS)T
for all S, T E Y = dom V. •
Example 7.1. Let either IF = IR or IF = C. C(Oj IF) will stand for all functions
defined on a set 0 C IR and with values in IF. Denote by An(Oj IF) the set of
all square matrices x = (aii)i,j=l, ... ,n of dimension n with entries xii E C(OjIF)
(n E N). For the sake of brevity, we shall consider the case when n = 2, 0 = [0, 1j.
Indeed, in this case one can see everything, what we want to show. Let then
X = A 2 ([O, 1jj IF). Write
(Dx)(t) = (:tXij(t)).. ,dom D = {x EX: xii E C 1 ([0,ljjIF)},

.,)=1,2
(Rx)(t) = (it Xij(S) dS). .

o .,)=1,2
,(Fx)(t) = (xd(0))i,i=1,2'
We shall write in the sequel: X~j = -itXij. Clearly, D E R(X), D E A(X) and
F is an initial operator for D corresponding to REnD, D satisfies the Leibniz
condition and the matrix e = (bij )i,j=1,2 is the unit of X. The kernel of D consists
of all scalar matrices of dimension 2.
Suppose that U = (uijki=1,2 E I(X). Clearly, not all entries vanish simultane-
ously and det U is invertible. Assume (again for the sake of brevity) that det U = e.
Then we have
and
_ ( UbU22 - (Inu 12)'

- Ull [U~l - U12 (In Ull )']
DLlu = (Du)u- 1
_ ( ul1u22 - (Inu 12)' ull[u12 -U21(l nU ll)'j)

- u22[u12 - U21(lnu22)'] Ul1U~2 - (Inu21)' .
(i) U21 = U12, i.e. matrix U is symmetric. Then
110 Chapter 7
(ii) U21 = U12 =f. 0, U22 = U11 =f. 0.

Clearly, such matrices commute each with another. Hence DLru = u- 1 Du =
(Du)u- 1 = DL1U = DLu. Write a = U11U12, b = ~ and J~ a(s) ds = ii(t). Then
we get
D Lu _ ( (In b)' a + (In b)' ) d L _ (Inb ii+lnb)

- a+(lnb)' (In b)' an u - ii In b In b
(up to a constant, i.e. scalar matrix). In this case we may apply Theorem 7.6.
(iii) U21 = U12 =f. 0, U22 = O:U11, 0: E IF \ {O, I}, U11 =f. 0.
Matrices of this form do not commute each with another. Nevertheless, with the
denotation used in Point (ii), writing J~ a(s)b(s) ds = (ab)(t), we get
L rU -_ ( In ab
" o:(ab) + In aba) , L lTU -_ ( ~In ab" o:(ab) + In ~)
.
In a; aba In a;
~
(ab) + In % o:(ab) + In
(iv) U12 = 0, i.e. u is a triangle matrix.

Triangle matrices of dimension 2 commute, so that in this case, as in (ii), we have
u- 1 Du = (Du)u- 1 and we may apply Theorem 7.6. 0
EXaIllple 7.2. Let C(O; X) will stand for all functions defined on a set 0 C IR
and with values in X. Let D E A(X). Denote by An(O;X) the set of all square
matrices x = (xii ki=l, ... ,n of dimension n with entries Xii E C(O; X) (n EN).
As in Example 7.1, for the sake of brevity, we shall consider the case when n = 2,
o = [0,1]. Let then X = A 2 ([0, 1]; X). Let D E R(X) satisfy the Leibniz
condition. Write Dx = (DXii )i,i=1,2 for x E X. Clearly, D E R(X) and D
satisfies the Leibniz condition. Suppose that X E Lg(D). It is not difficult to
°
verify that xDx - (Dx)x = for x E X if and only if
X11 = a, X12 = b E I(X), X21 = zlb, X22 = a+ Z2b, Zl, Z2 E I(X) n ker D.
°
Sufficiency is proved by checking. In order to prove the necessity, observe that the
condition xDx - (Dx)x = for x E X implies the following equations for entries:
(7.32)
(7.33)
(7.34)
Let X12,X21 E J(X). Equation (7.32) can be rewritten as x1"lDx12 = xi.lDx21,

which implies DLx21 = DLx12. Hence LX21 = L(ZlX12), where Zl E J(X)nker D,
i.e. X21 = ZlX12. Similarly, if X22 - Xu E J(X) then Equation (7.34) can be
rewritten as
Xl} DX12 = (X22 - xU)-l D(X22 - xu,
which implies DLx12 = DL(X22 - xu). Hence X22 - Xu = Z2X12, where Z2 E
J(X) n ker D. Writing Xu = a, X12 = b E J(X), we obtain the required form of
x. 0
Example 7.3. Let In(x), rn(x) (n E No) be defined by Formulae (7.15). Similarly,
as Formulae (7.16) for D E A(X) satisfying the Leibniz condition, we get for
X E J(X) n dom D
Indeed, let n E N be arbitrarily fixed and let X E J(X) n dom D. Since
the first of Formulae (7.16) implies that
Similarly, the second of Formulae (7.16) applied to D(xnx- n ) implies the second
of Formulae (7.35).
Let n E N, X E J(X) n dom D and let u = x-no Then u E J(X). Even more,
u E In(X). Formulae (7.35) can be then rewritten in the following manner:
We therefore conclude that for X E Lr(D)nLI(D)
for all X E J{X) n dom D, n E N.

Again by definitions, Formulae (7.36) can be rewritten as
for all x E J(X) ndom D, n E N

(cf. the proof of Theorem 7.5). o
112 Chapter 7
Example 1.4. The space X = C[O, T] with the usual pointwise multiplication
and with the operator D = -it
is a commutative Leibniz D-algebra with unit and
with the logarithmic and antilogarithmic mappings defined in the standard way:
Lu = In u and Ex = eX, provided (u, x) E C[n]. The same space X with the
multiplication defined by the convolution:
(x * y)(t) = lot x(t - s)y(s)ds for x,y E X,
and with the operator D is a commutative algebra without the unit. However, in
this case instead of the Leibniz condition for the product rule we have the Duhamel
condition D(x * y)(t) = (x * Dy)(t) + x(O)y(t) - y(O)x(t) for x, y E X which, by
the commutativity yields the product rule: D(x * y) = ~(x * Dy + Y * Dx) (cf. for
instance PR[8J, also Examples 2.3(v), 5.1(iii». Since CD = ~, we conclude that
logarithmic and antilogarithmic mappings do not exist.
Consider now the space X with the convolution as the multiplication and with the
operator M of the multiplication by argument: (Mx)(t) = tx(t) for x E X. It is
easy to verify that M satisfies the Leibniz condition, i.e. M(x * y) = x * My +
y * M x for x, y EX. The basic equation for logarithmic mappings induced by the
convolution * is M u = u * M x, where x = Lu. This equation can be written as
tu(t) = lot u(t - s)sx(s)ds.
One can show that under appropriate conditions imposed on u the last equation
has a unique solution. This solution x determines the logarithmic mapping, for
x = Lu by definition. Note that all algebras mentioned in this example have zero
divisors. 0
Example 1.5. Suppose that X E Lg(D) and (L, E) E C[n. Then, in general,
neither L nor E is linear. Indeed, E = L -1 is linear if and only if L is linear. On
the other hand, if CD = 0 then, by Theorem 2.4, L is not homogeneous, hence not
linear. Suppose now that X is a Leibniz algebra and L is linear. Then, by Corollary
7.8, L is of the exponential type and we have Ex+Ey = E(x+y) = (Ex)(Ey) for
x, y E dom n. Let y = e. Then Ey = E(O) = e and Ex + e = Ex + E(O) = E(x +
0) = (Ex)e = Ex. This implies e = 0, a contradiction. Since E is not additive, E
is not linear. Thus L is also not linear. In several other cases X ¢ Lg(D) (cf. for
instance, Theorem 5.1, Proposition 5.6). The following question arise: Do exist
X E Lg(D) with linear logaritmic and antilogarithmic mappings? Certainly, such
algebras X (if they exist) cannot belong to AD(a, c, d). 0
Example 7.6. Suppose that D E R(X) and X E Lg(D) has unit e. By Corollary
7.8 and Theorem 7.4, the following conditions are equivalent: (i) X is a Leibniz
algebra; (ii) logarithms are of the exponential type; (iii) a logarithm of u E X
exists if and only if u is invertible. These 3 equivalent conditions are fundamental
for the classical calculus of logarithms (and antilogarithms). 0
CHAPTERS
LINEAR EQUATIONS IN LEmNIZ ALGEBRAS
The main purpose of the present chapter is to find solutions of linear equations with
an operator D by means of already known properties of right and left logarithms
and antilogarithms.
Theorem 8.1. Suppose that X E LI(D) has unit e, D E R(X), (L/, Ez) E G[Oz),
a E X and Ra E dom nIl for an R E RD. Let x E leX). Then x E ker(D - a) if
and only if
(8.1) x = E/(Ra)z, where z E ker D is arbitrary.
Proof Sufficiency. Suppose that x is of the form (8.1). Then Dz = 0 and

Dx = D[E1(Ra)z]
= E/(Ra)(Dz) + [DE/(Ra)]z = (DRa)E/(Ra)z
=aE/(Ra)z = ax.
Necessity. By our assumption, Dx = ax. Hence DLlx = (Dx)x- l = axx- l = a,
i.e. Llx = Ra + z', where z' E leX) n ker D is arbitrary. Hence x = EI(Ra + z').
Let z = Elz'. By (7.14), we get
x = EI(Ra + z') = E/(Ra)Etz' = E/(Ra)z.
Corollary 8.1. Suppose that X E Lr(D)n LI(D) has unit e, D E R(X),

•
(Lr,Er ) E G[Or], (L/,Ez) E G[Oz) , a E X and Ra E dom 0;1 ndom nIl for
an R E RD. Let x E leX). Then x E ker(D - a) if and only if
(8.2) x = [Er( -Ra)]-l z, where z E ker D is arbitrary.
Proof By our assumptions and Theorem 7.3, [Er (-Ra)J-1 = E/(Ra). This, and
Theorem 8.1 together imply (8.2). •
Theorem 8.2. Suppose that X E Lr(D) has unit e, D E R(X), (Lr,Er ) E G[Or],
a E X and Ra E dom 0;1 for an R E RD. Let x E leX). Then x E ker(D - afo-)
if and only if
(8.3) x = zEr(Ra), where z E ker D is arbitrary.
113
114 Chapter 8
Proof. Sufficiency. Suppose that x is of the form (8.3). Then Dz = 0 and
Dx = D[zEr(Ra)]
= (Dz)Er(Ra)zDEr(Ra) = zE1(Ra)(DRa)
= zE/(Ra)a = xa.
Necessity. By our assumption, Dx = ax. Hence DLrx = x-I Dx = x-Ixa = a,

i.e. Lrx = Ra + z', where z' E ker D is arbitrary. Hence x = Er(Ra + z'). Let
z = E1z'. By (7.13), we get x = Er(z' + Ra) = (Erz')Er(Ra) = zEr(Ra). •
Corollary 8.2. Suppose that X E Lr(D)n LI(D) has unit e, D E R(X),
(L r , Er) E G[nr J, (L I, Ed E G[n!], a E X and Ra E dom n;:-l n dom nil for
an R E 'RD. Let x E leX). Then x E ker(D - a+-) if and only if
(8.4) x = z[EI( -Ra)t l , where z E ker D is arbitrary.
Proof. Our assumptions and Theorems 7.3,8.2 together imply (8.4) (cf. the proof
of Corollary 8.1). •
Theorem 8.3. Suppose that X E Lr(D)nL1(D) has unit e, D E R(X), (LI,EI) E
G[n!], (Lr, Er) E G[nrJ, a E X and Ra E dom nil n dom n;:-l for an R E 'RD.
Then the following identities hold for all x E dom D:
(8.5) Dx - ax = Er(Ra)D[EI( -Ra)x]; Dx - xa = D[xEI( -Ra)]Er(Ra).
Write for y E dom n;:-l n dom nil and Rl E 'RD
Then D - a, D - a+- E R(X) and Ra E 'RD-a, R a... E 'R D- a... independently of

the choice of Rl E 'RD.
ProoF. Since D Ra = a, for every x E dom D we have
D[Er ( -Ra)x]
= [DEr( -Ra)]x + Er( -Ra)Dx
= Er( -Ra)( -DRa)x + Er( -Ra)Dx = Er( -Ra)(Dx - ax).
Hence, by Theorem 7.3,
Dx - ax = Er(Ra)EI(-Ra)(Dx - ax) = Er(Ra)D[EI(-Ra)x],

which proves the first of the identities (8.5). The second identity follows in a
similar manner from Theorem 7.3, since
D[xEI( -Ra)]
= (Dx)EI( -Ra) + xDEI( -Ra)
= (Dx)EI(-Ra) + x(-DRa)EI(-Ra) = (Dx - xa)EI(-Ra).
Equations in Leibniz algebras 115
Again by Theorem 7.3, we get for y E dom n;:-l n dom nil and Rl = R,
(D - a)Ra y
= Er(Ra)D{EI( -Ra)Er (Ra)RdEI (-Ra)y])
= Er(Ra)DRdEI(-Ra)y] = Er(Ra)E/(-Ra)y = y,
which proves that D -a E R(X) and R a E 'RD-a independently of the choice of R l .

In order to prove that (D-a+-)Ra+- = I, write x = Ra+- y = RdyE1( -Ra)]Er(Ra).
Then, by Theorem 7.3,
i.e.
D[xEI(Ra)] = DRdyE/(-Ra)] = yEI(-Ra).
Hence (D - a+-)x = yE1( -Ra)Er(Ra) = y, which proves that D - a+- E R(X)
and that Ra+- is it right inverse independently of the choice of Rl E 'RD. •
Theorem 8.3 shall be used in order to determine general solutions of non-
homogeneous equations. Namely, we have
Theorem 8.4. Suppose that all assumptions of Theorem 8.3 are satisfied. Then
the equation
(8.7) Dx - ax = y, y E dom n;l ndom nil,

has the general solution of the form
(8.8) x = Xl + xo, where Xo = EI(Ra)z E ker(D - a), z E ker D is arbitrary
Proof. By Theorem 8.1, Xo = EI(Ra)z E ker(D - a) whenever z E ker D. By

Theorem 8.3, we find (D - a)xl = (D - a)Ra y = y, which proves that Xl is a
particular solution of Equation (8.7). •
Theorem 8.S. Suppose that all assumptions of Theorem 8.3 are satisfied. Then
the equation
(8.9) Dx - xa = y, y E dom n;:-l ndom nil,

(8.10) X = Xl + Xo, where Xo = zEr(Ra) E ker(D - a+-),
Xl = Ra+-y = R[yEI( -Ra))Er(Ra), z E ker D is arbitrary.
Proof. By Theorem 8.2, Xo = zEr(Ra) E ker(D - a) whenever z E ker D. By

Theorem 8.3, we find (D - a+-)xl = (D - a)Ra+- y = y, which proves that Xl is a
particular solution of Equation (8.9). •
116 Chapter 8
Theorem 8.6. Suppose that X E Lr(D)n L,(D) has unit e, DE R(X), (Lr' Er) E
G[Orl, (L" E,) E G[O,l, a, b E X and Ra, Rb E dom 0;:-1 ndom 0,1 for an REnD.
Then
(i) the equation
(8.11) (D - a)(D - b)x = y, Y E dom 0;1 n dom 0,1,
(8.12) x = Rb Yl = EI(Rb)R[Er ( -Rb)Y1] + E,(Rb)Zl, where
Y1 = Ray + EI(Ra)zo = E,(Ra)R[Er (-Ra)y] + E,(Ra)zo,

zo, Zl E ker D are arbitrary;
(ii) the equation
(8.13) (D - at-)(D - bt-)x = y, Y E dom 0;1 n dom 0,1,
(8.14) x = R b"'Y1 = R[Y1EI( -Rb)]Er(Rb) + zlEr(Rb), where
Y1 = Ra ... y + zoEr(Ra) = R[yE, ( -Ra)]Er(Ra) + zoEr(Ra),

zo, Zl E ker D are arbitrary.
Proof. (i) Write Y1 = (D - b)x. Then, instead of Equation (8.11), we have the
equation (D - a)Y1 = y. By Theorem 8.4,
(D - b)x = Y1 = Ray + E,(Ra)zo, where Zo E ker D is arbitrary.
(ii) Again by Theorem 8.4, we get x = Rb Y1 + E,(Rb)zl, where Zl E ker D is

arbitrary. We therefore conclude that x is of the form (8.12). In order to prove
(ii), apply Theorem 8.5. •
Note 8.1. Observe that for x E dom D2 we have
(D-at-)(D-bt-)x = D 2x-(Dx)a-D(xb)+xba = D2 x-(Dx)(a+b)+x(ba-Db).

In particular, if b = -a, a E dom D, then we get (D - at-)(D + a) = D2 - (a 2 +
Da)t- = D2 - a~ (cf. also Examples 8.1 and 12.4). 0
Corollary 8.3. Suppose that X E Lr(D)n LI(D) with unit e E dom 0;1 n
dom 0,1 is a complete linear metric space, either IF = IR or IF = C, DE R(X) is
closed, (L r , Er) E G[Orl, (L" E,) E G[Od, a E X and Ra E dom 0;:-1 n dom 0,1 n
XD for an REnD. Then Equation (8.7)
Dx - ax = y, y E dom 0;1 ndom 0,1

(8.15) x = Xl + Xo, where Xo = ze Ra E ker(D - a), z E ker D is arbitrary,
Proof. By our assumptions, the element DRa = a commutes with Ra. This, and
Theorems 8.4 and 7.6 immediately imply (8.15). •
Similarly, we have
Corollary 8.4. Suppose that X E Lr(D)n L/(D) with unit e E dom 0;1 n
=
dom 011 is a complete linear metric space, either IF IR or IF C, D E R(X) is =
closed, (LnEr) E G[Or]' (L1,Et) E G[Od, a E X and Ra Edam 0;1 ndom 011 n
XD for an R E RD. Then Equation (8.9)
Dx - xa = y, y Edam 0;1 n dom 011,
(8.16) x = Xl + Xo, where Xo = ze Ra E ker(D - at-), z E ker D is arbitrary,
Now we shall determine solutions of equations with polynomials in D having scalar

coefficients.
Proposition 8.1. Suppose that X E Lr(D)n L/(D) has unit e Edam 0;1 n
dam 011, (Ln Er) E G[Or]' (LI, Et) E G[Od, there are A E IF and R E RD such
that Ag = ARe E dom 0;1 n dam 011 and n E N is arbitrarily fixed. Then
Xo E ker(D - AI)n if and only if
(8.17)
n-l
Z = L zkRke E ker D n and Zo, ••. , Zn-l E ker D are arbitrary.
k=O
Proof, by induction. Observe that ker[D - (Ae)t-l = ker(D - AI). For n = 1

Formula (8.17) is proved by Theorem 8.1 and Corollary 8.1 for a = Ae (Theorem
8.2 and Corollary 8.2, respectively). Suppose Formula (8.17) to be true for an
arbitrarily fixed n E N. Consider the equation
(8.18)
By our inductive assumption, Equation (8.18) can be written as follows:
(D - AI)x = Xo, where Xo = E/(Ag)Z,

118 Chapter 8
n-1
Z = L zkRke E ker D n and Zo, ... , Zn-1 E ker D are arbitrary.
k=O
By Theorem 8.3,
x = El(Ag)R[Er ( -Ag)XoJ + El(Ag)Zn, where Zn E ker D is arbitrary.
By Theorem 7.3, Er( -Ag) = [El(Ag)J-1. Hence
x = El(Ag)R[Er (-Ag)EI (Ag)ZJ + EI(Ag)Zn

= EI(Ag)Rz
n-1
k=O
n-1
= EI(Ag)(L zkRkHe + zn)
k=O
n
= EI(Ag)(L zjRje + zn)
j=1
n
= EI(Ag)(L zjRje).
j=O
Clearly, L:;':-~ Ri Zj E ker Dn. Moreover, since X is a Leibniz D-algebra, we have

Ri zi = zjRie for j = 0, ... , n - 1, Zj E ker D (cf. PR[8]). We therefore conclude
that x = EI(Ag)Z. This, and Theorem 7.3 together imply that x = Z[Er (-Ag)J-1
(cf. Corollary 8.1). •
Let X be a linear space over an algebraically closed field IF of scalars. Recall
that an operator T E Lo(X) is said to be algebraic on X if there is a polynomial
P(t) E IF[tJ such that P(T)x = 0 for all x E X, i.e. P(T) = 0 on X. The
operator T is algebraic of the order N if deg P( t) = N and there is no polynomial
P'(t) E IF[tJ of degM < N such that P'(T) = 0 on X (we assume here and in the
sequel that any polynomial under consideration is normalized, i.e. its coefficient
of the term of the highest degree is equal 1). If it is the case then P( t) is said to be
a characteristic polynomial of T and its roots are called characteristic roots of T.
An operator T is algebraic on X of the order N with the characteristic polynomial
n
P(t) = II (t - tj)rj, tj -:j:. tk if j -:j:. k, r1 + ... + rn = N
j=1
if and only if X is the direct sum of the principal spaces of the operator Sh
corresponding to the eigenvalues t1, ... , tN:
and PI, ... , PN are disjoint projectors giving the partition of unit:
N
PjPk = 8jkPk (j, k = 1, ... , N), L Pj = I.
j=l
If rl = ... = r N, i.e. if the characteristic roots are single, then these projectors are
of the form
II
N
t - tk .
-- (J = 1, ... ,N)
tj - tk
k=l,k#j
and Xj = PjX are eigenspaces of T (cf. PR[6]' also PR(8)).

Theorem 8.7. Let IF be algebraically closed. Let
n
(8.19) P(t) = II (t - tjt j , rj ~ 1, rl + ... + rn = N,
j=l
tj ElF, tj =f. tk if j =f. k (j, k = 1, ... ,n).

Suppose that X E lr (D)n II (D) has unit e E dom n;l n dom nil, (Ln Er) E
G[nr], (L[, E l ) E G[nd, there is R E RD such that such that tjg = tjRe E
dam n;l n dam nil for j = 1, ... , n. Then Xo E ker P(D) if and only if
n rj-l n rj-l
(8.20) Xo = L [L EI(tjg)]Zjk Rke = L (L zjkRke)[Er(-tjg)tl,

j=O k=O j=O k=O
where Zjk E ker D are arbitrary (k = 0, ... , rj - 1; j = 1, ... , n).

Proof. The operator D is an algebraic operator on the space Xo = ker P(D) and
its characteristic polynomial is P(t). We therefore conclude that
X O =X I E!1 ... E!1Xn , where Xj=ker(D-tjI)r j (j=l, ... ,n).
This, and Proposition 8.1 together imply (8.20).

•
Lemma 8.1. Suppose that X E lr(D)n ll(D) has unit e Edam n;l ndom Oil
and u, v Edam Dn for an arbitrarily fixed n E N and A E IF. Then
120 Chapter 8
Proof. By the Leibniz formula, we have for u, v Edam Dn, A E IF, n E N':
=L
k=O
(~) (_l)n-k An- k Dk(uv)
=L
k=O
n
(~) (_l)n-k An- k ~
k
G) (Dk-ju)Djv
=~ ~ n!k! ( l)n-k\n-k(D k- j u)D j v

~ ~ k!(n - k)!j!(k - j)! - 1\
j=O k=j
_ n n.1 [ n (n - J')1. _ n-k n-k k-j ] j
- ~ j!(n _ j)! ~ (n _ k)!(k _ j)! (1) A DuD v
•
Proposition B.2. Suppose that all assumptions of Proposition 8.1 are satisfied.
Then the equation
(8.21)
and Xo is given by Formula (8.17).
Proof. By Proposition 8.1, EI(Ag) E ker(D - AI). By Proposition (8.1), Xo E

ker(D - AI) and it is of the form (8.17). Thus it is enough to check that Xl is a
particular solution of Equation (8.21). By Lemma 8.1, for every n EN we get
(D - u)nXI
= (D - u)n{Et(.~g)Rn[Er(-Ag)y)}
= ~ (;) [(D - u)n- j Et(Ag)]Dj Rn[Er( -Ag)y]
= EI(Ag)D nRn[Er( -Ag)y]

+ ~ (;)(D - u)n-I-j[(D - U)Et(Ag)]Rn-j[Er(-Ag)y]
= EI(Ag)Er(-Ag)y
=y.
•
the equation
(8.23) P(D)x = y, y E dom nr n dom nl ,

(8.24) x = Xl + xo, where Xo E ker P(D),
(8.25)
Proof. By Proposition 8.2, we have
n
= II (D - tjlriXI
j=l
n-l
= [II (D - Uri](D - urn {Et{tng)Rrn[Er(-tng)y)}
j=l
n-l n-2
= [II (D - tj1ri]Yn_1 = [II (D - tjlri](D - tj1rn-1Yn_1
j=l j=l
n-2
= [11 (D - tj 1r i ]Yn-2 = ... = (D - t1lr1Yl = Yo = y.
j=l
•
122 Chapter 8
Recall that X is a complete linear metric locally pseudo convex space over IF
(IF = ~ or IF = C) if is equipped with the topology induced by a sequence of
Pn-homogeneous F -norms:
Iltxlln = ItlPn Ilxlln far x E X, t E IF (n E N)
(cf. ROLEWICZ R[l]). In particular, complete linear metric spaces which are
either locally convex or locally bounded, hence also Banach spaces, are locally
pseudoconvex.
Definition 8.1. X is said to be a complete m-pseudoconvex algebra if it is an
algebra and a complete locally pseudo convex space with the topology induced
by a sequence {II . lin} of submultiplicative Pn-homogeneous F-norms, i.e. such
pseudonorms that
IIxylln :::; IlxllnllYlln for all x, y E X, n E N.
o
Corollary 8.5. Suppose that eitherlF = ~ orlF = C, X E Lr(D)nLI(D) with unit
e E dom 0;:-1 n dom Oil is a complete m-pseudoconvex algebra and (Lr,Er) E
G[Or], (L I, Ed E G[Od. Let D be closed. Let
XD = {x E CD(X) : xDx = (Dx)x}
(cf. Formula (7.22) in Theorem 7.6). Let 9 = Re and let >.g E dom 0;:-1 ndom Oil
for an R E RD and a >. E IF. Let the initial operator F corresponding to R be
multiplicative. Then >.g E XD, >. E 1JFR and
(8.26)
Proof. By our assumptions, Dg = DRe = e. Hence gDg - (Dg)g = ge - eg = O.

Clearly, >.g E CD(X). Indeed, since>. E 1JFR, the operator 1- >'R is invertible.
Recall that the Leibniz condition implies
(8.27) Rne = g~ far F E ML(X) (n E N)

n.
(cf. the author and VON TROTHA PRT[l]' also PR[8] and Formula (7.27)). This
imply that >.g E CD(X). Indeed, the series
00
L
n=O
is convergent. Thus we have

This, and Theorem 7.6 together imply (8.26).

Corollaries 8.5 and 8.3 (8.5 and 8.4, respectively) immediately imply
•
Corollary 8.6. Suppose that either IF = ffi. or IF = C, X E Lr{D)n L1{D) with
unit e E dom 0;:-1 ndom Oil is a complete m-pseudoconvex algebra, DE R(X) is
closed, (Lr,E r ) E C[Or], (Ll,Ct) E C[OdX and >.g = >.Re E dom 0;:-1 ndom Oil
for an R E RD and a >. E vJFR. Then >.g E XD and the equation
(8.28) (D - >.I)x = Y, Y E dom 0;1 n dom Oil,
(8.29) x = Xl + Xo, where Xo = ze Ag E ker(D - >.1), Z E ker D is arbitrary

Corollary 8.7. Suppose that either IF = ffi. or IF = C, X E Lr(D)n L1(D) with unit
e E dom 0;:-1 ndom Oil is a complete m-pseudoconvex algebra, P(t) is defined by
Formula (8.19), D E R(X) is closed, (L r , Er) E C[Or], (L 1, Ct) E C[Od and there
is R E RD such that tjg = tjRe E dom 0;:-1 n dom Oil for j = 1, ... , n. Then
t j 9 E X D for j = 1, ... , n and Equation (8.23) has the general solution x = Xl + Xo,
where
n rj-1 n rj-1
Xo =L (L etjg)zjkRke =L (L zjkRke)e-tjg,
j=O k=O j=O k=O
= etngRrn(e-tngYn_d, Yo = Y,
Xl
and Yj = etjg RTj (e-tjgYj_d for j = 1, ... , n - 1 (n ~ 2).
Zjk E ker D are arbitrary (k = 0, ... , rj - 1; j = 1, ... , n)

Putting in Theorem 8.3 a = >.e, by an easy induction we obtain
Then for all n E N
(8.30) (D - >'I)n = Er(>.g)DnEI(->.g) = El (>\g)DnEr (->.g) on dom Dn,
(D - >'I)n E R(X) and
(8.31) R~,n = Er(>.g)Rf E 1( ->.g), R~n = EI(>.g)Rf E r {->.g) E RD-AI

independently of the choice of Rl E RD.
This immediately implies
124 Chapter 8
Corollary 8.9. Suppose that all assumptions of Proposition B.1 are satisfied.
Then for all n E N Equation (B.21) has the general solution either of the form
(B.22) (as in Proposition B.1) or of the form
(8.32) x = Xo + x~, where x~ = Er(Ag)Rn[E, ( -Ag)y).
In the same manner, as Theorem 8.6 follows from Proposition 8.1, the following
theorem follows from Corollary 8.8:
Theorem 8.9. Suppose that all assumptions of Theorem B.7 are satisfied. Then
Equation (B.23) has the general solution either oEthe form (B.24) x = Xl +xo, where
Xo E ker P(D) and Xl is defined by Formulae (B.25), or of the form X = x~ + Xo,
where
(8.33)
and yj = Er(tjg)Wi [E,( -tjg)Yj-d for j = 1, ... , n - 1 (n ~ 2)
(cf. Formulae (8.32).
In general, even in Leibniz algebras, we cannot use any analogue of the results
already obtained in order to represent solutions of linear equations of higher order
as a product of a constant and an antilogarithm, as we did it until now. It is so,
because any power Dn, n ~ 2, of an operator D satisfying the Leibniz condition
does not satisfy this condition. Thus we cannot solve non-homogeneous equations
in a similar manner, as it was done for equations of the first order. We will show
the main difficulty for n = 2.
Proposition 8.3. Suppose that D E R(X) and R E RD. IE X E Lr(D) then

there are (L r,2, E r,2) E G[n r ,2] such that
(Er ,2 X ) (Er ,2Y)

(8.34) = E r,2{Y + R 2[(Er,2y)-I(D 2x)(Er,2Y)
+ 2(Er,2y)-I(Dx)(Er,2y}Dy + z]},
X = Lr,2U, Y = Lr,2V, U, v E J(X) n dom nr ,2, z E ker D2.
IE X E L/(D} then there are (L,,2, E , ,2) E G[n /,2) such that
(E, ,2 X)(E,,2Y)
(8.35) = E,,2n{X + R2[(E,,2X)(D2y)(E,,2X)-1
+ 2(E,,2y)-I(E,,2X)-I(Dx)(E,,2X)(Dy)E,,2Y + z]},
X = L,,2U, Y = L,,2V, U, v E J(X) n dom n,,2, z E ker D2.

Proof. Since D satisfies the Leibniz condition, we have CD = 1 and fD = O.

Thus, by the Leibniz Formula for u, v E dom D we have f D2 (u, v) = fg) (u, v) =
2(Du)(Dv). Hence
v- I u- I f(2)(u,v)
(8.36) = 2V- IU- I (Du)(Dv) = 2v- I (DL ru)Dv
=2(Er,2y)-1 (Dx)(Er,2y)Dy.
Further the proof is going on similar lines as the proof of Proposition 7.2. There
is no reduction similar to (8.36) in the case of the left multiplication. •
Proposition 8.4. Suppose that D E R(X) and R E RD. If X E Lr(D) then
there are (L r,2, E r,2) E G[f!r,2] such that
(8.37) ZEr,2X = E r,2{X + R2[2(Er,2X)-I(Dz)(Er,2X)Dx + z']},

z' = Lr,2Z, x = Lr,2U, u E leX) n dom f!r,2, Z E leX) n ker D2.
If X E Ll(D) then there are (Ll,2, El,2) E G[f!l,2] such that
(8.38) (El,2X)Z = El,2n{X + R2[2z-I(EI,2X)-I(Dx)(EI,2X)(Dz)z' + z']},

z' = LI,2Z, U E leX) n dom f!l,2, x = LI,2U, Z E leX) n ker D2.
Proof. Let x = Z E ker D2 in (8.34). Then D 2z = 0 and (8.34) implies (8.37).
Similarly, we put y = Z E ker D2 in (8.35). This yields (8.38). •
Theorem 8.10. Let n E N. Suppose that X E Lr(Dn), (Lr,n, Er,n) E G[f!r,n],
a E X, and Rna E dom f!;:~ for an R E RD. Then x is a solution of the equation
(8.39) Dnx = xa + y, y E X,
if and only if x = Xo + Xl, where Xo = Er,n(Rna + z) E ker(Dn - a+-), z E ker Dn

is arbitrary, and Xl = (Rv)xo, where v is any solution of the equation of order
n-1:
(8.40) L .
n-l ( n 1) (Djv)Dn-I-jxo = y.
j=O J +
Suppose that X E LI(Dn), (Ll,n, El,n) E G[f!l,n], a E X, and Rna E dom f!~~ for
an R E RD. Then x is a solution of the equation
(8.41) Dnx = ax + y, y E X,
if and only if x = Xo + Xl, where Xo = E1,n(Rna + z) E ker(Dn - a), z E ker Dn is

arbitrary, and Xl = xo(Rv), where v is a solution of the equation of order n - 1:
(8.42) L ..
n-l ( n ) (Dn-I-jxo)Djv = y.
j=O J +1
126 Chapter 8
Proof. Sufficiency is proved by checking.
Necessity. We are looking for a v such that Xl = xoRv is a solution of Equation

(8.39) and Xo is a solution of the homogeneous equation Dx = ax defined by
(8.39). Thus, by the Leibniz Formula, we get
y = DnXI - aXI
= Dn(xoRv) - aXoRv
= t
k=O
(~) (Dn-kxo)D kRv - aXoRv
= t
k=l
(~) (Dn-kxo)Dk-lv + aXoRv - aXoRv
=L
n-l
j=O
C: 1) (Dn-I-jxo)Djv,
i.e. Equation (8.42) is satisfied. If v is a solution to this equation then Xl = xoRv

is a solution, we are looking for. A similar proof for Equation (8.40). •
Note 8.2. Clearly, one can find solutions of Equation (8.40) if the operator
I - Rna is invertible (left invertible, right invertible, respectively, cf. PR[8], also
N[2]) , i.e. if a belongs to the set V.nn (X) defined by Formula (2.25). Similarly,
one can find solutions of Equation (8.39) if a belongs to the set W Rn (X) defined
by Formula (2.27). As it was shown by Propositions 2.14-2.17, these assumptions
assure the existence of the corresponding right and left antilogarithms. We have
only to remember that, by definition, a right (left) antilogarithm of order n for
the operator D is a right (left) antilogarithm of order 1 for the operator Dn. 0
An analogue of Theorem 7.2 is the following:
Proposition 8.5. Suppose that X E Lr(Dn)n Ll(Dn) for an n E N. Then

there are (Lr,n, Er,n) E G[Or,n], (Ll,nEl,n) E G[Ol,n] such that for all U E I(X) n
dom Or,n n dom Ol,n we have
n-l
(8.43) Dn(Lr,nu + Ll,nu-l) =- L (~) (D n- kLr,n_kU)D kLl,kU- I .
k=O
Proof. Since X is a Leibniz algebra, we have De = O. Write v = u- l . Then

uv = e. By the Leibniz formula,
= Dn(uv) = t
k=O
(~) (Dn-ku)Dkv
= (Dnu)v + uDnv + ~ (~) (Dn-ku)Dkv

= u(Dn Lr,nu)v + u(Dn LI,nV)v + ~ (~) (uD n- kLr,n_kU)(Dk LI,kV)V
=u[Dn Lr,nu + D nLI,nV + ~ (~) (D n- kLr,n_kU)(Dk LI,kV)]V.
k=1
Since U,V = u- 1 E I(X), we get

n-1
D nLr,nu + D n LI,nV + L (~) (D n- kLr,n_kU)D kLI,kV = 0,
k=1
which immediately implies (8.43).

Consider now the commutative case.
•
Theorem 8.11. Suppose that Condition [L] holds, (L, E) E G[O], a E X and
Ra E dom 0- 1 for an R E RD. Then x E ker(D - a) if and only if
(8.44) x = zE(Ra), where z E ker D is arbitrary.
Proof. The assertion is an immediate consequence either of Theorem 8.1 or of

Theorem 8.2, if we apply the fact that X is a commutative algebra. •
Corollary 8.10. Suppose that Condition [L] holds, (L, E) E G[O], there are
A E IF and R E RD such that Ag = ARe E dom 0- 1 . Then x E ker(D - AI) if and
only if
x = ZE(Ag), where z E ker D is arbitrary.
Proof. We apply Theorem 8.11 for a = Ae. Then Ra = ARe = Ag.

Proposition 8.6. Suppose that Condition [L] holds. Then Ag = Re E dom 0- 1
•
for every R E RD and A E vlFR and there are (L, E) E G[O] such that
E(Ag) = (I - AR)-1 Z = eAz E ker(D - AI) for all z E ker D.
Proof. Let R E RD be fixed. Elements of the form u = eAz = (I - AR)-1 Z are

well-defined for all z E ker D and (D - AI)u = D(I - AR)u = Dz = O. Moreover,
128 Chapter 8
Du = '\u = u'\e = u'\DRe = uD(,\g), which implies that'\g E dom n- 1 and there
are (L, E) E G[n] such that e A = u = E(,\g). •
Recall that elements of the form eAz are called exponentials for D (cf. PR[8]).
Indeed, these elements belong to ker(D - '\1), hence they are eigenvectors of D.
Corollary 8.11. Suppose that Condition [L] holds, (L,E) E G[n], R E RD,
,\ E VF R and F is an initial operator for D corresponding to R. Then
(8.45) FE(,\g) = e, FL[zE('\g)] = Lz far all z E kerD.
If F is almost averaging, i.e. F(zx) = zFx for all x E dom D, z E ker D, then
(8.46) F[zE(,\g)] =z far all z E ker D.
Proof. Since e E ker D, we have Fe = e. By Proposition 8.6, for u = E(,\g) we

have
FE('\g) = Fu = u - RDu = u - ,\Ru = (I - '\R)u = e.
Let z E ker D. Then Fz = z. Since L is of the exponential type and Lz E ker D,
we find:
FL[zE('\g)] = F[Lz + LE('\g)] = FLz + FLe = Lz + Le = L(ze) = Lz.

If F is almost averaging, i.e. F(zx) = zFx for z E ker D, x E X, then F[zE('\g)] =
zFE('\g) = ze = z. •
Note 8.3. If X is a D-algebra with unit e and dim ker D = 1, then all F E :FD
are almost averaging (cf. PR[9j, PR[lO]). 0
the general solution of the equation
(8.47) Dx - ax = y, y E dom n
is of the form
(8.48) x = E(Ra)R[E( -Ra)y] + zE(Ra), where z E ker D is arbitrary,
or, in an equivalent form,
(8.49) x = E(Ra) [RE(Ly - Ra) + z), where Z E ker D is arbitrary.
Proof. Either Theorem 8.4 or Theorem 8.5, the commutativity of the algebra
X and Theorem 8.11 together immediately imply (8.48). Since E(-Ra)y =
E(-Ra)ELy = E(Ly - Ra), we obtain from (8.48) the form (8.49). •
Observe that solutions given by Formula (8.49) in the classical case of the operator
D= ft
are of the same form as those obtained by the constants variation method.
Ra E J(X). If FE ML(X) is an initial operator for D corresponding to R then
the initial value problem for Equation (8.47) with the condition
(8.50) Fx = xo, where xo E ker D is given,
has a unique solution of the form
x
(8.51) = E(Ra)R[E(-Ra)y] +xOE(Ra)FE(-Ra)
= E(Ra)[RE(Ly - Ra) + XO FE( -Ra)].
ProoL If F E M L(X) n :FD corresponds to R, i.e. F R = 0, then
= F{E(Ra)R[E(-Ra)y] + zE(Ra)}
= [FE(Ra)]FR[E(-Ra)y] + F[zE(Ra)]
= (Fz)FE(Ra) = zFE(Ra).
By our assumption and Proposition 7.7.1(i), E(Ra) E J(X) and [E(Ra)]-l
E(-Ra). Since F E ML(X), we find
e = Fe = F[E(Ra)E( -Ra)] = [FE(Ra)][FE( -Ra)]

which implies that FE(Ra) E J(X) and z = xO[FE(Ra)]-l = xOFE(-Ra). This,
and Formula (8.48) together imply that the problem (8.47), (8.50) has a unique
solution of the form (8.51). •
Example 8.1. Suppose that DE A(X). Let P(D) = D 2 -pD+q, wherep,q E X.
We are looking for conditions for the polynomial P(D) to be factorized, i.e. to be
presented in the form: P(D) = (D - p')(D - q'), where p',q' EX (cf. Theorem
8.6 and Note 8.1). Clearly, for any p', q' E X we have
(D - p')(D - q') = D2 - p'D - Dq' + p'q' = D2 - p'D +q- q - Dq' + p'q'.

Put p' = p. Then we get (D - p)(D - q') = P(D) - Dq' + pq' - q. Hence
P(D) = (D-p')(D-q') if and only p' = pand -(D-p)q'-q = -Dq' -q+p'q' = O.
This last condition can be rewritten as the following equation with the unknown
q': (D - p)q' = -q. Suppose now that all conditions of Theorem 8.3 are satisfied
(with a = p, y = -q). Applying Theorem 8.4, we find
q' = El(Rp)R[Er { -Rp)( -q)] + EI(Rp)z, where z E ker D is arbitrary.
We therefore conclude that, under assumptions of Theorem 8.3, the problem of

factorization of the polynomial P{D) is solved in an explicit way (cf. SKORNIK and
130 Chapter 8
WLOKA SW[l] for more general questions concerning the factorization problem;
also Examples 12.3, 12.4). 0
Theorem 8.13. Let Condition [L] hold. Suppose that (L, E) E G[OJ, at, ... , an E
X and there is an R E 'RD such that Raj E dom 0- 1 (j = 1, ... , n). Then the
general solution of the equation
(8.52) [rr(D-aj)]X=Y, yEdomO,

3=1
is of the form
(8.53)
where
Yo = y, Yj = E(Raj)[RE(LYj_l - Raj)] + Zj (j = 1, ... , n)
ZI, ... , Zn E ker D are arbitrary.
Proof, by induction. For n = 1 Formula (8.53) is proved by Theorem 8.12, Formula

(8.49). Suppose Formula (8.53) to be true for an arbitrarily fixed n E N. Consider
the equation
n+l
(8.54) [II (D - aj)]x = y, where Ran+l E dom 0-1, y Edam D.
j=1
Write
(8.55)
Clearly, Yn satisfies Equation (8.52). So that, by our inductive assumption, Yn

is of the form (8.53). On the other hand, again by Formula (8.49), we get from
(8.55)
x = E(Ran+l)[RE(LYn - Ran+l)] + Zn+l' where Zn+l E ker D is arbitrary.
This proves that Equation (8.54) has all solutions of the required form. •
Proposition 8.1 immediately implies the following
Proposition 8.7. Suppose that [L] holds, (L,E) E G[OJ, there are>. E IF and
R E 'RD such that >.g = >.Re E dom 0- 1 and n E N is arbitrarily fixed. Then
Xo E ker(D - >-.J)n if and only if
(8.56)
n-l
Xo = zE(>.g), with Z = L zkRke E ker D n and arbitrary Zo, ... , Zn-1 E ker D.
k=O
Theorem 8.7 immediately implies

Theorem 8.14. Let IF be algebraically closed. Let
n
(8.57) P(t)=II(t-tjf;, rj2::1,rl+ ... +rn=N,
j=l
tj E F, tj i- tk if j i- k (j, k = 1, ... , n).

Suppose that Condition [L] holds, (L, E) E G[O] and there is an R E RD suell
that tjg = tjRe E dom 0- 1 (j = 1, ... , n). Then Xo E ker P(D) if and only if
n rj-l
(8.58) Xo = 2: [2: zjkRke)E(tjg)], where

j=O k=O
Zjk E ker D are arbitrary (k = 0, ... , rj - 1; j = 1, ... , n).
Similarly, Proposition 8.2 implies

Then the equation
(8.59) (D - >..I)nx = y, Y E dom 0
(8.60) x = Xl + Xo, where Xl = E(>..g)Rn[E(->..g)y],

and Xo is given by Formula (8.56).
Theorem 8.8 immediately implies
the equation
(8.61) P(D)x = y, y E dom 0
(8.62) X = Xl + Xo, where Xo E ker P(D),
(8.63) Xl = E(tng)Wn [E( -tng)Yn-l], Yo =Y

and Yj = E(tjg)Rr; [E( -tjg)Yj-l] for j = 1, ... , n - 1 (n 2:: 2).
Note 8.4. If IF = lR, it may appear pairs of conjugate imaginary roots: i>.., -i>..,
where>.. E IR \ {O}. To every such pair there corresponds the operator D2 + >..2 I.
This operator, particular properties oflogarithms and antilogarithms when IF = C
132 Chapter 8
and their relation to the real case will be considered in next sections. We should
point out that properties of right invertible operators in the complex case have been
recently studied in a series of papers by BINDERMAN (cf. Bin[1]-[19], BPR[1]-[2],
also Appendix). 0
Example 8.2. Suppose that Condition [L] holds and (L,E) E GR,dO] for an
R EnD. We recall that the Wronskian of the system {Xl. ... , XN} C dom D N -1
is the determinant
. 1
WD = WD(XI. ... , XN) = det(DJ- Xk)j,k=l, ... ,N
and that this determinant have the same properties as in the classical case, i.e.
we can prove the Wronski theorems for linear equations in D with coefficients
belonging to X (cf. PR[8], Section 6.3).
A system {X1, ... ,XN} E dom D N - 1 is said to be fundamental if and only if
W D(X1, ... ,XN) E leX) (cf. PR[8], Definition 6.3.3). If {Xl. .... XN} is a funda-
mental system of solutions to linear equation a(D)x = 0, where
N
a(D) = L akDk, and ao, ... ,aN E X, aN = e,
k=O
then the Liouville formula holds:
(8.64)
Formula (8.64) implies that DLWD = WD' l DWD = (-1)N aN_ 1. Hence we get
Let Xl = Dx. Suppose that either Xl E VA(X) or Xl E VReX). Then Propositions

2.15 and 2.16 respectively, implies FWD = FEx. We therefore conclude that
(8.65) WD = 0 if and only if FWD = 0
(cf. also PR[8], Theorem 6.3.4, with a different proof). o

Consider now left invertible operators in commutative algebras.
Proposition 8.9. Suppose that X E Lg#(D) is a Leibniz algebra, (L, E) E G[L],
a E X and Sa E dom E for an S E CD' Then X = E(Sa)v is a unique solution of
the equation
(8.66) (D - a)x = y, y E dom L
independently ofthe choice of S E CD only if E(Sa)Dv = y.

Proof. Since X is a Leibniz algebra, we have CD 1, JD = O. Suppose that

x = E(Sa)v, where v E dom D. Then
Dx = D[E(Sa)v]
= D[E(Sa)]v + E(Sa)Dv = [(DE(Sa]v + E(Sa)Dv
= E(Sa)(av + Dv) = E(Sa)(av + Dv) = ax + E(Sa)Dv,
which implies that x is a solution of Equation (8.66) only if E(Sa)Dv = y. If there
is a v E dom D which is a solution of the equation E(Sa)Dv = y, then a solution
of Equation (8.66) is x = E(Sa)v. The uniqueness of that solution follows from
the left invertibility of D. •
Note 8.5. In Proposition 8.8 X is a Leibniz algebra and DE A(X). We therefore
conclude that X has no unit (cf. Theorem 4.1). This is a reason, why we cannot
assume that E(Sa) is invertible. If we assume that -Sa E dom E then we also
conclude that E(Sa) is invertible, which again leads to a contradiction. In the
case of algebras with right invertible operators the invertibility of elements E(Ra),
where R E R D , follows from Theorem 7.3. This shows us that we cannot expect
too much results for left invertible and invertible operators, since they have trivial
kernels. 0
We shall consider now problems of a different kind.
Proposition 8.10. Suppose that X E Lgr(D), A E Lo(X), there is an R E RD
such that the operator I - RA is invertible and F is an initial operator for D
corresponding to R. Write D A = D - A and
(8.67) OA,r = {x E dom D: DAu = UDAX} for u E dom D.
Then
(i) DA E R(X), RA = (I - RA)-IR ERDA and FA = (I - RA)-IF is an initial
operator for D A corresponding to RA.
(ii) Selectors ofOA,r are ofthe form
LA,rU = (I - RA)-l[L r u - FLru - R(u- l Au) + z]
for u E I(X) n dom Or and arbitrary z E ker D.

(iii) FALA,rU = (I - RA)-l Z.
(iv) Ifz = F Lru then LA,rU = (I -RA)-I[Lru-R(u- 1 Au)] foru E I(X)ndom Or.
Moreover, if x = Lru, y = LA,rU, then y = (I - RA)-l[x - R(u-IAu)].
(v) If (LnEr) E GR,I[Or] then FALA,rU = O.
(vi) If (Lr,Er) E GR,I[Or] and (LA,nEA,r) E G[OA,r] then (LA,r,EA,r) E
GRA,dOA,r]'
Proof. (i) Clearly,
DARA = (D - A)(I - RA)-IR = D(I - RA)(I - RA)-IR = DR = I.

134 Chapter 8
Hence DA E R(X) and RA E'RDA' An initial operator for DA corresponding to

RA is
FA = 1- RADA = I - (I - RA)-lRD(I - RA) =
= (I - RA)-l[1 - RA - RD + RDRA] = (I - RA)-l(1 - RD) = (I - RA)-lF.
Hence ker DA = (I - RA)-l ker D.
(ii) Let U E I(X) n dom n r . By definition,
Hence for arbitrary z E ker D we find
LA,rU
= RA(DLru - u- 1 Au) + (I - RA)-l Z
= (I - RA)-l[R(DLru - u- 1 Au + z]
= (1 - RA)-l[RDLru - R(u- 1 Au) + z]
= (I - RA)-l[Lru - FLru - R(u- 1 Au) + z].
(iii) Since F2 = F, FR = 0 and Fz = z, we get
FALA,r
= (I - RA)-l F[Lru - FLru - R(u- 1 Au) + z]
= (I - RA)-l[FLru - F2L ru - FR(u- 1 Au) + Fz]
= (I - RA)-l Z.
By (iii), since z = FLru, we obtain (iv).

(v) By definition, if (Lr,Er) E GR,dn r ] then z = FLru = O. Then, by (iii),
FALA,rU = (I - RA)-lz = 0, which implies (v).
Suppose that all assumptions of Point (vi) are satisfied. Then, by (v), we conclude
that (LA,r, EA,r) E GRA,I[nA,r]' -
Similarly, we prove
Proposition 8.11. Suppose that X E Lg1(D), A E Lo(X), there is an R E 'RD
such that the operator I - RA is invertible and F is an initial operator for D
corresponding to R. Write D A = D - A and
(8.68) nA,l = {x E dom D: DAU = (DAX)U} for u E dom D.
Then the assertion (i) of Proposition 8.10 holds. Moreover,

(ii) Selectors ofnA,l are of the form
for U E I(X) n dom 01 and arbitrary z E ker D.

(iii) FALA,IU = (I - RA)-lZ.
(iv) If z = FLlu then LA,IU = (I - RA)-l{Lru - R[(Au)u- l ]} for u E I(X) n
dom 0 1. Moreover, if x = LIU, Y = LA,IU, then y = (I - RA)-l{x - R[(Au)u- l ]}.
(v) If(Lz,E I) E GR,l[Ozj then FALA,IU = O.
(vi) If (Lz, Ed E GR,l[f2 I ], (LA,/,EA,d E G[f2A,z] then (LA,,,EA,d E G RA ,l[f2 A,z].
Note 8.6. Observe that, in general, the operator D - A does not satisfy the
Leibniz condition, even if D does. 0
Theorem 8.16. Suppose that A E Lo(X), X E Lr(D)nLI(D) has unit e, there
is an R E RD such that the operator 1- RA is invertible and F is an initial
operator for D corresponding to R, DA = D - A, RA = (I - RA)-lR, FA =
(I -RA)-l F, f2 A,r and f2 A,1 are defined by Formulae (8.67) and (8.68), respectively,
(LA,r,EA,r) E G[OA,r], (LA,I,EA,I) E G[OA,z] , a E X and RAa E dom f2A~r n
dom f2A~' If A satisfies the Leibniz condition then the equation
(8.69) (D - A)x - xa = y, y E dom f2A~r n dom f2A~/l
has all solutions of the form
(8.70) x = Xl + Xo, where Xo = zAEA,r(RAa) E ker(D - A - a+-),
ZA = (I - RA)-lz, Z E ker D is arbitrary,
Xl = RA[yEA,I(-RAa)][EA,r(RAa)].
In particular, if a = 0 then
(8.71) x = (1 - RA)-l(RA + z), where Z E ker D is arbitrary.
ProoL By definition, D A = D - A. By our assumptions, both operators D and A
satisfy the Leibniz condition. So that D A satisfies this condition. Apply Propo-
sitions 8.10 and 8.11. Then Theorem 8.3 for the operator DA and induced an-
tilogarithmic mappings implies our conclusion. If a = 0 then, by our assumptions,
EA,r(O) = EA,I(O) = e, which implies (8.71). •
Similarly, we prove
Theorem 8.17. Suppose that A E Lo(X), X E Lr(D)nLI(D) has unit e, there is
an R E RD such that the operator 1- RA is invertible and F is an initial operator
for D corresponding to R, DA = D - A, RA = (I - RA)-lR, FA = (1 - RA)-l F,
Or and f21 are defined by Formulae (8.67) and (8.68), respectively, (LA,r, EA,r) E
G[f2A,r], (LA,z,EA,I) E G[f2A,z] , a E X and RAa E dom f2A~r n dom f2A~' If A
satisfies the Leibniz condition then the equation
(D - A)x - ax = y, y E dom OA~r n dom f2A~I'
has all solutions of the form
x = Xl + Xo, where Xo = EA,I(RAa)zA E ker(D - A - a),
ZA = (1 - RA)-l Z, Z E ker D is arbitrary,
Xl = [EA,1(RAa)RA[EA,r ( -RAa)y].
In particular, if a = 0 then x is of the form (8.71).
CHAPTER 9
TRIGONOMETRIC MAPPINGS AND ELEMENTS
In this chapter we shall consider particular properties of left and right logarithmic
and antilogarithmic mappings in algebras over C. In order to do it, we shall make
use of the following condition:
[Cln IF = C, n E N is arbitrarily fixed, X E Lgr(Dn) n Lgl(Dn),
and En = dom 0;,; n dom ot:"~ is symmetric, i.e. -x E En whenever x E En.
Clearly, if X E Lg(Dn) then En = dom 0;;-1.
Observe that, by Theorem 2.4, if either dom Or, 1 or dom 0 1,1 is symmetric, then
CD=j: O. By an easy induction we obtain the same conclusion for all n E N, since
CDn = cD.
Definition 9.1. Suppose that Condition [Clt holds and (Lr,Er ) E G[Or],
(L/, Ed E G[od. For ix E E1 we write
(9.1) Cx = ~[E/(iX) + Er(-ix)], 8x = ;i[E/(iX) - Er(-ix)].

In particular, if (L, E) E G[o] then
(9.1') Cx = ~[E(iX) + E(-ix)], 8x = ;i[E(iX) - E(-ix)].

The mappings C and 8 are said to be cosine and sine mappings, respectively,
or trigonometric mappings. Elements Cx and 8x are said to be cosine and sine
elements or trigonometric elements. 0
Proposition 9.1. Suppose that Condition [Clt holds. Let (Lr, Er) E G[Or],
(L/, Ed E G[od, (L, E) E G[o].Then trigonometric mappings C and 8 are well-
defined for all ix E E1 and
(i) the De Moivre formulae hold:
(9.2) E r (ix)=Cx+i8x, E I(-ix)=Cx-i8x !orixEE1 ;
(ii) C(O), 8(0) E ker D;
(iii) if X E Lg(D), i.e. Er = EI = E, then C and 8 are even and odd functions of
their argument, respectively, i.e.
(9.3) C(-x)=Cx, 8(-x)=-8x !orixEE 1 ;
and C(O) = z =j: 0, 8(0) = 0;
(iv) for all ix E E1
(9.4) (CX)2 + (8x)2 = ~[E/(iX)Er( -ix) + Er( -ix)E/(ix)].
136
Trigonometric mappings 137
Proof Let ix E El be arbitrarily fixed. By Condition [Ch, we have -ix EEl.

Thus the mappings C, S are well-defined.
(i) By definition,
Cx + iSx = ~[EI(iX) + Er( -ix)] + i :i [El(ix) - Er( -ix)] = El(ix).
A similar proof for Cx - iSx.

(ii) By Proposition 2.5, Er(O),EI(O) E ker D.
(iii) If X E Lg(D), i.e. Er = El = E, then Definition 9.1 immediately implies that
Cx is even and Sx is odd. Moreover, again by Proposition 2.5, E(ix) = z :j; O.
Then C(O) = !(z + z) = z and S(O) = ~(z - z) = O.
(iv) Again by definitions, we have
(CX)2 + (SX)2
= ~[(EI(iX) + Er( -ixW + 4~2 [(El(ix) - Er( -ixW
= ~{EI(iX)Er( -ix) + Er( -ix)EI(ix)

- [-EI(ix)Er( -ix)]- [-Er( -ix)EI(ix)]}
= ~[EI(iX)Er(-iX) + Er(-ix)EI(ix)].
•
Then the mappings C', S' defined as follows: C'x = C(x + z), S'x = S(x + z) for
ix E Eo, z E ker D also satisfy assertions (i)-(iv) of that Proposition.
Proof Let x' = x + z. Then ix' E Eo and Dx' = Dx. This, and Proposition 2.9
together imply the conclusion. •
Proposition 9.2. Suppose that Condition [Ch holds and (L, E) E G[O]. Then
for all ix E dom 0- 1 we have
(9.5) (CX)2 + (SX)2 = E(ix)(E(-ix)j
(9.6) DCx = -(Sx)Dx, DSx = (Cx)Dxj
(9.7)
(9.8)
138 Chapter 9
Proof. Let ix E dom 0- 1 be arbitrarily fixed. By Condition [Clt, -ix E dom 0- 1 .

By our assumption, X is commutative. Thus Formula (9.4) implies the equality
(9.5). By definitions, we also find
DCx = ~[DE(ix) + DE( -ix)]

1
= 2[E(ix)D(ix) + E(-ix)D(-ix)]
~i[E(iX) -
:i
= E(-ix)]Dx
= _i 2 [E(ix) - E( -ix)]Dx = -(Sx)Dxj
= :i
:i
DSx [DE(ix) - DE( -ix)]
= [E(ix)D(ix) - E( -ix)D( -ix)]
= i:i[E(iX) + E( -ix)]Dx
= ~[E(iX) + E( -ix)]Dx = (Cx)Dxj
(DCx)2 + (DSx)2 = [(-SX)Dx]2 + [(CX)Dx]2 = [(CX)2 + (Sx)2](Dx)2 j
D 2Cx = D[-(Sx)Dx]
= -cD[(DSx)Dx + (Sx)D 2x] -iD(SX, Dx)
= -cD[(Cx)(Dx)2 + (Sx)D x]- iD(SX, DX}j
2
D 2Sx = D[(Cx}Dx]
= cv[-(SX)(Dx}2 + (Cx}D 2x] + iD(Cx,Dx)
= -cD[(Sx)(Dx}2 - (Cx)D 2x] + iD(Cx,Dx).
Corollary 9.2. Suppose that all assumptions of Proposition 9.1 are satisfied and
•
X is a Leibniz D-aIgebra with unit e. Then the Trigonometric Identity holds, i.e.
(9.9) (CX)2 + (Sx)2 = e whenever ix EEl'
Proof. Since X is a commutative D-Leibniz algebra, Equality (9.5) implies (Cx)2+

(SX)2 = E(ix)E(-ix) = E(x - x) = E(O) = e. •
The results already obtained show us that cosine and sine mappings and elements
have all properties ofthe classical cosine and sine functions (d. also Example 9.4).
Relations between the Leibniz conditions and the Trigonometric Identity in the
commutative case have been studied in PR[18] and PR[19], in the noncommutative
case - in PR[25].
Proposition 9.3. Suppose that Condition [elt holds, (L, E) E C[O] and the
trigonometric identity (9.9) holds. Then
(i) E(ix), E( -ix) E J(X) and E( -ix) = [E(ix)]-1 for all ix E dom 0- 1 ;
(ii) E(O) = e, hence Le = 0;
(iii) e E ker D;
(iv) if X is an almost Leibniz D-algebra, i.e. if !D(X,Z) =0 for x E dom D,
z E kerD, then CD = 1;
(v) if CD = 1 then !D(u,e) = 0 for all u E J(X) n dom 0, i.e. 9D(U) =

u- 1!D(U, e) = O.
Proof (i) Suppose that the trigonometric identity (9.9) holds. By Formula (9.6),
for all ix E dom 0- 1
E(ix)E( -ix) = (Cx)2 + (SX)2 = e.
Hence E(ix), E(-ix) E J(X) and E(-ix) = [E(ix)J-1.

(ii) Put in (9.9) x = O. By Proposition 2.5, E(O) = z E ker D and z =J. o. We find
Z2 = E(O)E(O) = [C(0)]2 + [S(0)]2 = e.

Hence E(O) = z = e. Indeed, suppose that z = E(O) = -e. Then L( -e)
LE(O) = 0 and
Le = L[( -e)( -e) = 2CDL( -e) + R[( _e)-2 !D( -e, -e) = R!D(e, e).
Thus
= Le 2 = 2cDLe + R!D(e, e) = 2cDLe + Le
Le
which implies 2cDLe = O. Since, by our assumption, CD =J. 0, we get Le = 0, which
implies e = ELe = E(O) = -e, a contradiction. Thus z = e and Le = LE(O) = O.
This also proves (iii), since e = z E ker D.
(iv) Suppose that X is an almost Leibniz algebra. We already have proved that e E
ker D. This means that for all u E J(X)ndom 0 we have 9D(U) = u- 1!D(u,e) = 0,
i.e. the mapping 9D is constant. Since, by our assumption, X E Lg(D), Theorem
2.3 implies that CD = 1.
(v) Suppose that CD = 1. Since Le = 0, for all u E J(X) n dom n we have
!D(u,e) = uDRu- 1!D(u,e)
= uDRgD(U) = uD(1- cD)Lu
= (1 - cD)uDLu = (1- cD)Du.
This implies that 9D = O. Note that, by Theorem 2.3, the mapping 9D is constant,
but not necessarily zero. •
140 Chapter 9
Proposition 9.3 shows that commutative algebras with the trigonometric identity
are very" similar" to Leibniz algebras.
Open Question 9.1. Do exist non-Leibniz algebras with the trigonometric iden-
tity (9.9) ?
Corollary 9.3. Suppose that Condition [Clt holds, (L, E) E C[O], R E RD,
9 = Re and A E C. If Aig E dom 0- 1 then
(D2 + CDA 2I)C(Ag) = A[S(Ag)De + fD(S(Ag), e)],

(D2 + cDA2I)S(Ag) = -A[C(Ag)De - fD(C(>.g), e)].
Proof. Let x = >.g. By definition, we have Dx = >.Dg = >.DRe = >.e and
D X = >.De. This and Formulae (9.6), (9,7) together imply that
D 2C(Ag) = -CD[A 2 C(Ag) + AS(Ag)De] + fD(S(Ag), e),

D 2S(>.g) = -CD[A 2 S(>'g) - AC(>.g)De] + fD(C(Ag),e).
Corollary 9.4. Suppose that all assumptions of Corollary 9.3 are satisfied and
•
X is a Leibniz D-algebra. If Aig E dom 0- 1 then
(9.10) C(Ag), S(Ag) E ker(D 2 + >.2 I).
Proof. Here CD = 1, fD = 0 and De = O. Corollary 9.3 yields the conclusion. •

Proposition 9.4. Suppose that Condition [Ch holds, (L, E) E C[O] and ix E
dom 0- 1 • Then
(9.11) D2 E(ix) = -cDE(ix)[(Dx)2 - iD 2x] + ifD(E(ix), Dx).
Proof. Let ix E dom 0- 1 . Then, by definition,
DE(ix) = E(ix)D(ix) = iE(ix)Dx;

D2 E(ix)
= D[iE(ix)Dx] = CD i{[DE(ix)]Dx + E(ix)D 2 x} + ifD(E(ix) , Dx)
=CD i[iE(ix)(Dx)2 + E(ix)D 2 x] + ifD(E(ix), Dx)
= -cDE(ix)[(Dx)2 - iD 2x] + ifD(E(ix), Dx).
Corollary 9.5. Suppose that all assumptions of Corollary 9.3 are satisfied and
•
X is a Leibniz D-algebra. If Aig E dom 0- 1 then
E(Aig), E( -Aig) E ker(D 2 + A2 I).

Proof. By our assumptions, CD = 1, fD = 0 and x = )..ig Edam 0- 1 . Since

Dx = )"Dg = )"DRe = )..e and D 2x = )"De = 0, Formula (7.11) implies that
E()"ig) E ker(D 2 +)..2 J). The same arguments are used for -x. •
For noncommutative algebras we have
Proposition 9.5. Suppose that Condition [Ch holds, X is a Leibniz D-algebra
with unit e, (Lr' Er) E G[Or], (L/, E/) E G[Od, R E RD, 9 = Re and)" E C. If
)..ig E E1 then
C()..g) , S()..g) E ker(D 2 + )..2 J).
Proof. Let x = )..g. By our assumptions, Dx = )"Dg = )"DRe = )..e and D 2x =

)"De = o. Then
DC()..g)
= ~[DEl(i)..9) + DEr(-i)..g)] = ~i)..[eEl(i)..9) - Er(-i)..g)e]
= -~[El(i)..9) - Er(-i)..g)] = -)"S()..g).

Similarly, we prove that DS()..g) = )"C()..g). Hence
Proposition 9.6. Suppose that Condition [Ch holds, X is a Leibniz D-algebra

•
with unit e, (Lr.2, Er,2) E G[Or.2], (L/.2, E , .2) E G[01.2], R E R D , 9 = Re and
).. E C. If )..ig E E2 then
Then U E ker(D 2 +)..2 J) if and only if there are (L r ,2, E r,2) E G[Or.2], (Ll. 2, EI. 2) E
G[0/.2] such that
where z', Z" E ker D are arbitrary.
Proof. Sufficiency is proved by checking.

Necessity. Suppose that U E ker(D 2 + )..21). Then D 2 u = _)..2U. We are looking
for x such that D 2 x = -)..2e. Clearly, since X is a Leibniz D-algebra and 9 = Re,
we find
)..2g2 ,
X = --2- + z 9 + z, where z',z E ker D,
142 Chapter 9
i.e. z' 9 + z E ker D2. Hence

),,2g2 ,
u = Er,2X = Er,2(--2- + z 9 + g).
Clearly, D2X = _),,2e commutes with u. We therefore conclude that E,,2X = Er,2X,
Theorem 9.1. Suppose that Condition [C]2 holds, X is a Leibniz D-aIgebra with
•
unit e, (Lr,2, Er,2) E G[Or,2]' (L/,2, E/,2) E G[OI,2]' R E RD, 9 = Re, )" E C and
>.2 g2
)"ig EEl. Then - 2 E E2 and
(9.12) ker(D2 + ),,2 I)
={zC()"g) + zS()"g) : z, z E ker D}

= {C()"g)z + S()"g)z : z, z E ker D}
),,2g2
= {Er,2(--2- + z'g + z) : z', z" E ker D}
),,2g2
= {E/,2( --2- + z'g + z) : z',z" E ker D}.
Proof. Since X is a Leibniz algebra, by an easy induction we prove that
(9.12') Dn(zx) = zDnx, Dn(xz) = (Dnx)z for x E dom D, z E ker D, n E N.
Suppose that v E ker(D2 + ),,2 J). Then zv, vz E ker(D 2 + ),,2 J) for all z E ker D.
Indeed, by (9.12'),
This, Propositions 9.5 and 9.6 and similar arguments as those in the proof of
Proposition 9.7 imply the first 4 equalities (9.12). The 5th and 6th equalities
(9.12) are proved by Proposition 9.7. •
all solutions of the equation
(9.13)
are of the form
(9.14) x= Xl + Xo, where

Xl = El (-Aig)R{Er(Aig)EI (Aig)R[Er (-Aig)y]) ,

Xo E ker(D 2 + A2 I) is arbitrary.
(xo is determined by Theorem 9.1).

Proof. Applying Theorem 8.8 for n = 2, tl = -Ai, t2 = Ai, we get Xl in the
required form. Note that a symmetric form can be obtained if we change the role
of hand t2. •
Note 9.1. Clearly, we can express Xl in Formulae (9.14) by means of C(Ag) and
S(Ag) (cf. Formulae (9.12). However, since the algebras under considerations are
(in general) non-commutative, we do not have a reduction of terms leading to a
simpler formulae, as in the classical case. Indeed,
(u + iv)(u - iv) = u 2 + ivu - iuv + v2 for all u, vEX.
So that, if uv =f. vu then (u + iv) (u - iv) =f. u 2 + v 2 • Also we cannot use the
fact that the mapping C is even and the mapping S is odd (cf. also Proposition
9.1(iii)), which is here necessary. 0
Corollary 9.7. Suppose that Condition [Ch holds and X is a commutative
Leibniz D-algebra. Suppose that A E C, R E RD, g = Re and Aig E dom E l . If
(Ll,El ) E G[Oll and (L 2 ,E2 ) E G[02l then
(9.15)
= {ZC(Ag) +ZS(Ag): z,z E kerD} =

A2g2
= {(z"g + z')E2 ( - - ) : z',z" E ker D}.
2
Note 9.2. The assumption that Ai,-Ai E vcR ensures that Aig, -Aig E
dom 0- 1 = E l . In this case, -A 2 E vCR2 (cf. the following Example 9.1.)
o
Example 9.1. We shall show connections of the trigonometric mappings and
elements with the trigonometric operators and elements induced by a right inverse
of aD E R(X) in linear spaces (cf. PR[8], Sections 2.3 and 6.2). Suppose then
that Condition [Ch holds, R E RD and A E vcR. Let g = Re. Then there are
(L, E) E G[Ol such that
(9.16) E(Ag) = e>.z E ker(D - AI), where z E ker D is arbitrary.
Indeed, elements of the form u = e>.z = (I - AR)-l z are well-defined for all z E
ker D and called exponentials for D since they are eigenvectors for D corresponding
to the eigenvalue A: (D - AI)U = D(I - AR)u = Dz = 0 (cf. PR[8]). Moreover,
144 Chapter 9
Du = AU = UAe = uADRe = UD(Ag), which implies that Ag E dom 0- 1 and there

are (L, E) E G[O] such that e,\z = U = E(Ag). Clearly, Lu = Ag.
Suppose now that Ai, -Ai E vcR. The operators
(9.17)
are said to be cosine and sine operators, respectively. Let z E ker D. Then c,\z
and s,\z are said to be cosine and sine elements, respectively. It is not difficult to
verify that
Thus
c,\z= C(Ag); s,\z = S(Ag).
We therefore conclude that c,\ and s,\ have all properties listed in Proposition
9.1. Clearly, proofs in the book PR[8] are different, since they follows just from
definitions. Also the assumption made there was much stronger. Namely, we have
assumed that R is a Volterra operator, i.e. I - AR is invertible for all scalars A. 0
Example 9.2. As in Definition 9.1 suppose that Condition [Clt holds and
(Lr,E r ) E G[Or]' (Ll,E1) E G[Od. For ix E El write
ex = ~[Er(iX) + El(-ix)], Sx = ;i[Er(iX) - El(-ix)].
In particular, if (L, E) E G[O] then
ex = ~[E(iX) + E(-ix)], Sx = ;i[E(iX) - E(-ix)].
It is easy to verify that
ex = C( -x), Sx = -S( -x).
If X E Lg(D) then, by Proposition 9.1(iii), we get
ex = C( -x) = Cx, Sx = -S( -x) = Sx.

This shows that trigonometric mappings are uniquely determined by the choice of
right and left antilogarithms and antilogarithms. 0
Definition 9.2. Suppose that Condition [Ch is satisfied and the trigonometric
identity holds. Let u + iv E dom O. Write w* = u - iv and
iC(X) = {w = u + iv E dom 0 : ww* E 12 (dom On.

By definition, w E J(X). Clearly, (w*)* = wand (ie)* = -ie. We have ww* =
I I
(u + iv)(u - iv) = u 2 + v 2. Write w = (WW*)1/2. Let
JR(X) = {w E iC(X) : w* = w}.

It is easy to verify that I w I E JR(X) and

w-w*
u = w+w*
2
E JR(X); v = 2i E JR(X).
o
Note 9.3. We could define 0* = 0 and 101 = 0, which would require some little
modification in definitions of C(X) and JR(X). However, since 0 f/. dom n, this
modification is useless. 0
Proposition 9.8 Suppose that all assumptions of Definition 9.2 are satisfied and
(L, E) E GR,dn) for an R E RD. If w E C(X) then there is an x such that
ix E dom n- 1 and
(9.18) Lw = cD(L w I I + ix) + R[ I w 1-1 E( -iX)!D( I wi, E(ix))) ,

namely,
(9.19) x = -iL[u' + (e - (U ' )2)1/2J, where u ' = ~(w + w*) I w 1-1 .

In particular, if X is a Leibniz D-algebra then
(9.20) Lw = L I w I + ix.
I I I
Proof. Let u ' = u w -1 and v' = v Wi-I. Clearly, {U I )2 + (V ' )2 = e and
w' = u ' + iv' E dom n. Let x = -iL{u' + iV'), i.e. x is defined as in (9.19). By
definition,
w' = u ' + iv' = E[i( -i)L(u' + v'») = E(ix) = ex + iSx.

By Proposition 9.2, [E{ix»)-1 = E(-ix). We therefore conclude that
Lw = L[ w I I E(ix») = I
cD[L wi + LE(ix») + R[ I w 1-1 E{ix)-I) =
= cD{L w I I + ix) + R[ I w I E( -ix»).
Putting CD = 1, !D = 0, we get (9.20) with x defined by (9.19).
Example 9.3. Suppose that all assumptions of Definition 9.2 are satisfied. Let
•
X be a D-algebra over JR with unit e. Write
y = X + iX = {a + ib: a, bE X}.
Clearly, Y is a D-algebra over C with unit e and with the addition, multiplication
by scalars and multiplication of elements defined as follows: for a, b, c, d EX, A,
fLEJR
(a + ib) + (c + id) = (a + b) + i(c + d); (AifL)(a + ib) = (Aa - fLb) + i{Ab + fLa);
146 Chapter 9
+ ib)(e + id) = (ae - bd) + i(ad + be).

(a
The embedding a --t a + iO permit us to consider X as a subspace of Y. For an

arbitrary A E L(X) we write: A(a + ib) = Aa + iAb, where a + ib E Y. It is
easy to verify that IR(Y) = IR(X) = 12 (dom 0) nX. The space Y is said to be the
complex extension of the space X (cf. Example 3.3). 0
Definition 9.3. Suppose that all assumptions of Definition 9.2 are satisfied. An
element x defined by Formula (9.19) is said to be an argument of wand we write
I I
x = argw. The element w is said to be *-modulus of w. 0
By definition, if x is an argument then ix E dom 0- 1 .
Definition 9.3 and Proposition 9.8 immediately imply
Corollary 9.B. Suppose that all assumptions of Definition 9.2 are satisfied, X is
a Leibniz algebra and (L, E) E GR,1 [0) for an R E RD. If w E C(X) then there is
an argument x such that
Lw = L I w I + ix, i.e. x = -i(Lw - L I wi)·
X is a complete linear metric space over C, D is closed, (L, E) E G R,l [0) for an
R E RD, W E C(X) and x = argw. Ifx E [D(X) then also
x' = x + 27rike = argw for k E z.
Proof. For all .\ E C we have
(cf. Example 12.2(xii)). Since X is a commutative algebra, this and Theorem 7.6
together imply that E(.\e) = eAe. Let w = 27ri and let k E Z be arbitrarily fixed.
Suppose that x = argw E [D(X). Again by Theorem 7.6,
An immediate consequence of this proposition is

•
Corollary 9.9. Suppose that all assumptions of Definition 9.2 are satisfied, X is
a complete linear metric space over C, D is closed and (L, E) E G R,dO] for an
R E RD. Then for all wE C(X) and x = argw E [D(X) the mapping Ei and the
trigonometric mappings are 27re-periodic, i.e.
E[i(x + 27re)] = E(ix)j C(x + 27re) = eXj S(x + 27re) = Sx.

Note 9.4. The periodicity of elements e A , CA , SA which appear in Example 9.1

has been proved in another way, by means of properties of shifts induced by a
right invertible operator and its right inverse (cf. PR[8] , and Chapter 11; also
BINDERMAN, Appendix for generalizations of the mentioned shifts). 0
We shall modify now Condition [Ch in the following way:
[C]# IF = C, X E Lg#(D), the domain ofO- 1 is symmetric.
Clearly, here also CD =J 0 and Definition 9.1 works with [C]#. Assertions (i),
(ii), (iii) of Proposition 9.1 hold, provided that Condition [C]# holds. If it is the
case, assertion (iii) of that Proposition is: C(O) = S(O) = O. By Theorem 9.1,
the trigonometric identity (9.9) does not hold in the Leibniz case (since X has no
unit). Similarly, Definition 9.2 holds with Condition [C]# instead of Condition
[ClI-
Proposition 9.10. Suppose that Condition [C]# is satisfied, X has unit e, the
trigonometric identity holds and (L, E) E G[L]. If w E qX) then there is an x
such that ix E dom E and for S E .cD
Lw = cD(L w I I + ix) + S[ I w 1-1 E( -ix)fD( I wi, E(ix))],

namely (by Definition 9.2),
x = -iL[u' + (e - (U')2)1/2], where u' = ~(w + w*) I wi-I.

Proposition 9.11. Suppose that X is a complete linear metric space over JR,
Condition [L]# holds, X E Lg#(D), (L,G) E G[L] and D is closed. Then De x =
eX Dx for x E ED(X). (cf. Theorem 7.6).
Proposition 9.12. Suppose that all assumptions of Proposition 9.10 are satisfied,
X is a complete linear metric space over C, D is closed, w E C(X) and x = argw.
If x E ED(X) then also
x' = x + 27rike = argw for k E Z.

Corollary 9.10. Suppose that all assumptions of Proposition 9.10 are satisfied, X
is a complete linear metric space over C and D is closed. Then for all w E qX)
and x = argw E ED(X) the mapping Ei and the trigonometric mappings are
27re-periodic, i.e.
E[i(x + 27re)] = E(ix); C(x + he) = Cx; Sex + 27re) = Sx.

148 Chapter 9
Example 9.4. Suppose that Condition [C]1 holds, X E Lg(D) is a Leibniz D-

algebra with unit e and a (L, E) E G[n]. It is not difficult to verify that the
following equalities hold for ix,iy EEl:
C(x + y) = (Cx)Cy - (Sx)Sy; S(x + y) = (Sx)Cy - (Sy)Cx;
C(nx) + is(nx) = (Cx + iSx)n (n EN).
o
Example 9.5. Suppose that Condition [eb holds and X E Lg(D) is a Leibniz
D-algebra with unit e. Then nontrivial solutions of the d'Alembert functional
equation (called also cosine functional equation, cf. STETKJER St[l])
f(x + y) +
2
f(x - y) = f( x )f()
Y
. zy
f or zx, . E E2
are cosine mappings C induced by E = L -1, where L are invertible selectors of n,

i.e. f(x) = C(x) = HE(ix) + E( -ix)]. 0
Example 9.6. Suppose that Condition [C]1 holds, X E Lg(D) is a Leibniz D-
algebra with unit e and a complete m-pseudoconvex algebra, dom n
is dosed,
(L, E) E G[n], 9 = Re for an R E RD such that )..ig E El for all ).. E IR and
{Cn}nEZ C X. Clearly, by Proposition 7.7(i), (ii), we have n)..ig E El whenever
n E Z. If the series
L
00
(9.21) cnE(n)..ig)
n=-oo
is convergent, then we say that it is the Fourier series and that the function
L
00
(9.22) x()..) = cnE(n)..ig) E dom n

n=-oo
of the real variable).. has the Fourier expansion (9.21). Under the same assump-
tions, instead of the sequence {C}nEZ C X we may consider ao E X, {an}nEJII,
{bn}nEJII C X and the Fourier series of the form
L
00
(9.24) ao + [anC()..g) + bnS()..g)].

n=l
Open questions 9.2:
L
When
(9.23) x()..)E(n)..ig)d)" = Cn for all n E Z ?
Are Conditions (9.23) necessary and sufficient for the series (9.21) to be conver-
gent?
L L
When
(9.25) x()")C()..g)d)" = an (n E No), x()")S()..g)d)" = bn (n E N) ?
Are Conditions (9.25) necessary and sufficient for the series (9.24) to be conver-
gent?
When either {E(nAig)}nEZ = X or {ao,C(nAg),S(nAg): n E N} = X? 0
CHAPTER 10
SEMIGROUPS PROPERTIES OF SOLUTIONS TO LINEAR EQUATIONS
Let D E A(X). Let a E X. We denote by Xa any element of ker(D - a), i.e. by

definition, we have DXa = ax a. If X has the unit e and a = Ae for a A E IF then
we shall write Xa = x).,.
Proposition 10.1. Suppose that D E A(X) is a Leibniz algebra. If X has the
unit e then the set
(10.1) Xo(D) = {x)., E ker(D - AI) : A ElF}.
is a commutative group with unit e with respect to the addition in IF, i.e. for any
x)." xp. E Xo(D) there are x)." Xp. E Xo(D) such that x).,xp. = XMp., (X).,)-l = X_).,
whenever A, J.L E IF. If D E A(X) then the set
(10.1') X(D) = {xa E ker(D - a) : a EX}.
is a commutative semigroup with respect to the addition in X, i.e. for any Xa, Xb E
X(D) there is an Xa+b E X(D) such that XaXb = Xa+b whenever a, b EX.
Proof. First we prove that X(D) is a commutative semigroup. Suppose that
a, bE X and Xa, Xb E X(D). Since X is a commutative algebra, by definitions and
the Leibniz condition, we find
which implies that XaXb = Xa+b' The proof for Xo(D) is the same, since elements
Ae, J.Le commute with x)., and xl" Clearly, e = Xo E ker D is the unit of Xo(D).
Indeed, for an arbitrarily fixed x)., we have x).,e = X).,Xo = X)"+o = x)., = Xo+)., =
XOX)., = ex).,. Hence X).,X_)., = X_).,X)., = Xo = e, which implies (X).,)-l = X_).,. We
therefore conclude that Xo(D) is a commutative group with unit. •
In the sequel, whenever we shall consider the set X(D) and some induced sets, we
shall admit that D E A(X).
Proposition 10.2. Suppose that DE A(X) is a Leibniz algebra with unit e and
F is an initial operator for D. Then the set
(10.2) Xo(D;F) = {x)., E Xo(D): Fx)., = e}
is a commutative group with unit e with respect to the addition in IF and the set
(10.2') X(D;F) = {xa E X(D): FXa = e}

is a commutative group with unit e with respect to the addition in X.
149
150 Chapter 10
Proof. For the reasons used in the proof of Proposition 10.1, we prove only that
X(D; F) is a commutative group with unit. Let D E A(X). By the Leibniz
condition, we have De = O. Hence Fe = e. Let a = O. Then Xo E ker D which
implies Xo = Fxo = e, by our assumption that Xo E X(D; F). Let a E X be
arbitrarily fixed. By Proposition 10.1, XaX-a = X-aXa = Xa-a = Xo = e, which
implies X-a = (xa)-l. •
Corollary 10.1. Suppose that D E A(X) is a Leibniz algebra with unit e, D E
R(X) and F, F' are initial operators for D. Let the sets Xo(D; F), X(D; F) be
defined by Formulae (10.2), (10.2'), respectively. If F' is multiplicative then the
sets
F'Xo(D; F) = =
{u,x F'x,x : X,x E Xo(D; F)},
F'X(D; F) = {u a = F' Xa : Xa E X(D; F)}

are commutative groups with unit e with respect to the addition in IF and X,
respectively. In particular, if F' = F then FXo(D;F) = {e}, FX(D;F) = {e}.
Proof. Let D E A(X). Let a, bE X. Then, by our assumptions, ua+b = F'xa+b =
F'(XaXb) = (F'xa)(F'Xb) = UaUb. Moreover, Uo = F'xo = Xo = e and U- a =
F'x- a = F'(xa)-l = (F'Xa)-l = u;l. Clearly, by definition, FX(D;F) = {e}. A
similar proof for the set F'Xo(D). •
Proposition 10.3. (cf. Propositions 2.14, 2.15, also PR[8]) Suppose that
D E A(X), D E R(X) and F is an initial operator for D corresponding to an
REnD. If a E VR(X), where the set VR(X) is determined by Formula (2.25),
then the operator I - Ra is invertible and Xa = (I - Ra)-lz, where z E ker Dis
arbitrary and FXa = z, i.e. Xa = (1 - Ra)-l Fx a . Similarly, if D E A(X) and
oX. E VlFR then the operator 1 - oX.R is invertible and X,x = (I - oX.R)-l Fx,x.
Proof. By definition, we have
FXa = (I - RD)xa = Xa - RDxa = Xa - R(axa) = (1 - Ra)x a ,
which implies the required formula (cf. Formulae (2.25), (2.26) and (2.28)). A
similar proof for X,x. •
Let F'D be a set of multiplicative initial operators for D, i.e.
(fO C f).
Theorem 10.1. Suppose that D E A(X) is a Leibniz algebra with unit e, D E (X)
and F'D = {F-yhHo C FD n ML(X). Write
(10.3)
(10.3')
Semigroup properties 151
(with the convention admitted in Proposition 10.1). Then z8>.(p) E ker(D - >.I),
z8 a(P) E ker(D - a) for an arbitrary z E ker D and Fp[z8>.(p)] = z. In particular,
Fp8>.(P) = Fp8 a(p) = e.
Proof. Let D E A(X) and let x = z8 a(p), where z E ker D. Our assumptions and
the Leibniz condition together imply that
Dx = D[z8 a(p)]
= zD8 a(p) = zD[(Fpx-a)xa] = z(Fpx_a)Dxa
= z(Fpx_a)axa = aZ8 a(p) = ax.
Hence x E ker(D - a). Since Fp is multiplicative, by Theorem 8.11, we find
Fpx = qFp[z8 a(P)]

= (Fpz)[Fp8 a(P)] = zFp[(Fpx-a)xaJ
= z(F;x-a)(Fpxa) = z(Fpx-a)(Fpxa)
= zFp(x_axa) = zFpe = ze.
In particular, Fp8 a(p) = e. A similar prooffor 8>.(p) follows from Proposition 8.2
for n = 1. •
Theorem 10.2. Suppose that all assumptions of Theorem 10.1 are satisfied.
Write for p, q E ro, x>. E Xo(D; Fp), Xa E X(D; Fp)
(10.4)
(10.4')
Then for all A,1t E IF, a,b E X, p,q,r E ro
8>.(q,p) = [8>.(p,q)]-1; 8 a(q,p) = [8 a(p,q)]-1;

8>.(p,r)8>.(r,q) = 8>.(p,q); 8 a(p,r)8 a(r,q) = 8 a(p,q);
8>.(p,q)8 tt (p,q) = 8>.+tt(p,q); 8 a(p,q)8b(p,q) = 8 a+b(p,q).
Proof. Our assumptions and Proposition 10.2 together imply that for p, q E ro,
Xa E X(D; Fp) we have
8 a(P, q) = Fp8 a(q) = Fp[(Fqx-a)xa] = (FpFqx-a)Fpxa) = (Fpxa)(FqX_a);

8 a(p,p) = (Fpx-a)(Fpxa) = e;
=(Fqxa)(Fpx-a) = (Fpxa)-l(FqX_a)-l
= [(Fpxa)(Fpx_a)t1 = [8 a(p, qt 1;
152 Chapter 10
=(Fpxa)(Frx-a)(Frxa)(FqX-a)
=(Fpxa)[Fr(xax-a)(Fqx-a)
=(Fpxa)Fre(Fqx-a) = (Fpxa)e(Fqx_a)
= (Fpxa)(Fqx_a) = 0 a(P, q);
0 a(p, q)0b(P, q)
= (Fpxa)(Fqx-a)(FpXb)(Fqx_b)
= [Fp(XaXb)][Fq(x-aX-b)]
= (Fpxa+b)(Fqx-(a+b)) = 0 a+b(p,q).
A similar prooffor 0;,(P,q). •
Recall that for D E A(X) by definition, Erx = EIX = Ex for x E dom 0- 1 =
dom 0;:-1 = dom Oil.
Theorem 10.3. Suppose that X E Lr(D)n LI(D) has unit e, (Lr,Er) E G[Or]
and (LI,E I) E G[Ozl. Let El = dom 0;:-1 n dom Oil (cf. Section 9, Condition
[C]n). Let R E RD and let 9 = Re. Write
(10,5)
(10.6) O-l(R) = {a EX: Ra E Ed.

Then the sets
(10.8) E(R) = {E(Ra) : a E O-l(R)}
are commutative semigroups with respect to the addition in IF and X, respectively.

Proof. By Theorem 7.3, Er(-Ag) = [EI(Ag)]-l. This, Proposition 8.2 for n = 1
and Proposition 10.1 together imply (10.7). Let now D E A(X). By Theorem
8.11, any element E(Ra) belongs to the kernel of D - a. This, and Proposition
10.1 together imply (10.8). •
Note that for D E A(X) we have Eo(R) C E(R). Proposition 10.1 and Theorem
10.3 together imply
F is a multiplicative initial operator for D corresponding to R, Then the sets
FEo(R) = {FEr(Ag) : A E Ool(R)} = {FEI(Ag) : A E Ool(R)},
FE(R) = {FE(Ra) : a E O-l(R)}

are commutative semigroup with unit e with respect to the addition in IF and X,
respecti vely.
Proposition 8.2 and Theorems 10.1 and 10.3 together imply
Corollary 10.3. Suppose that X E Lr(D)n LI(D) has unit e, (Lr,Er ) E G[Or],
(LI,Et) E G[Ot] and F1 = {F,},Ho c FD n ML(X). Let Fp E F1 correspond
to an Rp E RD (p E rO). Let gp = Rpe. Then
(10.9) 8 A (P) = [FpEr( -Agp)]E(A9p) far A E Ool(Rp), p E roo

8 A (P,q) = [FpEr(-Agp)][FqEr(Agp)], far A E 0- 1 (Rp), p,q E ro;
(10.10)

Corollary lOA. Suppose that all assumptions of Corollary 10.3 are satisfied.
Then
8 A (p,q) = [FpEr(-Agp)][FqEr(Agp)], for gp = Rpe, A E O-l(Rp), p,q E ro;

8 a (p,q) = [FpE(-Rpa)][FqE(Rpa)), far a E 0- 1 (Rp), p,q E roo
Corollary 10.5. Suppose that all assumptions of Corollary 10.3 are satisfied,
A E 0 01 (Rp), a E O-l(Rp) and p E roo Then the equation
(10.11) (D - AI)X = y, Y E dom Or n dom 01,
has all solutions belonging to J(X) n dom Or n dom 0 1 of the form

(10.12)
where 9p = Rpe, z E ker D is arbitrary (P E rO);

while the equation
(10.13) (D - a)x = y, y E dom 0,
has all solutions belonging to dom 0 of the form
(10.14)
where Z E ker D is arbitrary (P E rO).

154 Chapter 10
Recall that by Theorem 7.4, if DE R(X) and X is a commutative Leibniz algebra,

then x E I(X) if and only if x E dom o.
Theorem 10.4. Suppose that all assumptions of Corollary 10.3 are satisfied.
Then all solutions of Equation (10.11) (Equation 10.12, respectively) satisfying
the initial condition
(10.14)
are of the form
(10.15)
or, respectively,
(10.16)
where 8.>.(p), 8 a (P) are defined by Formulae (10.3), (10.3'), respectively.

Proof Since Fp E ML(X), we have Rp(zx) = zRpx for all z E kerD, x E X,
P Era (cf. Lemma 6.2; also PR[8]). For the homogeneous equation our statement
is proved by Theorem 10.2. Thus it is enough to check that
are particular solutions of Equations (10.11) and (10.12) satisfying the conditions
Fpx' = 0, Fpx" = 0, respectively. By our assumptions,
Dx" = D{8 a (p)Rp[8_ a (p)yJ}

= [D8 a (P)]Rp[8_ a (p)y] + 8 a (p)DRp[8_ a (P)y]
= a8 a (p)Rp[8- a (p)y] + 8 a (P)8- a (p)y
= ax" + y.
Since FpRp = 0, we get
A similar proof for x'.

Theorems 10.1 and 10.3 well motivate the following
•
Definition 10.1. Suppose that all assumptions of Theorem 10.1 are satisfied.
Then 8.>.(p) (A E Ool(Rp)), 8 a (p), (a E O-l(Rp )), defined by Formulae (10.3),
(10.3') for all p E r a , are said to be fundamental solutions for the operators D - AI,
D - a, respectively. 0
Clearly, Formulae (10.9) and (10.10) determine also fundamental solutions, pro-
vided that the corresponding antilogarithms exist.
The following example show that classical fundamental solutions for an ordinary
differential equation of order one do coincide with fundamental solutions in the
sense of Definition 10.1. Namely, we have
Example 10.1. Let X = e[O, Tj (over ~) with the pointwise multiplication. Let
D = ft,
Rp = J;,
(Fpx)(t) = x(p) for x E X, P E [O,Tj. Let a E X be given.
Write
B(t,p)
t
= (Rpa)(t) = 1 a(s)ds for t,p E [O,Tj.
°
By definition, O(t,p) - O(s,p) = B(t, s) and O(p,p) = for p, t, s E [0, Tj. Hence
8 a(p)(t) = {[Fpe-RpajeRpa}(t) = e(J(t,p). Thus, by Theorem lOA, the equation
(10.16) x' - ax = y
has a unique solution x satisfying the initial condition x(p) = xo, namely,
x(t) = e(J(t,p) It e-li(s,p)y{s)ds + xOeli(t,p) = It eli(t,s)y(s)ds + xOeli(t,p).
We therefore conclude that 8 a (P) = eli(.,p) is the classical fundamental solution

of Equation (10.16). Clearly, if aCt) == A E ~, i.e. Equation (10.16) has a scalar
coefficient, we have O(t,p) = A(t - p) and 8,\(p)(t) = eA(t-p). 0
Recently, similar properties of resolvents to linear equations have been obtained
in a particular case by WYSOCKI, Wy[lj by means of the so-called Bittner Oper-
ational Calculus.
In general, if X is not a Leibniz algebra, the product rule for a D E A(X) implies
that D(XaXb) = cD(a+b)xaXb + fD(X a, Xb) for a, bE X. If DE A(X)) then we get
D(x,\xJL) = CD(A+/-L)X,\X JL + fD(X)..,X JL ) for A,/-L E IF. Thus, without any additional
information on the form of fD' similar results as these already obtained, cannot
be expected. In a particular case, if D(xy) = ')'(xDy + yDx) + d(Dx)(Dy) + axy
for x,y E dom D, where a,d,,), E X do not vanish simultaneously, then we get
D(XaXb) = [')'(a + b) + dab + aJxaxb for a, bE X.
We have shown (cf. LPR[2], also Chapter 5, in particular, Example 5.8) that
in several cases an operator D satisfying Condition (10.17) can be reduced by a
substitution to a mUltiplicative right invertible operator DO. If DO E M(X), i.e.
X E ADO (0, 0, e), then D(XaXb) = abXaXb. Hence XaXb = Xab. In this case one can
we obtain similar results with respect to the multiplication in X.
Note that the property Xa * Xb = Xa+b holds for Laguerre differential operators
of the second order with respect to the convolution "*" as a multiplication (cf.
KIERAT and SKORNIK, KS[l]). Clearly, these operators do not satisfy the Leibniz
condition.
Similar, but simpler, results can be obtained for a left invertible operator, since
in this case ker D = {O}. However, we have to take into account that an algebra
with a left invertible operator satisfying the Leibniz condition has no unit (cf.
Theorem 4.3).
CHAPTER 11
OPERATOR ehD
Recall some definitions and theorems which will be used in our subsequent con-
siderations.
By V(X) we denote the set of all Volterra operators belonging to L(X), i.e. the set
of all operators A E L(X) such that 1- AA is invertible for all scalars A. Clearly,
A E V(X) if and only if VIFA = IF \ {O} (d. Formula (2.28)).
Let X be a Banach space. Denote by QN(X) the set of all quasinilpotent operators
belonging to L(X), i.e. the set of all bounded operators A E Lo(X) such that
lim
n-+oo
v'IIAnxll = 0 for x E X.
It is well-known that QN(X) C V(X). If IF = C then QN(X) = V(X) n B(X),

where B(X) is the set of all bounded operators belonging to L(X).
Definition 11.1 (d. PR[13]). Let X be a complete linear metric space over a
field IF of scalars. Let A E L(X) be continuous. Let E C dom A c X be a
subspace. Let w be a non-empty subset of VJFA. The operator A E L(X) is said to
be w-almost quasinilpotent on E if
(11.1) lim An Anx
n-+oo
=0 for all A E w, x E E.
The set of all operators w-almost quasinilpotent on the set 0 will be denoted by
AQN(O;w). If w = VJFA then we say that A is almost quasinilpotent on O. The
set of all almost quasinilpotent operators on 0 will be denoted by AQN(O). 0
Lemma 11.1. Write
Then
(11.3) (t # 1; n E N)
and for n > mEN
sm,n(t)
(11.4)
= -(1 - t)-(m+l) [1 - t
k=m+l
(~) (1 - t)ktn-k]
+~ (n: k)t nH - m .
156
Operator ehD 157
Proof. Formula (11.3) is an immediate consequence of the well-known equality

n-I
1 - tn = (1 - t) L tk (t E JR.; n E N).
k=O
In order to prove (11.4), by (11.2) and by an easy induction, we get
s~~~ = v! nfv
)=0
(i: v) ti far all v E N.
Let xn(t) = 1 - t n for n E N and let y(t) = (1 - t)-I. Then, again by induction,
we obtain
y(k)(t) = k!(l- t)-(k+1) = k!yk+l(t) far kEN;
_ n! tn-v for v <n

xt)(t) = { (n - v)! - (n,vEN).
o for v >n
Let n >m ~ 1 be arbitrarily fixed. Then, by the Leibniz formula, for t =11,
dm 1- t n
---
dtm 1- t
dm
= dtm [xn(t)yn(t)]
= ~ (7)x~k)(t)y(m-k)(t)
=_ f
k=O
(m) n! tn-k(m _ k)!ym-k+1(t)
k (n - k)!
= _~ m!n!(m - k)! tn-k(1- t)-(m-k+1)
f::o k!(m - k)!(n - k)!
= -m!(l - t)-(m+1) f
k=O
(~) (1 - t)ktn-k
= -m!(l - t)-(m+1)[t (~) (1 - t)ktn-k

k=O
t (~)tn-k(l_ t)k]
k=m+1
n
= -m!(l - t)-(m+1)[(l_ t + t)n - L
t
k=m+1
= -m!(l - t)-(m+1)[l - (~) (1 - t)ktn-k].

k=m+1
158 Chapter 11
This implies that for n >m ~ 1

sm,n(t)
= ~ e:m)tk
t
k=O
= nfm (k+m) k
k=O m t + k=n-m
1
= _s(m)
m! O,n
+"~
)=0
m
1 d
= -
m
- nL:- t k +
1
L: m
(n + k) t n+k-m
m!dt m m
k=O k=O
= ~ d m
1-
m! dtm 1- t L....-
tn ~ (n m+ k) t n +k - m
k=O
=-(I-t)-(m+l)[I- t
k=m+l
(~)(I_t)ktn-kl+'f (n~k)tn+k-m,
k=O
which proves Formula (11.4). •
Theorem 11.1. (cf. PR[13]). Let n be a subspace of a complete linear metric
space X over IF. If A E L(X), n c dom A and 0 =J we vFA, then the following
conditions are equivalent:
(i) A is w-almost quasinilpotent on n;
(ii) for every A E w, x E n the series 2:::;"=0 An Anx is convergent and
L
00
(11.5) (1 - AA)-lX = An Anx (A E w, x En);

n=O
(iii) for every A E w, x E n, mEN the series 2:::;"=0 (n-:;'~~l)An Anx is convergent
and
(11.6) (1 - AA)-m x = ~ (n: r: ~ 1) An Anx (A E w, x E n, mEN).
Proof. (i)-t(ii) Suppose that A E AQN(n, w). Since A is continuous, by (11.3),

for A E w, x E E we have
x = x - lim An Anx
n-too
= n-too
lim (x - An Anx)
n-l
= lim (1 - AA)" Ak Akx
n--+-oo ~
k=O
n-l
= (/ - AA) lim "

n-too L...J
Ak Akx,
k=O
Operator e hD 159
which proves (11.5).

(ii)-+(i) Since the series L~=o An An x is convergent, we conclude that
lim An Anx =
n-+oo
o.
The arbitrariness of A E w, x E f! implies that A E AQN(f!iW).
(i)-+(iii) Since A E AQN(f!iW), Formula (11.4) implies that for all m ~ 2, n > m,
A E w, x E f!
Sm-l,n(AA)x
= (I - AA)-m[x - t
k=m
(~) (J - AA)k An- k An-kx]
tends to -(J - AA)-m x , as n -+ 00. We therefore conclude that the series in

question is convergent and that Equality (11.6) is satisfied. The case m = 1 is
already proved by (ii).
(iii)-+(i) Since the series
is convergent for all mEN, A E w, x E f!, putting n = 1, we get

lim An Anx = 0 for A E w, x E f!.
n-+oo
This means that A E AQN(f!iW). •

Clearly, by definition, AQN(f!iW) ~ V(f!). If X is a Banach space (IF = IR or
IF = q, E = dom A = X and VIF A = IF \ {O}, then
AQN(X) ~ QN(X) = V(X) n AQN(X).
For given DE R(X), R E RD we shall consider

• the space of smooth elements
Doo = n
kENo
dom Dk, where dom DO = Xi
• the space of D-polynomials
s = U ker Dni S = P(R) = lin {Rkz : z E ker D, kENo} c D oo ,

nEN
160 Chapter 11
which, by definition, is independent of the choice of an R E 'RD;

• the space of exponentials
E(R) = U ker (D - AI) =

AEvwR
= lin {(I - AR)-lZ: Z E ker D, A E Vf'R or A = O} c Doo,

which is independent of the choice of the right inverse R, provided that R is a
Volterra operator,
• the space of D-analytic elements in a complete linear metric space X (IF = C
or IF = R)
00
AR(D) = {x E Doo: x = "L...J RnFDnx} = {x E Doo: n~oo

lim RnDnx = O},
n=O
where F is an initial operator for D corresponding to an R E 'RD (cf. PR[9]).

Clearly, by definitions, we have S, E(R) C Doo. If X is a complete linear metric
space then S C AR(D) c Doo.
Definition 11.2. Suppose that X is a complete linear metric locally convex space
(IF = C or IF = R), D E R(X) is closed, ker D ¥ {O} and F is a continuous initial
operator for D corresponding to a right inverse R almost quasinilpotent on ker D.
Let A(IR) = IR+ or IR. If {ShhEA(R) C Lo(X) is a family of continuous linear
operators such that So = I and for h E A(IR) either
or
Sh(I - AR)-l F = eAh(I - AR)-l F for A E Vf'R,
then Sh are said to be true shifts. The family {ShhEA(R) is a semigroup (or group)
with respect to the superposition of operators as a structure operation. 0
The following theorems will be given here without proofs, since the theory of
functional shifts presented in Appendix is more general. However, in this Chapter
we shall make use of properties of true shifts for their connections with old classical
theorems.
Theorem 11.2. (cf. PR[13]). Suppose that all conditions of Definition 11.2
are satisfied, {ShhEA(R) is a strongly continuous semigroup (group) of true shifts
and either P(R) = X or E(R) = X. Then D is an infinitesimal generator for
{ShhEA(R), hence dom D = X and ShD = DSh on dom D. Moreover, the
canonical mapping /'i, defined as
(11. 7) K,X = {X/\(t)}tEA(R), where x/\(t) = FStx (x E X)

Operator ehD 161
is an isomorphism (hence separate points) and
Theorem 11.3. (cf. PR[13]). Suppose that all conditions of Definition 11.2 are
satisfied and {ShhEA(IR) is a family of true shifts. Then for all h E A(JR) and
x E AR(D) the series
is convergent,
(11.8)
and ehD maps AR(D) into itself.

Formula (11.8) yields the Lagrange-Poisson formula for a right invertible operator
D:
Definition 11.3. Suppose that X is a linear space over IF, D E R(X) and ker D :f:
{O}, F is an initial operator for D corresponding to an R E 'RD, A(JR) = ll4 or lR.
Let {ShhEA(IR) be a family of linear operators satisfying the condition
ShZ = Z for Z E ker D, hE A(JR).
The operator D.hR, where D.h = Sh - I, is independent of the choice of R E 'RD

(cf. PR[14]). The operator
1
..1(h) = hD.hR, where hE A(JR) \ {O}
is said to be the Bernoulli operator for D. o

Theorem 11.4. (cf. PR[14]). Suppose that all assumptions of Theorem 11.2 are
satisfied. Write
et 1 tn- t tn
= et -1 = L
_ 00 1 00
A(t) =- t = '""'
~ n!
-, B(t) Bnn! for t E JR,
n=l n=O
where Bn (n E No) are Bernoulli numbers (i.e. coefficients of the Taylor expansion
of the function
162 Chapter 11
1 1 1 1
B 2n+l = 0 for n EN, Bo = 1, Bl = -2' B2 = 6' B4 = - 30' B6 = 42' ...
and so on; all Bn are rational numbers). Then
ehD 1
(11.10) .J(h) = A(hD) = D- on AR(D),
.J(h)D = D.J(h) on AR(D),

the operator .J(h) is invertible and
(11.11) [.J(h)]-l = B(hD) = D on AR(D).

e hD -1
Moreover,
= Bo.J(h) + L ~~ hn-l~hDn-l
00
(11.12) 1 on AR(D) (h E A(JR) \ {OJ).

n=l
Formula (11.12) is the Euler-Maclaurin formula for a right invertible operator D.

Theorem 11.5. (cf. PR[14]) Suppose that all assumptions of Theorem 11.2 are
satisfied. Let h E A(JR) \ {OJ. Then the operator ~h is invertible on the space
= {x EX: L
00
AR(D; h) = S(h) n AR(D), where S(h) SkhX is convergent}

k=O
and
L
00
~hl = SR(h) = - Skh on AR(D; h),

k=O
where
is the summation operator for D.

Note that, in general, the summation operator is not invertible but it is right
invertible (cf. PR[14], Lemma 3.1).
Write for A E L(X)
An
= {x E dom A: L
00
(11.13) LA(X) (_I)n----=:. is convergent},

n=l n
An
=L
00
(11.14) In(I + A)x (-I)n-;-x for x E LA(X),

n=l
Operator ehD 163
An
L -,
00
(11.15) EA(X) = {x E dom A :

n.
x is convergent},
n=O
(11.16)
Clearly,
(11.17)
Theorem 11.6 (d. PR[15]) Suppose that all assumptions of Theorem 11.2 are
satisfied. Then on AR(D) n L~h (X)
1
= "hln(I + ~h) = L
00 ~n
(11.18) D (-It nhx for hE A(R) \ {O}.
n=l
Proof By our assumptions and Theorem 11.2, EhD(X) ~ AR(D). Moreover, if

x E AR(D) then ShX E AR(D) and
which implies the required formula (11.18).

Formula (11.18) is the Poisson-d'Alembert formula for right invertible operators.
•
Theorem 11.7. Suppose that X is a complete linear metric locally pseudoconvex
space. Let A E L(X) and let 0 C dom A be a subspace of X. Suppose that
A E AQN(O). If A E VIFA then 0 C L-AA(X), i.e.
An An
(11.19) In(I - AA)x = L
00
--x
n
for x E O.
n=l
Proof By Theorem 11.1, the series L:::"=o An An x is convergent for all x E 0,

A E VJFA. Since the space X is locally pseudoconvex, we conclude that also the
series
is convergent for all x E 0, A E VJFA. This, and Formulae (11.13), (11.14) together
imply that 0 C L-AA(X) and that Formula (11.19) holds. •
Note 11.1. A ZELAZKO'S example (d. for instance, R[l]) shows that out of
locally pseudoconvex spaces convergence of series appearing in Theorem 11.7 may
hold in a finite number of points only. 0
164 Chapter 11
Corollary 11.1. Suppose that all assumptions of Theorem 11. 7 are satisfied. If
y = In(I - "\A)x E E-AA(X) for an x E E and a..\ E 'Uf'A then
(11.20)
Proof. By definition, eY = (I - AA)x. Since..\ E VJFA, the operator 1- "\A is

invertible. This implies Formula (11.20). •
Proposition 11.1. Suppose that all assumptions of Theorem 11.2 are satisfied
and X is a Banach space. Write
If h E B(~h) then AR(D) c LAh (X).
Proof. Let II~II ~ a < 1. By our assumptions and Theorem 11.6, for x E AR(D)
we find
II In (I + ~h)xll
= II f:
n=l
(_I)n ~~ xii
n
~ f:
n=l
II~hlln Ilxli <
n
co n
< L
n=l
~llxli
n
= In (a + 1)lIxll·
Proposition 11.2. Suppose that all assumptions of Theorem 11.2 are satisfied
•
and X is a Banach space. Write for a, c > 0
D(a; c) = {x E Dco : IIDkxl1 < cak for kENo}.
IE Ihl < ~ln (1 +~) then
Proof. By our assumptions, c(e a1hl - 1) < 1. For x E AR(D) we have
II~hxll =
II(Sh - I)xll = II(ehD - I)xll
Ihl n n
= II L -, xII ~ cL
00 hnDn 00
_,a_llxll
n. n.
n=l n=l
< c(ea1hl - 1) < 1.

This, and Proposition 11.1 together imply our conclusion.
Another similar result can be obtained if the kernel of D is one-dimensional.
•
Namely, we have
Operator ehD 165
Example 11.1 (d. PR[lO)) Suppose that all assumptions of Theorem 11.2 are
satisfied and dim ker D = 1. Then there is a 0, 10hl < Ihl, such that
(11.21) il h = (Sh - I) = hSehD on dom D for a 0 E (0,1).
Formula (11.21) shows that for true shifts a theorem about the intermediate value
holds (provided that the kernel of D is one-dimensional). 0
Results of this section can be essentially generalized if we take into account the so-
called functional shifts induced by analytic functions belonging to H(O). Here the
generating function is et . The theory of functional shifts was recently developed
by BINDERMAN and is presented in Appendix. Various families of these shifts have
the structure of a commutative algebra (with respect to the superposition, as the
structure operation). So that, under appropriate assumptions, we may prove that
the function e hD is an antilogarithm induced by the operator V = d~ and the
corresponding logarithm is hD.
Example 11.2. Suppose that X = Co(l~) is the space of all functions bounded
and uniformly continuous for t E JR and D = ft.
Note that X with the pointwise
multiplication is also an algebra.
The operator D is not right invertible in X. Indeed, constants are bounded func-
tions and a primitive of a constant is not a bounded function: J;
cds = ct. Hence
the operator R = J~ does not preserve the space X. However, in the space X
there is defined a strongly continuous semigroup {Sh} hEIR defined by means of the
Gauss kernels:
(ShX)(t) = {
(27rh)1/2 1+
-00
00
e-(t-s)2/ 2h x (s)ds for h > 0,
(h E JR).
x(t) for h = 0,
The infinitesimal generator for this semigroup is the operator
(d. YOSIDA Y[1J, Section IX.5, Example 2). o

Example 11.3. Suppose that X is a complete linear metric locally convex space
(IF = C or IF = JR), D E R(X) is closed, ker D =I- {O} and F is a continuous initial
operator for D corresponding to a right inverse R almost quasinilpotent on ker D.
Let A(JR) = JR+ or JR, {ShhEA(IR) is a strongly continuous semigroup (group) of
true shifts and either P(D) = X or E(R) = X. All assumptions of Theorem 11.2
are satisfied. By our assumptions, the series
166 Chapter 11
is convergent for all .x E C, z E ker D and we shall write
Rn
eARz = 2: .x
00
n _, z for all >. E C, z E ker D.
n=O n.
The operator eAR is invertible and its inverse is e- AR . Theorem 11.2 implies that
tk
(Rkz)",(t) = k!z for kEN, z E ker D,
(SheARz)/\(t) = (eARz)/\(t + h) for >. E C, z E ker D.

We therefore conclude that
(eARz)(t) = z 2:
00 (>.t)n
(n!)2
for all >. E C, z E ker D.
n=O
o
CHAPTER 12
POWER MAPPINGS. POLYLOGARITHMIC FUNCTIONS.

NONLINEAR EQUATIONS
We begin with
Definition 12.1. Let X E Lgr(D) n Lg,(D). Write for A E IF:
EI,)..U = El(ALlu) if (Ll' Ed E G[Dd, U E dom D,.

H X E Lg(D) then we write
(12.1') E)..u = E(ALu) if (L,E) E G[DJ, u E dom D.
The mappings E r,).., El,).. and E).. are said to be of the power type with exponent
A or, if it does not lead to any misunderstanding, shortly, power mappings. 0
Note 12.1. We have seen that, without any additional assumptions, just by
definitions, left and right logarithms and antilogarithms of elements qe, where e
is the unit of X and q E Q, are well-defined (provided that CD =f. 0, d. Theorem
2.3). The questions connected with extensions of left and right logarithms and
antilogarithms to lR and <C in Leibniz algebras have been discussed in details in
Chapter 7 (d. Theorems 7.7, 7.8 and 7.9, Proposition 7.8 and Corollary 7.14.
o
Proposition 12.1. Suppose that either X E Lgr(D) n Lg,(D) or X E Lg(D),
(Lr,Er ) E G[DrJ, (Ll,EI) E G[Dd, (L,E) E G[D] and the mappings E r,).., El,)..,
E).. are defined by Formulae (12.1), (12.1'), respectively. Then for all A, /-l E IF
(i) Er,)..(dom Dr) C dom Dr, El,)..(dom Dl ) C dom Dl, E)..(dom D) C dom D;
(ii) LrEr,).. = ALr; LIEl,).. = ALl, LE).. = AL;
(iii)Er,)..Er,/J = Er,)../J' EI,)..EI,/J = EI,)..I" E)..E/J = E)..I"
i.e. the mappings E r,)..,
El,).., E).. are multiplicative functions of the parameter A.
Proof Let u E dom Dr, A E IF. By definitions,
Hence Er,)..u E dom Dr and LrEr,)..u = ALru. Again by definition, we get
Similar proofs for E l ,).. and E)...

•
167
168 Chapter 12
Theorem 12.1. Suppose that X E Lr(D)n Ll(D) and (Lr, Er) E G[Or],
(Ll' E I) E G[Od. Let El = dom Orndom 0 1• Then for all A E IF and U E J(X)n El
(i) EI,AU- 1 = (Er,AU)-l;
(ii) if X E L(D), (L, E) E G[O], u E dom 0, A E IF then EAu- 1 = (EAu)-l;
(iii) if D E R(X) then Er,A' EI,A' EA E M(X).
Proof. (i) Theorems 7.2 and 7.3 together imply that
(ii) follows immediately from (i), since X is commutative and, by Theorem 7.4,
uEX.
(iii) Let u, v E E1 , A E IF. Let x = Lru, Y = Lrv. Then Dx = DLru, Dy = DLrv

and Erx = u, ErY = v E J(X). Let D E R(X). By Proposition 7.2, there is an
R E RD such that
Er,A(UV) = Er[ALr(uv)]
= Er{ALrv + AR[v-1(DLru)v]}
= Er{Ay + AR[(Ery)-l(Dx)Ery]}
= Er(AX)Er(AY)
= Er(ALru)Er(ALrv)
= Er,A(U)Er,A(V).
Similar proofs for EI,A and EA.
Proposition 12.2. Suppose that X E L(D) and (L, E) E G[O]. Then for all A,
•
J.L E IF and u E dom 0
(12.2)
Proof. Suppose that X E A(D). By Corollary 7.8, L is of the exponential type.

By Theorem 7.4, u E J(X) whenever u E dom O. If A, J.L E IF then, by Proposition
2.6,
If J.L = -A then we get (EAu)(E_Au) = E(O· Lu) = E(O) = e. This implies that
EAu E J(X) and (EAu)-l = E_Au. •
Observe that Proposition 12.2 does not hold in the non-commutative case. In
general, we have
Power mappings 169
Proposition 12.3. Suppose that X E Lgr(D)nLgl(D) and (Lr,Er) E GR,dOrj,

(LI,Ed E GR,dOd for an R E RD. If A E IF and u,v EEl, Er,>.u, Er,>.v E J(X),
El,>. u, E l ,>. v E J(X) then there are Zr, Zl E ker D such that
= Er{CDALrv
+ R[CDA(Er,>.v)-lu-l(Du)(Er>.v)
+ (Er,>.V)-l (Er,>.u)fD(Er,>.u, Er,>.v)) + zr};
= Ec{CDALIU
+ R[CDA(EI,>.u)-1(Dv)v- 1 (EI>.u)
fD(E I,>.u,EI,>.v)(EI,>.v)-l(EI,>.U)) + zz}.
The proof follows immediately from Definition 12.1 and Theorem 2.2.
Corollary 12.1. Suppose that all assumptions of Proposition 12.3 are satisfied
and CD = O. Then the mappings E r,>., E l ,>. are not defined for A =P 1. If A = 1
then Er,l = EI,l = Jldom 0·
Proof Let u E dom Or. Then x = Lru E dom 0- 1 . Let A E IF. By Theorem
2.4, if AX E dom 0;1 then A = 1. Hence E r,>. is not defined for A =P 1. If A = 1
then we have Er,>.u = Er(ALru) = ErLru = u. The arbitrariness of U E dom Or
implies that E r,l = Jldom 0 •. A similar prooffor El,>.. •
Corollary 12.1 implies that for multiplicative D the mappings E r,>., E l ,>., E>. are
not defined (cf. Note 12.1).
Clearly, we can extend Definition 12.1 to left invertible operators. For the sake of
brevity, we shall consider only the commutative case, i.e. the class Lg#(D). We
get
Proposition 12.4. Suppose that X E Lg#(D}, (L, E) E G[L) and the mapping
E>. is defined by (12.1'). Let A E IF. Then E>.(dom L) C dom L and LE>. = AL
Proposition 12.5. Suppose that Condition [L)# holds and (L, E) E G[L). Let
A E IF. Then E>. E M(X) and DE>.u = A(E>'_lU)Du for u E dom L (cf. Proposi-
tion 12.2).
In general, we have
Proposition 12.6. Suppose that X E Lg#(D) and (L, E) E G[L). If A E IF and
u, v E dom L, E>.u, E>.v E J(X) then

170 Chapter 12
and CD = O. Then the mapping E>. is not defined for A =I- 1. If"\ = 1 then
E1 = Jldom !1 (cf. Corollary 12.1).
Definition 12.2. Suppose that either IF = JR or IF = C, X is a complete m-
pseudoconvex Leibniz algebra with unit e, either X E Lg(D) or X E Lg#(D),
e E dom 0- 1 and (L, E) E G[O] (Recall that for D E A(X) we have 0 = L).
Write
(12.3) Eb(X) = {u E dom L: "\Lu E ED(X) for (L,E) E G[OJ, ..\ E IF}
(cf. Definition 7.5) and
(12.4) u>' = e>.Lu for u E Eb(X), ..\ E IF.
The function u>' is said to be a power function. o

Proposition 12.7. Suppose that either IF = JR or IF = C, X E L(D) is a complete
m-pseudoconvex algebra with unit e E dom 0- 1 , (L, E) E G[O] and D is closed.
Then for ..\ E IF
(i) ifu E Eb(X), ..\ E IF, then e)..Lu E dom 0, e)..Lu = E>.u = u>' and Lu>' = "\Lu;
(ii) ifu E J(X) n Eb(X) then
(12.5)
(iii) in particular, if A = n E N then
u>' = un = U· .... u.
~
n-times
Proof. Since X is a Leibniz algebra, we have Dxn = nxn- 1Dx for all x E dom D,
n E N. (cf. Corollary 7.12). Since D is closed, by Theorem 7.6, De x = eX Dx.
Hence eX = Ex E dom 0 and LeX = LEx = x.
(i) Suppose that AU E ED(X) for a ..\ E IF, u E dom O. Then, by Proposition
12.2, u>' = E>. = E("\Lu) = e)..Lu and Du>' = De)..Lu = e>.Lu D("\Lu) = u>' D(..\Lu).
Hence u>' = e)..Lu E dom 0 and Lu>' = ALu.
(ii) If u E J(X) n EbE(X) then Du>' = DE>.u = "\(E>'_lU)Du = ..\u>'-1 Du.
(iii) Let ..\ = n E N. Since X is a Leibniz algebra, by (i), u>' = E>.u = E(ALu) =
E(nLu) = ELu n = un. •
Example 12.1. Let X = C(JR) with the pointwise multiplication and with the
topology induced by the uniform convergence on closed intervals. Let D = ft,
x E X, x(t) > 0 for t E JR. Then x>' = e>.lnx, Dx>' = "\X>.-1 Dx.
Consider the space X equipped with the topology induced by the uniform conver-
gence on compact sets and assume that D is closed. It is easy to verify that in
Power mappings 171
this case true shifts are of the form (S~x)(t) = x(rt), where h = In r, r E ll4 (cf.
Definition 11.2, Theorem 11.3). 0
In order to establish the relationship between the number e and the unit e of an
algebra under consideration it is very useful the following
Definition 12.2'. Suppose that X E Lg(D) and (L,E) E G[O]. Write
Y(O) = {(x,y) : x E dom 0, yLx E dom 0-1}
and
x Y = E(yLx) whenever (x,y) E Y(O).
By definition, Lx Y = yLx. Indeed, Lx Y = LE(yLx) = yLx. Let u = x Y for
(x, y) E Y and let y E I(X). Then
x = ELx = E(y-l LxY) = E(y-l Lu) = u y- 1

•
Clearly, x Y is a generalization of power mappings introduced by Definition 12.1

for scalar exponents, so that we call x Y a power mapping. Observe that x Y is
also a generalization of power functions introduced by Definition 12.2 for scalar
exponents. 0
Observe that, by definition, x e = x and x- e = X-I, since x e = E(eLx) = ELx = x
and x- e = E( -eLx) = E( -Lx) = X-I. Moreover, if u = x Y for (x, y) E Y(O) and
y E I(X), then x = u y - 1 •
Theorem 12.2. Suppose that X E L(D) has a unit e and (L, E) E G[O]. Then
the power function x Y has the following properties:
(i) if (x, a),(x, b) E Y(O) then a + bE Y(O) and xax b = xa+b;
(ii) if(x,a),(y,a) E Y(O) then (xy,a) E Y(O) and xaya = (xy)a;
(iii) if (x, y) E Y(O) then x Y E dom D and DxY = xY[(Dy)Lx + yx- 1 Dx], in
particular, if x E ker D then Dx Y = xY(Dy)Lx, ify E ker D then Dx Y = yx y- e Dx;
(iv) if (x,y) E Y(O) and x,y E kerD then x Y E kerD, in other words: a constant
to a constant power is again a constant;
(v) if(x,y) E Y(O) and a = Ey then xLa = a Lx ;

(vi) if (x, y) E Y(O) and a = Lx then (Ea)Y = yEa;
(vii) e Ae = e whenever>' E IF (cf. Example 7.5);
(viii) if x E dom 0 then XO = e (cf. Definition 7.5);
(ix) if(x,u),(xU,v) E Y(O) then (x,uv) E Y(O) and (XU)V = x uv ;
(x) if (x, y) E Y(O) then (x, -V) E Y(O), x Y E I(X) and (Xy)-l = x- Y;
(xi) if the logarithm L is natural (cf. Definition 7.6) then (eey = Ex whenever
x E dom 0- 1 ;
172 Chapter 12
(xii) if X is an m-pseudoconvex D-algebra and Ae E £D(X) for all A E IF (IF = IR

or IF = IC), then eAe = eAe, in particular, ee = ee.
Proof. (i) Assume that (x, a),(x, b) E T(O). Then, by the Leibniz condition,
xax b = E(aLx)E(bLx) = E(aLx + bLx) = E[(a + b)Lx] = xaH.
(ii) if (x, a),(y, a) E T(O) then
xaya = E(aLx)E(aLy) = E(aLx + aLy) = E[a(Lx + Ly)] = E[aL(xy)] = (xy)a.
(iii) Let (x,y) E T(O). Then, again by the Leibniz condition,
D(xY) = DE(yLx)
= E(yLx)D(yLx) = xY[(Dy)Lx + yDLx]
= xY[(Dy)Lx + yx- 1 Dx].
If x E ker D then Dx = 0 and D(xY) = xY(Dy)Lx. If y E ker D then Dy = 0 and

D(xY) = xYyx- 1 Dx = yxYx- e Dx = yx y- e Dx.
(iv) By (iii), if x, y E ker D then Dx Y = 0, i.e. x Y E ker D.
(v) Let (x, y) E T(O) and let a = Ey. Then = La and xLa = x Y = E(yLx) =
E[(LEy)Lx] = E[(Lx)La] = aLx .
(vi) If (x, y) E T(O) and a = Lx then x = Ea and (Ey)a = E[aLEy] =
E[y(LEa)] = (Ea)Y.
(vii) Let A E IF. Since X is a Leibniz algebra, we have Le = O. Hence e Ae =
E(AeLe) = E(O) = e.
(viii) For any x E dom 0 we have x O= E(OLx) = E(O) = e.
(ix) Let (x, u),(X U, v) E T(O). Then (XU)V = E(vLxU) = E[vLE(uLx)] =
E(vuLx) = XUV .
(x) Let (x, y) E T(O). Since X is a Leibniz algebra with unit e, we get x Y =
E(yLx) = E[-( -yLx)] = [E( -yLX)]-l = (X-y)-l. Hence (Xy)-l = x-Yo On the
other hand, by (vii), we find xYx- Y = x Y- Y = x O= e.
(xi) If the logarithm L is natural (cf. Definition 7.6) then L(ee) = elne = e·1 = e.
Hence (ee)X = E[xL(ee)] = Ex.
(xii) If all assumptions of Theorem 7.6 are satisfied, then eX = Ex. This, and (ix)
together imply that (ee)'" = Ex = eX. By the Leibniz condition, since X is an
m-pseudoconvex algebra and Ae E £D(X) for all A E IF, we have
eAe = ~ (Ae)n = ~ Anen = ~ An e = (~ An)e = eAe.

~ n! ~ n! ~ n! ~ n!
n=O n=O n=O n=O
In particular, if A = 1 then we get ee = ee. •
Power mappings 173
Definition 12.3. Suppose that X is a complete m-pseudoconvex algebra over IR

or C with unit e. Write
00 n
(12.6) Lipx = '"' ~ far x E X (p E No)
~ nP
n=l
whenever this series is convergent.
If p = ° then Liox is a geometric series and
Liox = (e - x)-l - e = -x(e - x)-l = x(x - e)-l.
If p ~ 1 then Lip is said to be a polylogarithmic function. 0

Example 12.2 (cf. ZAGIER Z[l]). Let X,D be defined as in Example 12.1. Then
we have
-In(l - t) for p=lj
(12.7) Lip(t) = L -n
00
n=l
tn
P
=
{
lt S S -ld S
L'Zp-l () for p ~ 2,
ItI < 1 (p E No).
o
o
Definition 12.4. Suppose that X E Lg(D) is a complete m-pseudoconvex algebra
over IR or C. Let p E No be arbitrarily fixed. Write
00 un
(12.8) £~)(X) = {u E dom 0: '"'
~ nJ
--:- are convergent (j = O,l, ... ,p)}.
n=l
o
Proposition 12.8. Suppose that X E L(D) is a complete m-pseudoconvex algebra
over IR or C, (L, E) E G[Ol, D is closed and pEN. If u E £~) (X) then
(12.9) Lilu = -L(e - u) and u = e - E(-LilU) E dom 0,
(12.10)
Proof. Since X is a Leibniz D-algebra, we have e E ker D. Let u E £~)(X). By

Theorem 7.4, u E J(X). Let x = LilU. Then
D
L
00 n 00 n-l 00
~= L ~Du = u-l(L un)Du = (e - U)-l Du.

n=l
n n=l
n n=l
Since D is closed, we conclude that
00 Dun
Dx = L - = (e - u)-l Du = -(e - u)D(e - u) = -DL(u - e),
n=l
n
174 Chapter 12
which implies (12.9). Since Li1u = x = -L(e - u), we find e - u = E(-x) and
u = e - E( -x). Similarly, for p 2:: 2 we get
D n n-l un-1
L n~ = L =L
00 00 00
= n:p n P - 1 Du
n=l n=l n=l
n
(L
00
= u- 1 n~-l) = (Li p_ 1 u)u- 1Du = (Li p_1u)DLu.

n=l
Corollary 12.3. Let pEN be arbitrarily fixed. Suppose that all assumptions of
•
Proposition 12.8 are satisfied, 9 = Re E cg) (X) for an R E RD and the initial
operator F corresponding to R is continuous. Then
(12.11)
Proof. We only have to prove that z = F Lipg = O. Since X is a Leibniz algebra,

by Formula (7.27), for n E N we find Fgn = F(n!Rne) = n!FRne = O. This
implies that
00 n ooF n
z = FLipg = F'"'
~nP
L = '"'
~ nP
-.!L = O.
n=l n=l
Note 12.2. Observe that both, Formula (12.10) and the second of Formulae
•
(12.11) are generalizations of (12.7). 0
Proposition 12.9. Suppose that X is a complete m-pseudoconvex algebra over
~ or ee, Condition [M] holds, (L, E) E G[O] and D is closed. Then
(12.12)
Proof. Let p E No. Since, by our assumption, D E ML(X), for u E cg)(X) we

have
Since D is closed, we get (12.12).

•
Then
1
L
00
(12.13) DLipg = e nP ' where 9 = Re, R E RD (p 2:: 2).
n=l
Power mappings 175
If F is a continuous multiplicative initial operator for D corresponding to R then
1
L
00
(12.14) Lipg = 9 nP ' where 9 = Re, R E RD (p E No).
n=1
Proof. By definition, Dg = e. This, and Proposition 12.9 immediately imply

(12.13). If F E FD corresponding to R is continuous and multiplicative, then
Fgn = (Fg)n = (FRe)n = 0 for n E Nand FLipg = O. This, and Formula (12.13)
together imply that
00
1 00
1 00
1
Lipg = R[e"
~nP
-J + FLipg = (Re) "~nP
- = g~nP
"
n=1 n=1 n=1
The functions Lip have been introduced and studied from the first beginning of
•
Mathematical Analysis, since they have several interesting properties, some of an
importance in applications to physics and engineering in the last few decades.
The results already obtained can be applied in order to solve some non-linear
equations. For instance, we have
Proposition 12.10. Suppose that X E Lgr(D) n LgI(D) is a Leibniz algebra,
(Lr,Er) E G[Or], (Ll,El) E G[Od, a,y E X, a E «:11\ {O} and Ra E dom Or n
dom 0 1 for an R E RD. Then
(i) x is a solution of the equation
(12.15) Dx = aEr(ax) + y, where Ry E dom On

only if the resolving equation
(12.16) (/ + aRy)u = z - aRa, where z E ker D is arbitrary,

has a solution. If u is a solution of Equation (12.16) for a z E ker D then x =
-a- 1 L1u is a solution of Equation (12.15);
(ii) x is a solution of the equation
(12.17) Dx = El(ax)a + y, where Ry E dom 0/,
only if the resolving equation
(12.8) (I + aRy ..... )u =z - aRa, where z E ker D is arbitrary,
has a solution. If u is a solution of Equation (12.18) for a z E ker D then x =

_a- 1Lru is a solution of Equation (12.17);
176 Chapter 12
(iii) ify = 0 then all solutions of Equation (12.15) are of the form x = -O'.- 1L , (z-
O'.Ra) , where z E ker D is arbitrary;
(iv) ify = 0 then all solutions of Equation (12.17) are of the form x = _0'.-1 Lr(z-
oRa), where z E ker D is arbitrary;
(v) if -a E vQRy and Z E ker D then all solutions of Equation (12.16) are of the
form u = LI(I + O'.Ry)-1(z - O'.Ra);
(vi) if -a E vtQRy+- and Z E ker D then all solutions of Equation (12.17) are of
the form x = -O'.Lr(I + O'.Ry+-)-1(z - O'.Ra);
(vii) if -1/0'. is an eigenvalue of the operator Ry then a necessary and sufficient
condition for the resolving equation (12.16) to have solutions is that there is a
Z E ker D such that z - O'.Ra E (I + O'.Ry)dom Or. If it is the case, then all
solutions of (12.16) are of the form u = (z - O'.Ry) (-1) + U-1/a, where by V( -1) we
denote an element of the inverse image of an element v by the mapping I + O'.Ry,
U-1/a E ker(l + O'.Ry) is arbitrary, hence Equation (12.15) has all solutions of the
form x = -0'.-1 L1u.
(viii) if -1/0'. is an eigenvalue of the operator RYt- then a necessary and sufficient
condition for the resolving equation (12.18) to have solutions is that there is a
z E ker D such that z - O'.Ra E (I + O'.Ry+-)dom 0 1. If it is the case, then all
solutions of (12.18) are of the form u = (z - O'.Ra)(_l) + U-1/a, where we denote
by v( -1) an element of the inverse image of an element v by the mapping I +O'.RYt-,
U-1/a E ker(I + O'.Ry+-) is arbitrary, hence Equation (12.17) has all solutions of
the form x = _0'.-1 Lru.
Proof (i) Let x be a solution of Equation (12.15). Set u = E,(-O'.x). Then
u E dom Or. By Theorem 7.3, u E I(X) and u- 1 = [E,( -O'.x)]-1 = Er(O'.x).
Hence
(Du)u- 1 = D[EI(O'.X)]Er(-O'.x)]
= -O'.(Dx)[E,(O'.x)]Er(-O'.x)
= -O'.Dx = -O'.(au- 1 + y)
= O'.(a + yu)u-l,
i.e. Du = O'.(a + yu). This implies that (I + Ry)u = -O'.Ra + z, where Z E ker D
is arbitrary, i.e. u should satisfy the resolving equation (12.16). Conversely, if u
is a solution of Equation (12.16) for a Z E ker D, then x = _0'.-1 LIU is a solution
of Equation (12.15). A similar proof for (ii).
(iii) Suppose that y = O. Then from the resolving equation (12.16) we find u =
z - O'.Ra, where Z E ker D is arbitrary. Hence x = -0'.-1 L1U = _0'.-1 LI(Z - O'.Ra).
A similar proof for (iv).
(v) Suppose that y =I 0, however -a E vtQRy, i.e. the operator I +O'.Ry is invertible.
Then from the resolving equation (12.16) we get u = (I +O'.Ry)-1(z-O'.Ra), where
Z E ker D is arbitrary. A similar prooffor (vi).
Power mappings 177
(vii) If -1/a is an eigenvalue of the operator Ry then the corresponding eigen-

vectors are elements of the kernel of the operator I + aRy. Solutions to the
resolving equation (12.16) exist if and only if there is a z E ker D such that
z-aRa E (I +aRy)dom Or. If they exist, then they are ofthe required form. Hav-
ing already found a solution u of Equation (12.16), we conclude that x = -a- 1Llu
is a solution of Equation (12.15). A similar proof for (viii). •
Theorem 7.7 immediately implies the following extension of Proposition 12.10 to
reals:
and, moreover, X is a complete m-pseudoconvex algebra over lR and L r , E r , L 1,
Er are continuous on their domains. Then all assertions of Proposition 12.10 hold
for a E JR.
In the sequel, for the sake of brevity,we shall consider non-linear equations in
commutative algebras.
Proposition 12.11. Suppose that Condition [L] holds, (L, E) E G[O], a E X
and (1 - a)Ra E dom 0 for an R E RD and an a E IF \ {I}. Then the general
solution of the equation Dx = aEOtx is of the form x = E 1/{l-Ot)[(1 - a)Ra + z],
where z E ker D is arbitrary.
Proof. Recall that x + z E dom 0 whenever x E dom 0, z E ker D. Indeed, let
(u,x) E graph 0 and z E kerD. Then Du = uDx = uD(x + z) which implies
(u, x + z) E graph O. By our assumptions,
= (1 - a)(E1- Ot _ 1x)Dx = (1 - a)(E_Otx)Dx

= (1 - a)(E_Otx)a(EOtx) = (1- a)(E_Otx)(EOtx)a
= (1 - a)ae = (1 - a)a.
Hence E 1- Ot x = (1 - a)Ra + z, where z E ker D is arbitrary. On the other hand,

y = E 1- Ot x = E[(1 - a)Lx], which implies

Proposition 12.12. Suppose that all assumptions of Proposition 12.11 are sat-
•
isfied and F is an initial operator for D corresponding to R. Then x is a solution
of the equation
(12.19) Dx = aEOtx + y, where y E dom 0,
if and only if x is a solution of the resolving equation:
(12.20) x = E 1/(1-Ot)[(I- a)R(a + yEOtx) + Fx].

178 Chapter 12
Proof. Let v = yE_Otx. Then y = vEOtx and Equation (12.19) can be written
in the form: Dx = (a + v)EOt. This, and Proposition 12.11 together imply that
x = E 1!(1_Ot)[(1-a)R(a+v)+z], where z = Fx E ker D. Thus x satisfies Equation
(12.19). On the other hand, if x satisfies (12.20), then it is easy to check that x is
a solution of Equation (12.19). •
Proposition 12.13. Suppose that either IF' = R or IF' = C, X is a complete m-
pseudoconvex algebra with unit e, Condition [L] holds, e E dom n- 1 , (L, E) E
G[O], D is closed, a E X and (l-a)Ra E £b(X) for an R E RD and an a E 1F'\{1}.
Then the general solution of the equation
(12.21)
is of the form: x = [(1 - a)Ra + zj1!(l-Ot), where z E ker D is arbitrary.

Proof. By Proposition 12.7, x Ot = EOtx = eOtLx. This, and Proposition 12.11
together imply that all solutions of Equation (12.21) are of the required form. •
In a similar manner (cf. Proposition 12.12) we obtain
and F is an initial operator for D corresponding to R. Then x is a solution of the
equation Dx = ax Ot + y, where y E dom n, if and only if x is a solution of the
resolving equation x = [(1 - a)R(a + xOty) + zj1!(l-Ot), where z = Fx E ker D.
Using notions introduced in Example 12.2, we can generalize Proposition 12.13 in
the following manner.
Proposition 12.13'. Suppose that Condition [L] hold, Ra E dom 0 for an
R E RD, (L, E) E G[O], bE ker D and e - b E J(X). Then the equation
(12.21')
for (x, b) E T(n) has the general solution of the form
x = [(e - b)Ra + z](e-b)-l , where z E ker D is arbitrary.
Proof. Since X is a Leibniz commutative algebra, we have e E ker D. Thus

e-b E J(X) implies that e-b E dom n. Moreover, R[(e-b)a] = (e-b)Ra E dom n.
If x is a solution of Equation (12.21') then (x, b) E T(n). Then
DE[(e - b)Lx]
= E[(e - b)Lx]D[(e - b)Lx] = E[(e - b)Lx](e - b)DLx
= E[(e - b)Lx](e - b)x- 1Dx = E[(e - b)Lx](e - b)x-1axb
= E[(e - b)Lx](e - b)ax b- e = (e - b)aE[(e - b)Lx]E[(b - e)Lx]
= (e - b)aE[(e - b + b - e) Lx] = (e - b)aE(O) = (e - b)ae
=(e-b)a.
Power mappings 179
This implies that
E[(e - b)Lx] = R[(e - b)a] + z = (e - b)Ra + z, where z E ker D is arbitrary,
i.e. (e - b)Lx = LE[(e - b)Lx] = L[(e - b)Ra + z]. Hence
x = ELx = E{(b - e)-l L[(e - b)Ra + z]) = [(e - b)Ra + z](e-b)-l,

By our assumption, since b E dom f!, we find that bE J(X). Again, by Example
•
12.2, for (b,x) E Y(f!) we have
bX = E(xLb) = E[(Lb)LEx] = (Ex)Lb.
We therefore conclude that the equation Dx = ab x can be solved in the same

manner as Equation (12.21'), taking Ex and Lb instead of b and x, because Lb E
ker D whenever b E ker D.
We shall see that several nonlinear equations may be reduced to the cases already
considered in Propositions 12.10-12.13 and Corollaries 12.5, 12.6.
Corollary 12.7. Suppose that either IF' = lR or IF' = C, X is a complete m-
pseudoconvex algebra with unit e, Condition [L] holds, e E dom f!-l, (L, E) E
G[f!] and D is closed. Then the generalized Riccati equation:
Dx = ax 2 + bx + c, where a,b,c E X, a E J(X),
has a solution x = u - ~a-l b if and only if u is a solution of the equation
(12.22) Du = au 2 +y,
i.e. if u satisfies the resolving equation u[Fu - R(yu 2 + a)] = e, where F is an

initial operator for D corresponding to an R E 'RD.
Proof It is easy to verify that Dx = a(x+ ~a-lb)2 +c- ~a-2b2. Let u = x+ ~a-lb.
Then
1 1 1 1
Du = Dx + D("2a-1b) = a(x + "2a-1b)2 + c - 4a-2b2 + D("2 a - 1b ) = au 2 + y.
In order to find solutions of (12.22), we apply Corollary 12.6 for 0: = 2. •

Proposition 12.14. Suppose that all assumptions of Corollary 12.7 are satisfied
and v is the already known solution of Equation (12.22). If the operator J - 2R(av)
is invertible for an R E 'RD and
(12.23) w = -[J + 2R(av)tl Ra + z E J(X) for a z E ker D
then u = v - w- 1 is also a solution of Equation (12.22).

180 Chapter 12
Proof. Let u = v - w'. By our assumptions, we have Dv = av 2 + y. We are

looking for w' such that u is a solution to Equation (12.22). We have
Dv = Du+Dw'
=
= au 2 + y + Dw' a( v - W')2 + y + Dw'
= av 2 - 2avw' + aw,2 + y + Dw',
which implies Dw' = 2avw' + aw'2. Suppose that w' E leX) and w = W,-I. Then
we get Dw = DW,-1 = _W,-2 Dw' = -(2avw,-1 + a) = -2avw - a which implies
D[l + 2R(av)] = -a. Hence w is of the form (12.23). •
In particular, if a E ker D, y = _a- 1 z 2 , z E ker D, Equation (12.22) can be
reduced by the substitution v = -yu-1 to the equation D 2v = v 2 - z2 solvable in
an explicit way. Namely, we have
Proposition 12.15. Suppose that Condition [L] holds, (L, E) E G[O), R E RD,
9 = Re and a E leX) n ker D. Then the special generalized Riccati equation
(12.24)
has particular solutions of the form: x = a, x = - a. If ag E dom 0- 1 and

E(2ag) + ker D c leX) then all solutions x of Equation (12.24), such that Dx E
leX), are of the form: x = a[z + E(2ag)][z - E{2ag)]-I, where z E ker D is
arbitrary.
Proof. Let R and a satisfy all our assumptions. Since a E ker D, Condition [L]
implies that Ra = aRe = ago Suppose that either x = a or x = - a. Then Dx =
o = x 2 - a 2, which proves that elements x = a, x = - a are particular solutions
of Equation (12.24). If ag E dom 0- 1 then Condition [L] implies E(ag)E(ag) =
E{2ag). Hence 2ag E dom 0- 1 .
Observe that the assumption that E{2ag) + ker D c leX) is not very restrictive.
Indeed, suppose that there is a z E ker D such that E{ag) = z. This implies that
Lz E ker D and a = aDRe = D{ag) = DLE(ag) = DLz = 0, a contradiction with
our assumption that a E leX).
We are looking for solutions x of Equation (12.24) such that Dx E leX). If it
is the case, then x 2 - a 2 = Dx E leX). Condition [L] implies that a2 E ker D.
Hence
2a = 2a(x2 - a 2)-IDx
= 2a(x - a)-lex + a)-l Dx
= [(x - a)-l - (x + a)-I]Dx
= (x - a)-l D(x - a) - (x + a)-l D(x + a)
= DL(x - a) - DL(x + a) = DL[(x - a)(x + a)-I],
which implies
L(x - a)(x + a)-l = L[(x - a)(x + a)-I] = 2Ra + z' = 2ag + z',
Power mappings 181
where z' E ker D is arbitrary. Since z = E( -z') = (Ez,)-l E ker D too, we

have z E J(X) and E(2ag + z') = (Ez')E(2ag) = z-l E(2ag). Thus we find
(x - a)(x + a)-l = z-l E(2ag), which implies (x + a)z-l E(2ag) = (x - a). Finally,
we get
x = -a[z-l E(2ag) + e][z-l E(2ag) - e]-l = a[E(2ag) + z][z - E(2ag)]-1 .
Example 12.3. Let either IF = IR or IF = C. Let D E A(X). Let P(D)

•
D2 - pD - q, where p,q EX. A necessary and sufficient condition for P(D) to be
factorized in the form: P(D) = (D - p)(D + p) is that Dp = p2 - q (cf. Example
8.1). Indeed,
P(D) = D2 - pD + Dp - p2 - q - Dp + p2 = (D - p)(D + p) _ (Dp _ p2 + q).

=
If q E h(ker D), i.e. there is an a E J(X) n ker D such that q a 2 , we may apply
Proposition 12.15. Observe that for a E ker D, q = a 2 , Condition [L] implies that
Dq = 2aDa = 0, i.e. q E ker D (cf. also SKORNIK and WLOKA SW[l]). D
Example 12.4. Let IF = C. Let D E A(X). Let P(D) = D2 - pD + q, where
p,q E X. A necessary and sufficient condition for P(D) to be factorized in the
form: P(D) = (D - p)2 is that Dp = p2 - q. Indeed,
P(D) = D2 - pD - Dp + p2 + Dp _ p2 + q = (D _ p)2 + (Dp _ p2 + q).

Similarly, as in Example 12.4, if -q E J2 (ker D), i.e. if there is an a E J(X) nker D
such that ql/2 = ia, then we may apply Proposition 12.15. D
Note 12.3. We have assumed in Proposition 12.15 that a E J(X), since in the case
a = 0 we have the equation Dx = x 2 . Any solution of this equation belonging to
J(X) is of the form x = (Z_g)-l, where z E ker D. Indeed, -e = -x- 2Dx = Dx- 1
which implies x-I = -Re + z = z - g, where R E RD and z E ker D. Recall that
a change of a right inverse R implies a change of the constant z only. Indeed, if
Rl E RD, Rl =f. 0 then R1u - Ru = Zl E ker D for all u E X (cf. PR[8]). Recall
also that Re E J(X) if and only if Re E dom 0 (cf. Theorem 7.4). D
Proposition 12.16. Suppose that all assumptions of Proposition 12.11 are satis-
fled, a = 2, a = d - e, v = z - Ra E J(X) for some z E ker D and Rv E dom 0- 1 .
Then all solutions of the equation
(12.25)
belonging to J(X) n dom 0 are of the form x = z'ER[z - R(d - e)t 1, where
z,z' E ker D.
Proof. Let u = Lx, i.e. x = Eu and
Dx = DEu = (Eu)Du = xDu; D 2x = (Dx)Du + xD 2u = X[(DU)2 + D 2u].
182 Chapter 12
Hence Equation (12.25) can be written as x 2[(Du)2 + D 2u] = dx 2(Du)2. Since

x E I(X), we obtain the following equation: (Du)2 + D 2u = d(Du)2, i.e. D 2u =
(d - e)(Du)2 = a(Du)2. Let v = Du. Then the last equation is of the form:
Dv = av 2. By our assumptions, arguing as before, we find v = [-Ra + zt 1 =
[z - R(d - e)]-l, where z E kerD. Since u = Rv + Zl, where Zl E kerD and
z' = EZI E ker D, we find
x = Eu = E(Rv + Zl) = (EZl)(ERv) = z' ERv = z' ER[z - R(d - e)t 1,
where z, z' E ker D.

Proposition 12.17. Suppose that either IF = =
IR or IF te, X is a complete m-
•
pseudoconvex algebra with unit e, Condition [L] holds, e E dom n- 1 , (L, E) E
G[nJ, D is closed and
N
Q(D) =L Qk Dk , Qk E L(X) (k = 0,1, ... , N - 1), QN = I.
k=O
Suppose, moreover, that the operator Q(I, R) = I:f=o QkRN-k is invertible for
an R E RD. Then x is a solution of the equation
[Q(D)x]" = aEax + y, a, y E X, 0, 'Y :j; 0,
if and only x satisfies for an Xo E ker Q(D) the resolving equation
The proof follows immediately from the form of the general solution to the equation
Q(D)x = y (cf. Theorem 5.3). •
Similar results can be obtained for operators of the form:
DM Q(D), Q(D)D M , Q(D), DM Q(D), Q(D)D M (M E N)
(cf. in particular, Theorems 5.3 and 5.4). Equations with operators of a more
general form can be solved by applications ofresults of NGUYEN VAN MAU N[2]).
Proposition 12.18. Suppose that Condition [L] holds, (L, E) E G[n], a E dom n
and f is a mapping of dom n into itself. Then a necessary and sufficient condition
for the equation
(12.26) Dx = f(aEx) +y, y E X,
to have a solution x = Lu is that u E dom n is the solution of the equation

(12.27) u = Ru[f(au) + y] + z, where Z E ker D is arbitrary, R E RD.
Power mappings 183
Proof. We are looking for solutions x E dom n- l . Let u = Ex. Then u E dom n.
By definition, Du = uDx = u[f(aEx) + y] = u[J(au) + y], which implies u =
Ru[f(ay) + y] + z, where Z E ker D is arbitrary. Clearly, if this equation has a
solution u E dom n then x = Lu is a solution of Equation (12.26). Conversely, if x
is a solution of Equation (12.26) then u = Ex E dom n satisfies Equation (12.27) .
Proposition 12.19. Suppose that Condition ([L] holds, (L, E) E G[nJ, a E

•
dom nand f is a mapping of dom n- l into itself. Then a necessary and sufficient
condition for the equation
(12.28) Dx = aEf(x) + y, y E X,
to have a solution x is that the equation
(12.29) v = aEf(Rv + z) + y, where Z E ker D is arbitrary, R E RD,
has a solution v E dom n. If it is the case, then x = Rv + z. In particular, if y =0

then every solution of Equation (12.29) belongs to dom n.
Proof. Write v = Dx. Then x = Rv+z, where z E ker D is arbitrary and R E RD.
Hence v = Dx = aEf(x) + y = aEf(Rv + z) + y. If this equation has a solution
v E dom n then x is a solution of Equation (12.28) and conversely. In particular, if
y = 0 then we find Lv = L[aE(f(Rv+z)] = La+LEf(Rv+z) = La+ f(Rv+z) E
dom n- l , which implies v E dom n. •
Proposition 12.20. Suppose that Condition [L] holds, (L, E) E G[n], 9 = Re
for an R E RD and >.-l,.clg E dom n for some >',J.L E IF \ {a}. Then the special
generalized Clairaut equation
(12.30) x = gDx + >.E(J.LDx)
has solutions Xl = J.L-lRL(>.-lJ.L-lg) +z; X2 = RZI + zo, where Z,Zl,ZO E kerD
are arbitrary. If(L,E) E GR,l[n] then Xl = J.L-lRL(>.-lJ.L-lg).
Proof. By our assumptions, >.-lJ.L-lg E dom n, hence is invertible. Moreover,
Dg = DRe = e. If we apply the operator D to the both sides of Equation (12.30)
then we get
Dx = (Dg)Dx + gD 2 x + >.DE(J.LDx) = Dx + gD 2 x + >,J.L[E(J.LDx)]D2 x,

i.e. [>,J.LE(J.LDx) + g]D 2 x = O. Clearly, if D 2 x = 0 then x = Xl = RZI + Zo,
where Zl,ZO E ker D are arbitrary. Let now >'J.LE(J.LDx) + 9 = O. Then E(J.LDx) =
_>.-lJ.L-lg and Dx = J.L- l L( _>.-lJ.L-lg) = J.L- l L(>.-lJ.L-lg). Hence x = X2 =
J.L- l RL(>.-lJ.L-lg) + z, where Z E ker D. If (L, E) E GR,dn] then Z = 0 and
X2 = J.L- l RL(>.-l J.L-lg). •
JR, Condition [L] holds, (L, E) E G[n], D is closed, 9 = Re E Eb(X) for an RD
and a, >. E JR \ {O}. Then the special generalized Clairaut equation
(12.31) x = gDx + >'(Dx)"', a E JR \ {O, I},

184 Chapter 12
has solutions Xl = REI /(a-1) (0'.-1 )..-lg) + Z; X2 = RZI + zo, where z, Zl, Zo E ker D
are arbitrary. If (L, E) E GR,l [flj then Xl = REl /(cr-1) (0'.-1 )..-lg).
Proof. Similarly, as in the proof of Proposition 12.20 we obtain from Equation
(12.31) the equation [)"O'.(Dx)cr-l + g]D 2x = O. The equation D 2x = 0 has all
solutions of the form X2 = RZI + Zo, where zo, Zl E ker D. Suppose now that
)"O'.(Dx)cr-l + 9 = O. Then (Dx)cr-l = _O'.-l)..-lg. Hence
and Xl is of the required form.

Example 12.5. Recall the so-called Schwarz derivative (or Schwarzian):
•
(12.32) S(f) = (f" / 1')' - !(fll / 1')2
2
defined for all twice continuously differentiable functions f of the complex variable
such that l' =F 0 and 1" / l' is again a differentiable function. It is well-known that
S(f) = 0 if and only if f is a homographic function, i.e.
f(z) __ az+b, h
were a, b, c, d E II....
tr'
an d a d - b..J.
c r O.
cz+d
o
We shall see that this property can be generalized. Namely, we have
Theorem 12.3. Suppose that Condition [Lj holds and (L, E) E G[fl]. Let
generalized Schwarzians Sn (n E N) be defined as follows:
(12.33) dom Sn = {x E dom D2 : Dx E dom fl, DLDx E I(X) n dom D}
and for all X E dom Sn (n E N)
Then X E ker Sn if and only if
~IZl(ZO - g)-n+1 + Z2, ifn ~ 2;

(12.34) x - { n-
-zlL(zo - g) + Z2, ifn = 1;
where 9 = Re, R E RD is arbitrarily fixed and Zo, Zl, Z2 E ker D are arbitrary.
Power mappings 185
Proof By Theorem 7.4, Dx E dom n implies Dx E J(X). By our assumptions,

u = DLDx = (DX)-1 D 2x E J(X) n dom D. Let n E N be arbitrarily fixed.
Sufficiency. Suppose that x is of the form (12.34). Since ZI, Z2 E ker D, it is enough
to verify that x = (zo - g)-l+n E ker Sn, where Zo E ker D is arbitrary. Indeed,
Condition [L] implies that D(zv) = zDv for all v E dom D, z E ker D. We have
Dx = D(zo - g)-n+l = (-n + 1)(zo - g)-n(-DRe) = (n -1)(zo _ g)-n,
D 2x = -n(n - 1)(zo - g)( -DRe) = n(n - 1)(zo _ g)-n+l.
Let n ~ 2. Then
1
u = (DX)-1 D 2x = --1 (zo - g)nn(n - l)(zo - g)-n-l = n(zo _ g)-I.
n-
Thus u 2 = n 2(zo - g)-2 and Du = -n(zo - g)-2( -DRe) = n(zo _ g)-2 = ~u2
which implies that x E kerSn . Let now n = 1. Then x = -L(zo - g) and
Dx = D[-L(zo - g)] = -DL(zo - g) = -(zo - g)-I( -DRe) = (zo _ g)-I,
D2x = -(zo - g)-2(-DRe) = ZI(ZO _ g)-2,

u = (DX)-ID 2x = (zo - g)-I(ZO - g)-2 = (zo - D)-I.
Hence Du = (zo - g)-2 = u 2, which implies that x E kerSI.

Necessity. If x E Sn then Du - ~U2 = Sx = O. Hence Du- 1 = -u- 2Du = -~e,
which implies u- 1 = -~g + ~zo, where 9 = Re, R E RD is arbitrarily fixed and
Zo E ker D is arbitrary. Thus u = n(zo - g)-I. By definition,
DL(zo - g) = (zo - g)-1 D(zo - g) = (zo - g)-I( -DRe) = -(zo _ g)-I.

By our assumption, v = Dx E J(X). Hence
DLv = v- 1Dv = u
=n(zo - g)-1 = -nDL(zo - g)
= D[-nL(zo - g)] = DL(zo _ g)-n,
which implies Lv = L(zo - g)-n + LZI = L[ZI(ZO - g)-n], where ZI E ker D is
arbitrary. Condition [L] implies that
D(zo - g)-n+l = (-n + l)(zo - g)-n(-DRe) = (n -1)(zo _ g)-no
If n ~ 2 then, by Condition [L],
1 1
Dx = v = ZI(ZO - g)-n = zID[--(zo
n-l
- g)-n+l] = D[--ZI(ZO - g)-n+l],
n-l
186 Chapter 12
which implies
1
x = --1Z1 (Zo - 9 )-n+l + Z2, where Z2 E ker D is arbitrary.
n-
Ifn = 1 then Dx = v = ZI(ZO-g)-1 = -D[ZIL(ZO-g)]. Hence x = -ZIL(ZO-g)+

Z2, where Z2 E ker D is arbitrary. We therefore have obtained Formulae (12.34) .
Consider now equations with logarithmic and antilogarithmic mappings of higher

•
order. However, in the case of order n 2:: 2 we cannot apply the Leibniz condition.
So that, instead of Proposition 12.10 we have the following statement with a
nonlinear resolving equation:
Proposition 12.22. Suppose that n E N, Condition [L] holds, (Ln, En) E G[OnJ,
a,y E X, a E Q \ {O} and Rna, Rny E dom On for an R E RD. Then x is a
solution of the equation
(12.35)
only if the resolving equation:
(12.36)
n-l
where Z = L Rkzk' ZI, .. ·,Zn-l E kerD are arbitrary,
k=O
has a solution U E dom On. If U is a solution of Equation (12.36) for a Z E ker D,
then x = -a- 1 Lnu is a solution of Equation (12.35).
Proof Let U = En(ax). Then u E dom On and x = a-I Lnu. By definition,
This implies Equation (12.36). If u is a solution of that equation belonging to

dom On then x = a-I Lnu is a solution of Equation (12.35). •
Theorem 7.7 immediately implies the following extension of Proposition 12.22 to
reals:
Corollary 12.8 Suppose that n E N, X is a complete m-pseudoconvex algebra
over JR, Condition [L] holds, (Ln, En) E G[OnJ, Ln and En are continuous, a, y E
X, a E JR \ {O} and Rna, RnJl E dom On for an R E RD. Then x is a solution of
Equation (12.35) only if the resolving equation (12.36) has a solution u E On. If U
is a solution of Equation (12.36) for a Z E ker D, then x = -a- 1 Lnu is a solution
of Equation (12.35).
CHAPTER 13
SMOOTH LOGARITHMS AND ANTILOGARITHMS
In Chapter 11 we have introduced the space Doc of smooth elements and its
subspaces S = P(R) of D-polynomials, E(R) of exponentials and AR(D) of D-
analytic elements. In addition, we will introduce other subspaces of Doc. Namely,
let D E R(X), ker D -:j:. {O} and let F be an initial operator for D corresponding
to an R. We shall consider the following subspaces of Doc:
• the space of singular elements
QR(D) = {x E Doc: V FDnx = O} = {x E Doc: V RnDnx = x};
nENo nEN
and, if X is a complete linear metric space and either IF = lR or IF = C,
• the space of D-paraanalytic elements
i.e. such elements that the nth terms in their Taylor expansions are tending to
zero (cf. PR[9]). Note that the inclusion AR(D) ffi QR(D) c PAR(D) is essential
(cf. ROGULSKI Rog[l], also PR[9]).
Theorem 13.1. Suppose that X E Lgr(D) (X E LgI, X E Lg(D)) is a complete
linear metric space over IF (lR or IF = C) , (Lr,Er ) E G[Or] (Ll,Et) E G[Ot],
(L,E) E G[OJ, respectively), D E R(X), there is an R E RD n AQN(ker D), F
is an initial operator for D corresponding to Rand 9 = Re. If >..g E dom Or
(>..g E dom Ol, >..g E dom 0, respectively) for>.. E VFR, then
Er(>"g), El(>"g), E(>..g) E ker(Dn - >..n I) far all n EN,
Er(>..g) = (I - >"R)-l FEr (>..g); El(>..g) = (J - >"R)-l FEl(>..g);

E(>..g) = (I - >"R)-l FE(>..g)
and Er(>..g), El(>"g), E(>..g) E AR(D), respectively.
Proof. Let>.. E vJFR. Observe that Er(>..g) E Doc. Indeed, by definition,
DEr(>..g) = Er(>..g)D(>..g) = >..Er(>..g)DRe = >..Er(>..g).
By an easy induction we find that DnEr(>..g) = >..nE(>..g) for all n E N, i.e.
Er(>..g) E ker(D n - >..n J) (n EN). Then
(I - >..R)Er(>..g)
= Er(>..g) - >..REr(>..g)
= (I - RD)Er(>..g)
= FEr (>..g) = Zr E ker D.
187
188 Chapter 13
Since>. E VJFR, we conclude that Er(>.g) = (I - >'R)-l Zr. By our assumption the
series 2:~=o >.n Rn z is convergent for all z E ker D and>' E VJFR. This implies
00
n=O
n=O
= (L
00
>.n Rn) FEr (>.g)

n=O
Hence Er(>.g) E AR(D). Similar proofs for EI(>'9) and E(>.g).

Note 13.1. Observe that a right inverse R satisfying assumptions of Theorem
•
13.1 is almost quasinilpotent on ker D: R E AQN(ker D) (cf. Definition 11.1 and
Theorem 11.1). 0
{ShhEA(IR) is a family of true shifts induced by R (cf. Definition 11.2). Then
ShEr(>.g) = eAh Er(>.g)j ShEI(>.g) = eAh El(>.g)j

ShE(>.g) = eAh E(>.g) (h E A(JR)).
Moreover, Er(>.g) = =
0 if and only if FhEr(>.g) 0 for all h E A(JR), El(>.g) 0 if=
and only if FhE/(>.g) = 0 for all hE A(JR), E(>.g) = 0 if and only if FhE(>.g) = 0
for all h E A(JR), where Fh = FSh.
Proof By Theorems 11.3 and 13.1, for all h E A(JR)
Thus
FhEr(>.g) = FShEr(>.g) = eAhFEr(>.g) = eAh(I - >.R)Er(>.g).
This and the invertibility of I - >'R together imply that Er(>.g) = 0 if and only if
FhEr(>.g) = 0 for all h E A(JR). Similar proofs for El(>.g) and E(>.g). •
Note 13.2. Clearly, the operators Fh = FSh (h E A(JR)) belong to :Fv, since they
are projections onto ker D and the corresponding right inverses are Rh = R - FhR
Smooth logarithms and antilogarithms 189
(h E A(~». One can prove that VFRh = VFR for all h E A(JR) (cf. BPR[2j, also
Appendix). 0
Theorem 13.2. Suppose that X E Lr(D) (X E LI(D), X E L(D») is a complete
linear metric space over IF (where IF = ~ or IF = C) and (Lr,Er ) E GR,l[Orj
(Ll,EI ) E GR,dOd, (L,E) E GR,dO], respectively) for a continuous R E 'RD. If
u E J(X) n AR(D) and u- l E AR(D) then Lru, L1u, Lu E AR(D), respectively.
Proof. Ifu,v E AR(D) then Du,Ru, DV,Rv E AR{D) (cf. PR[9]). By the Leibniz
formula, uv E AR(D). Let u E J(X) n AR(D) and let u- l E AR{D). Then
DLru = u- l Du, DLlu = (Du)u- l , DLu = u- l Du E AR(D). Let F be an initial
operator for D corresponding to R. By our assumptions, F Lru =
0, F Lzu 0, =
FLu = 0, respectively. This implies
Similarly, Ltu E AR(D), Lu E AR(D).

•
Corollary 13.2. Suppose that all assumptions of Theorem 13.2 are satisfied
and {ShhEA(IR) is a family of true shifts induced by R. Then ShLru = e hD Lru
(ShLtu = e hD Ltu, ShLu = e hD Lu, respectively) for all h E A(~).
Note 13.3. Similar results as Corollaries 13.1 and 13.2 can be obtained if instead
of true shifts we shall consider functional shifts introduced and studied recently
by BINDERMAN (cf. Appendix). In order to do this, we shall take instead the
function J(t) = et any function J E H(O), i.e. analytic in a domain 0 C C. 0
Theorem 13.3. Suppose that X E Lr(D)nLI(D) is a complete m-pseudoconvex
algebra over IF (where IF = ~ or IF = C), X has the unit e E dom 0;:-1 n dom Oil,
(Lr,Er ) E G[Or], (Lt,EI E G[Oz), D is closed, FE ML(X) is a continuous initial
operator for D corresponding to an R E RD and 9 = Re. Then for all>' E IF
(i) >.g E CD(X), Er(>.g) = Ez(>.g) = e Ag and Fe Ag = ej
(ii) Dne Ag = >.ne Ag and FDneA9 = >.ne for n E Nj
(iii) eAg E AR(D).
Proof (i) Since X is a Leibniz D-algebra with unit e, by (7.27), the series
(13.1)
is convergent for every ,X E IF. Hence e Ag E CD(X). By Theorem 7.6, Er(,Xg) =

Et(,Xg) = e Ag • Since for all n ? 1 we have Fgn = (Fg)n = (FRe)n = 0 and
Fe = e, we find
00 ,Xn n oo,Xn
Fe Ag =FL - g = L -Fgn=e.
n=O n! n=O n!
190 Chapter 13
(ii) is proved by an easy induction (d. also Theorem 13.1).

(iii) Since >..g E £D(X), by (ii), for all n E N we have
Hence, by (i), the series (13.1) is convergent. Thus e Ag E AR(D).

Theorem 13.4. Suppose that all assumptions of Theorem 13.3 are satisfied. Let
•
XD be defined by Formula (7.22). Then for all x E XD n QR(D) and >.. E IF we
have eX - e E QR(D). If, in addition, xg = gx then e Ag+X E AR(D) EB QR(D).
Proof Let x E XD n QR(D). Then xDx = (Dx)x, the series
00
L
n=O
:!n
is convergent and F Dnx = 0 for all n E No. Since F E M L(X) and Fx = 0, we

have Fxn = (Fx)n = 0 for n 2:: 1, which implies
00 n 00 1
(13.2) Fe x = FL..J
" ~ = Fe + "
n! L..J
-Fxn
n!
= e.
n=O n=l
We shall prove by induction that F Dne x = 0 for n E N. Indeed, by (13.2) and

Theorem 7.6, we get FDe x = F(e XDx) = (FeX)(FDx) = eFDx = O. Suppose
that F Dke x = 0 (k = 0,1, ... , n) for an arbitrarily fixed n 2:: 1. Then, by the
Leibniz formula,
FDn+le x = FDn(De X )
= FDn(e X Dx) = F t (~) (Dn-kex)Dk+lx
t
k=O
= (~) (FDn-kex)FDk+lx = O.
k=O
Hence y = e- e E QR(D), because FDny = 0 for n E No.

Let>.. E IF. By Theorem 13.3, e Ag E AR(D). Let x E XD and let xg = gx. Since
eX - e E QR(D), we get
e Ag = eAgex = eAg(e X - e) + e Ag E AR(D) EB QR(D).
Indeed, the product of an element belonging to Doo by an element of QR(D) again

belongs to QR(D) (cf. PR[9]). •
Note 13.4. If D E A(X} then xDx = (Dx}x for every x EX. Thus, instead of
the assumption that x E XD, it is enough to assume that x E £D(X) (d. Corollary
7.14). 0
Theorem 13.5. Suppose that all assumptions of Theorem 13.3 are satisfied,
FI =F F is a multiplicative initial operator for D corresponding to an Ri E nD
and 9 = Re E leX).
(i) If
(13.3)
then Lrg, Llg E PAR, (D);

(ii) if Ri is continuous and (13.3) holds then g~1 E PAR, (D);
(iii) if the series
(13.4)
is convergent then Lrg, L1g EARl (D);

(iv) if RI is continuous and the series (13.4) is convergent then g-n EARl (D) for
all n EN.
Proof. By our assumptions, RI =F R. Since FI E ML(X), we obtain for RI the
same property (7.27) as for R:
(RIe)n
RiZ = z--,
n.
- for Z E ker D, n E N.
Since 9 E leX), we conclude that FIg E leX). Indeed, by our assumptions,

e = FIe = FI(gg-l) = (FI9)(FI g-I). By definition, FRI = -FIR (cf. PRI8],
Section 2.2). Hence FRle = -FiRe = -FIg E leX). Let rn(x), In(x) (n E No)
be defined by Formulae (7.15). By Corollary 7.12, rn(g) = In(g) = (n + l)e for
n E No. This, and Theorem 7.6 together imply that DLrgn = DL19n = ng- l
(n EN.) .
(i) By an easy induction we prove
Hence for all n E No we have
R~FIDnLrg
= (FIDnLrg) (Rle)n = Fd(-l)n-l(n _ l)!g-nJ (Rle)n

n! n!
= (_l)n-l(n _ l)!(Flg-n) (RIe)n = (_1)n-1 (FIg)-n(Rle)n
n! n
= (_l)n-I (FIRe)-n(Rlet = (_l)n-1 (-FRle)-n(Rle)n
n n
( l)n-I-n 1
= n (FIRe)-n(Rle)n = --(FIRe)-n(Rle)n.
n
192 Chapter 13
Now, (13.3) shows that Lrg, L,g E PARI (D).

(ii) By (i), since Rl is continuous and (13.3) holds, we get
lim RrFIDng
n~oo
= DRI lim DnLrg
n~oo
=D lim R~+lFIDn+lLrg
n~oo
= O.
Hence g-1 E PARI (D). The same conclusion is obtained by means of the left
logarithm L , .
(iii) Since the series (13.4) is convergent, by (ii), the series
L
00 00
Rr F I DnL rg and L Rr F I DnL ,g

n=O n=O
are convergent. Thus Lrg, L,g EARl (D).

(iv) Since Rl is continuous, the space ARI (D) is D-invariant, i.e.
Hence, by (iii), g-1 = DLrg = DL,g EARl (D). Since the product of two elements
belonging to ARI (D) again belongs to ARI (D) (cf. PR[9]) , by an easy induction
we prove that g-n = (g-l)n EARl (D). •
Immediate consequences of Theorem 13.5 are
Corollary 13.3. Suppose that all assumptions of Theorem 13.5 are satisfied, Rl
is continuous,
N
P(t) = L Pk tk E JF[t]
k=O
and the series (13.5) is convergent. Then P(g-l) = g-np1(g) E ARI(D), where
N = degP(t) and
N
Pt(t) =L PN_k tk .
k=O
P(t) E JF[t]. Then P(Lrg), P(L,g) EARl (D).
Theorem 7.6 immediately implies for commutative algebras with left invertible
operators the following
JR, Condition [L]# holds, (L, G) E G[L] and D is closed. Then De" = e" Dx for
x E £D(X) (cf. also Theorem 13.4).
However, we have to remember that spaces S, AR(D) and QR(D) are well-defined
for right invertible operators with ker D ¥ {O}.
Again we shall see that in the mUltiplicative case we have a different situation.
Proposition 13.2. Suppose that either IF = IR or IF = C, D E ML(X), X E

Lgr(D) nLg) (D) with unit e E dom 0;:-1 ndom Oil is a complete m-pseudoconvex
algebra and (Lr,E r ) E G[Or], (LI,EI) E G[Od. If D is closed then Dx E CD(X)
whenever x E CD(X) and De'" = e D"'.
Proof Let x E CD(X). By our assumptions, De = e and
~ Dxnn! = ~ (Dx)n = eD"'.

L....t
n=O n=O
Dxn
L....t --;;! = 2:00
n=O
n!
Note 13.5. Observe that e'" is not an antilogarithm for a multiplicative D. If,
•
in particular, D E A(D), X is a complete m-pseudoconvex algebra over IF, where
either IF = IR or IF = C, Condition [M] holds, (L, E) E G[O] and D is closed, then
De'" = e D'" whenever x, Dx E CD(X). 0
and L is continuous. Then
(i) Lu E CD(X) and eLu = LeU = u for u E CD(X) n dom OJ
(ii) Du = e Du - u ifu, Du, Du - u E CD(X).
Proof By Proposition 2.6, Land E are multiplicative.
(i) By definition, if u E CD(X) n dom 0 then x = Lu E dom 0- 1 and
00 (L)n 00 L nOOn
U = Ex = eX = e Lu = ""'
L....t
_u_
n!
= ""'
L....t
= L ""'
~
n!
~
L....t n!
= Leu.
n=O n=O n=O
(ii) By (i), we have Du = DLe u . On the other hand, by definition, De u = eUDLe u .

Since e-ue u = eO = e, we find e- u = (e u)-l. By Proposition 13.2,
Example 13.1. Suppose that all assumptions of Proposition 13.3 are satisfied, F
•
is an initial operator for D corresponding to an R E RD and Z E ker D n CD(X).
We have De z = e Dz = eO = e and De Rz = e DRz = e Z • Hence
By an easy induction we get

n
e Rnz = (2: Rk)ez, F(e Rnz ) =e Z for Z E ker D, n E N.
k=O
o
194 Chapter 13
Definition 13.1. Suppose that X is a D-algebra with unit e Edam D, A E L(X)

and dam A J dam D. We write
n
(13.6) r:(u) = II Aiu for u E dam A, n EN.
i=1
o
Proposition 13.4. Suppose that Condition [M] holds and (L,E) E G[O]. Then
(i) Dn Ex = (Ex)r~(x) for all x Edam Dn n dam 0- 1 , n E N;
(ii) Ex E Doo for x E Doo n dam 0- 1 .
Proof. (i) by induction. For n = 1, by definition, DEx = (Ex)Dx = (Ex)rp(x).
Observe that r~(x) is well-defined for x Edam Dn. Suppose (i) to be true for an
arbitrarily fixed n E N. Let x E dom D n +1 n dom 0- 1 . Then
D n+1 Ex = D[(Ex)r~(x)]
n
= (DEx)[Dr*,(x)] = (Ex)(Dx)D II Dix
i=1
n n+1
= (Ex) (Dx) II Di+ 1 x = (Ex) II Dkx
j=1 k=1
= (Ex)r*'+1(x).
If x E Doo n dam 0- 1 then (i) implies (ii).

Proposition 13.5. Suppose that Condition [M] holds, (L, E) E G[O], F E
•
M L(X) is an initial operator for D corresponding to an R E RD and x E Doo n
dom 0- 1 . Then Ex E QR(D) if and only if FEx = O.
Proof. Since F E ML(X) and FEx = 0, by Proposition 13.4(i), we get
FDnEx = F[(Ex)r*,(x)] = (FEx)[Fr*,(x)] = 0 for n E N.
•
and X is a complete linear metric space over R If FEx = 0 then Ex AR(D). rt
Proof. It is an immediate consequence of Proposition 13.5, since AR(D)nQR(D) =
{O} (d. PR[9]) and Ex i= o. •
Proposition 13.6. Suppose that Condition [M] holds, (L, E) E G[O], F E
ML(X) is an initial operator for D corresponding to an R E RD and u E Doo n
dom O. Suppose that FLu = Fu = O. Then Lu E QR(D) if and only if u E
QR(D).
Proof. Let u E Doondom r! and let x = Lu. Then u = Ex. Since D,F E ML(X)
and Du = uDLu, we get for all n E N
FDnu = F[(Du)n)
(13.7)
= (FDu)n = [F(uDLu)t
= (Fu)n(FDLu)n = (Fu)n F(DLu)n
= (Fu)nFDnLu.
If Lu E QR(D) then FDnLu = 0 for all n E N. Hence FDnu = 0 for all n E N,

which implies u E QR(D). Conversely, if u E QR(D) then, by (13.7),
Since, by our assumption, ker D has no zero divisors, we find F Dn Lu = 0 for all
n E N, which implies Lu E QR(D). •
Proposition 13.7. Suppose that X is a complete linear metric space over JR,
Condition [M] holds, F, Fl E M L(X) are initial operators for D corresponding to
R,R1 E RD, Fl f:- F, 9 = Re E leX) and (L,E) E G[r!). If
(13.8)
then
(i) Lg EARl (D);
(ii) g-1 EARl (D), provided that Rl is continuous.
Proof. (i) Observe that Dgn = Dg-n = e for n E N. Indeed, for n E N we have
Dg n = (Dg)n = en = e and e = De = D(gng-n) = (Dgn)(Dg-n) = Dg-n.
Hence for n ~ 2 we have Dn Lg = Dn-l g l = e. We therefore obtain
and
Fte =e- R 1D 1e =e - RIC; F 1g- 1 = g-1 - R 1D 1g- 1 = g-1 - R 1e;
R 1F1g- 1 = Rlg- 1 - Rie and R~ FIe = R~e - R~+1e for n ~ 2.

Hence for all n E N
n=1
= R1 FIg- I + RlFle + ... + Rf' FIe
= RIg- 1 - Rie + Rie - Rre + ... + Rf' e - Rf'+le
= RIg- 1 - Ri"+1e.
196 Chapter 13
Condition (13.8) implies that the series
is convergent and
00
Lg = 2: R~FIDn Lg = FILg + RIg- I = FIg- I + RIg- I E ARt (D).

n=O
(ii) follows from (i) in a similar manner, as (iv) from (iii) in the proof of Proposition
13.3. •
Example 13.2. Let X, e, D, R be defined as in Example 5.3. Then D E M L(X)
and Condition [M] holds. Let U = {un}. If Un =J 0 for n E N then L{ Un} = {x n },
where
Xn = ~ for n ~ 2 and Xl E IF \ {O} is arbitrary,
Un-I
is a logarithm of u. Observe that all Xn =J O. It is easy to verify that E{x n } =
{Un} = UI bn(x)}, where
for n = 1;
otherwise;
and UI E IF \ {O} is arbitrary, is the corresponding antilogarithm. If Xn = >. =J 0

for all n E N then X = >.e, 'Yn(>.e) = >.n-l, Ex = UI {>.n-I}. 0
In general, if X is not an algebra, we cannot define left and right logarithmic map-
pings. However, Theorem 11.2 gives a possibility of an isomorphic transformation
of the problem in question to a more" classical" situation of spaces of functions of
a real variable with values in complete linear metric locally convex spaces. If it is
the case, then we have defined logarithmic and exponential function in the usual
way. Another possibility is the following (cf. PR[9j).
Definition 13.2. Suppose that X is a complete linear metric space, D E R(X),
F is an initial operator for D corresponding to an R E 'RD. Then
(13.9)
is said to be the space of coefficients of D-analytic elements.

Define the forward and backward shifts in the space XO:
(13.10)
where we assume Zn-k = 0 if n ~ k, FOxo = 8Fx = {8 In }Fx, 8 = {8In } (cf.

Example 5.3). 0
Since the space AR(D) is D-R invariant, i.e.
the operators DO, RO, FO are well-defined on XO and map XO into itself.
We assume that the topology in XO is determined by the coordinatewise conver-
gence induced by the topology in X. It is not difficult to verify that DO E R(XO),
FO is an initial operator for DO corresponding to Volterra right inverse RO of DO
and ARo(DO) = Xo.
The following question arises: when does a sequence {zn} C ker D determine a D-
analytic element, i.e. when does there exist an x E AR(D) such that Zn = FDn x
for all n E No?
The answer is given by
Theorem 13.6. Suppose that X is a complete linear metric space, F is continuous
initial operator for a closed D E R(X) corresponding to a continuous R E 'RD.
Write
00
(13.11)
L..J RnZn+k+l is convergent}.
V '"'
ZD ,R = {{Zn} C ker D: kENo
n=O
Then for every {Zn} E ZD,R there is a unique x E AR(D) such that
(13.12)
i.e. ZD,R is the space of coefficients of D-analytic elements belonging to AR(D).

Moreover, ZD,R = ARo (D), where DO, RO are defined by (13.1O).
Proof. Let {zn} E ZD,R. By our assumptions, the series
is convergent. Put
=L
00
x Rnzn+ 1 •
n=O
Then for all kEN we find
L
00
Dkx = Dk Rnzn+l
n=O
and
00 k-l 00 00
"L..J R m Zm+k+l = "L..J D k - n zn+l + "L..J R n - k zn+l - "DkR

L..J n Zn+l·
n=O n=O n=k n=O
198 Chapter 13
Since D is closed, we conclude that
=L
00
(13.13) Dkx R mZm+k+1 (k E No)

m=O
and this last series is convergent by our assumption. Thus
=L
00
FDkX FR mzm+k+1 = FZk+1 = Zk+1 far kENo,

m=O
since FR = O. Moreover, for all n E N we have
= Rn L L =L
00 00 00
RnDnx R mzn+m+1 = Rm+nZm+n+1 Rkzk+1'

m=O m=O k=n
Hence Rn Dnx -t 00 as n -t 00. This implies that x E AR(D).

Suppose now that there is another y E AR(D) such that FDny = Zn+1 for all
n E No. Then FDn(x - y) = 0 for all n E No, i.e. x - y E QR(D). Since
x - y E AR(D) n QR(D) = {O}, we conclude that x = y, i.e. the element x is
uniquely determined.
By our definition, if XO = {zn} E ZD,R, then XO E ARo(DO) = XO, where XO, DO,
no are defined by (13.9), (13.10).
On the other hand, if XO = {zn} E ARo(DO) = XO then, by definition of XO, there
is an x E AR(D) such that FDnx = Zn for all n E No. All series (13.13) are
convergent. We therefore conclude that XO E ZD,R. This finishes the proof. •
Example 13.3. Suppose that all assumptions of Theorem 13.6 are satisfied and
XO, DO, RO are defined by (13.9), (13.10). Define for a continuous R E 'RD
(13.14)
where F is an initial operator for D corresponding to R.

Then q, maps AR(D) into ARo (DO). Even more, q, is an isomorphism, for q,(X) =
XO. The inverse mapping q,-1 maps ARo(DO) onto AR(D), since ARo(DO) = ZD,R
by Theorem 13.6. Moreover, we have
i.e. q, is a D-R structure preserving isomorphism (d. VON TROTHA T[l]). Sup-
pose, moreover, that dim ker D = 1. It means that ker D = lin {d} for an ar-
bitrarily fixed d E ker D \ {O}. We thus may identify ker D with the field IF of
scalars by the mapping: Z = ad --+ a, where a E IF, Z E ker D. SO that we
can introduce in XO = ARo(D) the coordinatewise multiplication of elements and
multiplication by scalars. Clearly, with these operations, XO is a commutative

algebra and DO E M(XO). As in Example 13.2,
Zn
where Vn = - - for n 2': 2, Vl E IF \ {O} is arbitrary,
Zn-l
wh,,,e Wn = W,1n(Z;) = {ll, z,

for n = 1;
otherwise;
and Wl ElF \ {O} is arbitrary,

determine logarithms and antilogarithms induced by DO in xo. o
x"(t) = Ftx for x E X (cf. Formula (11.7)), i.e. x" : A(JR) ~ ker D. Suppose,
moreover, that dim ker D = 1. Similarly, as in Example 13.3, we may identify
ker D with the field IF of scalars. Our assumption on ker D implies that there
exists a scalar function ax : A(JR) ~ IF such that
(13.15) x"(t) = ax(t)d for all t E A(JR) (x EX).
The function ax is said to be a symbol function of the element x (cf. PR[lO]). Our
assumptions imply that
d
aDx(t) = dtax(t) for x E dom D,
aRx(t) = lot ax(s)ds, aFx(t), ashX(t) = ax(t + h)
for x E X, t, h E A(JR).
Suppose that X is a commutative algebra and the symbol function is multiplica-
tive, i.e. axy = axay for x, y EX. It is so, for instance, when F and Sh
(h E A(JR) are all multiplicative. Then the basic equation can be rewritten as
follows: a~ = aua~ for u,x n dom D, u E leX). Thus aLu = ax = In au + Inc,
c E IF is arbitrary.
Note that the notion of a symbol has been introduced firstly by S. G. MICHLIN in
1948 for multidimensional singular integral equations. 0
Example 13.5. Suppose that X E Lg(D)) is a complete linear metric space over
C, (L, E) E G[n], D E R(X), there is an R E RD n AQN(ker D), F is an initial
operator for D corresponding to R, 9 = Re, >"9 E dom n for>.. E vcR and {ShhEIR
is a family of true shifts induced by R (cf. Definition 11.2). Write
n m
(13.16) WeD) =L LakjDkS_hj,
k=O j=O
200 Chapter 13
0= ho < hl < ... < h m , akj E C (k = 0, ... ,nj j = 0, ... ,m),

n m
(13.17) W"(t) =L L akjtke- hjt .
k=O j=O
By Corollary 13.2, ShE(>.g) = eM E(>.g) for>. E vcR, hE JR. Thus we have
(13.18) W(D)E(>.g) = W"(>')E(>.g) for >. E vcR.
(13.19) E(>.g) E ker W(D) if and only if W"(>.) = o.

It means that the so-called quasipolynomial W"(t) plays an essential role when
we are looking for solutions of the equation
(13.20) W(D)x = o.
These >"s, which satisfy the equation W" (>.) = 0, are said to be characteristic roots
of the quasipolynomial W"(t). Observe that, in general, the quasipolynomial has
an infinite number of roots, Le. Equation (13.20) has infinite number of solutions
being antilogarithms.
On the other hand, if X is a complete linear metric locally convex space then, by
Theorem 11.2, we get
[W(D)E(>.g)]"(t) = W"(t)eAt z , where z = FE(>.g).

This means that to Equation (13.20) there corresponds in a unique way a
differential-difference equation of the form
n m
(13.21) L LakjX(k)(t-hj) =0.
k=O j=O
Again W"(t) is a quasipolynomial for Equation (13.21). Hence, by Theorem 11.2,

equations (13.20) and (13.21) have similar properties. We can study in this way
properties of solutions to Equation (13.21), in particular, their stability, in order
to have same informations about solutions of Equation (13.20) (cf. PR[7], PR[8]).
o
As a matter of fact, all results obtained in this section do not exhaust the subject.
They are examples only which are intended show how to examine the properties
of smooth logarithm and antilogarithms. In order to obtain more general results
we have to know the properties of non-Leibniz components in any particular case.
CHAPTER 14
REIMANN-HILBERT TYPE PROBLEMS IN LEIBNIZ ALGEBRAS
The following question has been posed by W. Wendland in 1995: Is it possible to

solve a problem of the Riemann-Hilbert type in algebras with logarithms? We
shall show that the answer is positive in Leibniz algebras.
Example 14.1. The classical Riemann-Hilbert problem can be stated, for in-
stance, in the following way.
Let nee be a domain with the boundary on = r which is an oriented closed
regular arc, i.e r = {z = z(t) : 0: :::; t :::; {3, z(o:) = z({3)}, where the func-
tion z E C l (0:, {3), is one-to-one, z' (t) =I- 0 for t E (0:, {3) and limt-H.+o z' (t) =
limt-+t3-o z'(t) =I- O. We have to find a function cI> piecewise analytic in the do-
mains n+ = nand n- = c \ n, bounded at infinity and such that its boundary
values cI>+ and cI>- satisfy the following condition:
(*) cI>+(t) = G(t)cI>-(t) + g(t) for t E f,
where functions g, G are given. The solution of the problem is well-known (cf. for
instance, MICHLIN Mi[l], POGORZELSKI P[I], MEISTER Me[I], WEGERT Weg[I],
WENDLAND Wen[1 l). In order to solve (*) we have to use properties of logarithmic
and exponential functions (cf. ANOSOV and BOLIBRUCH AB[Il) and of a singular
integral operator 8 defined by the Cauchy principal value of an integral, namely,
(**)
(8<p)(t) = ~ lim ( <p(7) d7 ~f ~ ( <p(7) d7 (t E r)
o-+olr\{zEc : Iz-tl<e} 7 - t
7rt 7rZ lr 7 - t
In the case, when g, G belong to the space HJ.L(f) of functions satisfying the Holder
condition on f with an exponent J.L, 0 < J.L < 1:
HJ.L(f) = {x E C(r) : Ix(t) - x(s)1 S Cit - slJ.L} (C E 114),
the singular integral operator 8 defined by (**) is an involution in the space
X = HJ.L(f): 8 2 = I on X. Thus there are disjoint projectors p+ and P- giving
the partition of unit and such that cI>+ = P+x, cI>- = P-x for an x E X (cf. PR[3] ,
also PR[6]). This x is to be found if we apply logarithms to the homogeneous
problem (*) (i.e. with 9 = 0). Then the non-homogeneous problem is solved by
a use of the already found solution to the homogeneous problem.
This problem can be formulated and solved in the same manner if f is a finite set
of closed regular arcs with a properly defined orientation of that system. 0
This is the way which we will follow in order to find solutions to a generalized
Riemann-Hilbert problem. Namely, we shall consider algebras with logarithms
induced by a linear operator D and with an involution 8. In commutative alge-
bras solutions will be obtained in a similar manner as in the classical case. In
noncommutative algebras we shall need some additional assumptions and some
modifications of considerations used in the commutative case. Recall
201
202 Chapter 14
Theorem 14.1. (cf. PR[3J; also properties of algebraic operators in Chapter 8).
Let X be a linear space over an algebraically closed field IF with an involution
S E Lo(X) (i.e. S2 = I, S f. 1). Then p+ = 1(1 + S) and P- = ~(I - S) are
disjoint projectors giving partition of unit: (pf)Z = p±; p+ p- = p- p+ = 0
and p+ + p- = I. Moreover, p+ - p- = Sand Sp± = p±S = ±P±. So that,
if we write X± = p±X, x± = P±x for x EX, then we have
This implies
Corollary 14.1. Let all assumptions of Theorem 14.1 be satisfied. If u E X,
u± = P±u and
(14.1) where v E X,
then u = Sv. In particular, if v = 0 then u = O.

Proof. Let (14.1) hold. Since Su = P+u - P-u = u+ - u- = v, we find u =
S2U = Sv. •
Lemma 14.1. Let D E A(X). Suppose that X E Lg(D) is a Leibniz algebra with
unit e and with an involution S E Lo(X) and (L, E) E G[OJ. Then the following
conditions are equivalent:
(i) Lu± C X± for all u Edam 0;
(ii) Ev± C X± for all v Edam 0- 1 ;
(iii) p±Lu± = ±Lu± for all u Edam 0;
(iv) P±Ev± = ±Ev± for all v Edam 0- 1 .
Proof. (i) -* (ii). Suppose that u± E X± are arbitrarily fixed and (i) holds, i.e.
there are v± E X± such that v± = Lu±. Then u± = ELu± = Ev± E X±.
(ii) -* (i). Suppose that v± E X± are arbitrarily fixed and (ii) holds, i.e. there
are u± E X± such that u± = Ev±. Then v± = LEv± = Lu± E X±.
(iii) -* (i). Suppose that u± E X± is arbitrarily fixed and (iii) holds. Then
Lu± E X±. The arbitrariness of u± implies (i). A similar proof for the implication
(iv) -* (i).
(i) -* (iii). Let LX± E X±. Let u E X be arbitrarily fixed. By our assumptions,
P+ Lu± = Lu± and P- Lu± = -Lu±, i.e. (iii) holds. A similar proof for the
implication (i) -* (iv).
Since we already have proved (iv) ++ (i), (i) ++ (ii) and (i) ++ (iii), we conclude
that (ii) ++ (iv) and (iii) ++ (ii). •
We admit the following condition:
Riemann-Hilbert problems 203
(A) Let IF be algebraically closed. Let D E A(X). Suppose that X is a Leibniz

algebra with unit e and with an involution S E Lo(X) such that X E Lg(D),
(L, E) E G[O) and
(14.2) L(X+ n dom 0) c X+, L(X- n dom 0) c X-.
In particular, (14.2) implies that X± n dom 0 c dom 0- 1 .

Lemma 14.2. Suppose that Condition (A) is satisfied. Then
OJ uv E X± whenever u, v E X± n dom 0;
Oi) (U±)-1 E I(X±) whenever u± E I(X±).
Proof (i) Let u,v E X± ndom O. By Lemma 14.1(i), Lu,Lv E X±. Since X is
a Leibniz algebra, we find L(uv) = Lu + Lv E X±. By Lemma 14.1(ii), we get
uv = EL(uv) E X±.
(ii) Let u± E X± be invertible. Since X is a Leibniz algebra, we conclude that
u± E dom 0 and L(u±)-1 = -Lu± E X±. Hence by Lemma 14.1(ii), (u±)-1 =
EL(u±)-1 E X±. •
Homogeneous Riemann-Hilbert Problem. Suppose that Condition (A) is satisfied.

Find an Xo E dom 0 such that
(14.3) xci = ax o , where a E dom 0 is given.

Theorem 14.2. Suppose that Condition (A) is satisfied. If a i ±e then the
homogeneous Riemann-Hilbert problem (14.3) has a solution
(14.4) Xo = E(P+ La) + E( -P- La) and x~ = E(±P±La) E J(X).

If a = ±e then the only solution of (14.3) is Xo = O.
Proof If a = e then the solution of (14.3) is given by Corollary 14.1. If a = -e
then we have Xo = xci + Xo
= -xi) + xi) = O. Suppose then that a i ±e.
Since X is a Leibniz algebra, the logarithmic mapping L is of the exponential
type. This, and Condition (14.2) together imply that Lxci = L(axo ) = La+Lxo .
Hence Lxci - Lxo = La. By our assumption (14.2), there is ayE X such that
y± = Lx~. Thus
Sy = y+ - y- = Lxci - Lxo = La,
which implies y = SLa. By Theorem 14.1,
Xo = Xo+ + Xo-
= ELxci + ELxci = Ey+ + Ey-
= E(P+y) + E(P-y) = E(P+ SLa) + E(P- SLa)
= E(P+ La) + E( -P- La).
204 Chapter 14
Since X is a Leibniz algebra, we find E( -P- La)E(P- La) = E(P- La-P- La) =
E(O) = e. Thus, by Lemma 14.1, we find xi) = E( -P- La) = [E(P- La)]-l.
Hence xi) E J(X). Similarly, xt = E(P+ La) = [E( -p+ La)]-l. Thus xci E J(X) .
Clearly, the solution (14.4) is dependent on the choice of selectors L.

•
Corollary 14.2. If all assumptions of Theorem 14.2 are satisfied then the solution
to the problem (14.3) can be written in the form Xo = (a + e)E( -P- La).
Proof. Since X is a Leibniz algebra with unit e and E(O) = e, Formula (14.4)
implies that
xoE(P- La)
= E(P+ La)E(P- La) + E(P- La)E( -P- La)
= E[(P+ + P-)La] + E[P-(La - La)]
= ELa + EP-O = a + E(O) = a + e.
Since E(P- La) E J(X) and [E(P- La)]-l = E( -P- La), we get the required
formula. •
a E J2 (dom n). Then the solution to the problem (14.3) can be written in the
form: Xo = (a~ + a-~)E(SLa~).
Proof. Since X is a Leibniz algebra, our assumptions and Corollary 14.2 together
imply that
Xo = (a + e)E( -P- La)

1 1 1
= (a + e)E[-2(J - S)La)] = (a + e)E(-2La + 2SLa)
= (a + e)(ELa-!)E(~SLa) = (a + e)a-!E(SLa!)
=(a! + a-!)E(SLa!).
Nonhomogeneous Riemann-Hilbert Problem. Suppose that Condition (A) is satis-

•
fied. Find an x E dom n such that
(14.5) x+ = ax- + b, where a E J(X} n dom n, bE X are given.
Theorem 14.3. Suppose that Condition (A) is satisfied. If a # ±e and Xo is a

solution of the homogeneous Riemann-Hilbert problem (14.3) then the nonhomo-
geneous Riemann-Hilbert problem (14.5) has a solution
(14.6)
1
x = Xo + 2(xoSYo + YoSxo}, where Yo = a-1bE(P-La) = a-1b(P-xo)-1.
If a = e then a solution to (14.5) is x = Sb. If a = -e then a solution to (14.5) is

x =b.
Proof. If a = e then the solution of (14.5) is given by Corollary 14.1. If a = -e
then x = x+ + X- = -x- + b + X- = b. Suppose then that a t- ±e. Since Xo is a
solution of the problem (14.3), Theorem 14.2 implies that x~ E J(X). Thus, again
by (14.3), a = xt(XO)-l. Suppose that Xl is a solution of (14.5). By Lemma 14.2,
we can write y± = (x~)-lxt. Let y = y+ + y-. Then
y+ = (+)-1
Xo Xl+ = (-)-1
Xo a -l( aX l- + b) = (-)-1
Xo Xl- + (Xo-)-1 a -lb = y - + Yo,
where
i.e.
We also have
Thus
Xl =
+ + Xl-
Xl
= y+xt + y-xo = xt(y+ + y-) - xty- + xoy-

= xty - (xci - Xo)y- = xt Syo - (Sxo)Y-
= xci Syo + (Sxo)Yo = i(xo + Sxo)Syo + (Sxo)i(yO - SYo)
1
= 2[xoSyO + (Sxo)SyO + (Sxo)yo - (Sxo)SYoJ
1
= 2[xoSyo + (Sxo)YoJ.
Clearly, X = Xo + Xl is also a solution of (14.5). •
Corollary 14.4. Suppose that Condition (A) is satisfied. If a t- ±e and Xl, x2
are two solutions to the non-homogeneous Riemann-Hilbert problem (14.5) then
their difference x = Xl - X2 is a solution to the homogeneous problem (14.3). If
a = e then the problem (14.5) has a unique solution X = Sb. If a = -e then (14.5)
has a unique solution X = b.
Proof. Let a t- ±e. By our assumption, xi = ax; + b, (i = 1,2). Then x+ =
xt - xt = aX l + b - (ax2" + b) = a(x 1 - x2") = ax-. If a = e then the conclusion
is given by Corollary 14.1. If a = -e then the conclusion follows immediately from
the corresponding part of the proof of Theorem 14.2. •
We should point out that an involution under consideration cannot be multiplica-
tive if Condition (A) is assumed. Namely, we have
Proposition 14.1. Suppose that X is a commutative algebra over a field IF with
a multiplicative involution S E Lo(X, i.e. S(xy) = (Sx)(Sx) for x,y E X. Then
x±y± E X+, x+y- ,x-y+ E X- for all x,y E X.
206 Chapter 14
Proof. Let x, y E X be arbitrarily fixed. Let, as before, x± = p±x, y± = P±y.

By our assumptions, we have
S(x±y±) = (Sx±)(Sy±) = (±x±)(±y±)= x±y±,
S(x+y-) = (Sx+)(Sy-) = x+ (-y-) = -x+y-,

S(x-y+) = (Sx-)(Sy+) = -x-y+ = -x-y+.
Corollary 14.5. Suppose that Condition (A) is satisfied. Then the involution S
•
is not multiplicative.
n
Proof. Suppose that the involution S is multiplicative.Let x, y E X dom ben
arbitrarily fixed. By Lemma 14.2, x-y- E X-. By Proposition 14.1, x-y- E X+.
Since X = X+ EBX-, we have X+ nX- = {OJ. Thus x-y- = O. The arbitrariness
of x and y implies x- = y- = 0 and X- = {OJ. Thus S = I, which contradicts
our assumption that S is an involution, i.e. S2 = I, however S :I I. Hence S is
not multiplicative. •
Corollary 14.5 is important, since there are several examples of functional-
differential equations with a multiplicative involution which can be also solved by
means of algebraic methods (cf. PR[3], PR[4], PR[5], PR[6], also next chapters).
In order to study the noncommutative case, instead of Condition (A) we shall
admit either the condition
(A)I Let IF be algebraically closed. Let D E A(X). Suppose that X is a Leibniz
algebra with unit e and with an involution S E Lo(X) such that X E Lg1(D),
uv E X± whenever u,v E X± and DS = SD on dom D, (LI,EI ) E G[nd and
(14.7)
or the condition
(A)r Let IF be algebraically closed. Let D E A(X). Suppose that X is a Leibniz
algebra with unit e and with an involution S E Lo(X) such that X E Lgr(D),
uv E X± whenever u,v E X± and DS = SD on dom D, (Lr,Er ) E G[nr] and
(14.8)
In particular, (14.7) implies that X± C dom nl, (14.8) implies that X± C dom nr •
Note 14.1. Let either Condition (A)I or Condition (A)r be satisfied. The assumed
condition that uv E X± whenever u, v E X± in the commutative case is proved
by Lemma 14.2. It is so, because in that case logarithms are of the exponential
type. Without this property, we cannot prove a lemma corresponding to Lemma
14.2.
We should point out also that here, in the noncommutative algebras, we shall
need essentially the property that D and S commute each with another. Thus
Condition (A) is not a particular case of Conditions (A)l and (A)r. 0
In the same manner as Lemmas 14.1, 14.2, we prove the following

Lemma 14.3. Let D E A(X). Suppose that X is a Leibniz algebra with unit e
and with an involution S E Lo(X) such that X E LgJ(D) and (L/, Ez) E G[Oz]
(X E Lgr(D) and (Lr' Er) E G[Or], respectively). Then the following conditions
are equivalent for all UI Edam 0 1, Ur E dam Or, VI E dam Oil, Vr Edam 0;:-1:
(i) Llu± E X± (Lru± E X±, respectively);
(ii) Elv± E X± (Erv± E X±, respectively);
(iii) p±Llu± = ±Llu± (P± Lru± = ±Lru±, respectively);
(iv) P±Elv± = ±Elv± (P±Erv± = ±Erv±, respectively).
Lemma 14.4. Suppose that either Condition (A)I or Condition (A)r is satisfied.
Then
(i) uv E X± whenever u, v E X± n dom 0 1 (X± n dom Or, respectively;
(ii) (U±)-l E I(X±) n dam 0 1 n dam Or whenever U E I(X±) n dom 0 1 n domO r .
In order to prove Lemma 14.4(ii), apply Theorem 7.2.
Homogeneous Riemann-Hilbert problem. Let either Condition (A)I or Condition
(A)r be satisfied. Let a E leX) n dam 0 1 (a E leX) n dom Or, respectively). Find
an Xo E X such that
(14.9) Xo+ = axo·

-
Theorem 14.4. Suppose that either Condition (A)I or Condition (A)r holds and
D E R(X). Then the Riemann-Hilbert problem (14.9) has a solution if and only
if there a U such that
(14.10) u+a - au- = Da (au+ - u-a = Da, respectively).
If it is the case then the solution, we are looking for, is
(14.11)
(xo = ErP+(Ru + z) + ErP-(Ru + z), respectively.)

Moreover, x~ = EIP±(Ru+z) E leX) (x~ = EIP±(Ru+z) E leX), respectively).
Proof Since X is a Leibniz algebra, we have
(14.12) DLl(UV) = u(DLlV)U- 1 + DLlU for u,v E leX) n dom 0 1,
(14.13)
(cf. Proposition 7.1').

Suppose that a E leX) n dom Ol. Sufficiency is proved by checking.
208 Chapter 14
Necessity. By (14.12), we get DL/x6 = DL/(axo ) = a(DL/xo)a- l + DL/a.

By Lemma 14.3, there is a y such that L/x~ = y±. This implies that Dy+ =
a(Dy-)a- l + DL/a. Thus (Dy+)a - a(Dy-) = (DL/a)a = Da. Since DS = SD
on dom D, we have DX± n dom D C X±. Indeed, if x E dom D, then
1 1
Dx± = DP±x = D-(l ± S)x = -(1 ± S)Dx = p±Dx E X±.
2 2
Thus there is a u such that Dy± = u±, which implies u+a - au- = Da, i.e.
Condition (14.10). Suppose now that u is already determined and D E R(X).
Let z E ker D. Since DSz = SDz = 0, we conclude that also Sz E ker D. Thus
we find y± = Ru± + z±, where z± E X± n ker D. Finally, since u = u+ + u-,
z = z+ + z-, we get y = y+ + y- = Ru + z and
Similarly, if a E l(X)ndom nr then, by (14.11), DLrX6 = DLra+a-l(DLrxo)a,

which yields to the second of Conditions (14.10). If this condition is satisfied by a
u, then a solution of the problem (14.9) is given by the second of Formulae (14.11).
Since X is a Leibniz algebra with unit, we have
(14.14) L/w = -Lrw- l for all w E leX) n dom D,
(14.15) E/( -y) = (Ery)-l whenever y E dom n;l, -y E dom nil

This implies that x~ E leX) in both cases (cf. also Lemma 14.4).
Theorem 14.5. Suppose that either Condition (A)/ or Condition (A)r holds. If
•
(14.16) (a- l X+a C X+, respectively)
then the Riemann-Hilbert problem (14.9) has a solution of the form (14.11), where
(14.17) u = DL/P+a - a-l(DL/P-a)a;
(u = a(DLrP+a)a- l - DLrp-la, respectively)
Proof. Let Condition (A)l be satisfied and let a E leX) n dom 0 1• Let (14.16)
hold, i.e. aX- a-I C X-. Let v- = au- a-I. Then (14.10) can be rewritten
in the following manner: u+ - v- = u+ - au-a- 1 = (Da)a- 1 = DLla. By our
assumptions,
u+ = P+DL/a = DP+Lla = DL/P+a;
v- = -P- DLa = -DP- L/a = -DL/P-a.
Thus au-a- 1 = v- = -DP- L,a = -DL/P-a, so that u- = -a-1(DL/P-a)a
and u = u+ + u- = DL1P+a - a-1(DL/P-a)a. A similar proof if Condition
(A)r is satisfied, a E leX) n dom Or and the second of Formulae (14.17) holds.
Indeed, writing v+ = a-Iu+a, we get v+ - u- = a-Iu+a - u- = a-I Da = DLra.

Hence u- = -DLrP-a, v+ = DLrP + a, and u+ = a(DLrP+a)a- l . This yields
u = a(DLrP+a)a- 1 - DLrp-Ia. Having already determined u, we can find
solutions to the problem (14.9). •
An immediate consequence of Formula (14.14) is
Corollary 14.6. Suppose that all assumptions of Theorem 14.5 are satisfied.
Then P-a E leX) n dom nr (p-a E leX) n dom nl, respectively) and
u = DLIP+a + a-I[DLr(P-a)-lja
(u = a(DLIP+a)a- 1 + DLr(p-a)-I, respectively).
Nonhomogeneous Riemann-Hilbert problem. Let either Condition (A)I or Condition

(A)r be satisfied. Let a E I(X)ndom nl, (a E I(X)ndom nr, respectively). Find
an x E X such that
(14.18) x+ = ax- + b, where b E X.
Theorem 14.6. Suppose that all assumptions of Theorem 14.5 are satisfied,
a:l ±e and Xo is a solution of the homogeneous problem (14.9). Then the nonho-
mogeneous Riemann-Hilbert problem (14.18) has a solution of the form
(14.19)
If either a = e or a = -e then (14.18) has a unique solution, namely x = Sb or

x = b, respectively.
The proof is the same as the proof of Theorem 14.3, if we take into account that
here the algebra under consideration is not commutative, so that Yo = (xo)-la-Ib,
since we cannot change the order of terms in products.
In a similar manner as in the commutative case we prove
Corollary 14.7. Suppose that either Condition (A)I or Condition (A)r is satis-
fied. If a :I ±e and Xl, X2 are two solutions to the non-homogeneous Riemann-
Hilbert problem (14.18) then their difference x = Xl - X2 is a solution to the
homogeneous problem (14.3). If a = e then the problem (14.18) has a unique
solution x = Sb (cf. Corollary 14.4). If a = -e then (14.18) has a unique solution
x =b.
Similarly, as in the commutative case, we get
Corollary 14.8. Suppose that either Condition (A), or Condition (A)r is satis-
fied. Then the involution S is not multiplicative (cf. Corollary 14.5).
The already obtained solutions to the Riemann-Hilbert problem can be used in
order to solve linear equations with involutions in Leibniz algebras with logarithms.
Namely, we have
210 Chapter 14
Example 14.2. Suppose that either Condition (A) or Condition (A}r or Condi-
tion (A}l is satisfied. Consider the equation:
(14.20) (aI + bS)x = y, where a, b, y E X.
Since aI + bS = a(P+ + P-) + b(P+ - P-) = (a + b)P+ + (a - b)P-, Equation

(14.20) can be rewritten as follows
(14.21) (a + b)x+ + (a - b)x- = y.
If a = ±b and b E I(X) then solutions to this equations exist if and only if

(a ± b)-ly E X±. If it is the case, then x± = (a ± b)-ly, x'f E X'f is arbitrary.
Suppose now that a+b, a-b E I(X) (I(X)ndom nr , I(X)ndom nl, respectively).
Then Equation (14.21) can be written as
i.e. we have a Riemann-Hilbert problem x+ = ax- +y with a = -(a+b)-l(a-b)

and y = (a + b)-ly. 0
Solutions to Equation (14.20) have been obtained earlier in another way under
other assumptions, for instance, that a, b commute (anticommute) with the invo-
lution S (d. PR[3]-PR[5]).
Theorem 14.7. Suppose that X E Lg(D), (L,E) E G[n], F is an initial operator
for D corresponding to an R E R D , a =I- ±e and xo, x~ are solutions of the homoge-
neous Riemann-Hilbert problem (14.9) induced by (L, E) and its m-normalization
(L', E') by R for an mEN, i.e. by Formulae (2.4),
xo = E(p+ La) + E( -P- La), x~ = E'(P+ L'a) + E'( -P- L'a).

Write
m-l
V =- L RiFDiLa.
i=O
Then
(14.22)
or, in equivalent forms
(14.23) x~ = WXo + (Sw)Sxo,
(14.24)
·
Proo.f B your assump t lOns, Xo+ = axo-,+
,xo = ax o,- . L et y ± = L' Xo± an d Iet
y = y+ + y-. In the same manner as in the proof of Theorem 2.2, we find
o
Sy = y+ - y- = L'xt - L'x = L'a. Thus y = SL'a. Clearly, by definition,
v E ker Dm. Since X is a Leibniz algebra, we get
~~± = E'y±
= E'(P±SL'a) = E'(±P±L'a) = E'[±P±(La+ v)]
= E'(±P±La± P±v) = E'(±p±v)E(±P±La)= E'(±v±)x~.
Having already proved Formula (14.22) and writing w = ME'(P+v) + E'(P-v)],

we find w± = tE'(v±) and
x~ = E'(v+)xt + E'(v-)xo
=E'(v+)~(xo + SXo) + E'(v-)~(xo - SXo)
=w+ (xo + Sxo) + w- (xo - SXo)
1
= "2[(w + Sw)(xo + SXo) + (w - Sw)(xo - Sxo)]
=wXo + (Sw)Sxo.
This proves Formula (14.23). We find
x~ = (a + e)E'(-P- La) = E'(-v-)x o= E'(-P-v)P-xo,

i.e. Xo is of the form (14.23).
•
(L, E) is already m-normalized by R, i.e. L' = L, E' = E. Then the homogeneous
Riemann-Hilbert (14.9) problem has a unique solution determined by Formula
(14.4).
Proof. By our assumption, v = O. Thus v± = O. Since X is a Leibniz algebra, we
have E(O) = e. Hence Xo = Xo. •
Definition 14.1. The greatest number of linearly independent solutions to the
homogeneous Riemann-Hilbert problem (14.9) is said to be its index and is denoted
by "'a. 0
Observe that if all conditions of Corollary 14.9 are satisfied, then the index is:
"'a = 1. If a = ±e then Xo = 0, so that "'e
= O. Clearly, without any additional
conditions imposed a priori on solutions, it may happen that "'a
= 00.
Antilogarithms E(v±) in Theorem 14.7 correspond in the classical case to entire
functions of the form eV ± , where v, hence also v±, are polynomials of degree m - 1.
Note 14.2. Another way for calculation of the index is indicated in Appendix
(Theorems A6.6, A6.12). 0
Write
(14.25) yj = E'(vj)x~, where Vi = _Ri FDi La (j = 0,1, ... ,m - 1)

212 Chapter 14
Then
m-l
(14.26) E'(v±) = II E'(vj).
j=O
If (L, E) is m-normalized by R then Yj = Xo (j = 0,1, ... , m - 1). In particular, if
F La = 0 then Yo = Xo. If (L, E) is not p-normalized for any p :s: m, then solutions
Yo, ... , Ym-l are linearly independent.
Proof By our assumptions, F Dj La = 0 for j = 0,1, ... , m - 1. Thus
Yj± -- E'(Vj±) Xo± -- E'(P±FDjLa ) Xo± -- E'(O) Xo± -- exo± -- Xo± , i.e. Yj = Xo·
In order to prove Formula (14.26), observe that X is a Leibniz algebra. Thus
E'(v±) = E'(P±v)
m-l m-l
= E'( - p± L RjFDjLa) = E'(p± L Vj)
j=o j=o
m-l m-l
= E'( L vj) = II E'(vj).

j=O j=O
If (L, E) is not p- normalized for any p :s: m, then v j as D- monomials of degree j,
respectively (j = 0, ... , m - 1) are linearly independent. Hence also antilogarithms
EVj U = 0, ... , m - 1) are linearly independent. •
On the other hand, one can prove that, under assumptions of Corollary 14.10, if
elements Yj = yj + Yj U = 0, ... , m - 1) are linearly independent, then (L, E) is
m-normalized by R.
Theorem 14.8. Suppose that Condition (A) is satisfied, (Li, Ei) E G[n) (i =
0, 1, ... , n) (n E N), a + eEl (X) and Xi are solutions of the homogeneous Riemann-
Hilbert problem (14.9) induced by (L i , E i ), respectively. Then
(14.27)
U:lij i,j=O,I, .... ,n).
Moreover, A ij , aij E ker D U:I ij i,j = 0,1, ... ,n).

Proof In order to prove that aij E kerD U :I ij i,j = O,I, ... ,n), we use the
facts that logarithms and antilogarithms of constants are again constants and that
logarithms are uniquely determined up to a constant, i.e. Lju - LiU = Z E ker D
whenever U E dom n. Clearly, the same arguments imply that Aij = Eiaij E ker D
j :I ij i,j = 0, ... , n. By Corollary 14.2, Xi = (a + e)Ei( -P- Lia) U :I ij i,j =
0,1, ... , n). Since X is a Leibniz algebra, we find
LjXj - LiXi
= Lj[(a + e)Ej ( -P- Lja)]- Ld(a + e)Ei( -P- Lia)]
= Lj(a + e) - Li(a + e) - P-(Lja - Lia)] = aij'
Since logarithms are determined uniquely up to a constant and a logarithm of a

constant is again a constant, there are Zij E ker D such that L j (x j) = Li (x j) -
L t-z-- t J- - L-x-
- L-x
tJ' Thus a-tJ- - t t -
- L(x
t J-) - L-z-
t tJ- - L-(x-)
t t · Hence
i.e. Xj is of the form (14.27).

In other words: solutions to the homogeneous Riemann-Hilbert problem induced
•
by two different logarithmic mappings are uniquely determined up to a constant
dependent only on the given element a and a constant dependent on a and the
choice of selectors of O.
Corollary 14.11. Suppose that all assumptions of Theorem 14.8 are satisfied
and constants Aij , aij (j # i; i,j = 0,1, ... ,n) are defined by Formulae (14.27),
(14.28), respectively. Then
(14.29) aji = -aij; Aij E leX) and Aji = AijI (j # i; i, j = 0,1, ... , n)
(14.30) aij =aOj -aOi; Aij = AOjAo/ (j 'Ii; i,j = 1, ... ,n)
m-I m-I
(14.31) L ai,HI = aOm ; II Ai,i+l = Aom if 1 ::; m ::; n ;
i=O i=O
Prool Let i,j (j # i; i,j = O,I, ... ,n) be fixed. The first of equalities (14.29),
(14.30) follow immediately from the definitions (14.27), (14.28). Then Aji =
Ei(aji) = E i ( -aij) = [E(aij]-I = AijI and
Formulae (14.27) and (14.30) together imply (by an easy induction) Formulae
(14.31). •
Corollary 14.12. Suppose that X E Lg(D), D E R(X), F is an initial operator
for D corresponding to an R E RD, (Lo,Eo) E GR,I[O], (Li,Ei ) E G[O] (i =
1, ... , n) (n EN), a + e E leX) and Xi are solutions of the homogeneous Riemann-
Hilbert problem (14.9) induced by (Li,Ei ) (i = 0,1, ... ,n), respectively. Then
(14.32)
Aj = AOj = Eo(aoj), aOj = FLj(a + e) - P-(Lja - Loa) (j = 1, .... , n)

(aij, Aij are defined as in (14.27), (14.28)).
Prool By our assumption, FLo(a + e) = FLoa = 0. Let j (j = 1, ... ,n) be fixed.
By Theorem 14.8, there is a Zj E ker D such that Xj = ZjAOjxO = zjAjxo. By
214 Chapter 14
definition, AOj = Eo(aoj), Since aOj E ker D, (Lo, Eo) E GR,dO], and Lja - Loa
is a constant, we find
aOj = Faoj
= F[Lj(a + e) - Lo(a + e) - P-(Lja - Loa)]
= FLj(a + e) - P-(Lja - Loa).
Since Loxj - Ljxj is a constant, say Lozil it follows that
LOZj = Loxj - Ljxj = F(Loxj - Ljxj),

which implies Zj = EoF(Loxj - Ljxj).
The assumption in Theorem 14.8 that (Lo, Eo) is I-normalized by R is not restric-
•
°
tive, since we may assume the same property for another fixed index io. The index
has been chosen just for its simplicity.
If D E A(X) then, under appropriate assumptions we can obtain similar results
to Theorem 14.7 and Corollary 14.9 for the homogeneous Riemann-Hilbert (14.3).
However, these statements are much more complicated than in the commutative
case, so that we shall not give them here.
This method without any essential change can be applied to problems with the
Hilbert transform and the cotangent Hilbert transform, also for singular integral
transform with the Cauchy kernel on a curve closed at infinity (cf. PR[2]).
Example 14.3. (cf. PR[5], PR[6]). Let X = HIL[0,211"] be the space of all 211"-
periodic functions satisfying the HOlder condition with an exponent JL E (0,1).
Clearly, X is a commutative algebra with the usual pointwise addition, multipli-
cation and multiplication by scalars. Let He be the so-called Hilbert cotangent
transform, i.e.
(14.33) (He)(t)
1
= 211" 10
r21T s- t
cot -2- x (s)ds,
where the singular integral He is understood as the Cauchy principal value (cf.
Example 14.1.). It is known that 1+ H; = K on HIL[O, 211"], where
10r
21T
1
(Kx)(t) = 211" x(s)ds, and dim K = 1.
So that in the space

X = {x E HIL[0,211"]: Kx = O}
the operator H = iHelx is an involution: H2 = I. Therefore we may apply
already obtain results in order to solve an induced Riemann-Hilbert problem (cf.
Examples 14.4 and 14.5).
Similarly, if we consider the quotient space Y = HIL[O, 211"]1 K X and the induced
operator 5 defined by means of the equality: H[x] = i[Hex], where [x] is the coset
induced by an element x EX, then we conclude that 52 = I. 0
Another approach is shown by

Example 14.4. Consider a function w holomorphic in a domain 0 c C. Write
u = ~w, v = ~w, where by ~w, ~w we denote the real and imaginary part of
w, respectively. It is known (cf. for instance, WEGERT, Weg[1]) that u, v are
conjugate harmonic functions satisfying the equation
v = -Hu + v(O), u = Hv + u(O) (0 EO),
where H is the so-called Hilbert cotangent transform, i.e.
.
(Hu)(e tt ) = -27r1 1211" . S -
u(e tS ) cot --ds
t
0 2
and that H2 = -I. Thus the operator H' = iH is an involution: H,2 = I. 0

Example 14.5. For 27r-periodic functions on ~ (cf. Example 14.3, also Weg[1])
consider the Hilbert cotangent transform (14.16) with the condition: (Hcu)(1) = O.
Since
Hc(sinkt) = coskt, Hc(coskt) = -sinkt (k EN),
we conclude that the transform
H=
- {OiHc for t = 1;
otherwise;
is an involution. Indeed, H2(coskt) = coskt, H2(sinkt) = sinkt (k EN). 0

Note that results for non-commutative algebras can be applied to problems and
equations with matrix functions.
Several interesting applications might be obtained for problems with functional
shifts, i.e. operators of the form f(hD), where D E R(X), h E C and f is
a function analytic in a domain 0 C C, introduced and examined recently by
BINDERMAN (cf. Appendix).
CHAPTER 15
PERIODIC PROBLEMS
Corollary 13.1 immediately implies the following

Proposition 15.1. Suppose that X E Lg(D) has the unit e and is a complete
linear metric space over F (IF = ~ or F = C), (L, E) E G[Ol, D E R(X) is
closed, {ShhEA(R) is a family of true shifts induced by an R E 'RDnAQN(ker D),
9 = Re, >..g E dom 0- 1 for>.. E vJFR. Then ShE(>..g) = e Ah E(>..g) whenever
>.. E vJFR, h E A(~).
Recall that ShD = DSh on dom D for all h E A(~) (d. Definition 11.2 and
Theorem 11.2).
Definition 15.1. (d. PR[7], PR[8]) Suppose that X is a linear space over C,
D E R(X) and S E Lo(X) commute with D: SD = DS on dom D. Let N E N
be arbitrarily fixed. If
(15.1)
then any element x E XSN is said to be SN -periodic.

Clearly, S is an involution of order N on the space XSN: SN = I on XSN (d.
Theorem 14.1). Thus there are N disjoint projectors Pj giving partition of unit,
i.e.
N
(15.2) PkPj = ~jkPj (j = 1, ... , N); L Pj = I,
j=O
such that
(15.3) SPj = PjS = cj Pj (j = 1, ... , N),
Formulae (15.2) and (15.3) together imply that
(15.4) XSN = X(l) ffi '" ffi X(N) , where X(j) = PjXSN (j = 1, ... , N)
(15.5) SX(j) = cjX(j) , where x(j) = PjX, x E XSN (j = 1, ... ,N).

Projectors Pj are of the form
(15.6)
216
Periodic problems 217
(cf. definition and properties of algebraic operators on pp. 117-118)

Theorem 15.1. Suppose that X E Lg(D) with unit e is a complete linear metric
space overC, (L, E) E c[n], D E R(X) is closed, {ShhER is a family of true shifts
induced by an R E RD nAQN(ker D), S-r is multiplicative for an r E JR, W = Nr
(N EN), 9 = Re and 2:ig E dom n- 1 whenever 2:i E vcR. Let
(15.7) Enr = Xsn- r = {x EX: S'!:.rx = x} = {x EX: S-nrx = x} =j:. 0 (n E N)
and let E(j) =PjE w =PjENr. where Pj are determined by Formula (15.6). If
U = E( - ~:i jg)v, where vEEr, i.e. S-rv = v (j = 1, ... , N), then U E E(j), i.e. if
S-r u = €Ju.
Proof. Let j (j = 1, ... , N) be fixed. By our assumptions and Proposition 15.1,
since W = Nr, we get
S_ru = S-r[E( -.\jg)v)
= [S_rE( - 27ri jg)J(S-r V )
W
-_ e21rijr/wE( - -27ri.)
Jg v
W
27ri.)
_ 21rijr/Nr E( --Jgv
-e
W
=e21rij/N u = €ju.
In order to prove the necessity of the condition given in Theorem 15.1 we need the
•
assumption that X is a Leibniz algebra. Namely, we have
Theorem 15.2 Suppose that X E L(D) with unit e is a complete linear metric
space over C, (L, E) E c[n], D E R(X) is closed, {ShhEIR is a family of true
shifts induced by an R E RD n AQN(ker D), S_r is multiplicative for an r E JR,
w = Nr (N EN), 9 = Re and 2:ig E dom n- 1 whenever 2:i E vcR. Let
E(j) = PjE w = PjENr' where Pj and ENr are determined by Formulae (15.6)
and (15.7), respectively. Then u = E(_21rijg)v, w,
where vEEr, i.e. S-rv = v
(j = 1, ... , N), only if u E E(j), i.e. S_ru = €Ju.
Proof Since X is Leibniz algebra, we have (Ex)-1 = E( -x) whenever x E
dom n- 1 . Suppose that u E E(j), j = 1, ... ,N is fixed. Then S-ru = €ju. Write
v = Ee: i jg)u. By Proposition 15.1,
S-r v = S_r[E(27ri jg)u)

w
= [S_rE(27ri jg))S-ru
w
= , 2 "()I 27ri
€J e 1r'J - r wE(-jg)u
w
= e21r'J"IN e- 2"
1r'Jr IN r 27ri
E(-jg)u = v.
w
218 Chapter 15
Hence u = E( - 2;i jg)v, where S-rv = V. •

An immediate consequence of Theorems 15.1, 15.2 and the decomposition (15.4)
onto the direct sum is
Then x E Ew if and only if
2 ..
L
N
x = E( - 7rlJ g)Vj, where S-rVj = Vj (j = 1, ... , N).
j=l W
Note 15.1. In the classical case of the space X = C(~) over C, when D = ft,
(ShX)(t) = x(t + h) for x E X, t,h E JR, we find that Ew , w = Nr, is the space of
of w-periodic functions. The operator S-r is an involution of order N on Ew for
(S~rx)(t) = x(t - Nr) = x(t - w) = x(t). Thus any w-periodic function x E X is
of the form
L
N
X= e21rijt/wVj where VI, ... , VN are r-periodic functions
j=l
(cf. PRR[2], also PR[7]). Functions of the form eAtv, where>. E C, V is a periodic
function, and their linear combinations are said to be exponential-periodic func-
tions. Elements of the form E(>.g)v, where>. E C, V is a periodic element, and
their linear combinations are said to be exponential-periodic elements. Recall that
in Corollary 15.1 X is a Leibniz algebra, hence antilogarithms are exponentials.
For an arbitrary MEN and a true shift Sh we denote by
XEP(h; >'1, ... , >'M)
= lin {E(>'jg)Vj : Vj E Xs h ; >'j E C; >'m ::f >'j + 2~k, m ::f j;
k EZ (m,j = 1, ... ,M)}
a space of exponential-periodic elements. By Corollary 15.1, x E Ew if and
only if x E XEp(h; - 2;i, ... , - 21r~N) (Sh = S-r). More details about spaces
of exponential-periodic functions and their applications can be found in PR[6],
PR[7]. 0
Theorem 15.1 and Corollary 15.1 will be used in order to generalize the Floquet
theorem. The classical Floquet theorem says that every linear ordinary differential
equation with periodic coefficients has at least one non-zero exponential-periodic
solution, i.e. a solution of the form eAtu(t), where u ::f 0 is a periodic function and
>. E C is properly chosen (cf. ARscoTT Ar[l], INCE In[l]).
Lemma 15.1. Suppose that X is a linear space over IF, D E R(X), ker D ::f {O},
F is an initial operator for D corresponding to an REnD and {ShhElR is a family
of R-shifts, i.e. a family of linear operators such that
k hk - j
(15.8) So = I, ShRkF = L, (k _ .)!RjF for kENo·
)=0 J
Then ShD = DSh on dom D whenever hE llt If
(15.9)
(15.10)
then Dh E R(X), Fh is an initial operator for Dh corresponding to the right inverse

Rh = R - FhR. Moreover, the operator Rg maps the space Eh onto the space E~l)
and
(15.11) D hR oh = I on E h, ROD
h h= I h ,z.e. D-
on E(l). h 1 = ROh'
Proof. It is not difficult to check that, by definition, R-shifts commute with D

on its domain (cf. PR[7]). By our assumptions, FR = 0, DR = I. Thus it easy
to verify that Dh E R(X) and that Fh is an initial operator for Rh whenever
hE A(JR). Suppose that h E A(JR) and x E Eh are arbitrarily fixed. By definition,
S-hX = x, hence Fhx = FS_hX = Fx. Write z = FhRx = FS-hRx E ker D.
Then we find
The arbitrariness of x E Eh implies the first of Identities (15.11). Suppose now

that x E E~l) is arbitrarily fixed. Then x E dom D, S-hX = x, FhX = FS-hX =
Fx. By definitions, F = I ~ RD on dom D, FR = 0 and F2 = F, moreover,
ShRFx = (R - h)Fx, so that
220 Chapter 15
Thus we get
R~DhX
1 1
= (I + Fh + hRFh)R(D - hFh)x
1 1
= (I + Fh + hRFh)R(Dx - hFhX)
1 1
= (I + Fh + hRFh)(RDx - hRFhX)
1 1
= (I + FS-h + hRFh)(x - Fx - hRFhX)
1 1
= x - Fx - hRFhX + FS-hX - FS_hFx - hFS-hRFhX+
1 1 1
+ hRFS-hFx - hRFS-hFhX - h2RFS-hRFhX
1 2 1
= x - Fx - -RFhX + Fx - F x - -F(RFx - hFx)
h h
1 2121 12
=x-Fx--FRFx+F x--RF x--RfRFx+-RF x
h h h2 h
1 1
= x - Fx + Fx - h + h RFx = x.
The arbitrariness of x E E~l) implies the second of Identities (15.11). Clearly,

Y = R~x E dom D. Indeed, since DF = 0, we find
1
Dy = D(R + FhR + hRFhR)X
1
=DRx + DFS_hRx + hDRFShRx
1
=x + hFS-hRx.
Moreover,
S-hY
1
=S-h(I + Fh + hRFh)Rx
1
=S_hRx + S_hFS_hRx + hS-hRFS-hRx
1
= S_hRx + FS_hRx + h(RFS-hRx - hFS_hRx) =
1
= S-hRx + FS_hRx + hRFS-hRx - FS_hRx
1
=S-hRx + hRFS-hRx.
Hence
u = Y - S-hY =
1
o - (S_h X - TtRFS-hRx
Rhx )
1 1
= Rx + FS_hRx + TtRx + TtRFS-hRx - S_hRx
1
+ TtRFS-hRx
= Rx - S-hRx + FS_hRx,
Du+D(y-S_hY) = DRx-DS_hRx+DFS_hRx = x = S_hDRx = X-S_hX = 0,
which implies u E ker D. On the other hand,
We therefore conclude that u = Fu = 0, Le. S-hY = Y and Y E E~l). The

arbitrariness of x E E~ implies that R~Eh E E~1). Since R~ is an inverse of D h ,
the corresponding initial operator vanishes on the domain of D h • •
Definition 15.2. Suppose that all conditions of Definition 15.1 are satisfied. Let
mEN. An operator A E Lo(X) is said to be Sm-periodic if smA = Asm. 0
Clearly, if D E A(X), Ax = ax for an a E X and S is multiplicative, then
sm(Ax) = Sm(ax) = (Sma)Smx. Thus A is Sm-periodic if and only if a E XSTn
(m EN).
Lemma 15.2. Suppose that all assumptions of Theorem 15.2 are satisfied. Write
K
(15.12) Q(D) =L Qk Dk , Qk E Lo(X), S-rQk = QkS-r (k = 0, 1, .. , K),
k=O
i.e. Qk are S_r-periodic. Suppose, moreover, that v E dom DK is an S_r-periodic
element and u = E(Ag)V E kerQ(D). Then
A = 21l'i j where j E Z is arbitrary.

w
Proof. By our assumption, S'.!:.ru E ker Q(D) whenever u E ker Q(D), n E No

Indeed,
K K
= L QkDkS':.ru = L QkS':.r Dku
k=O k=O
K K
=L S':.rQkDku = S':.r L Dku = S':.rQ(D)u = 0.
k=O k=O
222 Chapter 15
Since u E kerQ(D), we find
o= S~rQ(D)u
= Q(D)S~ru = Q(D)S~r[E(>'g)vl
= Q(D)[S~rE(>'g)lS~rv = Q(D)[e-AnrE(>.g)v
= e-AnrQ(D)[E(>.g)v] = e-AnrQ(D)u.
Thus for n =N we find eW =e- ArN = 1, i.e. >. is of the required form. •
Lemma 15.3. Suppose that all assumptions of Lemma 15.2 are satisfied. Let
>. = 2:i
j, where j E Z is arbitrary. Then Q(D + AI)v = O. Conversely, if
Q(D + >.I)v = 0, then u = E(>.g)v E ker Q(D).
Proof Necessity. Since DE(>.g) = >'E(>.g), we have
Du = D[E(>.g)v] = (Dv + >.v)E(>.g) = [(D + AI)v]E(>.g).

By an easy induction, we find
This implies that
[Q(D + >.I)v]E(>.g)
K
= L Qkf(D + AI)kv]E(>.g)
k=O
K
= L [E(>.g)v]
k=O
K
=L QkDku = Q(D)u = O.
k=O
Since X is a Leibniz algebra, E(>.g) E leX). Thus Q(D+AI)v = O. The sufficiency

of the condition is obvious. •
Theorem 15.3. (Generalized Floquet Theorem) Suppose that all assumptions
of Theorem 15.2 are satisfied. Let Dr, Fr , R~ and Q(D) be defined by Formulae
(15.9), (15.1O), (15.11), respectively. Write
(15.13)
K
(15.14) -
Q(t, s) = '~
"' -
Qkt k K-k
s , -
Q(t) = Q(t,
-
1).
k=O
If the operator Q(I, R~) is invertible in the space Er then the equation
(15.15) Q(D)x =0
has exponential-periodic solutions which are of the form
h S-rV = \ 27ri.
(15.16) x = E()..g)v, were V, /\ = -J (j E Z),
w
K k-1
(15.17) V = (R~)K[Q(I, R~)r1 L Qk L r-k+iz·
J'
k=O j=O
ZO, ... , ZK -1 E ker D are arbitrary.
Proof. By Lemma 15.3, Equation (15.15) has an exponential-periodic solution of

the form (15.16) if and only if)" = 2;i
j (j E Z) and v satisfies the equation
(15.18) Q(D + M)v = o.

According with our notations (15.13), (15.14), we have
Q(D + M)
m=O
=%;0 Qm~ (7) ..m-k D k

= t [t (7) ..
k=O m=k
m-kQm]D k
K
= L QkDk = Q(D).
k=O
Observe that Qo, ... , QK, as linear combinations of S_r-periodic operators, are
again S_r-periodic. We can rewrite Equation (15.18) in the following way:
(15.19) Q(D)v = O.
Since the operator Q(I, R~) is invertible in the space Er , we conclude that Equation
(15.19) has S_r-periodic solutions of the form (15.17). •
In order to reduce an equation with an involution of order N, in particular, an
equation with a true shift in the space of periodic elements, we need
224 Chapter 15
Lemma 15.4. (cf. PR[7]) Suppose that X is a linear space over C, DE R(X),
S E Lo(X) commutes with D: SD = DS on dom D, the operators Qkm E Lo(X)
are S-periodic, i.e. SQkm = QkmS (k = 0,1, ... , N - 1; N ~ 2; m = 0,1, ... , M).
Write
N-l M
Qm(S) = L Qkm Sk , Q(D,S) = L Dm+Ml (Ml E No))·
k=O m=O
If XSN =j:. 0 then

N
(15.20) Q(D,S) =L Q(D,cj)Pj on XSN, where c = eN,~
j=1
and projectors PI, .... , PN are defined by (15.6).
Proof. Since the operators D and Pj (j = 1, ... , N) commute, the decomposition
(15.4) implies that on XSN for j = 1,2, ... , N
Q(D,S)Pj
M M
= L Qm(S)Dm+Ml Pj =L Q(S)PjDm+Ml
m=O m=O
M M
=L (L L (L
~ ~
QkmSkpj)Dm+Ml = Qkmcjkpj)Dm+Ml
m=O k=O m=O k=O
M M
= L Q(cj)PjDm+Ml =L Q(cj)Dm+MIPj = Q(D,cj)Pj .
m=O m=O
Since
N
L Pj =1 on XSN,
j=1
we conclude that on XSN
N N N
Q(D,S) = Q(D,S) L Pj =L Q(D,S)Pj =L Q(D,cj)Pj .
j=1 j=1 j=1
Proposition 15.2. (cf. PR[7]) Suppose that all assumptions of Lemma 15.4 are
•
satisfied. Then the equation
(15.21) Q(D,S)x = y, YE XSN,
is equivalent in the space XSN to N independent equations
(15.22) Q(D,cj)xj = Yj, where Xj = Pjx,Yj = PjY E X(j), (j = 1, ... ,N),

and XU) are defined by the decomposition (15.4), Xj,Yj are defined by Formulae
(15.5).
Proof The definitions of the operators Pj and elements xi> Yj (j = 1, ... , N)
together imply that x = Xl + ... +XN, Y = Y1 + .. ·+YN and that this decomposition is
uniquely determined. Since the operators Qkm are S-periodic, we have SQkmXj =
QkmSXj = c:jQkmXj for j = 1, ... , N, k = 0,1, ... , N - 1, m = 0,1, ... , M. This
implies that the each operator Q(D, c: j ) preserves the space Xj (j = 1, ... , N).
Indeed, if Xj E Xj (j = 1, ... , N) then
M N-1 M N-1
Q(D,c:j)xj =L L QkmDm+M,pjx = Pj L L QkmDm+M,X E Xj.
m=O k=O m=O k=O
This, and Lemma 15.4 together imply that
Q(D, S)x
N N
= Q(D, S) L PjX = L Q(D, S)Pj
j=l j=l
N N
= L Q(D,c:j)Pjx =L Q(D,c:j)xj.
j=l j=l
Since
N N
y= L PjY= L Yj,
j=l j=l
the space XSN is the direct sum of spaces X(j) and Xj,Yj E X(j) (j = 1, ... ,N),
we conclude that Equation (15.21) is equivalent to the system (15.22) of N inde-
pendent equations. •
Theorem 15.4. (cf. PR[7]) Suppose that all assumptions of Lemma 15.4 are
satisfied. If each of Equations (15.22) has a solution Xj E Xj (j = 1, ... , N) then
Equation (15.21) has a solution x = Xl + .... + XN E XSN. Conversely, if Equation
(15.21) has an SN -periodic solution X then the jth Equation (15.22) has a solution
Xj = PjX E Xj.
Proof If Equation (15.21) has an SN-periodic solution then, by definition, X E
XSN. Hence SN x = x. Thus x = Xl + ... +XN, where Xj = PjX E Xj (j = 1, ... , N).
Proposition 15.2 implies that Equation (15.21) is equivalent to the system (15.22).
Thus the jth equation (15.22) has a solution Xj E Xj. Conversely, suppose that
the jth equation (15.22) has a solution Xj E Xj (j = 1, ... , N). Since PJ = Pj ,
Pj ::j:. Pk for j ::j:. k and Xj = PjX for an X E X(j, k = 1, ... , N) (cf. PR[6]), writing
x = Xl + ... + X N we find
N N
PjX = Pj L Xk =L PjPk X = PjX = Xj (j = 1, ... , N).
k=l k=l
226 Chapter 15
Proposition 15.2 implies that x E and that x is a solution of Equation (15.21)

•
XSN
Theorem 15.4 has several applications. We may use this theorem when S = Sh is a
true shift (under appropriate additional assumptions). Some of these applications
are quite far from classical differential-difference equations (cf. PR[6], PR[7]).
For instance, this method can be used for a reduction of a stochastic differential-
difference equation to a stochastic differential equation in order to find its periodic
solutions (cf. WILKOWSKI Wi[l]). In a similar manner one can consider the
operator
M
Q(D,S) = DMl 2: Qm(S)Dm.
m=O
Another possibility is given by

Theorem 15.5. Suppose that all assumptions of Theorem 15.2 are satisfied.
Consider a nonlinear equation
(15.23)
where the mapping

G Xx ... xX--tX
'-....-'
m-times
is continuous in each variable and all numbers Wl, ••• , Wm are commensurable, i.e.
there exist an r E IR \ {O} and '171, ... , 17m E Z such that Wj = '17jr (j = 1, ... , m). We
admit Wo = '170 = O. Let N be a common multiple of positive integers l'17d, ... , 117m I
and let W = Nr. Then Equation (15.23) has a solution x E Ew if and only if the
following system of N equations without shifts
(15.24) - = -
D Xj 27rij
-Xj
- + G-j (-Xl, ... ,XN
-) (j = 1, ... ,N),
W
has a h-periodic solution, where h = -r, and
(15.25) -(-Xl, ... XN

G - ) -- G(~
~ e21rkf/OE(27rik)-Xk, ..• , ~
---;;;-g ~ e ---;;;-g
21rkf/mE(27rik )- )
,Xk ,
k=l k=O
If it is the case then
N 2 ..
(15.26) X = '"' 7rZJ_
~ E(-g)xj,
j=O W
where (Xl, ... , XN) is a h-periodic solution of the system (15.24), where h = -r.
Proof. Acting on both sides of Equation (15.23) by projectors PI, ... , PN and
applying Theorem 15.2 we can rewrite this equation as a system of N equations
without shifts. Since, by Theorem 15.2, Xl, ... , XN are h-periodic functions, where
h = -r, for v = 1, ... , N we have
pj](X1, ... ,XN)

N-1
1 "~ c- kvG(S-krX1,
=N - ... ,S-krXN)
k=O
N-1
= L ~c-kVG(X1' ... ,XN) = PvG(X1, ... ,XN)
k=O
Theorem 15.5 can be use also in order to study linear equations with shifts, in
•
particular, differential-difference equations with periodic coefficients. In a similar
manner one can consider other periodic problems (cf. PRR[2], also PR[6], Section
V.5). Note that in Theorem 15.5 we obtain a system of equations which, in general,
are not independent.
Consider again the space of exponential-periodic elements
(15.27)
lin {E(Ajg)Vj : Vj E Xs h ; Aj E C; Am f; Aj + 27r~k , m f; j; k E Z (m, j = 1, ... , M)}

defined in Note 15.1 for an arbitrary MEN and a true shift Sh. We have
Theorem 15.6. Suppose that all assumptions of Theorem 15.2 are satisfied. Let
Aj = 2:i
j (j = 1, ... , N), and let h = -r. Then Sh is an algebraic operator on the
space XEP(h; AI, ... , AN) with the characteristic polynomial
N
pet) = II (t - tj), where tj = eA;h (j = 1, ... , N)
j=l
(with single characteristic roots).
Proof. Suppose that x = l:f=o E(Ajg)Vj E XEP(h; Aj, ... , AN) is arbitrarily fixed.
Recall that Vj E Er (j = 0, ... , N). By our assumptions,
N N
=L Sh[E(Ajg)Vj] =L [ShE(Ajg)](ShVj)
j=l j=l
N N
=L eA;hE(Ajg)Vj =L tjE(Ajg)Vj.
j=l ;=1
228 Chapter 15
By an easy induction we show that
N N
S~x = S~ 2: E(Ajg)Vj = 2: tJE(Ajg)Vj (k EN).
j=1 j=1
If we write P(t) = nf=1 (t - tj) = 2:~=o Pktk, where PN = 1, then we have

P(tj) = 0 for j = 1, ... , N. Thus
N N N
= L PkS~X =L Pk L tJE(Ajg)Vj
j=1 k=O j=1
N N N
= 2: (2: PktJ)E(Ajg)Vj = 2: P(tj)E(Ajg)Vj = O.
j=1 k=O j=1
The arbitrariness of x implies that P(Sh) = 0 on XEP(h; AI, ... , AN). The roots
of the polynomial P(t) are single because, by our assumptions, Am :j:. Aj + 2""ik
whenever m :j:. j, k E Z, which implies that tm :j:. tj whenever m :j:. j (j,m =
1, ... ,N). •
Then X = XEP(h; AI, ... , AN) is a direct sum of the eigenspaces of the operator
Sh corresponding to the eigenvalues tl, ... , tN:
(15.29)
We shall consider now combinations of exponential-periodic elements and D-

polynomials. However, in that case Sh is an algebraic operator with multiple char-
acteristic roots. Note that polynomial-periodic solutions of differential-difference
equations firstly have been studied by WLODARSKA-DvMITRUK (cf. WD[1]; a
generalization for polynomial-exponential-periodic elements and shifts induced by
right invertible operators was given in PR[7]).
Lemma 15.5. Let D E R(X), ker D :j:. {O} and let F be an initial operator for
D corresponding to an R E RD. Let {ShhEIR C Lo(X) be a family of R-shifts,
i.e. such operators that So = I and for h E ~
(15.30)
(cf. Lemma 15.1) and Formula (15.8). Then for all h E JR, z E ker D, k, n E N, we
have
k
(15.31) ShnRkz =~
L..J h(n)
k,m Rmz ,were
h f or J. = 0, 1, ... , k
m=O
(15.32) h(l) _ 1 hk-j

k,j - (k _ j)! '
k
(15.33) hk(n:t- l )
,1
=~
L....J
hen) h(l).
k,m m,j'
hk,k
(nH) -- 1 'f n >
t _ 1.
m=l
Proof, by induction. Let h E JR, z E ker D and kEN be arbitrarily fixed. For
n = 1 (15.31) follows just from the definition (15.30). Suppose then (15.31) holds
for an arbitrarily fixed n E No By (15.32) and (15.33), we get
m=O m=O
m=O j=O j=O m=j

k
- ~ h(nH) Rj
- L..J k,j Z,
j=O
which was to be proved. Moreover, by definition, hili = 1. By an easy induction

we show that hk~t) = hk~k = 1 for all n E N. ' •
Proposition 15.3. Suppose that all assumptions of Lemma 15.5 are satisfied.
Write
M M
(15.34) P enH) (t) =~
L..J PM,m t ,
m p(O)(t) = P(t) = L Pmtm,
m=O m=O
M
(15.35)
(0) (nH) _ ~ (n) hen) (nH) - 1
PM,m =Pm, PM,j - L..J PM,m m,j' PM,M - ,
m=j
pmEC, PM=l (j,m=O,l, ... ,M; nENo),

230 Chapter 15
where h~!j are defined by Formulae (15.32), (15.33). Then pt~M = 1 for MEN,
n E No and
(15.36) Sl:p(n)(R) = p(n+l)(R) for all hE JR, z E ker D (n E No).
Proof, by induction. Let hER and z E ker D be arbitrarily fixed. Lemma 15.5
and Formulae (15.35) together imply that
M M M M
SI: P(R)z =L PM,jRj z =L [L Pmh~!jlRj z =L p<f),jRj z = p(l) (R)z.
j=O j=O m=j j=O
Suppose Formula (15.36) to be true for an arbitrarily fixed n > 1. By our assump-
tions and Lemma 15.5,
SI:P(R)z
M M
= SI: L Pm Rmz =L PmSI:Rmz
m=O m=O
M M M M
= L PmL h~!jRjz=L L Pmh~!jRjz

m=O j=O j=O m=j
M
= L PM,jRjz,
j=O
where
M
PM,j =L Pmh~!j (j = 0,1, ... , M)
m=j
and PM,M = PMht~M = PM = 1. This, and Formulae (15.35) together imply that
SI: p(n+l) (R)z

M
= Shn '"' (n+l)Rj Z
~ PM,j
m=O
j=O m=j
M
= L pt~2) Rjz = P(n+2)(R)z.
j=O
By definition, P~M = PM,M = 1. Suppose that pt~) = 1 for an arbitrarily fixed

1 Th enPM,M(n+2) (n+l)h(n) 1 •
n>. =PM,M M,M = .
Proposition 15.4. Suppose that all assumptions of Lemma 15.1 are satisfied.
Then for every h E ~ and kENo the operator Sh is algebraic on the space
ker Dk+l with the characteristic polynomial P(t) = (t - l)k+l, i.e.
Proof. Let h E ~ and kEN be arbitrarily fixed. Since Sh is an R-shift, for all
z E ker D we have Sz = z, i.e. (Sh - I)z = o. Indeed, ShZ = ShRoz = ROz = z.
Clearly, ker Dk+l = lin {Rkz : kENo}. Let p(O)(t) = t-1, i.e. M = 1, Pl,O = -1,
Pl,l = 1. By Lemma 15.5,
m=O m=O
k-l k-l
= (Sh - I) L h1~~Rmz = L h1~~(Sh - I)Rmz
m=O m=O
(k-2) /l-
- "h(l) "h(l) RV
- L..t k,/l-+l L..t 1'+1 ,v z
1'=0 v=o
k-2 k-2
= "["
L..t L..t h(l)
JL+l,v ]RVz .
k,JL+l h(l)
Writing
for j = 1, ... , kj v = 0,1, ... , k - j,
we get
(15.37)
v=O
In a similar way, acting on both sides of Equality (15.37) by operator Sh - I, we

find after j steps (j = 1,2, ... , k)
k-j
(Sh - I)j = "L..t h,u-l)
k,v R z and
V
(Sh - I)k+l Rk z = (Sh - I) [(Sh - I)k Rk z] = (Sh - I)h~~;;l) z = h1~;;1) (Sh - I)z = o.
•
232 Chapter 15
Corollary 15.3. Let D E R(X), ker D ::j:. {o} and let F be an initial operator
for D corresponding to an R E RD. Let X E A(D). Let {ShhEIR C Lo(X) be a
family of multiplicative R-shifts (cf. Lemma 15.5). Then
i.e. Sh is an algebraic operator on the space
lin {vRkz : z E ker D, sf:v = v (k E No)}
with the characteristic polynomial P(t) = (t N _1)k+l and with the characteristic
roots c: j = e21rij/N (j = 0,1, ... , N), each of multiplicity k + 1.
Proof. Let k, N E N, z E ker D, h E JR, v E XSN be arbitrarily fixed. Then
h
Sf:(vRkz) = (Sf:v)(Sf:Rk z ) = vSf:Rkz. Thus (Sf: - I)(vRkz) = vSf:Rkz-
vRk z = v(sf: - I)Rkz. By an easy induction we prove that (sf: - I)m(vRkz) =
v(Sf: - I)m Rk z for mEN. This, and Proposition 15.4 together imply that
(sf: - I)k+1(vR k z) = v(sf: - I)k+1R k z = v(I + ... + Sf-l)k+1(Sh - I)k+1 Rkz.
The arbitrariness of z E ker D and v E XSN implies that Sh is an algebraic operator

h
on the space under consideration with the characteristic polynomial (t N - 1)k+l.
Since all Nth roots c: j of unity are roots of the polynomial t N - 1, we conclude
that they are characteristic roots of Sh, each of the multiplicity k + 1. •
An immediate consequence of Corollary 15.3 is
Corollary 15.4. Suppose that all assumptions of Corollary 15.3 are satisfied.
Write for M, no, ... , nM, nj i- nk if j i- k, h E JR
(15.38)
M
= {u = L vmRmzm: Zm E ker D, S~mvm = Vm (m = 0,1, ... , M)}.
m=O
Then
(15.39) P(Sh) = ° on Xpp(h; no, ... , nM), where P(t) = II

M
m=O
(tnm - I)M+1,
i.e. Sh is an algebraic operator on the space Xpp(h; no, ... , nM) with the character-
isticpolynomialP(t). The characteristic roots arec:t,. = e~ (m = 0,1, ... ,M; j =
O,I, ... ,nm - 1), each of the multiplicity M + 1. Thus Xpp(hjno, ... ,nM) is
a direct sum of principal spaces X jm such that (Sh - c:t,.I)M+1 = on X jm
(j=O,I, ... ,n m -l; m=O,I, ... ,M).
°
Elements of spaces Xpp(h; no, ... , nM) are said to be D-polynomial-periodic ele-
ments.
Theorem 15.7. Suppose that X E L(D) has the unit e and is a complete linear
metric space over C, (L, E) E G[O], D E R(X) is closed, {ShhER is a family
of multiplicative true shifts induced by an R E RD n AQN(ker D), 9 = Re,
M, K j , N j E N and Ajg E dom 0- 1 whenever Aj E vcR (Aj 1= Ak + 27ril/h if
j 1= k; I E Z; j, k = 0,1, ... , M). Then Sh is an algebraic operator on the space of
D-polynomial-exponential-periodic elements:
(15.40) XPEP(h·, A··

J, K·· J, M)
J' N··
M Kj
= {x = L (L VjkRkZjk)E(Ajg): Zjk E kerD, Vjk E X s :/' k = O,I, ... ,Kj}
j=O k=O
with the characteristic polynomial
II (t Nj -
M
(15.41) P(t) = t JNj )K j+l, h
were tj = e Ajh (.J = 0 , 1, ... , M) .
j=O
The characteristic roots of the polynomial P(t) are

21fi
(15.42) tjm=tjcj, wherecj=e Nj (m+O,I, ... ,Nj-l; ,j=O,I, ... ,M)
of the multiplicity K j + 1, respectively.

Proof. Let A E vcR, h E JR, K, N E N, Z E ker D, v E XSN be arbitrarily fixed.
Let w = VE(A)RK z. Since, by our assumptions, Sh is a multiplicative true shift
h
and sf: v = v and sf: = SNh, we find
sf: (VRK z)
= sf: w = (sf: v)[sf: E(Ag)]sf: RK z
= ve ANh E(Ag)Sf: RK Z = eANhvE(Ag)Sf: RK Z
= eANhvSf: RK z.
Thus
(sf: - e ANh I)w = eANhvE(Ag)(Sf: - I}RK z,
and, by Proposition 15.4, we get
(15.43)
Suppose now that u is an arbitrary element of XPEP(h; Aj; Kj; Nj; M), i.e. u is
of the form
M Kj
U = L (L
VjkRkZjk)E()..jg),
j=O k=O
234 Chapter 15
where h, M, Aj, K j , N j , Vjk, Zjk are described in Formula (15.40). The assumption
that Aj :f:. Ak + 27f'il / h whenever j :f:. k, l E Z, asserts us that antilogarithms
E(Ajg) are pairwise different and linearly independent. It is so because, by our
=
assumption that Aj E vcR (j 0,1, ... , M), we have E(Ajg) =
e)..;zj, Zj E ker D
and exponentials e)..;zj are linearly independent (cf. PR[8)). Let pet) be defined
by Formula (15.41). Formula (15.43) implies that
M M K,
= [II (8:; - tf;)K;+l] L (L vlk Rkzlk)E(A19)
j=O 1=0 k=O
M K, M
=L {L [II (8:; - tf;)K;+l]vlkRkzlk}E(Alg) = O.
1=0 k=O j=O
•
Then a principal space corresponding to the root 1 of the multiplicity k + 1 is
lin {Rjz: Z E kerD, j = O,I, ... ,k}.
We can solve now equations with shifts in the spaces Xpp(h; no, ... , nM) and
XPEP(h; Aj, Kj; Nj; M) in the same manner as we did it in the space of periodic
elements and in XEP(h; Aj, ... , AM) (cf. for instance, Proposition 15.2, Theorem
15.4).
Note 15.2. Theorems 15.6, 15.7 and Corollary 15.2 are proved for Leibniz alge-
bras. A modified proof could be used for quasi Leibniz algebras, i.e. commutative
algebras with the product rule D(xy) = xDy+yDx+d(Dx)(Dy) for x, y E dom D,
where d:f:. 0 is a scalar, i.e. if X E AD(O,e,de) (cf. Example 5.1 and Corollary
5.4). We should only to remember that in this case logarithms (provided that they
exist) are not of of the exponential type, but satisfy the following equation
DL(uv) = D(Lu + Lv) + d(DLu)(DLv) far u, v E dom 0, (L, E) E 0[0].
Thus
(Ex)(Ey) = E{x + y + dR[(Dx)(Dy)] + z} where Z E ker D, x = Lu, y = Lv.

We cannot use similar arguments for simple Duhamel algebras, i.e. algebras with
the product rule D(xy) = xDy, what can be written D(xy) = HxDy + yDx) for
x, y E dom D (cf. Example 5.1), since those algebras belong to AD(O, ~e, 0). By
Theorem 5.1, logarithmic (hence also antilogarithmic) mappings in that case do
not exist. 0
Example 15.1. Suppose that all assumptions of Example 13.5 are satisfied, i.e.
X E Lg(D)) is a complete linear metric space over C, (L, E) E 0[0], D E R(X),
there is an R E RD n AQN(ker D), F is an initial operator for D corresponding
to R, 9 = Re, Ag E dom 0 for A E vcR and {8h lhEIR is a family of true shifts
induced by R (cf. Definition 11.2). Let WeD), W"(t) be defined by Formulae

(13.16), (13.17), respectively, i.e.
n m
WeD) =L LakjDkS_hj,
k=O j=O
0= ho < hI < ... < hm, akj E C (k = 0, ... , nj j = 0, ... , m),

n m
W"(t) =L L akjtke- hjt .
k=O j=O
By (13.19, E(>..g) E ker WeD) if and only if W"(>..) = o. It means that in order to
determine solutions of Equation (13.20)
W(D)x = 0,
which are of the form E(>..g), it is enough to find characteristic roots, i.e. zeros of
the quasipolynomial W"(t). According to Definition 9.2, assume that Condition
[eh is satisfied, i.e. -x E dom 0- 1 whenever x E dom 0- 1 and the trigonometric
identity holds. Let w = u + iv E dom 0, w* = u - iv and let
C(X) = {w = u + iv E dom 0 : ww· E 12 (dom On.
By definition, w E leX). Recall that a sufficient condition for the trigonometric

identity to be satisfied is that X is a Leibniz algebra.
Suppose that, in addition, D is closed and (L, E) E G R,t[O]. By Corollary 9.B,
for all w E C(X) and x = argw E £D(X) the mapping Ei is 21Te-periodic, i.e.
E[i(x + 21Te)] = E(ix).
Suppose then that all the listed conditions are satisfied, a >. satisfies the equation
W"(>..) = 0 and >..g E £D(X). Then E(>..g) = E[i( ->..ig)] is 21Te-periodic, i.e.
E(>..g) = E(>..g + 21Tike) , whenever k E Z. 0
It should be mentioned that R-shifts Sh defined by Formula (15.B) correspond to
R-shifts S-h studied in PR[7], PR[B]. This change of sign is not essential, however,
it is convenient in order to have a unified approach to different questions considered
here. Some results of this chapter are true also in the case when h E A(JR) = JR+.
CHAPTER 16
EQUATIONS WITH MULTIPLICATIVE INVOLUTIONS OF ORDER N
In Chapter 14 we have considered non-multiplicative involutions (cf. Corollaries

14.5 and 14.8). In Chapter 15 we have studied periodic problems with multiplica-
tive involutions of order N 2': 2. However, neither in this book nor elsewhere
equations with multiplicative involutions of order N 2': 2 have been studied to the
full. So that, we shall present here some results never published before and we
shall apply them to equations with logarithms and antilogarithms.
Proposition 16.1. Suppose that X is a commutative algebra over C, S E M L(X)
is an involution of order N 2': 2, i.e. SN = I and c: = e27ri / N . If
N-I
(16.1) a(t) =L aktk, where ao, ... , aN-I EX
k=O
then
N
(16.2) sna(S) = L c:jn[sna(c:j)]Pj for all n E N,
j=1
where PI, ... , PN are disjoint projectors giving the partition of unity and such that
SPj = PjS = c:j Pj (j = 1, ... , N) (cf. Definition 15.1).
Proof. Let S be a multiplicative involution of order N 2': 2. By an easy induction
we prove that
snpj = c:jnPj (j = 1, ... ,N) and sn E ML(X) for all n E N.
By Formula (15.3),
N
a(S) = a(S) L Pj
j=1
N N N-I
= L a(S)Pj =L L akS k Pj
j=1 j=1 k=O
N N-I N
=LL akc:jkPj =L
j=1 k=O j=1
Hence
N
(16.3) a(S) = L a(c:i)pj .
j=1
By (16.3), for any fixed n E N we get
236
Multiplicative involutions 237
N N
= sn L a(ei)Pi =L sn[a(ei)Pi] =
i=1 j=1
N N
=L [sna(ej))snpi = L ein[sna(ein))Pj.
j=1 j=1
•
and X has the unit e E X N = PN X. Then Se = e, Sx E J(X) whenever x E J(X)
and (SX)-l = Sx- 1.
Proof By definition, Se = eN e = e21ri e = e. Suppose that x E J(X). Then
(Sx)(Sx- 1) = S(xx- 1) = Se = e. Hence Sx E J(X) and (Sx)-l = Sx- 1. •
Theorem 16.1. Suppose that X is a commutative algebra over C, S E M L(X)
is an involution of order N 2:: 2, e = e21ri / N and a(t) is defined by (16.1). Then
the equation
(16.4) a(S)x = y, y E X,
is equivalent to the system

N
(16.5) L ejk[Ska(ei)]Xi = Sky (k = 0,1, ... , N - 1)
i=l
of N equations, where Xj = PjX (j = 1, ... , N). The determinant of the system

(16.4) is
(16.6)
a(e)
eSa(e)
=
Proof Acting on both sides of Equation (16.4) by the operators S, ... ,SN-1 and
applying Proposition 16.1, we get the system (16.5). Clearly, the determinant of
that system is of the form (16.6). •
Note 16.1. Suppose that all assumptions of Theorem 16.1 are satisfied. If there
is an index j (1 ~ j ~ N) such that a(e i ) = 0, then ~(a) = O. In other words, if
~(a) ¥- 0 then a(ei ) ¥- 0 for j = 1, ... , N. 0
238 Chapter 16
Corollary 16.1. Suppose that all assumptions of Theorem 16.1 are satisfied. If
~(a)E leX) then Equation (16.4) has a unique solution of the form
N
(16.7) X= L Xj, where
j=l
N N-l
Xj = [~(a)tl L L (_l)i+k+l~jk(a)Sky (j = 1, ... ,N)
j=l k=O
and by ~jk(a) is denoted the minor determinant obtained by cancelling the jth
row and the (k+1)th column in ~(a) (j = 1, ... , N; k = 0, ... , N - 1).
Corollary 16.2. Suppose that all assumptions of Theorem 16.1 are satisfied. If
a(c i ) = 0 for j :j:. m and a(c m ) E leX) (j, m = 1, ... , N) then Equation (16.4) is
reduced to the equation
(16.8)
A necessary and sufficient condition for Equation (16.8) to have a solution is that
(16.9)
If this condition is satisfied then the general solution of Equation (16.8) is
N
(16.10) X = [a(cm)tly + L Xi, where Xi are arbitrary.
i=l,i#m
Proof. By Proposition 16.1 and our assumption, a(S)x = a(cm)Pmx = a(cm)x m .

Thus Equation (16.4) can be rewritten as (16.8). Clearly, Equation (16.8) has a
solution if and only if the condition (16.9) is satisfied. If it is the case, then any x
of the form (16.10) is a solution of (16.8), since there is no condition imposed on
Xi (j = 1, ... N; j :j:. m). •
Let e2rrijN (N E N, N 2: 2). Denote by VN(c) the Vandermonde determinant of
numbers ci (j = 1, ... , N), i.e.
(16.11) VN(c) = det (c ik )._

J-l, ... ,N,. k-O,
_ ... ,N-l .
Clearly, VN(c) :j:. 0, because ci :j:. c k if j :j:. k (j, k = 1, ... , N).

a(ci ) (j = 1, ... , N) are S-periodic, i.e. SaCci) = a(c j ) (j = 1, ... , N). Then
N
(16.12) ~(a) = VN(c) IT a(c j ),
j=l
hence Equation (16.4) has a unique solution if and only if a(c: i ) 1:- 0 for j = 1, ... , N.
This solution is of the form (16.7).
Sa(c: i ) - a(c:i)S = 0 (j = 1, ... , N). Then ~(a) is of the form (16.12).
Proof By our assumptions,
~(a)
= det '_ .. "N,,_

[c:ika(c:i)Sk] J-1, k-0, ... ,N-1
N N-1 N
= Vn(c:) [II a(c: j )] [II Sk] = VN(C:) II a(c: j ),
j=l k=O i=l
for S ... SN-1 = SN(N-l)/2 = I.
•
Sa(c:i ) - a(c:i)S E J (j = 1, ... , N), where J is a proper two-sided ideal in X. Then
~(a) is a sum of a determinant of the form (16.12) and an operator T E J.
Proof. All assumptions of Corollary 16.4 are satisfied in the quotient algebra XI J .
•
a(c: j ) E Xi (j = 1, ... , N). Then Equation (16.4) is equivalent to the system
(16.13) a(c:j)Xi = Yi (j = 1, ... , N)
of N independent equations. Moreover, if a(c: j ) E I(X) for j = 1, ... , N, then

the system (16.13) has a unique solution Xj = [a(c: i )]-lYi (j = 1, ... , N), so that
Equation (16.4) has a unique solution of the form
N N
(16.14) X= L [a(c: j )]-lYj= { L [a(c:j)]-l Pi }y.
j=l i=l
Proof By Proposition 16.1 and our assumptions, for k = 1, ... , N we have

N N
Yk = Pk[a(S)x] = Pk L a(c:j)PjX = L a(c:i)pkPjx = a(c:k)PkX = a(c:k)xk'
j=l j=O
We therefore have obtained the system (16.13). If a(c: j ) E I(X) for j = 1, ... , N,
then we find Xj = PjX = [a(c: j )t 1Yi> which yields Formula (16.14). •
Note 16.2. The condition a(c: j ) E I(X) (j = 1, ... , N) is the classical one for
equations with involutions of order N (without the assumption that they are mul-
tiplicative). The same property have solutions given by Formula (16.14). 0
We shall consider now algebras belonging to Lg(D) and multiplicative operators
commuting with D.
240 Chapter 16
Proposition 16.3. Suppose that X E Lg(D), S E M L(X), DS = SD on dom D

and (L,E) E G[O]. Then (Su,Sx) E graph 0 whenever (u,x) E graph 0 and
DLSu = SDLu, ESx = SEx.
Proof Let (u,x) E graph O. Then x = Lu and Du = uDx. Write x = SLu.
Then DSu = SDu = S(uDx) = (Su)DSx, which implies x = Sx and Dx = DSx.
Thus DLSu = Dx = DSLu = SDLu. Since u = Ex, Su = Ex, we get ESx =
Ex = Su = SEx. •
By an easy induction we obtain
Then
Note 16.3. The condition DS = SD on dom D implies that Sz E ker D whenever

z E ker D. Indeed, if z E ker D then DSz = SDz = O. 0
Corollary 16.8. Suppose that X E Lg(D), S E M L(X), DS = SD on dom D
and (L,E) E GR,dO] for an R E RD. Let F be an initial operator for D corre-
sponding to R. If Sz E ker D whenever z E ker D, then FSLu = O.
Proof By our assumption and Proposition 16.2, FSLu = FLSu = 0 for all
u E dom O. •
Proposition 16.4. Suppose that X E Lg(D), S E ML(X) is an involution of
order N, DS = SD on dom D and (L,E) E G[O]. Then
(16.16)
for all (u,x) E graph 0 (k = 0,1, ... , N -1).
Proof By our assumption, SN = I. By (16.15), for all (u, x) E graph 0,

k = O,l, ... ,N -1 we have SN-kDLSk = SN-kSkDLu = SNDLu = DLu and
SN-kESkx = SN-kSkEx = Ex. •
Theorem 16.2. Suppose that X E Lg(D), S E M L(X) is an involution of order
N, DS = SD on dom D and (L,E) E G[O]. Let
N-l
(16.17) A(S) =L akESk, where aO, .. ·,aN-l EX.
k=O
Let a(S) and ~(a) be defined by Formulae (16.1), (16.6), respectively. Let
~jk(a) be the minor determinant obtained by cancelling the jth row and the
(k+1)th column in ~(a) (j = 1, ... , N; k = 0, ... , N - 1). If ~(a) E I(X) and
[~(a)J-l ~jk(a)Sky E dom 0 (j = 1, ... , N; k = 0, ... , N - 1), then the equation
(16.18) A(S)x = y, Y E X,
has a unique solution

N N-1
(16.19) X = L{[~(a)t1 L L (_1)i+k+1~jk(a)Sky}.
j=l k=O
Proof. By our assumptions and (16.15), we find
N-1 N-1
A(S) =L akES k =L akS kE = a(S)E.
k=O k=O
By Corollary 16.1,
N N-1
Ex = [~(a)l-1 L L (_1)i+k+1 ~jk(a)Sky,
j=1 k=O
which implies (16.19), since x = LEx.
Example 16.1. Suppose that all assumptions of Proposition 16.3 are satisfied
•
and X is a Leibniz algebra with unit e. Then
N
SEx = [E(II ej Pjx)]1 IN.
j=1
Since X is a Leibniz algebra, by Formulae (16.3), (16.15) and Proposition 7.7(iii),
we get
N N
SEx = ESx = E(~ L ejpjx) = [E( II ejpjx)] liN.
j=l j=l
Let N = 2, i.e. S2 =I and e = -1. Then S = p+ - P- , I = p+ + P-. Hence
which implies
o
N = 2. Then e = -1 and
~(a) = I-Sa(-I)
a(-I) a(1) I
Sa(l) = a(-I)Sa(l) + a(I)Sa(-I) E J(X).
Observe that a(±I) = ao ± al. By Theorem 16.2, we find
x = L{[a( -1)Sa(l) + a(I)Sa(-I)t1[ySa(l) - a(I)Sy]}.

242 Chapter 16
If X is a Leibniz algebra then we get
x = L[ySa(1) - a(1)Sy]- L[a(-1)Sa(1) + a(1)Sa(-1)].

o
Recall that in Chapter 15 devoted to periodic problems we have considered involu-
tions of order N commuting with the operator D. We shall pass now to a different
kind of problems.
Definition 16.1. Let X E Lg(D). Then S E Lo(X) is said to be a Carleman
operator of order N E N, N 2:: 2) if S E M L(X) is an involution of order N such
that DS = dSD on dom D, where d E J(X) \ {e}. If N = 2 then S is called briefly
a Carleman operator. 0
Note 16.4. The case d = e corresponds to the already examined case of operators
commuting with D. However, we shall see later that d ¥- e leads to problems of a
quite different nature (cf. also Example 2.8, where we have considered an operator
A E ML(X) such that DA = dAD, dE J(X) \ {e}, without the assumption that
A is an involution of order N). 0
Definition 16.1 and Corollary 2.13 implies
Theorem 16.3. Suppose that X E Lg(D), (L, E) E G[n] and S is a Carleman
operator of order N. Then (Su, Sx) E graph n whenever (u, x) E graph nand
(16.20) LSu = SLu, ESx = SEx for (u,x) E graph n.
Proof. Let (u,x) E graph n, (L,E) E G[n]. Then x = Lu and Du = uDx. By

our assumptions, since S is multiplicative, we find DSu = dSDu = dS(uDx) =
d(Su)SDx = (Su)dSDx = (Su)DSx, which implies SLu = Sx = LSu. Since
u = Ex, we have SEx = Su = ELSu = ESLu = ESx. •
By an easy induction, taking into account the property SN = J, we get
Then (Sku, Skx) E graph n whenever (u, x) E graph n (k = 1, ... , N - 1) and
(16.21) LSku = Sk Lu, ESkx = Sk Ex for (u, x) E graph n, k = 1, ... , N -1.

Then S is invertible and (S-1 U ,S-1 X ) E graph n whenever (u,x) E graph n.
Proof. By our assumption, S is an involution of order N. Thus S is invertible
and S-1 = SN-1. The rest of the proof is an immediate consequence of Corollary
16.9. •
Definition 16.2. For an arbitrary function h defined on a set n and with values
in n we write
hn+1 (t) = h[hn(t)] for n E N and hI (t) = t.

A function h defined on a set 0 and with values in 0 satisfies the Carleman

condition of order m if there is an integer m ~ 2 such that
(16.22) hm=:t on 0, hk(t)~t fork=O,l, ... ,m-l.
If m = 2 then we say briefly that h satisfies the Carleman condition. A function

satisfying the Carleman condition of order m is said to be a Carleman function of
order m. A Carleman function of order 2 is called briefly a Carleman function. 0
In other words, the function h generates a cyclic group of automorphisms of the
set 0 of order m.
Note 16.5. A Hilbert boundary value problem for analytic functions with trans-
formed argument was firstly considered by CARLEMAN (cf. Ca[l], Ca[2]). Thus
in all papers concerning the corresponding singular integral equations Condition
(16.22) (with m = 2) is called the Carleman condition and these equations are said
to be of Carleman type. The identity (16.22) considered as a functional equation
is called the Babbage equation (cf. BABBAGE B[l], KUCZMA Ku[l)). F'unctional-
differential equations of Carleman type were examined, in particular, in PR[4],
PR[6]. Often Carleman functions are called Carleman shifts. 0
Example 16.3. (cf. PR[4]) It is easy to check that the following functions satisfy
the Carleman condition:
(a) h(t) = -t on IR
(a reflection of argument, cf. Example 16.4);
(b) h(t) = ~ on (0, (0);
-at for t > 0;

(c) h(t) = { 0 for t = 0; where a E IR \ {O} is arbitrary;
_it for t < 0;
a
(d) h(t) = { C 1/ k for t ~ 1; where kEN is arbitrary;

t- k for 0 < t $ 1;
(e) h(t) =c - t on IR, where c E IR is arbitrary.
It should be pointed out that in examples (a)-(e) there is only one fixed-point a
of the function h, i.e. a point a such that h(a) = a. Namely, a = 0 in Examples
(a) and (c), a = 1 in Examples (b) and (d), a = c/2 in Example (e).
The following function satisfies the Carleman condition of order m:
(f) h(t) = c:kt for t E C, where c: = e27ri / m

244 Chapter 16
(a rotation of argument, cf. Example 16.5). o

Definitions 16.1 and 16.2 together imply
Theorem 16.4. Let X = C(O), where either 0 C IR or 0 C C, and let D = -It.
Let X E Lg(D). Then any differentiable Carleman function h of order m ~ 2 on
o such that h'(t) :J 0 for all tEO induces on X a Carleman operator of order m
defined by the equality
(16.23) (Sx)(t) = x[h(t)) for tEO (with d = h' E leX))
Proof. By definition (16.23), the operator S is multiplicative and (smx)(t) =

hm(t) = t for all tEO, Le. S is an involution of order m. Moreover, by our
assumptions,
(DSx)(t) = :tX[h(t)) = h'(t)x'[h(t)) = d(SDx)(t),

where d = h' E leX).
Now we shall examine properties of Carleman functions of order m.
•
Lemma 16.1. (cf. PR[4)) Every Carleman function of order m ~ 2 on a set 0 is
one-to-one.
Proof. Let hm(t) == t on 0 and let t,s E 0, t :J s. Suppose that h(t) = h(s).
Hence h2(t) = h[h(t)) = h[h(s)] = h2(S). After m - 1 steps we get
which contradicts our assumption that t:J s. Then h(t) :J h(s). •

Lemma 16.2. (cf. PR[4)) If h is a Carleman function on IR with a fixed-point a
then the function ho(t) = h(t + a) - a is a Carleman function with the fixed-point
o (and conversely, h(t) = ho(t - a) + a is a Carleman function with a fixed-point
a, provided that 0 is a fixed-point of ho.)
Proof. Let h[h(t)] == t on IR and let h(a) = a. Then h(O) = h(a) - a = 0 and for
every t E IR
ho[ho(t)) = ho[h(t+a) -a] = h[h(t+a) -a+a]-a = h[h(t+a)]-a = t+a-a = t .
Lemma 16.3. (cf. PR[4)) Let h be a continuous Carleman function on IR. Then
•
(i) h is decreasing;
(ii) the derivative h' exists almost everywhere and h(t) = J: h'(s)ds;
(iii) h has a unique fixed-point ah;
(iv) if h is odd then h(t) = -to
Proof. (i) By Lemma 16.1, h is one-to-one, hence strictly monotone. Suppose that
h is increasing. Since, by definition, h(t) t
t, there is a to such that h(to) ::j:. to.
Then either h(to) > to or h(to) < to. The first inequality implies that ho =
h[h(to)) > h(to) > to, which contradicts our assumption. A similar proof the
second inequality. Hence h is decreasing.
(ii) Since h is continuous and monotone, its derivative exists almost everywhere
and h(t) = J: h'(s)ds, as it is well-known from Calculus of one real variable.
(iii) Our assumptions and (i) immediately imply that
(16.24) lim
t-++oo
h(t) = -00, lim
t-+-oo
h(t) = +00.
Since h is continuous and decreasing on ~ and Formulae (16.24) hold, there exists
a unique point ah such that h(ah) - ah = O.
(iv) Suppose now that h is odd, i.e. h( -t) = -h(t) for t E R Then h(O) = -h(O),
which implies h(O) = O. Thus 0 is the unique fixed-point of h. Suppose that
h(t) ::j:. -to If h(t) > -t then t = h[h(t)] < h( -t) = -h(t), since h is a decreasing
function. This implies that -t > h(t) > -t, a contradiction. For the inequality
h(t) < -t we get a contradiction in a similar manner. Hence h(t) = -to •
An immediate consequence of Lemmas 16.2 and 16.3 is
Lemma 16.4. (d. PR[4]) Every continuous Carleman function on ~ has the
graph symmetric with the respect to the line x = t.
In the sequel we will denote the superposition of two functions x and y by x 0 y,
i.e. by definition,
(x 0 y)(t) = x[y(t)],
provided that this superposition exist.
Theorem 16.5. (d. PR[4]) Every continuous Carleman function on ~ with
a fixed-point a is the superposition of a homeomorphism H of ~ onto itself, a
reflection and of the inverse homeomorphism H-I,
(16.25) h = H- I 0 (-H), where H = hI 0 h2'
hI is a homeomorphism of~ onto itself such that hI (0) = 0 and h2(t) = t - a is

a shift. Conversely, every function of the form (16.25) is a continuous Carleman
function on ~ with the fixed-point a.
Proof. Let h be of the form (16.24). Since h;;l(t) = t + a, we can write
For an arbitrary homeomorphism hI with the property hI (0) = 0 the function

246 Chapter 16
is a continuous Carleman function with the fixed point O. Indeed,
and ho(O) = hII[-hl(O)] = hII(O) = O. By Lemma 16.2, we conclude that h is a

continuous Carleman function with the fixed-point a.
Conversely, suppose that h is a continuous Carleman function on JR with the fixed-
point a. Let hl(t) = t - ho(t), where ho(t) = h(t + a) - a. By Lemma 16.3,
ho, as a continuous Carleman function, is decreasing, hence the function hI is
increasing. Since hI is also continuous, it is a homeomorphism of JR onto itself and
hl (0) = -ho(O) = O. Moreover,
hl[ho(t)] = ho(t) - ho[ho(t)] = ho(t) - t = -hl(t).

Hence ho(t) = hII[-hl(t)] and
h(t) = hII[-hl (t - a)] + a = h21 hIl [-(hI h2)(t)],
0 0 0
where h2(t) = t - a. Writing H = hI h2, we obtain the required formula (16.25)

0
for h. •
Corollary 16.11. (cf. PR[4]) Every continuous Carleman function on an open
arc L is of the form: h(t) = H-l[-H(t)], where H = hI 0 h2 0 h3, h3 is a homeo-
morphism L onto JR, h2 (s) = s - a, a is the fixed-point of the function h3 0 h 0 h3 1 ,
and hI is a homeomorphism oflR onto itself such that hI (0) = O. Conversely, every
function of this form is a Carleman function on L with the fixed-point h3 l (a).
Proof Let h3 be a homeomorphism of the arc L onto JR and let s = h3(t). Then
the function
(16.26) Hl(s) = (h3 ohoh3 1)(S) = (h3 og)(t), s E JR,
is a Carleman function on JR, and is therefore of the form
(16.27)
a is the fixed-point of Hl and hl is a homeomorphism of JR into itself. Formulae

(16.26) and (16.27) together imply that h(t) = H-I [-H(t)], where H = hl oh2oh 3.
Since h2(a) = a - a = 0, we have h21(0) = O. Writing b = h3 1 (a), we find
h(b)
= h3 1 0 h21 0 h I 1 { - h1 0 h2 0 h3[h3 1(a)]}
= hI1 0 h21 0 h3 1[-(h1 0 h2)(a)]
=h3 1 h21 h11[-h 1(0)] = (h3 1
0 0 0 h21 0 hll)(O)
=(h3 1 0 h21)(0) = h3 1(a) = b.
Hence b is the fixed-point of h.
•
Example 16.4. It is easy to check that the function
11" b+a
x(t) = tan b _ a (t - -2-)
is a homeomorphism of the open interval (a, b) onto JR.. The function x(t) = In tis
a homeomorphism of the positive open half-axis onto JR.. The function x(t) = -t
is a homeomorphism of (-00,0) onto (0, +(0).
By definition, each open arc on a real or complex plane( as homeomorphic with an
open interval) is homeomorphic with JR.. 0
Theorem 16.6. (cf. PR[4]) Continuous functions satisfying Carleman condition
of order m > 2 on JR. do not exist.
Proof. Suppose that there exists a function h such that h(t) ~ t and
(16.28) hm(t) == t for an m > 2 on R

Let m be an odd number. If h is decreasing then we have
t < s {:} h(t) > h(s) {:} ... {:} hm-1(t) < hm-1(s) {:} t = hm(t) > hm(s) = s,
a contradiction. If h is increasing then there is a to such that h(to) > to and from
the identity (16.27) we find
again a contradiction. Thus m cannot be odd. Let m = 2k, k > 1. Since (16.28)
holds, hk is a Carleman function and therefore decreasing. •
Corollary 16.12. (cf. PR[4]) Continuous functions satisfying the Carleman
condition of order m > 2 on an open arc L do not exist.
Proof. Suppose that hm(t) == t on L for an m > 2. Let H be a homeomorphism
of L onto JR. of the type described in Corollary 16.12. We define on JR. the function
h* = H 0 h 0 H- 1 of s = H(t) E JR., tEL. Then
h; = h* 0 h* = (H 0 h 0 H- 1 ) 0 (H 0 h 0 H- 1) = H 0 h 0 h 0 H- 1 = H 0 h2 0 H- 1,
h:n = h* 0 h:n-l = (H 0 h 0 H- 1 ) 0 (H 0 hm - 1 0 H- 1) = ... = H 0 h m 0 H- 1

and
h:n(s) = (H 0 h m 0 H-l)(S) = (H 0 hm)(t) = H(t) = s for s E JR.,
which contradicts Theorem 16.6.

We turn next to the question of differentiability of Carleman functions. An im-
•
mediate consequence of Definition 16.2 is
248 Chapter 16
Proposition 16.5. Let L be either JR. or an open interval or the positive or

negative half-axis. If h is a continuous Carleman function on L then the derivative
of the function h 0 h exists almost everywhere and (h 0 h)' = 1 a.e..
Proposition 16.6. (cf. PR[4]) If h is a differentiable Carleman function on JR.
then h'(t)h'[h(t)] == 1 and h'(t) i- 0 for all t E R
Proof. By Proposition 16.5, 1 = {h[h(t)]}' = h'(t)h'[h(t)] for all t E R This
implies that h' (t) i- 0 for t E JR.. •
Proposition 16.7. If h is a Carleman function on an interval (a, b) such that
h(t) = H-l[-H(t)] and H is a diffeomorphism of (a, b) onto JR., then
(i) the derivative h' exists and is continuous at each point t E (a, b);
(ii) h'(t)h'[h(t)] == 1;
(iii) h'(t) i- 0 for all t E (a, b).
Proof. (i) Let H be a diffeomorphism of (a, b) onto R Since both derivatives H'
and (H-l)' exist and are continuous by assumption, it follows that h' also exists
and is continuous.
(ii) Since H' i- 0, we have
h'(t)
= {H-1[-H(t)]}' = H'[H~l(S)]ls=-H(t) [-H(t)]

1 , H'(t)
H'{H-l[-H(t)]} [-H (t)] = - H'[h(t)]"
Therefore at each point t E (a, b)
" H'(t) [ H'[h(t)]] H'(t) H'[h(t)]

h (t)h [h(t)] = - H'[h'(t)]· - H'{h[h(t)]} = H'[h(t)]· h'(t) = 1.
(iii) By (ii), h'(t) i- 0 for all t E (a, b). •

Example 16.5. Suppose that X E Lg(D), (L, E) E G[flJ, S E M L(X) is an
involution of order N ~ 2, c = e2rri / N and a permuting operator, i.e. DS = cSD
on dom D . Then S is a Carleman operator of order N. So that (cSu,cSx) E
graph fl whenever (u,x) E graph fl and L(Sku) = SkLu, E(Skx) = SkEx for all
(u,x) E graph fl, k = 0, ... ,N -1. 0
ft
Example 16.6. Suppose that X = C(JR.), D = and (Sx)(t) = x( -t) for x EX,
t E JR.. Clearly, S is the operator of reflection and a multiplicative involution
for S2 = I. Hence here N = 2 and c = -1. Since (DSx)(t) = [x( -t)]' =
-x'(-t) = -(SDx)(t) , S is a permuting operator, hence a Carleman operator.
In this particular case we have d = -e E I(X). The equality DS + SD = 0 on
dom D shows that Sis anticommuting with D. 0
Example 16.7. Suppose that all assumptions of Example 16.5 are satisfied and
N = 2, i.e. e = -1. Then the operator D and S are anticommuting (cf. Example
16.6). Consider the equation
(16.29) A(S)x = y, where A(S) = aoE + alES, ao,abY E X.
By Theorem 16.3, this equation can be rewritten as aoEx - alSEx = y. Putting

ah = ao and a~ = -aI, we find a'(±I) = a(=Fl) and
, _I
~(a) -
a(l)
-Sa(l) Sa(-I) 1-
a( -1) - a(I)Sa(-I) + a(-I)Sa(l) -
- ~(a),
where ~(a) is determined by (16.4). If ~(a) E leX) the a unique solution of

Equation (16.29) is
x = LEx = L{[~(a)rl[ySa(-I) +a(-I)SyJ}.
o
Example 16.8. Let 0 C C be a domain invariant under the rotation through
the angle 2;/, i.e. eO = 0, where e = e27ri / N (N EN). Consider the algebra
It
X = C(O), the operator D = and the operator S of rotation through the angle
2;/: (Sx)(t) = X(et) for x E X. Clearly, S is a (multiplicative) permuting operator
for N 2: 2, since (DSx)(t) = [x(et)l' = eX'(et) = e(SDx)(t) (cf. Example 16.5). 0
Consider now equations of the Carleman type.
Theorem 16.7. Suppose that X E Lg(D) and S is a Carleman operator of
order N 2:: 2 on X, i.e. S E ML(X) is an involution of order N 2:: 2 on X and
DS = dSD on dom D, where d E leX) \ {e}. Let
N-I
a(S) = L akS k , where ao, ... , aN-l EX.
k=O
Then the equation
(16.30) Dx = a(S)x + y, y EX
is equivalent to the system of N equations
(16.31) DSjx = aj(S)x + (dS)jy (j = 0, ... ,N -1),
N-I
(16.32) where aj(S) =L aj,kSk and
k=O
(dS)jak-o if k ~ N -1;
aj,k ={ j J
(dS) ak-j-N if k 2:: Nj
(j,k == O, ... ,N -1).
250 Chapter 16
Proof By an easy induction, taking into account the property SN = I, we prove

that
(16.33) DSj = (dS)jD on dom D (j = 0, ... ,N -1).
Clearly, (dS)j E M L(X) (j = 0, ... , N - 1. Then for k = 0, ... , N - 1, again by an

easy induction, we obtain
Acting on both sides of Equation (16.30) by operators dS, ... ,(dS)N-l and using
Equalities (16.34) we obtain the system (16.31). Conversely, if we put in (16.31)
j = 0 then we get Equation (16.30). •
An immediate consequence of this Theorem is
aj,k (j, k = 0, ... , N - 1) are determined by Formulae (16.32). If the determinant
(16.35) ~(a) = det(aj,k)j,k=O, ... ,N_l
of the system (16.31) is invertible, then Equation (16.30) has a unique solution,
which is of the form
N-l
(16.36) X = [~(a)rl L (_1)k+1~k(a)(dS)ky,
k=O
where by ~k(a) we denote the minor determinant of ~(a) obtained by cancelling

the 1st column and the kth row.
Example 16.9. If all assumptions of Theorem 16.7 are satisfied, N = 2 and
then the equation

Dx = aox + a1Sx + y
has a unique solution
o
Theorem 16.8. Suppose that X E Lg(D), S is a Carleman operator on X, i.e.
S E ML(X), S2 = I and DS = dSD on dom D, where d E I(X) \ {O}, and
there is an initial operator F corresponding to an R E RD such that FS = F. If
x E dom D2 is a solution of the equation
(16.37) Dx = Sx + y, y E dom D
then x satisfies the equation
(16.38)
with the condition
(16.39) FDx = Fx+Fy.

Conversely, if x satisfies Equation (16.38) and Condition (16.39) and the operator
1- R 2 d is invertible, then x is a solution of Equation (16.37).
Proof Necessity. Let x E dom D2 be a solution of Equation (16.37). Then,
by our assumption, Condition (16.39) is satisfied. Indeed, FDx = F(Sx + y) =
FSx + Fy = Fx + Fy. Moreover,
D 2x = D(Dx)
= D(Sx + y) = DSx + Dy = dSDx + Dy = dS(Sx + y) + Dy
=dS 2x + dSy + Dy = dx + dSy + Dy,
i.e. x is a solution of Equation (16.38).
Sufficiency. Suppose that x is a solution of Equation (16.38) satisfying Condition
(16.39) and the operator I - R 2d is invertible. We write u = Dx - Sx - y and
we show that u = 0, i.e. Equation (16.37) is satisfied. Observe that u satisfies the
equation Du + dSu = 0 with the condition Fu = O. Indeed,
Du = D(Dx - Sx - y) =
D 2x - DSx - Dy = D 2x - dSDx - Dy
=dx + dSy - dSDx = dS(Sx +y - Dx) = -dSu,
i.e. Du + dSu = O. Moreover,
Fu = F(Dx - Sx - y) = FDx - FSx - Fy = FDx - Fx - Fy = O.

Write v = Su, i.e. u = Sv. We have
dS(Dv + Sv) = dSDv + dS 2v = dSDv + dv = DSv + dv = Du + dSu = O.

Since d E I(X), we conclude that Dv + Sv = O. Moreover, Fv = FSu = 0 and
FDv = FDv = -FSv = _FS2u = -Fu = O. Since D 2v = -DSv = -dSDv =
-dS(-Sv) = dS 2v = dv, by the Taylor Formula, we find v = Fv + RFDv +
R2 D 2v = R2 D 2v = R 2dv. Hence (I - R 2 d)v = 0, which implies v = 0, as the
operator I - R 2d is invertible. Thus Dx - Sx - y = u = Sv = 0, i.e. Equation
(16.37) is satisfied. •
Note 16.6. Theorem 16.8 has been proved for functional-differential equations
of the Carleman type (cf. PR[4], PR[6j, PR[8]). Clearly, the operator 1+ R 2 d in
the case of spaces of continuous functions and the operator D = -it
is invertible,
252 Chapter 16
but this fact was not used in an explicit way. Particular properties of functions
considered there assure that Equation (16.37) is equivalent to Equation (16.38)
together with Condition (16.39). 0
Corollary 16.14. Suppose that all assumptions of Theorem 16.8 are satisfied,
(L 2, E 2) E G[02]' 1- R 2d is invertible, R 2d E dom 0- 1 and y = O. Then
(16.40) x E ker (D - S) if and only if x = E2(R 2 d + z),

where z E ker D2 is arbitrary.
In order to solve nonhomogeneous equations, i.e. equations with y ¥- 0, we need
some additional informations about the structure of the algebra under considera-
tion, in particular, about the product rule.
Consider now equations of order M ~ 2.
Theorem 16.9. Suppose that X E Lg(D), S is a Carleman operator on X, i.e.
S E ML(X), S2 = I and DS = dSD on dom D, where dE I(X) \ {O}, there
is an initial operator F corresponding to an R E 'RD such that F Dk S = F Dk
(k = 0,1, ... ,M -1; M ~ 2) and
M
(16.41) A(D) =L akDk, where ao, ... , aM E X,
k=O
M
(16.42) A(D) = L (Sak)(Dd- 1 )kd.
k=O
If x E dom D2M is a solution of the equation
(16.43) A(D)x = Sx + y, y E dom DN-l,
then x satisfies the equation
(16.44) A(D)dA(D)x - dx = dSy + A(D)y
with the condition
(16.45) FDk A(D)x = FDkx + FDky (k = 0, 1, ... , M -1).

Conversely, if x satisfies Equation (16.44) with Condition (16.45) and the operator
M
(16.46) A(I,R) = L RM-kak
k=O
is invertible, then then x is a solution of Equation (16.41).

Proof. Necessity. Let x E dom D2M be a solution of Equation (16.31). Then, by
our assumption, Condition (16.45) is satisfied. Indeed, for k = 0,1, ... , M - 1, we
have FDk A(D)x = FDkSx + FDky = FDkx + FDky. Since d E I{X), by an
easy induction we prove
(16.47)
(16.48) dSA{D) = A(D)dS.
Indeed, since S is multiplicative and Formulae (16.47) hold, for all x E dom DM
we have
dSa(D)x
M M
=dS L ak Dkx = d(Sao)Sx + L {Sak)dSDkx
k=O k=l
M
=d(Sao)Sx + L (Sak)(Dd- 1 )kdSx = A(D)dSx.
k=l
Identity (16.48) immediately implies the following identity:
(16.49) A(D) = dSA(D)Sd- 1 .

Indeed, since S2 = I, we have A(D) = A(D)dSSd- 1 = dSA(D)Sd- 1 •
Since x is a solution of Equation (16.41) and (16.49) holds, we get
dx + dSy = dS(Sx + y) =
dSa(D)x = A(D)Sx = A(D)[A(D)x - y]
= A(D)dA(D)x + dx - A(D)y,
which implies Equation (16.44).
Sufficiency. Suppose that x satisfies Equation (16.44) and Conditions (16.45). By
(16.48), we can rewrite Equation (16.44) as follows:
dSA(D)SA(D)x - dx = dSy + dSA(D)Sy,

i.e.
dSA(D)SA(D)x - dS 2x = dSy + dSA(D)Sy,
and dS[A(D)SA(D)x - Sx] = dS[y + A(D)Sy]. Since d E I(X) and S2 = I, we
obtain the following equivalent equation:
(16.50) A(D)SA(D)x = Sx + y + A(D)Sy = A(D)x + A(D)Sy.

254 Chapter 16
Write u = A(D)x - Sx - y. By our assumption, for k = 0,1, ... , M
From Equation (16.49) we find
0= A(D)[SA(D)x - x - Sy] = A(D)S[A(D)x - Sx - y] = A(D)Su.

Let v = Su, i.e. u = Sv. Then A(D)v = 0. Recall that the operator A(D) satisfies
the following identity:
M-l
(16.51) A(D)(I - L RkFDk) = A(I,R)DM on dom DM.
k=O
(cf. PR[8], Formula (3.2.13)). By (16.45), we have FDkv

0,1, ... , M - 1) and
= FDkSu = ° (k =
M-l
A(I,R)DMv = A(D)v - 2: RkFDkv = A(D)v = 0.
k=O
Since the operator A(I, R) is assumed to be invertible, we conclude that DM v = 0.

Thus, by the Taylor Formula,
M-l
V = 2: Rk F Dkv + RM DM V = 0.
k=O
This implies that A(D)x - Sx - y = u = Sv = 0, i.e. x satisfies Equation (16.41) .
•
Sak = ak (k = 0,1, ... , M). Then A(D) = A(Dd- 1).
Proof. By definitions (16.41), (16.42) and our assumption,
M M
A(D) =L (Sak)(Dd-1)k =L ak(Dd-1)k = A(Dd- 1).
k=O k=O
•
Sak = akS (k = 0,1, ... , M). Then SA (D) = A(D) and A(D) = dA(D)Sd- l .
Proof. By definitions (16.41, (16.42) and our assumption,
M M M M
SA(D) =S L ak Dk =L (Sak)SD k = L ak S2 Dk =L ak Dk = A(D).
k=O k=O k=O k=O
This, and (16.49) together imply that A(D) = dSA(D)Sd- 1 = dA(D)Sd- l . •
Using Theorem 16.9 (Theorem 16.8, respectively) and Theorem 16.3, we may solve
equations of the form
A(D)x = Sx + H(Ex, ESx), A(D) = Sx + H(Lx, LSx)

in Lg(D), where (L,E) E G[O], H is a given nonlinear mapping and S is a Car-
leman operator, by a reduction to equations without S. Similarly, using Theorem
16.7 and Corollary 16.9, we may solve equations of the form
Dx = a(S)x + H(Ex, ... , ES N - 1 ), Dx = a(S) + H(Lx, ... , LS N - 1 ),

with a Carleman operator of order N 2:: 2,
Note 16.7. Suppose that A(D) is defined by Formula (16.41), D E R(X) and F
is an initial operator for D corresponding to an R E RD. Let H be a mapping (in
general, nonlinear) of X into itself. Consider the equation
(16.52) A(D)x = H(x)

together with the initial conditions
(16.53) FDkx=xkEkerD (k=O,1, ... ,M-1).
Write
M-l
XM =L Rkxk.
k=O
Clearly, XM E ker DM. Suppose that the operator A(I, R) is defined by Formula
(16.46). Formula (16.51) implies that
L
M-l
H(x) = A(D)x = A(I, R)DM x + A(D) Rk F Dkx = A(I, R)DM x + XM.
k=O
If the operator A(I, R) defined is invertible then
We therefore conclude that Equation (16.52) is equivalent to the equation
(16.54)
which in the classical case is an integral equation (cf. PR[8]). o

Clearly, this chapter does not exhaust questions connected with Carleman opera-
tors of arbitrary order and, in general, with multiplicative involutions. It should
be treated rather as a guide into this subject.
CHAPTER 17
REMARKS ON THE FRACTIONAL CALCULUS
In this section we will show how one can introduce elements of the fractional
calculus (cf. for instance, KIRYAKOVA Kir[l]), by means of antilogarithms. We
begin with
Proposition 17.1. Suppose that X is a Leibniz D-algebra with unit e and F is
a multiplicative initial operator for D corresponding to an R E RD and 9 = Re.
Then
(17.1) D m 9n = (n _n!m)!g n-m f ar a II n, m E ~T
l~O, n ~ m,
(17.2) R m 9n =( n! )Ign+m f araII n,mEl~o·

~T
n+m.
Proof. We have proved that Dgn = ngn-1 for all No (cf. Corollary 7.4(i), also
PR[8]). By an easy induction, we get (17.1).
Since F is multiplicative, X is a Leibniz D-algebra, by (7.27), Rne = ~ for all
n E N. This implies that
nl
Rmgn = Rm(n!Rne) = n!Rn+me = ( . )Ign+m far all m,n E N.
n+m.
Note 17.1. Suppose that all assumptions of Proposition 17.1 are satisfied and
•
9 E J(X). Then, by Proposition 7.9(iii), Fg- 1 does not exists, since F E M L(X).
o
Proposition 17.2. Suppose that IF = JR, X is a complete linear metric space,
X E L(D) and has the unit e E dom 0, (L, E) E G[OJ, F is a multiplicative initial
operator for D corresponding to an R E RD, 9 = Re E J(X) and aLg E dom 0- 1
for every a E R Then for all a E 1I4
(i) gO = Eog = E(aLg);
(ii) Dgo = ag o - 1;
(iii) 9 E £b(X) implies gO = eoLg ;
(iv) Dmgo = r(~(~t~~)go-m for all mEN, where r is the Euler function;
(v) g-O = Eog = E( -aLg) = [E(ag)]-1 = (gO)-1.
Proof. Recall again that 9 E J(X) if and only if 9 E dom O. Assertions (i)-(iii)
and (v) are immediate consequences of our definitions and assumptions. Point (iv)
is proved by an easy induction. •
Definition 17.1. Suppose that IF = JR, X is a complete linear metric space,
X E L(D) and has the unit e E dom 0, (L, E) E G[O], F is a multiplicative initial
256
Fractional calculus 257
operator for D corresponding to an R E RD, 9 = Re E leX) and aLg E dom n- 1

for every a E 114. Then for all a, (3 E ll4
(17.3) D/3 <> _ f(a + 1) <>-/3. Ri3 <> - f(a + 1) <>+/3

9 - f(a+l-{3)g , 9 - f(a + 1 + (3)g .
D
Corollary 17.1. Suppose that all conditions of Definition 17.1 are satisfied. Then
DV <> = {0 if 1/ > a, for all a,1/ E ll4 U {o}.

9 R-v g <> if 1/ ~ a,
Write
(17.5) peg) = lin {gP : p E Z \ {O} } for 9 E leX).
Clearly, g-P E leX) whenever 9 E leX), p E Z.

Corollary 17.2. Suppose that all conditions of Definition 17.1 are satisfied and
the set peg) is defined by (17.5). Then
(17.6)
Write
(17.7) Po(g) = lin {gP : p E Q \ {O}; Lg} for 9 E leX).
the set Po (g) is defined by (17.7). Then
(17.8)
Note 17.2. By Corollaries 17.2, 17.3, the operator D is invertible on the sets
peg), Po(g), respectively, so that ker D = {O}. On the other hand, we know that
in a Leibniz D-algebra with unit we have ker D =j:. {O} (cf. Theorem 4.3). It is
so, since the unit e does not belong to these sets. Indeed, gO = e and we consider
integers (rational numbers, respectively) different than zero. If F is an initial
operator for D corresponding to Rand {3 = 1 then we get for all a E lR:
Fg<> = (1 - RD)g<>
_ <> _ R r(a + 1) <>-1 _ <> _ f(a + 1) R <>-1
- 9 f(a + 1 - l)g - 9 f(a) 9
_ <> _ f(a + 1 r(a) <>-1+1 _ <> _ <> _ 0
-g f(a) f(a-l+l+l)g -g 9 - .
258 Chapter 17
On the other hand, Fe = (I - RD)e = e. Thus e E ker D.

In order to extend this construction to complexes we introduce the following
Definition 17.2. Suppose that IF = C, X is a complete linear metric space,
X E L(D) and has the unit e, (L, E) E G[OJ, F is a multiplicative initial operator
for D corresponding to an R E RD, g = Re E I(X) and aLg E dom 0- 1 for
every a E R Suppose, moreover, that dom 0- 1 is symmetric, i.e. -x E dom 0- 1
whenever x E dom 0- 1 (cf. Condition [C]n in Chapter 9). Then for all a = iJL,
JL E JR, we put gip. = Eip.g = E(iJLLg). By our assumption, -iJLLg E dom 0- 1
whenever iJLLg E dom 0- 1 (cf. Definition 12.1 and Example 12.2). 0
gC>. E dom 0 whenever a:j:. 0. Then
Df3 gC>. E dom 0, whenever (3 < a;
Rf3 gC>. E dom 0, whenever (3 ~ - a.
Example 17.1. Suppose that X = C(JR), D = ft, R = J~. It is well-known that
t (t s)k-1
(17.9) (RkX)(t) = Jto (; _ I)! x(s)ds for x EX (k EN).
This formula motivated at least one way for introducing the fractional calculus.
Namely, the so-called Riemann-Liouville fractional integral of order a is defined
as follows:
[t (t S )C>.-1
(17.20) (RC>.x)(t) = Jto f(a) x(s)ds for x E X (Re a E 114)·
The integral (17.20) coincides with the integral (17.19) for a = kEN. We admit
RO = I. Thus we may write
In particular, DC>. Ra. x = x for Re a E 114 and all x EX, i.e. DC>. RC>. = I. Putting
a = 0, we get for x E dom Dk = Ck[O, 1]
o
APPENDIX
FUNCTIONAL SHIFTS
ZBIGNIEW BINDERMAN
A1. Functions of a right invertible operator.
We shall preserve the denotations admitted in the preceding chapters. Let H(n)
be the space of all functions analytic on a set n ~ C. In our subsequent considera-
tions, the field of scalars is always IF = Co Denote by lK( aj r) disks on the complex
plane, i.e.
lK(ajr) = {h E C: Ih - al < r}, 0 < r ~ +00.
For the sake of brevity, we shall write JK,. = lK(O, r).
In this section lK will stand either for the unit disk lKl = lK(Oj 1) or for the complex
plane C.
Note that, by our assumption, there exists always a ..\ E lK such that
ker(D - ..\1) :j: 0 whenever D E R(X).
For instance, 0 E lK.

Suppose that a function f E H (lK) has the following expansion:
L
00
f(h) = akh k for all hE lK.

k=O
Let X be a complete linear metric space and let D E R(X). Write
(A1.l) SI(D) = S}O)(D),
Sr(D) = n
nENo
S}n) (D).
Proposition A1.1. Suppose that X is a complete linear metric space, D E R(X)

and ker D :j: {OJ. Then
259
260 Functional shifts
(i) the space S of all D-polynomials is a subset of Sr(D);

(ii) ker(D - >.1) c Sr(D) for all >. E K
Proof. (i) Let xES, h E lK be arbitrarily fixed. By our assumptions, there are
REnD, m E No, Zk E ker D (k = 0, 1, ... ) such that
Then for n E No we find
=L
00
Y akhk Dk+nx =
k=O
m m
L L L L
00 00
= akh k Dk+n Rj Zj = akhk Dk+ n Rj Zj

k=O j=O k=O j=O
for n > mj
m-n-k
L
p=o
for m ~ n.
Hence
for n > mj
(A1.2)
for m ~ n,
where z~ = E;=~n-k aphPzk+n+p E ker D (k = 0,1, ... ,m - n). This implies that
the series E~o akhk Dk+nx is convergent for all n E No, i.e. x E Sr(D).
(ii) Let>. E lK and let x E ker(D - >'1), h E lK be arbitrarily fixed. By our
assumption, Dx = >.x. The equalities Dm x = >.mx for all m E No (obtained by
an easy induction) together imply for all n E No
00 00 00
(A1.3) L akh k Dk+ n x =L akh k>.k+nx = >.n[L ak(>.h)k]x = >.n f(>.h)x.

k=O k=O k=O
Write
•
00
(A1.4) f(hD)x = L ak hk Dkx for x E S,(D) :::> S,(D), hE K

k=O
We shall need the following obvious

Appendix 261
Lemma ALl. Suppose that all assumptions of Proposition Al.I are satisfied
and D is closed. Let h E l!{ be arbitrarily fixed. Then
L ak hk Dkx = L
00 00
(i) D ak hk Dk+l x for all xES f(D) n SY) (D);

k=O k=O
L L
00 00
(ii) D n akh kDkx = ak hk Dk+ nx for all n EN, x E Sr(D);

k=O k=O
(iii) [Dn f(hD)](Sr(D)) c Sr(D).
Formula (Al.2) and Lemma Al.1 imply, respectively,
Corollary ALl. Suppose that all assumptions of Proposition AI.I are satisfied.
Let xES, h E l!{ be arbitrarily fixed. Then
L
00
(i) ak hk Dk+nx E S for all n E No, in particular, f(hD)x E S;

k=O
(ii) if the operator D is closed then
L
00
D n f(hD)x = ak hk Dk+nx E S, i.e. D n f(hD)S C S for all n E No.

k=O
Lemma A1.2. Suppose that all assumptions of Proposition AI.I are satisfied,
R E RD and A E vcR. Then A is an eigenvalue of the operator D and the
corresponding eigenvectors are
(Al.5) eAz = (J - AR)-l z, where z E ker D are arbitrary
(cf. Proposition 8.6 and Example 9.1, also Theorem 13.3).

Lemma Al.2, Proposition Al.l(ii) and Formula (Al.3) together imply
Corollary Al.2. Suppose that all assumptions of Proposition Al.I are satisfied
and A E l!{nvcR. Let eAz = (I - AR)-lz, Z E kerD (cf. (Al.5)). Then
(i) EA(R) = lin {eAz : z E ker D} C Sr(D);
L
00
(ii) ak hk Dk+ neAz E EA(R) for all n E No, z E ker D;

k=O
(iii) f(hD)eAz = f(Ah)eAz for all h E l!{, z E ker D;
(iv) if the operator D is closed then Dn f(hD)eAz = An f(Ah)eAz for all n E N,
hE l!{, z E ker D (cf. Proposition 8.6).
Corollary A1.3. Suppose that all assumptions of Proposition AI.I are satisfied
and l!{ = C. Then
E = U ker(D - AI) = E(R) c Sr(D).

Aec
Clearly, if there is an operator REnD n V(X) then, by definitions, E(R) E

Sr(D).
In the sequel we shall admit the following convention: if a theorem (propo-
sition, corollary, lemma, definition, formula, respectively) is communicated by
PRZEWORSKA-RoLEWICZ then we will enumerate it by 18.n, instead of A1.n, ... ,
A6.n. We can do so because, as a matter of fact, Appendix is the chapter 18.
Corollary A1.2 and Proposition 8.1 (for n = 1) together imply
Proposition 18.1. Suppose that X E Lr(D)n L,(D) has unit e E dom 0;1 n
dom Oil, X is a complete linear metric space, kerD :j:. {OJ, (Lr,Er ) E G[Or],
(L"E, ) E G[O,], and there are A E C and REnD such that Ag = ARe E
dom 0;1 n dom Oil. Then
(i) x,\ = E/(Ag)Z = z[Er( -Ag)t 1 E ker(D - AI), where z E ker D is arbitrary;
(ii) f(hD)x,\ = f(Ah)x,\ for all h E OC, z E ker D;
(iii) if the operator D is closed then Dn f(hD)x,\ = An f(Ah)x,\ for all n E N,
hE OC, z E ker D;
(iv) in particular, if X E L(D), (L, E) E G[O], then the assertions (i)-(Ui) hold for
x,\ = ZE(Ag), z E ker D.
Let X be a complete linear metric locally bounded space. Recall (cf. ROLEWICZ
R[I]) that
• for a p, 0 < p ::; 1, there is a p-homogeneous F -norm equivalent to the original
one (AOKI-RoLEWICZ theorem);
• for every p, 0 < p < Po = lo~o:(~J(")' where c(X) is the modulus of concavity of
the space X, there is a p-homogeneous F -norm II II equivalent to the original one
(ROLEWICZ theorem).
Proposition A1.2. Suppose that X is a complete linear metric space with a
p-homogeneous F-norm 1111 (0 < p ::; 1), D E R(X) and ker D :j:. {OJ. Let
= {x E Doo : lim sup VIIDnxll::; 1 if OC = OC1 and {Dnx} is bounded if OC = C}.

n-too
Then Xl (D) C Sf(D).
Proof. Let x E Xl (D) and let f E H(OC) be arbitrarily fixed. Let P1 =
lim sup VIIDnxl1 for OC = OC1 . Let M > 0 be such that IIDnxll < M for all
n-too
n E No if OC = C.
For all n E No, hE OC we have Ilanh nDnxll = lanlPlhlnpllDnxll. Hence
lim sup Vllanhn Dnxll =

n-too
{O0 < P1 ::; 1 for OC = OC
for OC = C.
1; (h E OC)
Appendix 263
This implies that the scalar series E~=o lanlPlhlnplIDnxll is convergent for all
h E OCl (for all h E ee, respectively). Since X is a complete linear metric space
with an invariant metric J.t(x, y) defined by the equality: J.t(x, y) = Ilx - yll for all
x, y EX, we conclude that the series E~=o anh nDnx is absolutely convergent for
allhEK •
Note ALl. Suppose that all assumptions of Proposition A1.2 are satisfied. Then
Indeed, let xES be arbitrarily fixed. Then there is an N E N such that Dn x = 0

for all n > N. This implies that x E Xl(D). Clearly, Dx E X1(D) whenever
x E Xl(D). Let R E RD be arbitrarily fixed. Since DnRx = Dn-l x for all
x E Doe, n E N, we conclude that Rx E Xl(D) whenever x E Xl(D). 0
A2. Functional shifts.
We still assume that X is a complete linear metric space and f E H(lK), where
lK = lKr (0 < r :::; 00). Suppose that D E R(X). We shall study families
T"K = {T"hhEIK C Lo, where the operators T"h coincide with the operator-
valued functions f(hD) on the set S,(D).
Clearly, by our definitions, Proposition Al.I and Corollary Al.3, we have
S,E(R) c Sj(D) c S,(D) C Dco C dom D.
Definition A2.1. Let X be a complete linear metric space and let f E H(lK)
have the expansion
L
00
(A2.1) f(h) = anh n far all h E lK.

n=O
Suppose that D E R(X) and ker D =I O. A family T"IK = {T"hhEIK C Lo(X) is

said to be a family of functional shifts for the operator D induced by the function
f if
(A2.2) T"hX = f(hD)x far all h E lK, x E S,(D),
where the operator f(hD) and the set S,(D) are defined by Formulae (Al.4),
(Al.l), respectively. 0
Remark A2.1. By the definition of the set S, the sum E~=o anh n Dnx for
h E lK, xES has only a finite number of members different than zero. Thus some
further results for S are true if we assume only that X is a linear space (without
any topology). 0
In the sequel we shall assume (for the sake of brevity only) that either lK = C or
lK is the unit disk lKl .
Remark A2.2. Some further results for the sets ker(D - >.J) (>. E lK) and E are
true if X is only a linear topological space.
o
Let G be a subset of an open set 0 C C, let 9 E H(O) and let DE R(X), R E RD.
A family Tg,G = {Tg,hhEG C Lo(X) was named by
• PRZEWORSKA-RoLEWICZ a family of R-shifts (D-shifts on the set E(R) for
R E RD n V(X» if g(h) = eh and either G = lR or G = Il4 (cf. PR[7], PR[8]);
• BINDERMAN a family of complex R-shifts for an entire function 9 (cf. Bin[l]),
functional shifts for an analytic function 9 (cf. Bin[5]).
Appendix 265
Families of R-shifts and D-shifts were defined in two different ways. In general,
they do not coincide. If D-shifts and R-shifts coincide, then in PR[13] are called
true shifts (cf. also Definition 11.2). Note that the definition of a family of R-shifts
is independent of the choice of a right inverse R (cf. Bin[5]).
Shifts for a right invertible operator D induced by a function analytic in a ring and
a right inverse of D are studied in Bin[12], Bin[13] and Bin[15]. Shifts induced by a
member of a subalgebra of the algebra of all functions continuous on the unit circle
are considered in Bin[16]. For the entire collection, see also Bin[2]-Bin[4], Bin[6]-
Bin[ll], Bin[14], BPR[l], BPR[2], PRUDNIKOV Pru[l], PR[14], PR[15], PR[23].
It will be shown that, if Tg,h have behaviours as the operator-valued function ehD
=
(h E IG) has on a set containing either S P(R) or E(R), then Tg,G is a family of
R-shifts on P(R) and a family of D-shifts on E(R).
An immediate consequence of the definition is
Proposition A2.1. Suppose that X is complete linear metric space, D E R(X),
ker D =f:. {O} and Tf,'{t;. = {T"hhElK is a family of functional shifts induced by a
function f E H (lK). Then
(i) the operator T"h ofthe functional variable f E H (lK) and the complex variable
hE lK, as an operator acting in linear space Lo(S,(D)), is linear, i.e.
TA/+pg,h = >"T"h + JLTg,h for all f, g E H(lK), >.., JL E C;
(ii) for all h E lK the operators T"h are uniquely determined on the set S,(D);
(iii) if S,(D) = X and T"h are continuous for h E lK then T"h are uniquely
determined on the whole space X .
Theorem A2.1. Suppose that X is complete linear metric space, D E R(X),
ker D =f:. {O}, F is an initial operator for D corresponding to an R E RD and
TO{ = {T"hhElK is a family of functional shifts induced by a function f E H(lK).
Then
(i) for all h E lK, kENo and Z E ker D
k k
(A2.3) T"hRk F =L ajh j Rk-j F, i.e. T"hRk Z =L ajh j Rk-j Z;
j=O j=O
(ii) if a family S"lK = {S"hhElK satisfies the condition
k
S"hR~z = L ajhjR~-jz for all hE lK, kENo, z E kerD, where Rl E RD,
j=O
then
Proof. (i) Let x E X be arbitrarily fixed. By the definition, Rk Z E S for all

kENo, where z = Fx E ker D. Then for all h ElK, kENo we have
00
=L
j=o
k 00
=L ajhjRk-jz+ L
j=o j=k+1
k k
j
=L ajh Rk-j z =L ajhj Rk-j Fx.
j=O j=O
The arbitrariness of x implies the first of Formulae (A2.3).

(ii) Let xES and h E IK be arbitrarily fixed. By our assumptions, there exist
m E No, Zk E ker D (k = 0,1, ... , m) such that
Then for mEN and h E IK

m
S"h X = S"h L R~Zk

k=O
m m k
= L S"hR~Zk =L L ajhiRk-jzk
k=O k=O j=O
m k m m
= L L aihiDjR~zk = L L aihiDi(R~Zk)
k=O i=O i=O k=i
m m m
= ao L R~ Zk +L aj h j Dj (L R~ Zk)
k=O j=1 k=j
m m j-I
=aox+ L ajhjDj(L R~Zk- R~Zk) L

j=1 k=O k=O
m m j-I
= aox + L aj hi Dj x - L aj h j L Dj R~ Zk
j=1 j=1 k=O
=L
m
ajh j Djx = T"h X.

j=O
If m = 0 then x E ker D and S"hX = aox = T"hX for all h ElK. •

Appendix 267
Proposition A2.2. Suppose that all assumptions of Proposition A2.1 are satis-
fied. Then for all h E I[( the operators T"h commute on the set S with the operator
D.
Proof Let h E 1[(, kEN, z E ker D be arbitrarily fixed. By Theorem A2.l,
k
=D L ajh j Rk-j z - T"hR k- l Z
j=O
k-l k-l
- " aJ·hjRk-j-l z - ~
- ~ " aJ·hjRk-l-j z -- 0 .
j=O j=O
For k = 0 we have DT"hZ = D(aoz) = aoDz = 0 and T"hDz = T"hO = O. This

implies that DTj,hX = T"hDx for all XES, hE 1[(. •
Lemma Al.l implies

Proposition A2.3. Suppose that all assumptions of Theorem A2.1 are satisfied
and the operator D is closed. Then for all h E I[( the operators T"h commute on
the set S,(D) n S?)(D) with the operator D.
Clearly, by definition and Proposition Al.l(i), S c S,(D) n SJ1)(D).

Proposition A2.4. Suppose that all assumptions of Proposition A2.1 are satis-
fied. Let ker(D - >..I) f. {O} for a >.. E K Then for all h E I[(
(A2.4) T"hX = f(>"h)x for all x E ker(D - >..I)
and the operators T"h commute on ker(D - >..1) with the operator D.
Proof Formula (A2.4) is an immediate consequence of (Al.3). Let x E ker(D->..I)
be arbitrarily fixed. Formula (A2.4) together with the equality Dx = >..x imply
for all h E I[(
DT"hX - Tj,h Dx
= D[f(>"h)x] - T"h(>"X) =
f(>"h)Dx - >"T"hX
= f(>"h)>"x - >..f(>"h)x = O.
Propositions 18.1 and A2.4 together imply

•
Theorem 18.1. Suppose that X E Lr(D)n L1(D) has unit e E dom n;l n
dom nIl, X is a complete linear metric space, ker D f. {O}, (L r , Er) E G[Or]'
(L 1, Ed E G[Od, there are >.. E C and R E RD such that >..g = >..Re E dom n;l n
dom nIl and T"K = {T"hhEK is a family of functional shifts induced by a
function f E H (I[(). Then for all h E 1[(, z E ker D
In particular, if X E L(D), (L,E) E G[O], then
T"h[zE(Ag)] = f(Ah)zE(Ag) far all hE OC, z E ker D.
and the operator I - AR is invertible for a A E OC \ {O}. Then for all h E OC
(i) T"he).,F = f(Ah)e).,F, where e)., = (I - AR)-l j
(ii) The operators T"h commute on the set E).,(R) (defined in Corollary A1.3(i))
with Dj
(iii) FT"he).,Z = f(Ah)z for all z E ker D.
Proof. (i) and (ii) follow from Lemma A1.2 and Proposition A2.4.
In order to prove (iii), assume that z E ker D and h E OC are arbitrarily fixed. By
(i) and properties of an initial operator for D, we get
(FT"he).,]Z = F[J(Ah)e).,z]
= f(Ah)Fe).,z = f(Ah)(I - RD)e).,z
= f(Ah)[e).,z - R(Ae).,z)]
= f(Ah)(I - AR)(I - AR)-l z = f(Ah)z.
Either Proposition A2.4 or Proposition A2.5(i),(ii) implies

•
Corollary A2.1. Suppose that all assumptions of Theorem A2.1 are satisfied and
OC = C. Then for all h E C
(i) T"h X = f(Ah)x for all x E Ej
(ii) the operators T"h commute on E with D.
Clearly, Corollary A2.1 holds if the operator D has a Volterra right inverse R.
Consider now the space
A(D) = U AR(D) c Doo

RERD
of D-analytic elements (cf. Chapter 11).
ker D -:f. {O} and T,,'K. = {T"hhE'K. is a family of continuous functional shifts for
D induced by a function f E H(OC). Then
(i) if D is closed then A(D) C S,(D)j
(ii) if an operator R E RD is continuous then AR(D) C S,(D).
Proof. Let x E A(D) be arbitrarily fixed. Then there exists an initial operator F
for D corresponding to an R E RD such that x = ~~=o RnFDnx .
Appendix 269
(i) By our assumptions and Theorem A2.1, for all h E OC we have
=T/,h L L
00 00
R n F Dnx = T/,hRn F Dnx

n=O n=O
n
=L L =L L
00 0000
akh k R n - k F Dnx ak hk R n - k F Dnx

n=O k=O k=O n=k
L =L L
00 00 00 00
= ak hk L R n- k F Dnx akh k Dk Rk R n - k F Dnx

k=O n=k k=O n=k
k-l
L L L
00 00 00
= akhkD k RnFDnx=aox+ LakhkDk(x- RnFDnx)

k=O n=k k=l n=O
k-l
L L
00 00
=Lk=O
akhkDkx -
k=l
akh k
n=O
DkRnFDnx
k-l
L L
00
= f(hD)x - ak hk Dk-nFDnx
k=l n=O
k
=
00
f(hD)x - L ak hk
k-l m=l
L
DmFDk-mx = f(hD)x.
(ii) Fix x E AR(D). By the continuity of R, we have Dkx E AR(D) for all kENo
(cf. PR[9], Theorem 3.2). This, and Theorem A2.1(i) together imply for all h E OC
=T/,h L =L L
00 00 00
R n F Dnx akh k R n - k F Dnx

n=O k=O n=k
=L L L
00 00 00 00
ak hk RmFDm+k x = akh k L RmFDm(Dkx)

k=O m=O k=O m=O
=L
DO
ak hk Dkx = f(hD)x.
k=O
Proposition A2.6. Suppose that all assumptions of Theorem A2.2 are satisfied.
•
If F is a continuous initial operator for D corresponding to a continuous R E RD
then T/,h map AR(D) into itself for all h E K
Proof. Let x E AR(D) and h E OC be arbitrarily fixed. The continuity of the
operators Rand F implies that D is closed (cf. PR[9]). By Theorem A2.2 (either
(i) or (ii)) , T/,hx =
f(hD)x. Since Dk E AR(D) for all kENo (cf. PR[9]) , by
definition,
n=O
=L L
00 00
RnFD n akhkDkx
n=O k=O
=L L
00 00
akhkRnFDn(Dkx)
n=O k=O
00 00
k=O n=O
L
00
= ak hk Dkx = T"h X •
k=O
•
Appendix 271
A3. Isomorphisms of spaces of functional shifts.
It is well-known that the set H(OC) is a commutative algebra with the following
operations:
(f + g)(h) = f(h) + g(h), (af)(h) = af(h), (fg)(h) = f(h)g(h),

where f,g E H(OC), a E C, hE II(.
Let X be a complete linear metric space. Let D E R(X). Let either OC = C or

OC = K 1 . Let T(OC) be the set of all families of functional shifts for D induced by
functions f E H (OC), i.e.
(A3.1) T(OC) = {T/,K : f E H(OC)}.

ker D f. {O}, T/,K = {T/,hhEK are families of functional shifts induced by func-
tions f E H(OC) and T(OC) is defined by Formula (A3.1). Write
Ty(OC) = T(OC)ly = {T/,Kly : f E H(OC)} far Yen /EH(K)

SleD).
If Y is the set S of all D-polynomials then
(i) the set Ts(OC) is a commutative algebra with the operations
(A3.2)
where f, 9 E H(OC), a E Ci
Oi) the algebras H(OC) and Ts(OC) are isomorphic and T : f -t T/,Kls is an algebra
isomorphism from H(OC) onto Ts(OC).
Proof. (i) It is enough to show that the multiplication of two families of functional
shifts for D is well defined. Suppose then that we are given families T/,K, Tg,K,
where f,g E H(OC) are arbitrarily fixed and
=L
00
=L
00
f(h) anh n , g(h) bnhn (h E OC).

n=O n=O
Let z E ker D, kENo, hE OC, R E RD be arbitrarily fixed. Then
k
=T"h(Tg,hR k z) = T"h '~
"'
bk-jh k -3. R3. z
j=O
k k j
=L bk_jhk-jT"hRjz =L bk_jhk- j L aj_mhj-mRmz
j=O j=O m=O
k j k k
= L L bk_jaj_mhk-m Rmz =L [L aj_mbk_jjh k- mRmz
j=O m=O m=O j=m
k
= '"'
~ Ck-m hk-mRmz,
m=O
where n
C n= L apbn_p (n E No).
p=O
On the other hand, if a = {an}, b = {b n }, then the sequence c = {cn } is a

convolution of a and b, i.e.
n
c=a*b={L apbn_p}.
p=O
Now, by the Cauchy theorem about the multiplication of two series, we conclude
that w = fg E H(OC) and
L
00
w(h) = cnhn.
n=O
(ii) the neutral elements and units of the commutative algebras H(OC) and Ts(lK)
are
O(h) == 0 on lK, ORkz = 0, l(h) == 1 on lK, IRkz = Rkz,
respectively (k E No, z E ker D, R E RD).
Observe that
TO,h = 0, T-/,h = -T/,h, T1,h = I (h ElK).
Moreover, if f (h) :f 0 on lK then
which implies that T1//,h = (T/,h)-l (h ElK).

Appendix 273
By definition, if f, 9 E H(OC) and T/,h = Tg,h on S for all h E OC then f = 9 on OC.

It is well-known that, if f, 9 E H(OC), f(h n ) = g(h n ) (n = 1,2 ... ) and the sequence
{h n } has a limit point in OC, then f(h) == g(h) on K Thus, if f,g E H(OC)
and T/,h = Tg,h for all h E OC, then f =
9 on OC. We therefore conclude that
T : f -+ Ts(OC) is an algebra isomorphism of H(OC) onto Ts(OC). •
Theorem A3.2. Suppose that all assumptions of Theorem A3.1 are satisfied. Let
Y = ker(D - AI) =f. {O} (A E OC). Then
(i) the set Ty(OC) is a commutative algebra with the operations defined by Formulae
(A3.2);
(ii) if A =f. 0 then the algebras H(OC) and Ty(OC) are isomorphic and T : f -+ T/,Kly
is an algebra isomorphism from H(OC) onto Ty(OC).
Proof. (i) It is enough to show that the multiplication of two families of functional
shifts for D restricted to the set Y = ker(D - AI) is well defined. Suppose then
that we are given families T/,h, Tg,h, where f, 9 E H(OC). Let x E Y \ {O} be
arbitrarily fixed. Then Dx = AX. By (A2.4), for all h E OC we get
(T/,hTg,h)X
(A3.3) =T/,h(Tg,hX) = T/,h[g(Ah)x] = g(Ah)T/,hX
= g(Ah)f(Ah)x = T/g,hX.
The proof of (ii) is going along the same lines as the proof of Theorem A3.2(ii) . •
Similarly, we prove
OC = C and let Y = E = E(R). Then
(i) the set Ty (OC) is a commutative algebra with the operations defined by Formulae
(A3.2);
(ii) ifE =f. ker D then the algebras H(OC) and Ty(OC) are isomorphic and T : f -+
T/,Kly is an algebra isomorphism from H(OC) onto Ty(OC).
Write for an operator D E R(X)
(A3.4) SK(D) = n
/EH(K)
Sr(D).
where Sr(D) is defined by Formula (Al.I). Proposition Al.I and Corollary Al.3
together imply that
S c SK(D),
ker(D - AI) c SK(D) for A E OC,
E = E(R) c SK(D) for OC = Co

Let x E Soc(D) and f,g E H(OC) be arbitrarily chosen and let

00 00
f(h) = L anh n , g(h) =L bnhn for all hE K

n=O n=O
By definitions, T/,oc, Tg,oc, T/g,oc E T(OC) , the values T/,hX, Tg,hX, T/g,hX are
uniquely determined for all h E OC, where
00 n
(fg)(h) = L (L an_kbk)h n for all h E K
n=O k=O
Suppose that the operator D is closed. Then for x E Soc(D), hE OC
00 m 00 m
=L (L am_nbn)hmDmx =L L am_nbnhmDmx
m=O n=O m=O n=O
00 00 00 00
=L L akbnhk+nDk+nx = L bnhnDn(L akhkDkx)

n=O k=O n=O k=O
00
=L bnhnDn(T/NC) = Tg,h(T/,h X) = (Tg,hT/,h)X.

n=O
This immediately implies

If D is closed then
T/,h(Soc(D)] C Soc(D) for all f E H(OC), hE K
Similarly as Theorem A3.2, we prove

Y = Soc(D). Then
(i) the set Ty (OC) is a commutative algebra with the operations defined by Formulae
(A3.2)j
(ii) the algebras H(OC) and Ty(OC) are isomorphic and T : f -t T/,ocl y is an algebra
isomorphism from H(OC) onto Ty(OC).
Let TC(OC) C T(OC) be the set of all continuous functional shifts for D E R(X)
induced by H(OC). An analogue of Theorems A3.1-A3.4 is
Y = TAR (D) = TC(OC)IAR(D)' where R E RD is continuous. Then
(i) the set Y is a commutative algebra with the operations defined by Formulae
(A3.2)j
Appendix 275
(ii) the algebras H(OC) and Y are isomorphic and T : f ~ Y is an algebra isomor-
phism from H (OC) onto Y.
Proof. (i) It is enough to show that the multiplication of two families of continuous
functional shifts for D restricted to AR(D) is well-defined. Suppose then that we
are given families Tn". , Tg,K E TC(OC), where f, 9 E H(OC) are arbitrarily fixed and
=L =L
00 00
f(h) ak hk , g(h) bk hk for all hE K

k=O k=O
Theorem A2.2 implies that AR(D) is a subset of the sets S,(D), Sg(D), Sfg(D).
Let F be an initial operator for D corresponding to the continuous R. Let x E
AR(D) \ {O} and h E OC be arbitrarily fixed. Proposition Al.I implies that
(T"hTg,h)X
00
= T"h(Tg,h X) = T"h(Tg,h L RnFDnx)

n=O
n
L L L
00 00
= T"h Tg,hRnFDnx = T"h bn_khn-kRkFDnx

n=O n=O k=O
n
L L
00
= bn_khn-kT"hRk F Dnx
n=O k=O
n k
=L L L
00
bn_khn - k ak_jh k- j Ri FDnx

n=O k=O j=O
n k
=L L L
00
ak_jbn_khn-jRjFDnx
n=O k=O j=O
n n
=L L (L
00
ak_jbn_i)hn-iRiFDnx
n=O j=O k=j
n
L L
00
= cn_jRi FDnx,
n=O j=O
where
n
Cn = L apb n _ p (n EN).
p=O
Since w = fg E H(OC) and
=L
00
w(h) (h E OC), cnh n

n=O
by the Cauchy theorem about the multiplication of two series we get
L L
00 00
(T"hTg,h)X = Tw,hRnFDnx = Tw,h RnFD n = Tw,h X.

n=O n=O
Hence Tw,lK = T/g,lK E TC(II().

(ii) is proved in the same manner as Theorem A3.1(ii).
The following functional equations are referred to as the Cauchy equations:
•
(A3.5) J(t + s) = J(t) + J(s),
(A3.6) J(ts) = J(t) + J(s),
(A3.7) J(t + s) = J(t)J(s),
(A3.8) J(ts) = J(t)J(s)

Continuous solutions of these equations are well-known. Namely, they are func-
tions: ct, e loga t, ce at , eta, respectively, where c, a are arbitrary scalars (for the
second function it should be a > 0) (cf. C[1J, KUCZMA Ku[2]).
Theorem A3.1 immediately implies
If II( = C and J satisfies Equation (A3.5) for all hl,hz E II( then T/,Kis is an
Abelian group with respect to the addition of operators, i.e.
and the function J satisfies Equation (A3.6) for all hI, hz E 11(. Then T/,lKis is a
commutative semigroup for II( = 11(1 and an Abelian group for II( = C with respect
to the addition of operators, i.e.
and the function J satisfies Equation (A3. 7) for all hI, hz E 11(. If II( = C then
T/,lKis is an Abelian group with respect to the superposition of operators,i.e.
and the function J satisfies Equation (A3.8) for all hI, hz E K Then T!,I{is is a
commutative semigroup for II( = 11(1 and an Abelian group for II( = C with respect
to the superposition of operators, i.e.
Appendix 277
Remark A3.1. Suppose that all assumptions of Theorem A3.I are satisfied.
Propositions similar to Propositions A3.2-A3.5 for the family T'.Kls
hold for the
families T'.Kly,where
ker(D - )../) if A E IK;

y _ { E = E(R) if IK = C;
- SK(D) if D is closed;
AR(D) if R is continuous.
This follows from Theorems A3.2, A3.3, A3.4, A3.5, respectively. o

We shall study now limit properties of families of functional shifts. Let nc IK.
Write
A = P E C : An c IK} and EA = U
ker(D - )../).
AEA
Proposition A3.6. (d. BPR[I]) Let D E R(X) and let ker D -:j:. {o}. Let T,.o,
TI'.o be families of functional shifts for the operator D induced by the functions
r
f, E H(IK), respectively. Then
(i) if ho Ene IK is a limit point of n then
I
lim
03h-+ho
ho _ h(T'.ho - T,.h)X = TI'.hoDx = DTI'.hoX for all xES U EA;
(ii) if n C IK is an open set then
Definition A3.1. (cf. BPR[I]) Let 0 Ene IK be the limit point of the set n.
A family To = {ThhEO C Lo(X) has the limit property on a set Y C X if for
every x E Y there exists lim -hI (Th - To)x. If it is the case, then the operator A
h-+O
defined as
(A3.9) Ax = h-+O
lim -hI (Th - To)x for x EY = dom A
is said to be an infinitesimal generator for the family To. o
Note that the infinitesimal operator A (if exists) is well-defined. Clearly, if X
is a locally convex complete linear metric space, then any strongly continuous
semigroup To = {ThhEO of linear operators is a family with the limit property
and its infinitesimal generator satisfies Condition (A3.9) (d. HILLE and PHILLIPS
HP[I]).
Proposition A3.6 leads us to the following
Corollary A3.1. (d. BPR[I]) Suppose that T,.o is a family of functional shifts
for D induced by a function f E H(IK) and 0 E n is a limit point of n. Then
this family has the limit property on the set SUEA and j'(O)D is its infinitesimal
generator.
Theorem A3.6. Suppose that OC = c, n c OC is the interior of a spinal semi-
module (i.e. an additive semigroup which contains a ray from the origin and an
open set intersected by this ray, cf. HILLE and PHILLIPS HP[l]) containing the
positive real axis, D E R(X), ker D =f:. {O} and T/,n = {T/,hhEO is a family of
functional shifts for D induced by a function f E H(C), f ~ O. Then T/,n is
a commutative semigroup with respect to the superposition of operators on the
set Y = S U EA (Y = Sn(D) if D is closed, Y = AR(D) if R is continuous,
respectively), i.e.
if and only if there is an a E C such that f (h) = eoh . If it is the case then
(A3.1O) l~ ~(T/'h - T/,o)x = aDx for all x E Y,
i.e. the operator aD is the infinitesimal generator of semigroup T/,n on Y.

Proof. Let Y = S U EA. Sufficiency has been proved by Propositions A3.2-A3.4.
Necessity (cf. BPR[l]). Suppose that T/,n is a semigroup. It follows from proper-
ties of functional shifts considered on the set SUEA that in this case the function
f should satisfy the functional equation: f(t + s) = f(t)f(s) for all t, sEn. It is
well-known (cf. HILLE and PHILLIPS HP[l], Lemma 17.3.1) that either f
n or there exists a complex number a such that f(h) = eoh for all hEn. This
0 on =
implies that f(h) = eoh for all h E C. Proposition A3.6(i) implies the equality
(A3.1O). We therefore conclude that aD is the infinitesimal generator of the semi-
group under consideration. Similar proofs for the sets Sn(D) if the operator D is
closed and for AR(D) if the operator R is continuous (cf. Bin [4], [5]). •
Clearly, Theorem A3.6 holds also in the case when either n = OC or n ~ OC contains
the interior of a spinal semi-module containing the positive real axis.
In a similar manner we obtain the following
Theorem A3. 7. (cf. BPR[l]) Suppose that X is a complete linear metric locally
bounded space and all assumptions of Theorem A3.6 are satisfied. Then T/,n is a
commutative semigroup with respect to the superposition of operators on the set
Xl (D) if and only if there is an a E C such that f(h) = eoh (h E K) and Formula
(A3.1O) holds on Xl (D). If it is the case then aD is the infinitesimal generator of
the semigroup TJ,n.
Note A3.1 Functional shifts may be defined as operators induced by an analytic
function f : OCp -+ JR, where OCp = (-p, p), 0 < p ~ +00. The particular case,
when OC = JR, n ~ OC is either JR or JR+, f(h) = eh, has been considered in Chapter
11. Clearly, also in that case results analogous to Theorems A3.6, A3.7 can be
obtained. 0
Appendix 279
For perturbed operators we get

Theorem A3.8. (cf. BPR[l]) Let DE R(X), ker D # {O}. Suppose that Tj,K is
a family of functional shifts for D induced by a function f E H (1K), the operator
A E Lo(X) maps S U EA into itself, R E RD and
(A3.11) Tj,h = Tj,h(1 + RA) - Tj,oRA for all hE 11 C 1K.
If zero is the limit point of 11 then
(A3.12) lim (Tjl h - T jl o)x = f(O)(D

03h-tO' ,
+ A)x for all xES U EA,
i.e. the family Tj,K has the limit property on the set S U EA, its infinitesimal
generator is f(O)(D + A) and T;,K acts in the following manner:
Tj,hx = (f(hD) - f(O)](I + RA)x + f(O)x for all xES U EA, hE 11.
If, in addition, f (0) # 0 then the family TJ,f!{ = TIl f' ,oT;,f!{ has the limit property
on S U EA and its infinitesimal generator is D + A. Moreover,
1
T7,h X = reO) ([f(hD) - f(O)](1 + RA)x + f(O)x} for xES U EA, hE lK.
Proof. Suppose that the family Tj,o is defined by Formula (A3.12). Then T;,o =
Tj,o. Let xES U EA be arbitrarily fixed. By Proposition A3.6, we get
1 (Tj,h
lim -h I I
- Tj,o )
x
h-tO
1
= lim -h[Tj,h - Tj,o
h-tO
+ (Tj,h - Tj,o)RA]x
= l~ ~(Tj'h - Tj,o)x + l~ ~(Tj'h - Tj,o)RAx

= f(O)Dx + f(O)DRAx = f(O)(D + A)x.
Let now xES n EA be arbitrarily fixed. By definition and our assumptions, for
all h E IK we have
T;,h X
= Tj,h(I + RA)x - Tj,oRAx = f(hD)(1 + RA)x - f(O)RAx
= [f(hD) - f(O)](1 + RA)x + f(O)x.
The conclusion if f (0) # 0 is obvious. •
Note A3.2. Suppose that D E R(X), R E RD, A E Lo(X) and the operator
1 +RA is invertible. Recall that DO = D+A E R(X) and RO = (I +RA)-I R E RD
(cf. PR[8]). Recall also that the sum of a closed operator and a continuous operator
is again closed. So that, if D is closed and A is continuous, then DO is closed. 0
More details about the limit property, perturbations of functional shifts and ex-
amples can be found in BPR[l I.
Theorems A3.1, A3.2, A3.3, A3.4, A3.5, respectively, immediately imply the fol-
lowing
Theorem 18.2. Suppose that all assumptions of Theorem A3.1 are satisfied. Let
(18.1) Y = {~~r(D -)../) if A E lK;

E = E(R) iflK = C;
Then functional shifts TJ,h satisfy the Leibniz condition in the following sense:
DT
(fg)"h
x = DT J'g+Jg',h x = DT Jg',h x + DT gJ',h X
for x E Y; hE lK; f,g E H(lK).
Theorem 18.3. Suppose that all assumptions of Theorem A3.1 are satisfied,
Y is determined by Formulae (18.1) and f, 1/ f E H(lK). Then Y E Lg(D) and
(L, E) E G[OI, where
DLT x = T x = DT x for x E Y,
J,K f' / J,K In J,K
i.e. for R E RD we have

LT
J,K
x = RT J'/J,K x + z, where Z E ker D is arbitrary.
Suppose that TJ,K = {TJ,hhEK is a family of functional shifts for D E R(X)

induced by a function f E H(lK) such that f(O) f:. O. Let F be an initial operator
for D corresponding to an R E RD. Write
(A3.13)
It is easy to verify that Fh are projections onto ker D, hence they are initial
operators for D corresponding to the right inverses Rh (h ElK), respectively. We
shall see that functional shifts preserve essential properties of R, namely, its regular
values.
Lemma A3.1. Suppose that D E R(X) and R E RD. Then RnDn[SJ(D)1 c
SJ(D) for all n E N.
Proof. Let n E N, x E SJ(D) be arbitrarily fixed. By the Taylor formula,
L
00
ajhjDiRnDnx
j=O
n-l
=L L L
00 00
aihiDix - ajhiDi RkFDkX

i=O j=O k=O
L L
n-l k
= f(hD)x - ajhi Rk-i FDkx.
k=O i=O
Appendix 281
The last term is a polynomial of degree n - 1 hence has no influence on the

convergence of the series in question. Thus RnDnx E SleD). The arbitrariness of
x implies the assertion. •
Lemma A3.1 immediately implies
Proposition A3.1. Suppose that DE R(X), ker D :f; {O} and T/,K = {T/,hhEK
is a family of functional shifts for D induced by a function f E H(OC). Let ker(D-
M) :f; {O} and let ,xOC c OC (,x E C). Let R E RD. Then for all h E OC, n E N we
have Rn[ker(D - ,xI)] C SleD) and for all x E ker(D - M)
n-l
(A3.14) T/,hRnx = ,x-n[J(,xh)x - L ,xkT/,h Rk Fx], (,x:f; 0).

k=O
Theorem A3.9. (cf. BPR[2]) Suppose that D E R(X), ker D :f; {O}, F is an
initial operator for D corresponding to an R E RD, Tj,K = {Tj,hhEK, Tl/j,K =
{Tll/,hhEK are families of functional shifts for D induced by functions f, 1/ f E
H(][{) , respectively, and Fh,Rh (h E OC) are defined by Formulae (A3.13). Let
ker(D -,xI) :f; {O} and let,xOC C OC (,x E C). Suppose, moreover, that f(h) :f; 0 for
hE oc. Then
(A3.15) R'hx = f(,xh)T1I/,hRnx for all x E ker(D-M), n EN, hE OC (,x:f; 0).
Proof, by induction. Let h E OC, x E ker(D - M) be arbitrarily fixed. Take into

account that the function 1/ f has the expansion
L
00
[f(h)]-l = bkh k for all hE OC,

k=O
where the coefficients bk (k E No) are determined by the following recurrence

relations:
1-1
bo = a Dl , bl = -aDl L al_jbj (I EN).
j=O
If n = 1 then, by (A3.14), we have
T1I/,hRx = ,x-I [J-l (,xh)x - Tl//,hFx] = ,x-I [J-l(,xh)x - aD1Fx].
On the other hand, again by (A3.14) and the obvious equality Rx = ,x-I (x - Fx),
we obtain
RhX = Rx - FhRx
= Rx - a Dl FT/,hRx = Rx - a Dl FP-l[f(,xh)x - aoFx]}
= ,x-I (x - Fx) - aolp-l[f(,xh)Fx - aoF2 x]}
o
= ,x-IX - ,x-lFx - ,x-la l f(,xh)Fx + ,x-I Fx
= ,x-I [x - a Dl f(,xh)Fx].
Hence Rhx = f(>"h)Tl/j,hRx, i.e. Formula (A3.15) holds for n = 1.

Suppose Formula (A3.15) to be true for an arbitrarily fixed n ~ 1. By the our
assumption and Formula (A3.14), we have
Rn+lx
h
= Rh(RJ:x) = Rh(f(>"h)Tllf,hRnx] = f(>..h)(R - FhR)Tl/f,hRnx
= f(>"h)[RT1If,hRnx - FhRTllf,hRnx]
n-l
= f(>..h)p-n R[r1(>..h)x - L >.kTllf,hRk Fx]
k=O
n-l
- >.-nar/FTf,hR[rl(>'h)x - L >.kTllf,hRkFx]}
k=O
n-l k
=>..-n f(>'h)[r1(>'h)Rx - R L >.k L bjhj Rk-j Fx
k=O j=O
n-l k
-aolrl(>'h)FTf,hRx+ar/FTf,hR L >.k L bjhjRk-jFx]
k=O i=O
L L
n-l k
=>..-nf(>'h)U-1(>'h)>..-1(x-Fx)- >.k bj hi R k+1-iFx
k=O j=O
o
- a f-1(>"h)>.-1 F[f(>'h)x - Tf,hFxj
1
n-l k
+a oL
1 >..kF L bjhjTf,hRk+1-jFx}
k=O j=O
k
L L
n-l
= >.-(n+l) f(>.h)(f-l(>.h)x - rl(>'h)Fx - >.k+l bjh j R k+1-j Fx
k=O j=O
- a o Fx + a o rl(>'h)aoF x
1 1 2
n-l k k+l-j
+ a o L >.k+l L bjhj F L
1 aphP Rk+l-j-p Fx]
k=O j=O p=o
k
L L
n-l
=>.-(n+1) f(>.h)[f-l(>.h)x - f-l(>.h)Fx - >.k+l bjh j R k+1-j Fx
k=O j=O
n-l k
- ao l Fx + rl(>'h)Fx + a o1 L >.k+l L bjhjak+1_jhk+l-j Fx]
k=O j=O
n-l k
=>.-(n+1) f(>.h)(f-l(>.h)x - a o Fx - L
1 >.k+l L bjh j R k +1-j Fx
k=O j=O
n-l
- ao1ao L >.k+ 1bk+1hk+1 Fx].
k=O
Appendix 283

R~+1x =
L L
n-I k+I
= A-(n+1) I(Ah)[rI(Ah)x - a l Fx -o Ak+1 bjh j R k+1-i Fx]
k=O j=O
n-I
= I(Ah)A-(n+I) [J-I(Ah)x - o
a l Fx - L Ak+1TII/,hRk+1 Fx]
k=O
n-I
= I(Ah)A-(n+1)[!-l(Ah)x - boFx - L Ak+1TI I/,h Rk +1 Fx]
k=O
n
= I(Ah)A-(n+1)[r I (Ah)x - TII/,h Fx - L APTII/,hRPFx]
p=l
n
= I(Ah)A-(n+1) [rI(Ah)x - L AkTII/,hRkFx]
k=O
This proves Formula (A3.I5) for all positive integers. •

Corollary A3.2. (cf. BPR[Ij) Suppose that all assumptions of Theorem A3.9
are satisfied. Then
T/,hRhX = I(Ah)Rnx lor all n E N, hE OC, x E ker(D - AI).
Theorem A3.10. (cf. BPR[2j) Suppose that D E R(X), ker D :j; {O} and
R E nDnAQN(ker D;w). Let T/,¥:. = {T/,hhE¥:., T II /,¥:. = {TII/,hhEK be families
of functional shifts for D induced by the functions 1,1/1 E H(OC), respectively. If
the operator TII/,h is continuous for some hE OC then Rh E nD nAQN(ker D;w).
Proof By our assumptions, for A E w and Z E ker D
lim AnRhz
n-too
= lim Anl(O)TII/hRnz=/(O) lim TII/h(AnRnz)
n-t(X) ! n-too'
= I(O)Tll/,h(nl~~ An Rnz) = I(O)TII/,h(O) = o.

By an easy induction, we prove that
(A3.I6) Rhz = I(O)Tll/,hRnz lor all n E No, hE OC, z E ker D,
The continuity of the operators Rh (h E OC) on ker D follows from Formula (A3.I6).
Suppose now that A E wand there is an h E OC such that the operator I - ARh
is not invertible on ker D. It means that there is a z E ker D \ {O} such that
z E ker (I - ARh), i.e. z - ARhZ = O. Hence we get
z = rI(O)/(O)z = rl(O)FT/,hZ = Fhz = Fhz - AFhRhZ = Fh(I - ARh)Z = 0,

a contradiction with our assumption that z :I O. This proves that Rh E

AQN(ker D;w) for all h E K
In a similar way to Theorem A3.1O (by an application of Theorem A3.9) we prove
•
Theorem A3.1l. (cf. BPR[2]) Suppose that D E R(X), ker D :I {OJ, ker(D -
>.1) :I {OJ, ,XlK C lK, (,X E q, R E RD n AQN(E>..;w) and T/,K = {T/,hhEK'
TII/,K = {TI//,hhEK are families of functional shifts for D induced by functions
f, l/f E H(lK), respectively. If the operator TII/,h is continuous for some hE K
then
Rh E RD n AQN[ker(D - >.1);w).
The list of similar results for other subsets of S/(D) can be easily prolonged.
Appendix 285
A4. Functional shifts induced by operators of complex differentiation.
The purpose of the present section is to give formulae for functional shifts induced
by an operator D of generalized complex differentiation on the linear span S =
P(R) of all D-monomials. The choice of functions under consideration depends
on algebraic summation formulae. Results obtained here will permit us to derive
explicit forms of functional shifts in much more complicated cases. This section
contains results of Bin[5]'[8]'[9J.
Let X = H(OC), where OC is either the unit disk OCl or C. Recall that H(OC)
is a complete linear metric space with the topology of uniform convergence on
compact sets. Denote by Xl C X the set of all polynomials defined on OC and with
coefficients in C, i.e. Xl = crt], (t E OC). Clearly, Xl is a linear space over C with
the usual pointwise addition and multiplication by scalars. Let U = {Uk} C C be a
sequence such that Uk f:. 0 for all kENo. Define linear operators D and R, called
operators of generalized complex differentiation and integration, respectively, as
follows:
(A4.1) Dl = 0, Dt Hl = ~tk (k EN),

Uk+l
The operators D and R E Lo(Xt} are uniquely determined on Xl and dim ker D =
1, codim RX I = 1 (cf. FAGE and NAGNIBIDA FN[l]). The kernel of D consists of
all constant functions on][{, Since there is a one-to-one correspondence between
all constant functions belonging to X and complex numbers, we can identify here
constants functions with scalars if it does not lead to any misunderstanding. We
can prove (cf. Bin[6]) that
(i) D E R(Xt}, R E RD;
(ii) (Fx)(t) = [(J - RD)xJ(t) = x(O) for t E OC, x E Xl;
(iii) S = P(R) = Xl and Xl = P(R) = X.
In the sequel we shall study only such operators D and R for which the condition
1·1m
n--+oo
(-!£n =
n --
Un+l
1
is satisfied. These operators can be extended to linear continuous operators defined

on the whole space X and mapping X into itself, i.e. to operators belonging to
Lo(X) (cf. HAPLANOV H[lJ, FN[l]).
Example A4.1. Take a = =

{ak}, where ak 1 for all kENo. Then the operators
D and R determined by Formulae (A4.1) are of the form:
x(t) - x(O)
for t::j: OJ
(Dx)(t) = {
t
(A4.2) (Rx)(t) = tx(t)
x(t) ~ x(O) It=o = x'(O) for t = OJ
for x EX, t E ]I{.
The operator D is called the Pommiez operator (cf. Example 5.4). In this case
S = peR) = lin {Rkl: kENo} = lin {1,t,t 2 , ... }.

One can prove that for x EX, n E N
t-n[x(t) - ~ ~x(kl(O)l for t::j: OJ

(A4.3) (Dnx)(t) = { k=O .
x(n) (0)
n!
for t = o.
Hence for all x E X we have
00 00 (nl(o) 00 (nl(o)
" (RnFDnx)(t)
L...J
="
L...J
_x-Rnl
n!
="
L...J
_x_tn
n!
= x(t) (t ElK).
n=O n=O n=O
This implies that AR(D) = Doo = X. If]l{ = IKl then it is easy to verify that the
operator J - )"R is invertible for all ).. E IKl and
= <C then for every).. E <C \ {O} the operator is not invertible in X. Indeed,
If ]I{
consider the equation (J - )"R)x = y with yet) == 1. Then a solution of this
equation is x(t) = 1/(1- )..t). Clearly, x is not holomorphic at the point t = 1/)".
Moreover, if ]I{ = ]l{l then ker(D - AI) ::j: {O} if and only if A E ]l{l and
ker(D - AI) = E>.(R) = {1-a At : a E <C}

-
for all A E ]l{l·
Hence
Appendix 287
Let OC = OC1 and let f(h) = 1-1 h' Let x E X, h E OC1 be arbitrarily fixed. By
(A4.3),
00
n=O
=~ hn ~ t k x(k) (0) =~
x(k) (0) ~ hntk-n
~ tn ~ k! k!
~ ~
n=O k=n k=O n=O
1L
tk+l - hk+1 x(k) (0)
L
00
h-t ~ fortoJh;
_ k=O
T
- 00 (k)(O)
(k+1)h k fort=h.
k=O
Hence
tx(t) - hx(h) for t oJ h'
L
00 {
hn(Dnx)(t) = t- h '
n=O x(h) + hx'(h) for t = h.
This proves that S,(D) = X. Moreover, this shows that the family TKI
{ThhEKt> where
tx(t) - hx(h)
for t oJ hj
(nx)(t) = { t- h
x(h) + hx'(h) for t = hj
is the unique family of functional shifts for D induced by the function f(h) =
h ~ l' The operators Th (h E OCt} are continuous (cf. Bin[5]). 0
In Example A4.1 we have proved the following
Proposition A4.1. Let OC = OC1 and let f E X = H(OC): f(h) = h _1 l' Let
T"K = {T"hhEOC be a family of functional shifts for the Pommiez operator D
(A4.2) induced by f. Then
tx(t) - hx(h) for t oJ h'

(A4.4) (ThX)(t) ={ t - h ' (h E OC, x E X)
x(h) + hx'(h) for t = hi
is the unique family of continuous functional shifts for D induced by f.
Remark A4.1. Denote by X* the space of all linear continuous functionals de-
fined on the space X. According to LINCHUK Lin[!], we conclude that a continuous
operator V E Lo(X) commutes with the Pommiez operator D on X if and only if

there exists a functional x* E X* such that
(A4.5) (Vx)(t) = x*[(Tj,hX)(t)] (h E K, x EX).
Set xh, = X*Tj,h and x*(h k ) = bk E C (k E No). Then, by Equality (A4.5), we get
00 00
(A4.6)
Thus a continuous operator V E Lo(X) commutes with D if and only if Formula

(A4.6) holds, where the series is convergent for all x E X in the topology of X. 0
Proposition A4.2. Let K = OCl . Let
h
f EX = H(K) : f(h) = ao + (1 _ h)2' where ao E C is arbitrarily fixed.
Let Tj,K = {Tj,hhEK be a family of functional shifts for the Pommiez operator D
defined in (A4.2) induced by the function f. Then for all x E X, hE OC
aox(t) + h h~ tx'(h) + (h ~t)2 [x(t) - x(h)] for h #t E OCj

(Tj,hX)(t) ={ h2
aox(h) + hx'(h) + 2'x"(h) for t = h.
Proof By our assumption, the function f has the expansion
=L
00
f(h) ak hk = ao + 1h + 2h2 + 3h3 +... for hE K

k=O
We will show that Sj(D) = X. Let x E X, h E OC be arbitrarily fixed. The
equality
~ k kbn-k nan+! - (n + l)an b + bn+!
~ a =a (a _ b)2 for a, bEe, a # b, n E N
and Formula (A4.3) together imply that

00
n=l
=~ n hn ~ t k X(k) (0) =~x(k)(O) ~ n h n tk

~ tn ~ k! ~ k! ~ tn
n=l k=n k=l n=l
tk+l - (k + l)hkt + khk+! x(k)(O)
1
00
_ ~ h (h _ t}2 -k-!- for t # hj
- ~ k(k+1)h kx (k)(O)
~ 2 k! for t = h.
k=l
Appendix 289
Hence for t =I- h E IK we have

00
n=l
h ~ 2 x(k) (0) k-l x(k) (0) k
= (h _ t)2 L..J [h (k _1)!h + t-k!-t ]
f:
k=l
_t [h x(k) (0) hk-1 + x(k) (0) hk]

k=l (k - 1)! k!
h2 th
= h _ t x'(h) + (h _ t)2 [x(t) - x(h)]
and for t = h E IK
k=l
h2
= hx'(h) + -x"(h).
2
This proves that SI (D) = X. Moreover, this also shows that for all x EX, h E IK
+L
00
ao(Ix)(t) nhn(Dnx)(t) =
n=l
= {aox(t) + hh~ tx'(h) + (h :tt)2 [x(t) - x(h)] for h =I- t E IKj

h2
aox(h) + hx'(h) + '"2x"(h) for t = h.
Clearly, Tn.. is the unique family of functional shifts for D induced by the function
f and the operators Th (h E IK) are continuous (cf. Bin[lO]). •
Example A4.2. Take a = {fi}. Then the operators D,R E Lo(X) determined
by Formulae (A4.1) have the form
(A4.7) D = i, R= {t.
dt 10
In this case
(A4.8)
and every element xES can be written (in a unique manner) in the form
n n
(A4.9) x(t) =L bkt k =L ck Rk1 , (n E No),
k=O k=O
where Ck = k!b k , bk are scalars (k E No). 0

The results of Bin[l],[6] imply (cf. also books of PRZEWORSKA-RoLEWICZ
PR[7],[8])
Proposition A4.3. Let X = H(OC) and let OC = C. Let Tj,K = {Tj,dhEK be
a family of functional shifts for the operator D = -it
induced by the exponential
function, cosine, hyperbolic cosine, sine, hyperbolic sine, respectively. Then for all
x EX, h, t E OC
x(t + h) if f(h) = eh;

Hx(t + ih) + x(t - ih)] if f(h) = cos h;
(Tj,hX)(t) = ~[x(t + ih) - x(t - ih)] if f(h) = sinh;
Hx(t + h) + x(t - h)] if f(h) = coshh;
Hx(t + h) - x(t - h)] if f(h) = sinhh.
Remark A4.2. (cf. FN[l]) If
L
00
T = f3kDk, where 13k E C (k E No),

k=O
is an isomorphism of H (K), then for all x E H (K), t E OC
(f3x)(t + h) if OC = C, where 13, hE C;

(A4.10) (Tx)(t) = {
(,x)(t) if OC = OC1 , where, E C.
o
Proposition A4.3 and Remark A4.2 together imply
Corollary A4.1. LetOC = C. Let Tj,K = {Tj,hhEK be a family offunctionalshifts
induced by the exponential function. Let an operator T of the form (A4.1O) be an
isomorphism of the space H(OC). Then there are 13, hE C such that T = f3Tj,h'
Proposition A4.4. Let f E X = H(OC): f(h) = ao + heh, where ao E C is
arbitrarily fixed. Let Tj,K = {Tj,hhEK be a family of functional shifts for the
operator D = -itinduced by f. Then for all xES, h E OC
(A4.U) (Tj,hX)(t) = aox(t) + hx'(t + h) (t E OC).

Appendix 291
Proof The function f has the following expansion:
hk
=L = ao + L
00 00
f(h) ak hk (k _ I)! far hE ][{.

k=O k=1
Let h E ][{ and xES of the form (A4.9) be arbitrarily fixed. By (A4.8), for t E ][{
we have
n n n k
=T/,h L ckR k l = L ckT/,hRkl = L Ck L a j h j R k- j l
k=O k=O k=O j=O
n k
= L [L ajhj Rk-jl + aoRkl] + aoCo
k=1 j=1
n n k k'
-_ aoCo + ao '~
" Ck R k 1 + '~
" bk '~
" (. _ l)!(k
. _ .)! h j t k-j
k=1 k=1 3=1 J J
n k N
= aox(t) + (; bk ~ G)ihjtk-i = aox(t) + {; bkkh(h + t)k-1,
since the following formula holds:
This implies the required formula (A4.11). •
Proposition A4.5. Let ][{ = C and let f E X = H(][{): f(h) = ao + a1h + h 2e h ,

where ao, al E C are arbitrarily fixed. Let T/,K = {T/,hhEK be a family of
functional shifts for the operator D = ft
induced by f. Then for all XES, h E ][{
(A4.12)
Proof The function f has the expansion:
hk
=L = ao + a1 h + L
00 00
f(h) akh k (k _ 2)! far all hE IK.

k=O k=2
Let h E ][{ and xES of the form (A4.9) be arbitrarily fixed. By (A4.8), for t E ][{
we have
n k
=L Ck L ajh j R k- j 1
k=O j=O
1 k n 1
= L Ck L ajh j R k- j l +L Ck L ajh j R k- j l
k=O j=O k=2 j=O
n k
+ L Ck L aj h j R k- j 1
k=2 j=2
n
=boao +b1(aot+alh) + L bk(aot k +alhktk- l )
k=2
~ ~ k! 'k'
+ L...J bk L...J (' _ 2)!(k _ ,)!h3 t -3
k=2 j=l J J
n n
= ao L bk tk + alh L bkktk- 1
t t
k=O k=l
+ bk (~)j(j - 1)h t j k- j
k=2 j=2 J
L
n
= aox(t) + alhx'(t) + bkk(k -1)h 2 (h + t)k-2,

k=2
since the following formula holds:
This implies the required formula (A4.12).

Proposition A4.6. Let lK = C and let I E X = H(lK): I(h) = ao + heh + h 2eh,
•
where ao E C is arbitrarily fixed. Let T/,K = {T/,hhEK be a family of functional
shifts for the operator D = c1t induced by I. Then for all xES, h E lK
(A4.13) (TJ,hX)(t) = aox(t) + (h - t)x'(O) + tx'(t + h) + ex"(t + h) (t ElK).
Proof The function I has the expansion:
I(h) = {;
00
ak hk = ao + t;
00 khk
(k _ 1)! lor all hE lK.
Appendix 293
Let h E OC and xES of the form (A4.9) be arbitrarily fixed. By (A4.8), for t E OC
we have
n k
= L Ck L ajh j R k- j l
k=O j=O
1 k n k
= L Ck L ajh j R k- j l +L Ck L ajh j R k- j 1
k=O j=O k=2 j=l
n
= boao + b1(aot + h) + ao L bkt k
k=2
n k .k'
+ "L..t bk "L..t (. _ l)!(k
J . _ .)! h j tk- j
k=2 j=l J J
By the formula
t
k=O
(~) k 2akbn- k = nb(a + nb)(a + b)n-2 for all a, bEe, 1 < n E N,
for t E OC we get
n
= aox(t) + hx'(O) + t L kbk(h + kt)(h + t)k-2
k=2
n
= aox(t) + hx'(O) + t L kbk(h - t)k-l
k=2
n
+ t2 L bkk(k - l)(h + t)k-2
k=2
= aox(t) + hx'(O) + t[x'(t + h) - x'(O)] + t 2x"(t + h),
which implies (A4.13).
Example A4.3. Take the sequence a = {ad, where ao = 1, ak = (k~l)! for all
•
kEN. Then the operators D, R E Lo(X) determined by Formulae (A4.1) have
the form
(A4.14) (Dx)(t) = x'(t)+x'(O)- x(t) ~ x(O) , where x(t) ~ x(O) It=o = x'(O),
(Rx)(t) =t (t x(s) - x(O) ds far x E X, t E OC

10 s
(cf. Example A4.1j also Bin[6]). In this case
1 for k = 0
(A4.15) Rk1 ={ tk
(k _ 1)! for k E Nj
and every element xES can be written (in a unique way) in the form
n n
(A4.16) x(t) =L bktk = L ck Rk1 , Co = bo, Ck = (k-1)!bk' bk E C (k EN).
k=O k=O
o
Proposition A4.7. Let OC = C and let X = H(OC). Let T/,'ft{ = {T/,hhE'ft{ C
Lo(X) be a family of functional shifts for the operator D determined by Formula
(A4.14) induced by the exponential function f(h) = e h • Then for all XES,
h, t E OC
x(t + h) - x(O) + (h x(s) - x(O) ds + x(O) for t =I -hj

(T/,hX)(t) = { t +h fo s
x(O) - hx'(O) + ( x(s) - x(O) ds for t = -h.
10 s
It is obvious that
Similarly, as in Proposition A4.3, as a consequence of Proposition A4.7 (cf.

Bin[ll]) we can obtain explicit formulae for functional shifts induced by cosine,
hyperbolic cosine, sine and hyperbolic sine functions. Here, as an example, write
a formula for functional shifts for D induced by the hyperbolic sine function.
Namely, for xES, h, t E OC we have
1[x(2h) - x(O) - 2hx'(O)] +~

21_t x(s) - x(O) ds
s
for t = hj
ih
4
(T/,hX)(t) = { -h h
i[x(O) - x(-2h) - 2hx'(O)] +~ x(s) ~ x(O) ds for t = -h
2 _ x(t + h) - x(t - h) _ ~ x(t + h) - x(t - h) - 2x(O)
(T x)(t) = _ t
/,h t2 _ h2 2 t 2 - h2 2 +
+ (h x(s) - x(O) ds otherwise.
l-h s
o
Appendix 295
Taking into consideration Formulae (A4.15), (A4.16), in a similar way as Propo-

sition A4.4, we prove the following
Proposition A4.S. Let OC = C. Let f E X = H(OC) : f(h) = ao + he h , where
ao E C is arbitrarily fixed. Let Tj,JI( = {Tj,hhEK C Lo(X) be a family offunctional
shifts for the operator D determined by Formula (A4.14) induced by f. Then for
all xES, h, t E OC
aox( -h) + x(h) - x(O) - ~2 xl/ (0) for t = -hj

{
ht x t + h ) -h x 0
aox(t)+x(h)-x(O)+-h[x'(t+h)- ( ()) for t f. -h.
t+ t+
In a similar way as Proposition A4.5, we prove

Proposition A4.9. Let OC = C. Let f E X = H(OC) : f(h) = ao + a1h + h 2e h ,
where ao, al E C are arbitrarily fixed. Let Tj,K = {Tj,hhEK C Lo(X) be a family
offunctional shifts for the operator D determined by Formula (A4.14) induced by
f. Then for all xES, h, t E OC
=aox(t) + alhx'(O) + a1ht[ x(t) -t x(O) Y

+ h2[X(t) - x(O))'1 + h 2 t[x(t + h) - x(O))1/
t t=h t +h '
x(t) - x(O) ,
where, as before, t It=o = x (0).
In a similar way as Proposition A4.6, we prove
Proposition A4.10. Let OC = C. Let f E X = H(OC) : f(h) = ao + (h + h2)e h ,
where ao E C is arbitrarily fixed. Let Tj,K = {Tj,hhEK C Lo(X) be a family of
functional shifts for the operator D determined by Formula (A4.14) induced by f.
Then for all XES, h, t E OC
aox(-h) + x(h) - x(O) _ h 2 xl/(O) _ h 3 xll/(O) for t = -hj

2 3
= aox(t) + x(h) - x(O)+
t 2 (h-t)[x(h+t)-x(0)y t 3 I/(h )
+ h+t h+t + h+t X +t for t f. -h.
A5. Euler-Maclaurin type formulae.
The purpose of the present section is to give some analogues of the Euler-Maclaurin
Summation Formula (cf. MILLER Mil[l], PR[I]j also Chapter 11) induced by
functional shifts. We shall give examples of applications of formulae obtained for
few elementary functions (cf. Bin[14]).
Assume that X is a complete linear metric space, OC = OC(O, r), 0 < r :S: 00 and a
function f E H (OC) has the expansion
=L
00
(A5.1) f(h) ak hk for all h E lK.

k=O
Suppose that D E R(X), ker D ::J {O} and TO.. = {T/,hhEIK is a family of func-
tional shifts induced by the function f E H(OC). Consider the Pommiez type
operators defined as follows:
h-l(T/,h - T/,o) for 0 ::J h E OCj

(A5.2) s:}JT/,h ={ on X,
DT/(l) ,0 for h = OJ
h-l(f(h) - f(O)] for O::J h E OCj

(A5.3) (Pf)(h) ={
f(I)(O) for h = OJ
where, as usually, we write f(n) = d~: (n EN).

Observe that the function P f E H(OC) has the expansion
L
00
(A5.4) (Pf)(h) = akHh k for all h E OC.

k=O
This implies
(A5.5) P E Lo[H(OC)].
Note that the operator (Pg)(hD) for the function g(h) = eh is called the Bernoulli
operator for D (cf. PR[14]' also Chapter 11).
Using Proposition A2.2, it is easy to prove
Proposition A5.I. Suppose that DE R(X), ker D ::J 0 and T/,K = {T/,hhEIK is
a family of functional shifts for D induced by a function f E H(OC). Then
(A5.6) s:}JT/,h = (P f)(hD)D on S for all h E oc.

Appendix 297
The following theorem is fundamental for our subsequent considerations in the

present section.
Theorem AS.I. Suppose that D E R(X), ker D =I {O}, T/,K = {T/,hhEK is a
family offunctional shifts for D induced by a function f E H(IK), f(1)(O) =I 0 and
f(h) =I f(O) for hE IK \ {O}. Then the operator (Pf)(hD) is invertible on the set
S for all h ElK \ {O} and [(Pf)(hD)]-l = g(hD), where
L
00
g(h) = bkh k E H(IK),

k=O
the coefficients bk (k E No) are determined by the recurrence relations
all forn=Oi
(A5.7) b - {
L
n-1
n - -all bk a n -k+1 for n E N.
k=O
Proof. By our assumptions, the function
h[l(h) - f(O)]-l for h E IK \ {O}i

(A5.8) g(h) = {
[1(1) (0)]-1 for h = Oi
belongs to H (1K) and
(A5.9) (Pf)(h)g(h) = 1 far all hE lK.
Theorem A3.1 implies that the operators (P f) (hD) (h E 1K) are invertible on S
and [(P f)(hD)]-l = g(hD). Taking into account the expansion of the function g,
by Formulae (A5.4) and (A5.9), we get
[L HL L
00 00 00
1 = (Pf)(h)g(h) = ak+1 hk bkhk] = ckh k far all hE 1K,

k=O k=O k=O
where
L
00
em = bk am-k+1 far m E No.

k=O
This implies Formulae (A5.7). •
Theorem A5.1 and Formula (A5.4) together imply
Theorem AS.2. Suppose that all assumptions of Theorem A5.1 are satisfied.
Then Tpf,K is invertible in the algebra Ts(lK) and its inverse is (Tpf ,K)-l = Tg,K E
Ts (1K) , where the function 9 is defined by Formula (A5.B).
We are now ready to give the main theorem of this section, which is an immediate
consequence of Theorem A5.2 and Formula (A5.6).
Theorem A5.3. Suppose that all assumptions of Theorem A5.1 are satisfied.
Then the following Euler-Maclaurin type formulae hold on the set S:
+L
00
(A5.1O.a) 1= bo(Pf)(hD) bnhn(pf)(hD)Dn,

n=l
+L
00
(A5.1O.b) 1= bo(Pf)(hD) bnhn~Tf.hDn-l,

n=l
where 0 'I h E lK, bk (k E No) are determined by Formulae (A5.7).
Example A5.1. (cf. Theorem 11.4 and Formula (11.12), also PR[14]) Let lK = C
and let f(h) = eh • Then the Euler-Maclaurin Formula (A5.10.b) has the form
1= (PeS)ls=hD + f:
n=l
Bn h:~l (e hD - I)D n - 1 for hE lK \ {O},
where Bn are Bernoulli numbers. 0

Example A5.2. The expansions of elementary functions used here can be found
in GRADSTEIN and RYZHlK GR[l]' PRUDNIKOV Pru[l].
(a) Let lK = lK.r/2 and let f be the tangent function. Then for h ElK \ {O}
tan h =~ 22n(22n - 1) IB Ihn- 1 (P tan)(h) = h- 1 tan h,

L (2n)! 2n ,
n=l
00 22n
g(h) = [(Ptan)(h)]-l = h(tanh)-l = 1- ~ (2n)! IB2nlh2n,
where, as before, B 2n are Bernoulli numbers. In this case Formula (A5.1O.b) has
the form
I = tan(hD) _ ~ 22n IB Ih 2n - 1 tan(hD)D2n-l (O....L h E lK)

hD ~ (2n)! 2n T'
(b) Let JK. = lK.r /2 and let f be the hyperbolic tangentfunction. Then for s E JK.\ {O}
00 22n(22n-l _ 1)
tanh(s) =~ (2n)! B2ns2n-l,
Appendix 299
where B 2n are Bernoulli numbers. In this case Formula (A5.1O.b) has the form
1= tanh(hD) +~ 22n B h 2n - 1 tanh(hD)D 2n - 1 (0 4 h E OC).

hD ~ (2n)! 2n , -r
(c) Let OC = OC and let f

7T be the sine function. Then for h E lK. \ {O} we have
(P sin) (h) = h- 1 sin hand
2(2 2n - 1 1)
g(h)
h
sm = h esc h = 1 + L
= --:--h
00
n=1
()'
2n.
-
IB2n lh 2n ,
where B 2n are Bernoulli numbers. In this case Formula (A5.1O.b) has the form
(d) Let OC = OC7T and let f be hyperbolic sine function. Then for h E lK. \ {O} we
have (Psinh)(t) = t- 1 sinht and,
t 00 2(2 2n - 1 - 1)
g(t) = sinht = 1- ~ (2n)! B 2n t 2n ,
where B 2n are Bernoulli numbers. In this case Formula (A5.10.b) has the form
1= sinh(hD) _ ~ 2(2 2n - 1 - 1) B h2n- 1 sinh(hD)D2n-1 (h E OC \ {a}).

hD ~ (2n)! 2n
n=1
o
Remark A5.1. Suppose that all assumptions of Proposition A5.1 are satisfied
and f(h) i- 0 for hE K Then, by Theorem A3.1,
L
00
1= Ck hk f(hD)D n on S (h E lK. \ {O}),

k=O
where Ck are determined by the recurrence relations:
Co = ao-1 , far n E N.
o
Assume that either lK. = C or lK. is the unit disk lK.1 . In the sequel we will show
that the Euler-Maclaurin type Formulae (A5.1O) hold for the sets
• ker(D - >.1), A E lK,
• Soc(D) = n S';(D),
gEH(OC)
• AR(D), R E RD.
Let DE R(X) and let ker(D - >.1) =f. {O} for h E lK\ {O}. We rewrite the Pommiez
type operator s,p defined by Formula (A5.2) as follows
J(hD) - J(O)I for h E lK \ {O};

(A5.11) s,pTI,h = { h
J(1) (O)D for h = o.
The equality Dx = AX for X E ker(D - >.1) and Formula (A2.4) together imply
that on ker( D - AI) we have
J(Ah) - J(O) 1 for hE lK \ {O};

(A5.12) s,pTI,h = { h
Af(1) (0)1 for h = O.
Clearly, Formula (A5.12) implies that the operator D commutes with s,p on the
set Tker(D-'xI)(lK) (d. Theorem A3.2).
A result similar to Proposition A5.1 is
Proposition A5.2. Suppose that D E R(X), ker D =f. {O} and TI,oc = {TI,hhEOC
is a family of functional shifts for D induced by a function J E H(lK). Suppose,
moreover, that ker(D - >.1) =f. {O} for a A E lK. Then
(A5.13) s,pT/,h = (PJ)(hD)D = (DPf)(hD) on ker(D - >.1) Jor all hE lK,
where the Pommiez operator P E Lo(H(lK)) is defined by Formula (A5.3).

Proof. Observe that Formula (A5.3) and Corollary A1.3 together imply that
ker(D - AI) c SPI(D), where SPI(D) is determined by Formula (A1.I). By
definition and Formula (A5.4), we get on ker(D - )..I) for all h E lK \ {O}
= (Pf)(hD)D = [L = h- 1 L
00 00
(DPf)(hD) anhn-lDn-l]D anhnD n =

n=l n=l
= h-1[J(hD) - J(O)/] = h-1[J()"h) - = s,pTI,h.

J(O)]I
By definition, Formula (A5.3) holds on ker(D - AI) also for h = O. •
Lemma A5.1. Suppose that D E R(X), ker D =f. {O}, R E RD and TI,K =
{TI,hhEOC is a family of functional shifts for D induced by a function J ElK. Then
(A5.14) (PJ)(hD) = s,pT/,hR on SPIeD) Jor all h E lK.

Appendix 301
Proof. Let x E Spf(D) and hE lK\ {O} be arbitrarily fixed. By (A1.l) and (Al.4),
(Pf)(hD)x
L =L
00 00
= ak+1hkDkX ak+lhkDkDRx
k=O k=O
= h-1[f(hD)Rx - aoRx] = h-1[f(hD) - f(O)I]Rx = I.l3Tf,h.
Clearly, Formula (A5.l4) holds also for h = O.
Observe that, by the last equality, Rx E Sf(D) whenever x E SPf(D). Conversely,
x E Spf(D) whenever Rx E Sf(D). •
A consequence of Lemma A5.l is
Proposition A5.3. Suppose that all assumptions of Lemma A5.1 are satisfied.
Then Equality (A5.14) holds on the set Soc.(D) determined by Formula (A3.4).
Clearly, by definitions and Corollary A1.3, we have
Applying Theorem A3.2, in a similar way as Theorem A5.l (cf. Theorem A5.2),
we prove
Theorem A5.4. Suppose that D E R(X), ker D ¥- {O} and Tf,oc. = {Tf,hhEK
is a family of functional shifts for D induced by a function f E H (lK). Let Y =
ker(D - >d) ¥- {O} for a A E lK. Let f(1)(O) ¥- 0 and f(h) ¥- f(O) for h E
Iy
lK \ {O}. Then Tpf,oc. is invertible in the algebra Ty(lK) = T(lK) and its inverse is
(Tpf,oc.)-l = Tg,oc. E Ty(lK), where the function 9 is determined by Formula (A5.8).
Theorem A5.4 and Proposition A5.2 together imply
Theorem A5.5. Suppose that all assumptions of Theorem A5.4 are satisfied.
Then Formulae (A5.1O.a), (A5.lO.b) hold on ker(D - >d).
Remark A5.2. Clearly, by Theorem A3.2, if all assumptions of Proposition A5.2
are satisfied and f(h) ¥- 0 for h E lK, then Formula (A5.11) holds on ker(D - AI).
o
Corollary A5.1. (cf. PR[8]) Suppose that all assumptions of Proposition A5.2
are satisfied, R E RD and A E lK n vcR. Then
E).,(R) = (I - AR)-l ker DC ker(D - >d).
Corollary A5.l implies that Theorems A5.4, A5.5 and Remark A5.2 hold also for
E).,(R). If J[( = C then Theorems A5.4, A5.5 and Remark A5.2 hold for the set
E= E(R). Clearly, if D has a Volterra right inverse R then Corollary A5.l holds
for all A E C (cf. PR[8]).
Definition (A3.4) of the set Soc.(D), Formula (A5.5) and Proposition A2.5 together
imply
Proposition A5.4. Suppose that all assumptions of Theorem A2.1 satisfied and
the operator D is closed. Then Formula (A5.13) holds on the set SK(D).
Applying Theorem A3.4, in a similar way as Theorem A5.I (cf. Theorem A5.2),
we prove
Theorem AS.6. Suppose that X is complete linear metric space, D E R(X) is
closed, ker D -:j:. {O}, F is an initial operator for D corresponding to an R E RD,
T/,K = {T/,hhEK is a family offunctiona1 shifts induced by a function f E H(IK),
f(1)(O) -:j:. 0 and f(h) -:j:. f(O) for h E IK \ {O}. Let Y = SK(D). Then Tp/,K is
invertible in the algebra Ty and (Tp/,K)-l = Tg,K E T y , where the function 9 is
defined by (A5.B).
Theorem AS.7. Suppose that all assumptions of Theorem A5.6 are satisfied.
Then Formulae (A5.lO.a), (A5.lO.b) hold on the set SK(D).
Remark A5.3. Suppose that all assumptions of Theorem A2.I are satisfied, the
operator D is closed and f(h) -:j:. 0 for hE lK. By Theorem A3.4, Formula (A5.H)
holds on SK(D). 0
Proposition A5.5. Suppose that X is complete linear metric space, D E R(X),
ker D -:j:. {O}, R E RD is continuous and T/,K = {T/,hhEK is a family of continuous
functional shifts for D induced by a function f E H (1K). Then
(A5.I5) ~T/,h = (Pf)(hD)D on AR(D) far all hE 1K.
Proof. Fix x E AR(D) and set y = Dx. The continuity of R E RD implies that
AR(D) c SPIeD), D[AR(D») c AR(D). The first inclusion follows from Theorem
A2.2 and Formula (A5.5), the second one is proved in PR[9], Theorem 3.2. This,
and Formula (A5.4) together imply that for h E IK \ {O} we get
[(Pf)(hD)D)x = (Pf)(hD)y
= L: = L:
00 00
anh n - 1Dn-1y anh n- 1D n- 1Dx

n=l n=l
n=l
=h-1[f(hD) - f(O)I]x = ~T/,hX.
For h = 0 Formula (A5.I5) follows from the definition.
Theorem A5.8. Suppose that all assumptions of Proposition A5.5 are satisfied,
•
f(1)(O) ¥ 0 and f(h) -:j:. f(O) for h E IK \ {O}. Then Th,K is invertible in the
algebra TAR (D) and [Th,K]-l = T;,K E TAR(D)' where the function 9 is defined by
(A5.B).
Appendix 303
Theorem A5.9. Suppose that all assumptions of Proposition A5.5 are satisfied,
f(l)(O) f=. 0 and f(h) f=. f(O) for h E lK. \ {OJ. Then Formulae (A5.1O.a) and
(A5.1O.b) hold on AR(D).
Remark A5.4. Suppose that all assumptions of Proposition A5.5 are satisfied
and f(h) f=. 0 for h E lK.. By Theorem A5.S, Formula (A5.11) holds on AR(D). 0
A6. Differential and integral properties.
In the present section algebraic, differential and integral properties of functional

shifts defined on the set S of D-polynomials are studied. In particular, gener-
alizations of the Cauchy theorem and Cauchy integral formula for shifts under
consideration are given (cf. Bin[4],[5]).
Let X be a linear space over C and let P be a polynomial in h E C of the form
m
P(h)=L xj(h-a)j, wherexjEX{j=O,I, ... ,m), mEND, aEC.

j=o
Write
o form=O;
8 {~m
(8h P) (h) = jXj(h - a)j-1 for m ~ 1;
The operators
8 8- 1
8h'
8h
are said to be the first formal derivative and the formal integral of P with respect to
h, respectively. If these operators are applied n times, then the resulting operators
are denoted by
8h n '
8h n '
and are called the nth formal derivative and the nth formal integral, respectively.
Let OC = OC(a;r) (0 < r::; 00). Write for j E H(OC), hE OC
where, as usually, j(n) = d~n (n E N) and j(O) = j.

It is well-known that j(m) E H(OC) for any m E Z. It is easy to verify the following
fundamental properties:
(A6.1) j(k+1) = (j(k))(l), j(-k-l) = (j(-k))(-l), (j(-k-1))(1) = j(-k) (k END),

[J(k+1)(h)](-l) = j(k)(h) - j(k)(a) (h E OC).
Appendix 305
Definition A6.I. Suppose that D E R(X) and ker D :I {O}. Let {l ~ C be a

simply connected domain, lK = lK(a;r) ~ {l (0 < r ~ 00) and let f E H(lK):
=L
00
(A6.2) f(h) ak(h - a)k . far all h ElK.

k=O
We say that T;:i= {TJ:~hEOC (p E Z) of linear operators belonging to is a L(X)

family of extended functional shifts for D induced by f if the following conditions
are satisfied for all n E N, h E lK, xES:
L
00
ak(h - a)k Dkx for p = 0;

k=O
L
00
T;:~x = (k + l)(k + 2) ... (k + n)an+k(h - a)k Dn+kx for p = n;

k=O
00
L (k _ n + l~(k~ n + 2) ... k (h - a)k Dk-nx for p = -no

k=n
We admit here and in the sequel
If it does not lead to any misunderstanding, then we shall call extended functional
shifts briefly functional shifts. 0
Remark A6.I. By definition of the set S, the sums in Definition A6.1 consists
only of a finite number of terms different than zero. 0
In a similar way as Theorem A2.1, we prove
Theorem A6.I. Suppose that all assumptions of Definition A6.1 are satisfied,
T;:i (p E Z) is a family of extended functional shifts for Dinduced by f
and is F
an initial operator for D corresponding to an R E RD. Then
(i) for all h E lK, kENo, n E N
L
k
aj(h - a)i Rk-j F for p = OJ
j=O
o for p = n > kj
k
L j(j - 1) ... (j - n + l)aj(h - a)j-n Rk-j F for p = n ~ kj
j=n
k
" aj (h _ a)i+n Rk-j F for p = -nj
f::o (j + l)(j + 2) ... (j + n)
(ii) for all h E ][{, kENo, z E ker D and n E N
L
k
aj(h - a)jRk-jz for P = OJ
j=O
o for p = n > k;
L
k
j(j - 1) ... (j - n + l)aj(h - a)j-n Rk-j z for p = n $ kj
j=n
L
k
aj (h - a)i+n Rk-j z for p = -nj
j=O (j + l)(j + 2) ... (j + n)
w;;i
(iii) if a family = {W;;~hEK c L(X), p E Z, satisfies for an Rl E RD and
for all h E ][{, kENo, z E ker D, n E N on the set S the conditions
L
k
aj(h - a)j R~-j z for p = 0;
j=O
o for p = n > kj
L
{p)Rkz - k
W f,h 1 -
j(j - 1) ... (j - n + l)aj(h - a)j-n R~-j z for p =n $ kj
j=n
k
" aj (h _ a)i+nRk-jz £or p
~ (j
j=O
+ 1)(j + 2) ... (j + n) 1 = - n',
then
W{p) -
f,h -
T{P)
f,h
on S far all h E ][{, p E Z.
Remark A6.2. Observe that the definition of the family (p E Z) depends T;;i
essentially on the disk][(' For example, let f(h) = h (h E q. Then the function
has the Taylor expansion
h for h E ][{' = ][((O; 00);

f(h) = { 1 + (h - 1) for h E ][{" = ][((1; 00).
Hence for z E ker D and R E RD we get
hz if Tf,h E Tf,K';
Tf,h Rz = { Rz + (h - l)z 1'f T f,h E T f,K" .
o
An immediate consequence of the definition are
Proposition A6.I. Suppose that all assumptions of Theorem A6.1 are satisfied.
Then
Appendix 307
(i) the operators Tj:;:), mE Z, h E OC, are uniquely determined on the set 8;
(ii) if X is a complete linear metric space, S = X and Tj:;:) (m E Z) are continuous

for h E OC, then they are uniquely determined on the whole space X.
Then the operators Tj:;:) (m E Z) of the functional variable 1 E H(OC) and the
complex variable h E OC, as operators acting in the space Lo(8), have the following
properties:
(i) T~7~1'9,h = >'T):;:) + /LT;:;:) for all I,g E H(OC), >',/L E C, hE OC;
(ii) for 1 E H(OC), hE OC, and n E N
T(-n) _ 8- 1 T(-n+1) _ 8- n T .
',h - 8h ',h - 8h n ',h,
(iii) the operators T):;:) commute with D on the set 8 for all h E II{.
We admit here and in the sequel that all assumptions of Definition A6.1 are satis-
fied and TjZ:?,K (m,p E Z) is a family of extended functional shifts for D induced
by I.
Corollary A6.1. Let n E N. Then for all hE OC we have on the set 8
T(m)
,(n-1) ,h for mEN;
T,(m) - {
(n) -1 h - -m
[J 1, T(m) _ l(n-1)( ) (h - a) I
for -m E No.
,(n-1),h a (-m)!
Proof. Let l(h) == 1 for hE OC. For all h E OC we have on S
o for mEN;
T (m)
1,h =
{ (h - a)-m
(-m)! I for -m E No·
Formula A6.1 and Proposition A6.2(i) together imply that on 8 for all h E OC we
have
Lemma A6.1. Let n E N. Then T);hn) = T,(-n),hRn on 8

•
for all hE OC, R E 'RD.
Proof. Let hE lK, z E ker D, kENo, R E RD be arbitrarily fixed. By Theorem

A6.1(i),
k+n .
[Tf(-n),h R nJRk Z -- T f(-n),h R k+ n Z -_ " "
L...J .
aJ - n (. _ 1)(' a)3
(h -
_ 2). Rk+n-j z,
.
J=n
J n + J n + ...J
T (-n)R k
f,h Z
~ (h - a)i+n k- .
=L...J a· R Jz
j=O J (j + 1) ... (j + n)
k+n ).
= "" J. n (h - a J RHn-j
~ a - (j _ n + l)(j _ n + 2) ... j Z
J=n
= [Tf(_n),hRnjRkz.
Lemma A6.1 and the equality [j(-m)j(-n) = f(-m-n) for m, n E No together imply
•
Proposition A6.3. Let m, n E No. Then
T f(-m),h
(-n)
= Tf(-m-n),h R n S
on, h E lK, R E RD are arbitrary.
An easy consequence of our definitions is

Lemma A6.2. Let n E N. Then Tj~n) = T;~~~~) on S for all h ElK.
Proposition A6.4. Let n E N, m E No. Let R E RD. Then for all h E lK
(-m-l) _ [T(-m)
T f(n),h (n - I)! (h )mljR S
- f(-n-l),h - m! an-l - a on .
Proof. Observe that f(q)(a) = q!a q and Ti"q) = (q!)-l(h - a)Q[ on S for all
q E No, where, as before, l(h) == 1 on lK. Fi~ h E lK and R E RD. Lemma A6.2
and Corollary A6.1 together imply that on S
T (-m-l) _ IT'(-m) R _ [T(-m) (n - I)! (h )mlj

f(n),h - .L[f(n)]-l,h - f(-n-l),h - m! an-l - a R.
Lemma A6.3. Let n E N and R E RD. Then for all h E lK

•
(A6.3)
Proof, by induction. Let h E lK, Z E ker D, R E RD be arbitrarily fixed. Observe

that
(A6.4) T(n) - T(n+l) (n EN).

f(1),h - f,h
Appendix 309
Indeed, fix an n E N. Then for k ~ n
[ T(n+l) R]Rkz
f,h
= TthH ) RkH z
k+1
= L j(j - 1) ... (j - n)aj(h - a)j-n-1 Rk+1-j z
j=n+1
L
k
= (j + l)j ... (j + 1 - n)aH1 (h - a)j-n Rk-j z
j=n
=E
k
j(j - l) ... (j - n + l)[(j + l)ajH](h - a)j-n Rk-jz
j=n
(n) Rk
T !(I),h
= Z.
By Theorem A6.1(ii), for n > k ~ 0 we have [Tj~h+1) R]Rkz = 0 and Tj~»,hRkz = O.

If n = 1 then for kENo
[T(1) R]Rkz
j,h
k+1
= TX2Rk+1 z =E jaj(h - a)j-1 RkH-j z
j=1
k
= '
~ " ' k-JZ=Tj(I),hR
(j+l)aH1(h-aFR ' k z.
j=O
Suppose Formula (A6.3) to be true for an arbitrarily fixed n ;::: 1. Then, by (A6.4),
on S
T j(n+I),h = 7'
.L(j(l)](n),h = T(n)
j(l),h
Rn
= T(nH)RRn
!,h = T(n+1)Rn
j,h
+ 1.
Lemma A6.3 and the equality [j(m)](n) = j(m+n) for n E Nand m E Z together
•
imply
Proposition A6.5. Let n E N, mE Z, R E 'RD. Then for all hE lK
Theorem A6.2. Let q,m E No, p E Z. Then for all h E lK we have on S
(A6.5)
(A6.6) n qT(m) - nq+mT

!(p),h - f(p+m),h'
Proof. Let h E 1K, R E 'RD be arbitrarily fixed. Propositions A6.2(iii) and A6.5
together imply that on S for q, mE No and p E Z we have
D qT(m) - DqT(m) DmR m - Dq+mT(m) R m - Dq+mT

j(p),h - j(p),h - j(p),h - j(p+m),h,
=T(m+ q) Dm+qRm+q
j(p-q),h
- Dm+qT(m+ q)
- j(p-q),h
-- Dq+mTj(p-q+m+q) ,h
-- Dq+mTj(p+m),h'
Theorem A6.3. Let q E No, m,p E Z. Then for all hE IK Formula (A6.5) holds
•
on S. If, in addition, q + m ~ 0 then for all h ElK Formula (A6.6) holds on S.
Proof. Fix h E IK and R E 'RD. Propositions A6.2(iii) and A6.3 together imply
that
D qT(m)
j(p),h = DqTj(p+m),h R- m = T j(p+m),h DqR- m .
Again by Proposition A6.(iii) and by Theorem A6.2, for q ~ -m we get
-- T j(p+m),h Dq+m -- Dq+mTj(p+m),h

- T(q+m) _ T(q+m)
- j(p+m-q-m) ,h - j(p-q) ,h'
For q < -m, by Proposition A6.3, we have on S
q) T(m+ q)
T j(p+m),h D qR-m = T j(p+m),h R-m-q = T(m+
f(p+m),h = j(p-q),h'
Remark A6.3. Otherwise (i.e. if q, -m E No, PEN), a formula for DqT;;;,~,:-l)

•
(h E 1K) can be found by means of the recurrence relations in Proposition A6.4.
For all h E IK we have on S
(-m-l) - T(-m)
DTj(-p),h (p - 1)1 (h )m]
- j(-p-l),h - ml ap-l - a .
o
Proposition A6.6. Let n E N. Then for all h E IK
Appendix 311
Proof. Let h E OC, R E RD be arbitrarily fixed. Propositions A6.(ii), (iii) and

A6.5 together imply that on S
In T _ T(n) _ T(n)DnRn - DnT(n)Rn - DnT

Jh n i,h - i,h - i,h - i,h - i(nl,h'
Proposition A6.7. Let n E N. Then for all h E IK, R E RD

•
15- 1 T(-n) - T(-n) R S
Jh i,h - i(1l,h on.
Proof Fix h E IK, R E RD. Proposition A6.2(ii), (iii) and Theorem A6.3 together
imply that
0- 1 T(-n) - T(-n-l) - DT(-n-l) R - T(-n) R S

Jh i,h - i,h - i,h - i(-1l,h on.
Example A6.1. We admit all assumptions of Definition A6.1 with a = 0 and

•
IK = <C. Let q E Z and let R E RD. Write fh = eh . Suppose that we are given
families
)q) - T(q) c(q) - T(q) s(q) - T(q)
'-I{ - .,lK' lK - cos,iK' K - sin,lK'
h (q) - T(q) h(q) - T(q)

C lK - cosh,iK' S iK - sinh,iK'
Now we compute derivatives and integrals for these functional shifts. The equality
TPZ = 0 on S for h E IK and Theorem A6.2 together imply that on S for all h E IK
we' have
(A6.7)
C~I) = -Ds h , s~l) = DCh, ch~l) = Dshh, sh~l) = Dchh.

By Lemma A6.1 and the equality T-1,h = I on S for h E IK, we get
S~-I) = (I - ch)R, Sh~-I) = (Chh - I)R.
By Proposition A6.6 and Formulae (A6.7), we conclude that formal derivatives

and formal integrals of functional shifts under consideration are on S of the form
(for all h E IK)
(A6.9) (n EN),
By Proposition A6.2 and Formulae (A6.8) we have on S for all h E IK
(A6.1O)
o
Let an operator T(rn) mapping the space X into the set of operators defined on OC
and with values in X be defined by means of the equality:
T ,(rn) x = {T(rn)}
',h X hEIK
j or x E X .
Write for all h E IK, x E X
(A6.11)
This definition immediately implies

Proposition A6.8. Let m E Z. Then the function Tj(x; h) of the complex
variable hE OC and with values in X defined by Formula (A6.11) has the following
properties:
(i) Tjrn) (Ax + J.-LY; h) = ATjm) (x; h) + J.-LT;m)(y;h) for all X,y E X, A,J.-L E Cj
(ii) for all xES
(iii) if OC1 = IK(b; rI) ~ lK then T,(x; h) = TjO) (x; h) has the Taylor expansion
~ (h - b)k (k)
T,(x;h)=L...J k! T, (x;b) jarhEOC1,xES.
k=O
Lemma A6.4. Let X be a complete linear metric space and let X' be the space
of all linear functionals defined on X. Let D E R(X), ker D =I {O}. Let kENo,
z E ker D, x' E X', R E RD be arbitrarily fixed. Then x' [T,(Rkz; .)] as a complex-
valued function is holomorphic in IK and
Proof. Let h E ][{, R E RD be arbitrarily fixed. Let {(n} be a sequence such that
h + (n E lK for all n E N and lim (n = O. Write t = h - a, tn = t + (n' By
n-too
Theorem A6.1(ii), we have for k ~ 1
Appendix 313
k k
= l!..m
n~oo
(;:;-1 [x' (2: ajt~Rk-j 2: ajtj Rk-j z)]
Z -
j=O j=O
k k
= J~~ (;:;-1[2: ajt~x'(Rk-jz) - 2: ajtjx'(Rk-jz)]
j=O j=O
k
= J~~ (;:;-1 [2: aj(t~ - tj)x'(Rk-j z)]
j=l
On the other hand,

X'[T,(I) (R kZj h)]
k
= x' [D 2: (j + l)aj+1 (h - a)j Rk-j z]
j=O
k k
= x' [2: (j + l)aj+1tj R k-j-1 z] = x'[2: jajt j - 1Rk-jz]
j=l j=l
k
- . J·t j - 1x '(Rk-j Z.
" Ja
- ~ )
j=l
If k = 0 then (by Theorem A6.1 for p = k = 0) we have T,(zj h + (n) = ao =
T,(zj h), which implies
lim (;:;-1 [T,(z; h + (n) - T,(z; h)] = 0 and x'[DT,(1)(z; h)]
(n---+O
= x'(O) = o.
Proposition A6.9. Suppose that all assumptions of Lemma A6.4 are satisfied
•
and x E 5, x' E X' are arbitrarily fixed. Then x'[T, (x; .)] as a complex-valued
function is holomorphic in ][{ and
d~ x'[T,(x; h)] = x' [DT,(l) (x; h)] (h ElK).
Proof Since x E 5, there are n E No,

n
Zk E ker D (k = 0,1, ... , n) such that
x= 2: Rkzk. Since
k=O
n n
x'[T,(xj h)] = x'[T,(2: Rkzk)] = x'[T,(Rkzk)], 2:
k=O k=O
by Lemma A6.4,
d~ x'[T,(xj h)]
d
2:
n
= dh X' [T,(R k Zk)]
k=O
= t d:
k==O
x' [T,(Rkzk j h)] = t
k=O
x'[DT,(l)(Rkzk)]
n
= x' [DT,(1) (2: RkZkjh)) =x'[DT,(l)(xjh)].
k=O
Lemma A6.5. Suppose that all assumptions of Lemma A6.4 are satisfied. Then
•
the mappingT,(Rk zj ') is strongly holomorphic in][{, i.e. for every hE ][( the limit
T,(Rk Zj h) - T,(Rk Zjw )
lim
w-th h- w
exists and
(A6.12) d k k
dh T,(R Zj h) = DT,(l)(R Zj h) far all h E ][{.
Proof Let h E ][{, R E RD be arbitrarily fixed. Let {(n} be a sequence such that
h + (n E ][{ for all n E N and lim (n = O. Write
n--+oo
(h + (n - a)j - (h - a)j .
tn,j = (n aj (J = 0,1, ... ,kj n EN).
Then lim tn j = j(h - a)j-1 aj (j = 0,1, ... , k) and

n~oo '
lim (;l[T,(R kZjh+(n) -T,(Rkjh)]
n-too
k
- 2: aj(h-a)iRk-jz]
j=O
= n--+oo
lim "t
k
L.J n ,)
·Rk-j Z
j==O
=" [lim
k
L.J n-too
tn ,)·]Rk-j Z
j==O
k
= 2: jaj(h - a)j-1 R k - 1-j Z
j=1
= DT,(1) (Rkzj h).
Appendix 315
Formula (A6.12) is true also for k = 0 (cf. the proof of Lemma A6.4). •
Lemma A6.5 implies (cf. the proof of Proposition A6.9)
Proposition A6.10. Suppose that all assumptions of Lemma A6.4 are satisfied
and xES is arbitrarily fixed. Then the mapping Tf(x;·) is strongly holomorphic
in OC and
d
dh Tf (x; h) = DTf (1) (x; h) for all h E K
Propositions A6.9 and A6.10 immediately imply

Theorem A6.4. Suppose that all assumptions of Lemma A6.4 are satisfied and
xES, x' E X' are arbitrarily fixed. Then
(i) for all x' E X' the function Xl [T;m) (x; .)] is holomorphic in OC and
d~ x' [T;m) (x; h)] = x'[DTj~;(x; h)] (h E OC);
(ii) the mapping T}m) (x; .) is strongly holomorphic in K and
d T(m)( . h) - DT(m)( . h) - DT(m) x - T(m+1) (h E OC);

dh f x, - fell X, - j(l),h - f,h x
(iii) for x' E X'
Theorem A6.4 implies

Theorem A6.5. Suppose that all assumptions of Lemma A6.4 are satisfied. Then
(i) ThemappingT;m)(x;.) is continuous on OC, i.e. lim T;m)(x;h n ) = T;m)(x;h)
hn-+h
for every fixed xES;
(ii) for all n E N
d~n Tf(x; h) = DnTf(n) (x; h) = Tt) (x; h) (h E OC).
In order to study integral properties of T(x; h) in a complete linear metric space,

we introduce subspaces Xo = ker D and
n
Xn={x=L RkZk:Zk EkerD\{O}} CkerD n +1 cS (nEN),
k=O
where R E RD is arbitrarily fixed. Clearly, if xES then there is an n E N such

that x E X n .
Definition A6.2. Let an arc r c ][ be given. Let X be a complete linear metric

locally convex space. A function 9 : r -* X is said to be integrable over r if there
is ayE X such that
x*y = l (x*g)(()d( for all x* E X*,
where X* is the space of all continuous linear functionals over X. Elements y

satisfying this equality will be denoted by
l g«()d(.
Since the space X* is total, i.e. X* separates points of X (cf. Lemma 4.1), the
element y is uniquely determined and called the integral of 9 over r. 0
Proposition A6.11 Let r c ][ be an oriented arc with the initial point band
the final point c. Let X a complete linear metric locally convex space. Then for
every xES the integral fr T,(x; h)dh exists. If x E Xn then
where
(c - a)k+l - (b - a)k+l
Ck = ak k +1 (k = 0,1, ... , n).
Proof. Let
n
X =L Rk Zk E Xn C S, x· E X*
k=O
be arbitrarily fixed. By definition and Proposition A6.9, we have
r x*[T,(x; h)]dh
lr
n n
=L ckx*(Dkx) = x*(L ck Dkx ).
k=O k=O
o
Appendix 317
Recall that the index of awE C over a rectifiable closed arc f c IK is
1
IndrCw) = -2'
ITt
i
r
d(
-(-.
- w

Theorem A6.6. Let X be a complete linear metric locally convex space and let
xES be arbitrarily fixed. Then the Cauchy Theorem and the Cauchy Integral
Formula hold for the mapping TJ(x; .), i.e. we have
(i) if f is a rectifiable closed arc in IK such that
IndrCw) 1. { ~ =0
= -2 far all w t/. IK,
ITt lr (- w
£ £T~~;~)
then
Tf(x; ()d( =0 and Tf(x; h) = 2~i d(
for an arbitrary hE IK \ f which satisfies the condition Indr(h) = 1;

(ii) iffl and f2 are rectifiable closed arcs in IK such that Indrl(w) = Indr2(w)
for every w t/. IK then
Theorem A6.6 can be extended for the whole space. For instance, we have
Corollary A6.2. Suppose that all assumptions of Theorem A6.6 are satisfied.
IES = X and {AhhEK C Lo(X) is a family of uniformly continuous operators
I
such that Ah s = Tf,h for h E IK, then
We shall consider now an indefinite integral of the mapping TJ(x;,) for xES.
Theorem A6.7. Suppose that all assumptions of Theorem A6.6 are satisfied. For
all XES, R E RD
Proof. Let xES, R E RD be arbitrarily fixed. The there is an n E N such that

x E X n . Proposition A6.11 implies that
By definit.ion, since Ex E S, we get
TjC-I) (Rx; h)
(h - a)k
=L
n+1 k
ak-l k D Rx
k=l
=L
n+l
ak-l (h - a)kk-l Dk-1X
k=l
(h _ a)k+l
L
n
= ak k 1 Dkx
k=O +
= lh Tj(x; ()d( (h E OC).
The equality T;-l) (x; h) = Tj(-l)(Rx;h) for hE II{ follows from Lemma A6.1. •
Propositions A6.6 and Proposition A6.1O together show relations between the
formal derivative, the strong derivative and the operator D. Namely, we have
For all h E OC
J d
c5h Tj,h = dh Tj,h = DTjCI) ,h on S.
Propositions A6.7 and Theorem A6.7 together show relations between the formal
integral, the integral and a right inverse of D.
For all h E OC
Extensions of Corollaries A6.3 and A6.4 to the whole space X can be obtained in
a similar way as Corollary A6.2.
Similarly, as Theorems 18.2 and 18.3, we get
Corollary 18.1. Let j, g E H (OC). Then the Leibniz condition holds in the
following sense:
Corollary 18.2. Suppose that all assumptions of Theorem A6.6 are satisfied. Let
j, 1/ j E H(K). Then for all h E OC
{} d
c5h 1inj,h = dh 1inj,h = DT(lnf)(I),h = DT/'Ij,h on S,
Appendix 319
i.e.
Remark A6.4. In general, we do not obtain a result similar to Corollary 1B.1,

since the following equalities hold for all f,g E H(IK), hE 1K:
(n)T(m) - Dn+mT
T J,h g,h - f(n)g(n),h if m,n E Ill;
= {TJ(P),hTg(q)'hDP+q if p + q ~ 0;
Tf(p) ,hTg(q) ,h R - p- q if p + q > 0;

if p+ q ~ 0;
if p+q > 0;
(q)T(P)
T J,h g,h -- T J(q),h R-qTg(p),h DP'f "T "T
Z P E 1'1, -q E 1'1;
(P)T(-q)
T J,h g,h -- T f(-p),h RPTg(-q),h Rq'f
Z p,q E "T
1'1.
o
Example A6.2. Let the families fO, Co, so, ch o , shn be defined as in Example
A6.1 and let R E RD be arbitrarily fixed. For all xES we have
d I (fh - I)x
d h ch h-Ox = lim h = Dcox = Dlx = x,
- h-too
d d d d
dh Shxlh=O = Dx, dh ShhXlh=O = Dx, dh Chxlh=o = 0, dh ChhXlh=o = 0
(cf. Corollary A6.2, Formula (A6.9)). By Corollary A6.3 and Formula (A6.1O),
for all xES we have
o
Now we shall study canonical mappings (cf. PR[7]) generated by a family of
extended functional shifts.
Definition A6.3. Suppose that F is an initial operator for D E R(X) corre-
sponding to an R E RD. A mapping I\;j~~) of X into the set of operators defined
on IK and with values in ker D is said to be canonical for a family of extended

functional shifts Tjr~) ,iK: (p, q E Z) if
(q) - {FT(q) } far x EX.

"'f(p)x - f(p),h X hEiK:
We shall write for all x EX, h E IK, p, q E Z
if q = 0;
o
In the sequel we admit that all assumptions of Definition A6.3 are satisfied. From
this definition it follows immediately
Proposition A6.l2. Let p,q E Z. Then the function ",j~~)(x;.) of a complex
variable h E IK and with values in X defined in Definition A6.3 has the following
properties:
(i) for all X,y E X, >',/.L E C
(ii) for all f,g E H(IK), )..,/-L E C, x E X
",(q) (x' h)
>'f+l'g'
= )..",(q)
f(x' ' 'h)'+'1J",(q)
9
(x' h)
, •
Proposition A6.l3. The restriction of the canonical mapping ",j~~) to the set S
of all D-polynomials has the following properties:
OJ for all I E H (IK) and xES
=L
00
(A6.13) "'f(x; h) (h - a)n F Dnx (h E IK)j

n=O
(m)( . ) _ {m!amFDmX for mE No;

(A6.14) "'f x, a - 0 ~T
J:
lor -m E 1''1;
(ii) "'f(x; h) = "'g(Xj h) if and only if I = g, where I, 9 E H(IK), hE IK, x E Si

(iii) ker/'i,fls = {O} ifand only ifall coeflicientsan (n E No) in the Taylor expansion
(A6.2) are not equal zero for f E H(IK);
(iv) /'i,f separates points of S (i.e. /'i,f(x; h) = ",,(y; h) for f E H(IK), hE IK, if and
only if x = y, where x, yES) if and only if an ¥ 0 for all n E No.
Appendix 321
Proof (i) Formula (A6.3) is a consequence of Definition A6.3. Formula (A6.14)

follows from Proposition A6.8(ii). Assertion (ii) follows immediately from Theorem
A3.1. In order to prove (iii), consider the equation /'i,f(x; h) = 0 (h E JK) with a
fixed xES \ {o}. Let am =f. 0 be a coefficient in the expansion (A6.2) such that
aj = 0 for j < m. Then there are n E No, Zk E ker D \ {O} (k = 0,1, ... , n) such
that x = L:~=o Rkzk E X n . If n < m then, by (A6.13), we have
n
/'i,l(x; h) =L ak(h - a)k FDkx = 0 (h E JK).
k=O
If n ~ m then, again by (A6.13),
n n
= L ak(h-a)kFDkx= L adh-a)kFDkx
k=O k=m
n n
=L ak(h-a)kFDk(L Rjzj )
k=m j=O
n n
= L ak(h - a)k F( L Rj-k Zj )
k=m j=k
n n
= L ak(h-a)k(L FRj-k Zj )
k=m j=k
n n
= L ak(h - a)k FZk =L ak(h - a)kzk'
k=m k=m
where am =f. 0, Zj =f. 0 (j = m, m + 1, ... , n). This shows that am+l = ... = an = 0
for n > m implies that /'i,f(x; h) = 0 (h E JK) for x =f. O. We therefore conclude
that 0 =f. x rt. ker "'I if and only if ak =f. 0 for all n E No.
Assertion (iv) follows immediately from (iii) and Proposition A6.12(i). •
In a similar way we prove for a family Tjr~) ,II< (p, q E Z) we prove
Theorem A6.B. Let XES, x' E X' be arbitrarily fixed. Then x' [/'i,}m) (x; .)], as
a complex-valued function, is holomorphic in JK and
Theorem A6.9. The canonical mapping ",}m) (x; .) (m E No) is strongly holo-
morphic in JK and for all h E JK, xES
~",(m)(x' h) = tl;(m+l) (x' h)

dh I ' f "
d~n Kf(Xi h) = Dnff,f(n) (Xi h) = K}n) (Xi h) (n EN).
Theorems A6.8 and A6.9 together imply

Corollary A6.5. Let X E 5, x' E X', m E No be arbitrarily fixed. Then
Theorem A6.10. Let X E 5 be arbitrarily fixed. Then the canonical mapping

K}m) (Xi .) (m E No) is strongly continuous on K
According to Definition A6.3, by (A6.13) (cf. Proposition A6.11), we prove

Proposition A6.14. Let r c IK be an oriented arc with the initial point band
the final point c. Let X a complete linear metric locally convex space. Then for
every xES the integral fr Kf(Xi h)dh exists and
where
(c - a)k+l - (b - a)k+l
Ck = ak k +1 (k E No).

Theorem A6.11. Suppose that all assumption of Proposition A6.14 are satisfied
and r c IK is a rectifiable closed arc. Then the Cauchy theorem holds for the
canonical mapping Kf(Xi') whenever x E 5.
Theorem A6.12. Suppose that all assumption of Proposition A6.14 are satisfied.
Let x E 5 be arbitrarily fixed. Let r C lK be a rectifiable closed arc such that
Indr(w) 1
= -2'
1rZ
1 =
r
-(d(
-
- w
0 for all w ~ K
Then
(n)( . ) _ n!
Kf x, h - 21ri r
1 «( _ Kf(X;()
h)n+l d( (n E No
)
(cf. Chapter 14).

For an indefinite integral of the restriction of K f to S we have
Theorem A6.13. Suppose that all assumptions of Proposition A6.14 are satisfied.
Let x E 5 be arbitrarily fixed. Then
Appendix 323
Clearly, this last result does not contradict Formula (A6.14).

Theorem A6.14. (Taylor expansion) Suppose that all assumptions of Proposition
A6.14 are satisfied. Let a disk OC1 = OC(b, rt) C OC and the function Tf(x; h) be
given. Then
=L
00 (h - b)k k
"'f(x; h) k! "'~ ) (x; b) where hE OC1 , xES.
k=O
In particular, if OC1 = OC then, by (A6.14) we get (A6.13).

Theorem A6.15. Let xES be arbitrarily fixed and let f, 9 E H(OC). If
"'f(x; h n ) = "'g(x; h n ) for h n E OC (n E N) and {h n } has a limit point in OC,
then "'f(x;h) = "'g(x;h) on OC.
The above results can be extended for the whole space X (cf. Corollary A6.2).
For instance, we have
Corollary A6.6. Suppose that all assumptions of Proposition A6.14 are satisfied.
IfS = X and {AhhEIK C Lo(X) is family of uniformly continuous operators such
that Ahl s = "'f(·;h) for h E OC, then
where r is defined in Theorem A6.12.

Spaces invariant under functional shifts can be considered in a similar manner as
spaces of periodic elements in Chapter 15 (cf. Bin[8],[1O]).
REFERENCES
D. V. ANOSOV and A. A. BOLIBRUCH
AB[I] The Riemann-Hilbert Problem. Aspects of Mathematics, 22. Vieveg &
Sohn, Braunschweig/Wiesbaden, 1994.
A. ANTONEVICH
Ant[l] Linear Functional Equations. Operator Approach. Birkhauser Verlag,
Basel-Boston-Berlin, 1996.
P. ANTOSIK and J. MIKUSII'~SKI
AM[I] On Hermite expansions. Bull. Acad. Pol. Sci., 16(1968), 787-79l.
C. ApOSTOL and L. A. FIALKOV
AF[I] Structural properties of elementary operators. Canadian J. Math., 38(1986),
1485-1524.
M. N. ARASLANOV and Yu. L. DALETSKII
AD[l] A composition logarithm in the class offormal operator power series. Funct.
Anal. Appl., 26(1992), 121-124.
M. F. ARSCOTT
Ar[l] Periodic Differential Equations. Pergamon Press, Oxford, 1964.
Ch. BABBAGE
B[l] Des equations functionelles. Ann. Math. Pura Appl. 12(1821/22),73-103.
H. BART, T. ERHARDT and B. SILBERMANN
BES[l] Logarithmic residues in Banach algebras. Integral Equations Operator
Theory, 19(1994), 135-152.
Z. BINDERMAN
Bin[l] Complex R-shifts for right invertible operators. Demonstratio Math.,
23(1990), 1043-1053.
Bin[2] Initial operators for generalized invertible operators. Commentationes
Math., 31(1991), 25-37.
Bin[3] On some properties of complex R-shifts. Demonstratio Math., 25(1992),
207-217.
Bin[4] Cauchy integral formula induced by right invertible operators, Demonstratio
Math., 25(1992), 671-690.
Bin[5] Functional shifts induced by right invertible operators. Math. Nachrichten,
157(1992), 211-214.
Bin[6] Some remarks on polynomials generated by generalized invertible operators.
Opuscula Math., 12(1993), 13-24.
Bin[7] On singular boundary value problems for generalized analytic functions.
Sci. Bull. Lodz. Technical Univ. Math., 23(1993), 11-16.
Bin[8] Periodic solutions of equations of higher order with a right invertible oper-
ator induced by functional shifts. Annales Univ. M. Curie-Sklodowska, 46(1993),
9-22.
324
References 325
Bin[9] Some properties of operators of complex differentiation and shifts. Sci. Bull.
Lodz. Technical Univ. Math., 24(1993),5-18.
Bin[10] On periodic solutions of equations with right invertible operators induced
by functional shifts. Demonstratio Math., 26(1993), 535-543.
Bin[ll] A unified approach to shifts induced by right invertible operators. Math.
Nachrichten, 161(1993), 239-252.
Bin[12] On some functional shifts induced by operators of complex differentiation.
Opuscula Math., 14(1993), 45-57.
Bin[13] Some remarks on sequential shifts induced by right invertible operators.
Functiones et Approximatio, 22(1993), 71-84.
Bin[14] Applications of sequential shifts to an interpolation problem. Collect.
Math., 44(1993), 47-57.
Bin[15] Some fundamental properties of functional R-shifts for right invertible
operators. Commentationes Math., 33(1993), 9-22.
Bin[16] A note of functional R-shifts for right invertible operators. Discussiones
Math., 14(1994),63-76.
Bin[17] On summation formulas induced by functional shifts of right invertible
operators. Demonstratio Math., 28(1995), 301-314.
Bin[18] On isomorphisms of spaces of functional R-shifts for right invertible oper-
ators. In: Different Aspects of Differentiability. Proc. Intern. Conf. Warszawa,
September 1993. Ed. D. Przeworska-Rolewicz. Dissertationes Math. 340,
Warszawa, 1995, 23-36.
Bin[19] Remarks on functional R-shifts for right invertible operators. In: Proc.
Intern. Conf. Different Aspects of Differentiability II. Warszawa, September 1995.
Integral Transforms and Special Functions, 1-2, 4(1996), 39-48.
Z. BINDERMAN and D. PRZEWORSKA-RoLEWICZ
BPR[l] The limit property and perturbations of functional shifts. Demonstratio
Math., 28(1995), 835-85l.
BPR[2] Almost quasinilpotent right inverses. In: Proc. Intern. Conf. Different
Aspects of Differentiability II. Warszawa, September 1995. Integral Transforms
and Special Functions, 1-2, 4(1996), 23-38.
T. CARLEMAN
Ca[l] Sur la resolution de certaines equations integrales. Arkiv Math. Astron.
Fys., 16(1922), 19.
Ca[2] Sur la theorie des equations integrales et leurs applications. Verhandl. In-
ternation. Math. Kongress, 1. Zurich, 1932, 138-15l.
A.L. CAUCHY
C[1] COUTS d' analyse de l'Ecole Poly technique. Ire partie, Analyse algebrique.
Paris, 1821; Oeuvres, (2)3, Paris, 1897.
J. B. CONWAY and B. B. MOREL
CM[l] Roots and logarithms of bounded operators on Hilbert space. J. Funct.
Analysis, 70(1987), 171-193.
C. DAVIS
Da[l] Explicit functional calculus. Algebra and Appl., 6(1973), 193-199.
326 References
A. DI BUCCHIANICO
DB[l] Banach algebras, logarithms, and polynomials of convolution type. J. Math.
Anal. Appl., 156(1991), 253-273.
DB[2] Polynomials of Convolution Type. Ph.D. Thesis, Groningen, 1991.
I. H. DIMOVSKI
Di[l] Convolutional Calculus. Kluwer Acad. Publishers, Dordrecht, 1990 (1st ed.
Publish. House of the Bulg. Acad. Sci., Sofia, 1982)
V. A. DITKIN and A. P. PRUDNIKOV
DP[l] 'fransformations Integrales et CaIcul Operationnel. Mir, Moscou, 1982;
Russian ed. 1972.
PINGPING DONG
Do[l] Minorations de combinaisons lineaires de logarithmes p-adiques de nombres
algebriques. Dissertationes Math. 343, Warszawa, 1995.
Z. DUDEK
D[l] On decomposition of quasi-Leibniz D-algebras. Demonstratio Math., 14
(1981),745-757.
D[2] On multiplicative operators in commutative algebras. Ibidem, 23(1990), 921-
928.
M. K. FAGE and N. I. NAGNIBIDA
FN[l] An Equivalence Problem of Ordinary Linear Differential Operators. (In
Russian), Nauka, Novosibirsk, 1987.
C. G. FRASER F[l] The calculus of algebraic analysis: some observations on math-
ematical analysis in the 18th century. Arch. Hist. Exact. Sci., 39(1989), 317-335.
H. GANGL
G[l] Families of functional equations for polylogarithms. Preprint 95-101, Max-
Planck Institut flir Mathematik, Bonn, 1995.
I. M. GELFAND and R. MACPHERSON
G MP [1] Geometry of Grassmannians and a generalization of the dilogarithm. Ad-
vances in Math., 44(1982), 279-312.
I. G RADSTEIN and I. RYZHIK
GR[l] Tables of Integrals, Sums, Series and Products. (In Russian), Nauka,
Moskva, 1971.
W. HAACK and W. WENDLAND
HW[l] Partial and Pfaflian Differential Equations. Pergamon Press, Oxford-New
York, 1972.
M. HANAMURA and R. MACPHERSON
HMP[l] Geometric construction of polylogarithms, II. In: Functional Analysis on
the Eve of 21th Century. In Honor of the Eightieth Birthday of 1. M. Gelfand.
Eds. S. Gindikin, J. Lepowsky and R. Wilson. Vol. II (Progress in Mathematics
132). Birkhauser, Boston-Basel-Berlin, 1996, 215-282.
References 327
M. G. HAPLANOV
H[l] Linear Transformations of analytic spaces (In Russian). Dokl. AN SSSR,
80(1951), 21-24.
E. HILLE and R. S. PHILLIPS
Functional Analysis and Semigroups. AMS Colloquium Publications, v. 31, Prov-
idence, R.I., 1957 (reprinted 1985).
E. L. INCE
In[l) Ordinary Differential Equations. (Longmans, Green 1926), London, 1927.
H.N.JAHNKE
J[l) Algebraic analysis in Germany, 1780-1840: some mathematical and philosoph-
ical issues. Historia Math., 20 (1993), no. 3, 265-284.
I. KAPLAN SKY
K[l) An Introduction to Differential Algebra. Hermann, Paris, 1957.
D. KARAYANNAKIS
Ka[l) Logarithms of isometries and their spectral decompositions in Hardy spaces.
Ann. Mat. Pura Appl., 165(1993), 1-2l.
W. Z. KARWOWSKI and D. PRZEWORSKA-RoLEWICZ
KPR[l) Green operator for linear boundary value problems with a right invertible
operator DN. Math. Nachrichten, 152(1991), 21-34.
KPR[2) General linear boundary value problems for polynomials in a right invert-
ible operator. Demonstratio Math., 25(1992), 325-340.
W. KIERAT and K. SKORNIK
KS[l) A remark on solutions of Laguerre differential equations. In: Different
Aspects of Differentiability. Proc. Intern. Conf. Warszawa, September 1993. Ed.
D. Przeworska-Rolewicz. Dissertationes Math. 340. Warszawa, 1995, 137-141.
V. KIRYAKOVA
Kir[l] Generalized Fractional Calculus and Applications. Pitman Research Notes
in Mathematics 301. Longman, Harlow-London, 1994.
I. I. KOLODNER
Kol[l) On exp (tA) with A satisfying a polynomial. J. Math. Anal. Appl.,
52(1975), 514-524.
H. KORNACKI
Kor[l) Non-Leibniz components in non-commutative algebras with unit. Demon-
stratio Math., 18(1986), 483-497.
M. KUCZMA
Ku[l) Functional Equations in a Single Variable. PWN-Polish Scientific Publish-
ers, Warszawa, 1968.
Ku[2) An Introduction to the Theory of Functional Equations and Inequalities.
Cauchy's Equation and Jensen's Equation. PWN-Polish Scientific Publishers and
the Silesian University, Warszawa-Krak6w-Katowice, 1985.
328 References
J.P. LAGRANGE
L[l] TMorie des Fonctions Analytiques con tenant Les Principes du Calcul Differ-
en tiel, degages de toute consideration d'infiniment petits, d'evanouissans de limites
et de fluxions, et reduit a l'analyse algebriques de quantites finies. 2nd revised
and enlarged ed., Mme Ve Courcier, Imprimeur-Librairie pour les Mathematiques,
Paris, 1813 (1st ed. 1797).
H. LAUSCH, D. PALLASCHKE and S. ROLEWICZ
LPaR[l] Universal derivations in spaces of functions with values in non-commuta-
tive Banach algebras. In: Proc. Intern. Conf. Different Aspects of Differentiability
II. Warszawa, September 1995. Integral Thansforms and Special Functions, 1-2,
4(1996), 115-120.
H. LAUSCH and D. PRZEWORSKA-RoLEWICZ
LPR[l] Pseudocategories, paraalgebras and linear operators. Math. Nachrichten,
138(1988), 67-87.
LPR[2] Some functional equations appearing in Algebraic Analysis. Opuscula
Math., 6(1990), 111-122.
S. J. LEE and M. Z. NASHED
LN[l] Algebraic and topological selections of multivalued linear relations. Ann.
Scuola Norm. Sup. Pisa Cl. Sci., (4) 17(1990),111-126.
L. LEVIN
Lev[l] Structural Properties of Polylogarithms. AMS monographs, Providence,
R.I., 1991.
N. LINCHUK
Li[l] Representation of commutants of the Pommiez operator and their applica-
tions. (In Russian). Mat. Zametki, 44(1988), 794-802.
D. E. LOEB
L[l] The iterated logarithmic algebra. Advances in Mathematics, 86(1991), 155-
234.
L[2] The iterated logarithmic algebra. II. Sheffer Sequences. J. Math. Anal.
Appl., 156(1991), 172-183.
D. E. LOEB and G.-C. ROTA
LRo[l] Formal power series of logarithmic type. Advances in Mathematics,
75(1989), 1-118.
G. LUMER and M. ROSENBLUM
LR[l] Linear operator equations. Proc. Amer. Math. Soc., 10(1959), 32-4l.
E. MEISTER
Me[l] Randwertaufgaben der Funktionentheorie. B. G. Teubner, Stuttgart, 1983.
N. MERCATOR (or: N. KAUFMANN)
Mr[l] Logarithmo-technia. Londini, 1668.
References 329
M. MICHALSKI
M[l] Derivatives of noninteger order and their applications. Dissertationes Math.
328, Warszawa, 1993.
S. G. MICHLIN
Mi[l] Integral Equations and their Applications. (in Russian), 2nd ed. OGIZ,
Moskva-Leningrad, 1949.
J. B. MILLER
Mil[l) The standard summation operator, the Euler-MacLaurin sum formula and
the Laplace transformation. J. Australian Math. Soc., 39(1985),376-390.
J. NAPIER (or: NEPER)
Np[1] Mirif1ci logarithmorum canonis descriptio. Lugduni, 1614.
NGUYEN VAN MAU
N[l] Generalized algebraic elements and singular integral equations with trans-
formed argument. Warsaw Univ. of Technology Publications. Warszawa, 1989.
N[2] Boundary value problems and controllability of linear systems with right
invertible operators. Dissertationes Math. 316, Warszawa, 1992.
W. POGORZELSKI
P[1) Integral Equations and their Applications. (in Polish; 4 volumes, Warsaw,
1953-70) English Ed. (in one volume vol. 1-3), Pergamon Press and PWN-Polish
Scientific Publishers, Oxford-Warszawa, 1966.
S. D. POISSON
Poi(1) Memoires sur les equations aux differences melees. Journal de l'EcoIe PoIy-
technique, 6, Cahier 13 (1806), 126-147.
T. POULSEN and H. STETKJER
PS(1) Functional equations on abelian groups with involutions. Matematisk Insti-
tut, Aarhus Univ. Preprint Ser. 1995, N° 16. .
A. PRINGSHEIM and G. FABER
PF[l] II C 1. Algebraische Analysis. In: Encyklopiidie der Mathematischen Wis-
senschaften mit Einschluss ihrer Anwendungen. Zweiter Band Analysis. Eds. H.
Burkhardt, M. Wirtinger, R. Fricke, R. Hilb. Teubner, Leipzig, 1909-1921.
A. P. PRUDNIKOV
Pru[l) On the continuation of the ideas of Heaviside and Mikusinski in Opera-
tional Calculus. In: Different Aspects of Differentiability. Proc. Intern. Conf.
Warszawa, September 1993. Ed. D. Przeworska-Rolewicz. Dissertationes Math.
340, Warszawa, 1995, 237-287.
A. P. PRUDNIKOV, Yu. A. BRICHKOV and O. I. MAL'CEV
PBM[l] Integrals and Series. Complementary Chapters. (In Russian). Nauka,
Moskva, 1986.
D. PRZEWORSKA-RoLEWICZ
PR[l) Probleme non-lineaire d'Hilbert pour Ie systeme infini de fonctions. Annales
Polan. Math., 5(1958), 293-301.
330 References
PR[2] Sur l'integrale de Cauchy pour un arc ferme it l'infini. Annales Polon. Math.,
8(1960),155-171.
PR[3] Sur les equations involutives et leurs applications. Studia Math., 20(1960),
95-117.
PR[4] Generalized linear equations of Carleman type. Bull. Acad. Polon. Sci.,
20(1972), 635-639.
PR[5] A characterization of commutators with Hilbert transform. Studia Math.,
44(1972), 27-30.
PR[6] Equations with Transformed Argument. An Algebraic Approach. PWN-
Polish Scientific Publishers and Elsevier Scientific Publish. Compo Warszawa-
Amsterdam, 1973.
PR[7] Shifts and Periodicity for Right Invertible Operators. Research Notes
in Mathematics, 43. Pitman Advanced Publishing Program, Boston-London-
Melbourne, 1980.
PR[8] Algebraic Analysis. PWN-Polish Scientific Publishers and D.Reidel. War-
szawa-Dordrecht, 1988.
PR[9] Spaces of D-paraanalytic elements. Dissertationes Math. 302, Warszawa,
1990.
PR[10] Advantages of one-dimensional kernels. Math. Nachrichten, 149 (1990),
133-147.
PR[11] Property (c) and interpolation formulae induced by right invertible oper-
ators. Demonstratio Math., 21(1988), 1023-1044.
PR[12] Commutators with right invertible operators. J. Math. Anal. Appl.,
158(1991),414-426.
PR[13] True shifts. J. Math. Anal. Appl., 170(1992), 27-48.
PR[14] Generalized Bernoulli operator and Euler-MacLaurin formula. In: Ad-
vances in Optimization. Proc. The 6-th French-German Conf. on Optimization,
Lambrecht(Landau), FRG, June, 1991. Lectures Notes in Economics and Math.
Systems, 382. Springer-Verlag, Berlin-Heidelberg-New York, 1992, 355-368.
PR[15] The operator exp(hD) and its inverse formula. Demonstratio Math.,
26(1993), 545-552.
PR[16] D-algebras with logarithms. Math. Nachrichten, 161(1993), 321-344.
PR[17] Remarks on commutators with operators having either finite nullity or
finite deficiency. Linear Algebra and its Applications, 208/209(1994), 199-207.
PR[18] Logarithmic and antilogarithmic mappings. Dissertationes Math. 337,
Warszawa, 1994.
PR[19] Consequences of Leibniz condition. In: Different Aspects of Differentia-
bility. Proc. Intern. Conf. Warszawa, September 1993. Ed. D. Przeworska-
Rolewicz. Dissertationes Math. 340, Warszawa, 1995, 289-300.
PR[20]. Introducing to Different Aspects of Differentiability. In: Different As-
pects of Differentiability. Ed. D.Przeworska-Rolewicz. Dissertationes Math. 340.
Warszawa, 1995,9-12.
PR[21] Some applications of logarithmic and antilogarithmic mappings to nonlin-
ear equations in Leibniz D-algebras. Workshop on Transform Methods and Special
Functions. Sofia, August 12-17, 1994. Eds. P. Rusev, 1. Dimovski, V. Kiryakova.
SCTP, Singapore, 1995,238-252.
References 331
PR[22] Antilogarithmic mappings and properties of resolvents to linear equations

in Leibniz D-algebras. Demonstratio Math., 28(1995), 771-780.
PR[23] Functional stability of linear systems. Demonstratio Math., 28(1995), 449-
464.
PR[24] Conditions for interpolation by polynomials in right invertible operators
to be admissible. Math Balcanica, NO 4, 9(1995), 353-363.
PR[25] Non-commutative algebras with logarithms. Preprint N° 544, Institute of
Mathematics, Polish Acad. Sci., Warszawa, August 1995.
PR[26] Short story of the term" Algebraic Analysis". In: Proc. Intern. Conf. Dif-
ferent Aspects of Differentiability II. Warszawa, September 1995. Integral Trans-
forms and Special Functions 1-2, 4(1996), 211-220.
PR[27] A priori determined solutions of linear equations. Math. Japonica, 2,
44(1996),395-412.
PR[28] Two centuries of Algebraic Analysis. Preprint NO 572, Institute of Math-
ematics, Polish Acad. Sci., Warszawa, April 1997.
PR[29] Riemann-Hilbert type problems in algebras with logarithms. (to appear)
D. PRZEWORSKA-RoLEWICZ and S. ROLEWICZ
PRR[1] Equations in linear spaces. PWN-Polish Scientific Publishers, Warszawa,
1968.
PRR[2] On periodic solutions of non-linear differential-difference equations. Bull.
Acad. Polon. Sci., 16(1968), 577-580.
PRR[3] The only continuous Volterra right inverses in Ce[O, 1] for the operator it
are J:. Colloq. Math., 51(1987), 279-283.
PRR[4] On a representation of continuous Volterra right inverses to the deriva-
tive in the spaces of infinitely differentiable functions. J. Austral. Math. Soc.,
41(1986), 138-142.
PRR[5] On the spectrum ofright inverses. Demonstratio Math., 23(1990), 739-751.
D. PRZEWORSKA-RoLEWICZ and H. von TROTHA
PRT[1] Right inverses in D-R algebras with unit. J. Integral Equations, 3(1981),
245-259.
J. S. ROGULSKI
Rog[l] Contributions to the theory of paraanalytic functions. Demonstratio Math.,
21 (1988), 1003-1022.
S. ROLEWICZ
R[1] Metric Linear Spaces. 2nd extended ed. PWN-Polish Scientific Publishers
and D. Reidel, Warszawa-Dordrecht, 1985.
S. ROMAN and G.-C. ROTA
RR[1] The umbra} calculus. Advances in Mathematics, 27(1978), 95-188.
K. SKORNIK and J. WLOKA
SW[l] Factoring and splitting of s-differential equations in the field of Mikusinski.
Integral Transforms and Special Functions, 3,4(1996), 263-274.
332 References
P. SMERL
S[I] The functional equation of multiplicative derivation is superstable on standard
operator algebras. Integral Equations Operator Theory, 18(1994), 118-122.
S[2] Isomorphism of standard operator algebras. Proc. Amer. Math. Soc.,
123(1995), 1851-1855.
J. H. M. STEENBRINK
Sb[I]Logarithmic embeddings of varieties with normal crossings and mixed Hodge
structures. Math. Ann., 1, 301(1995), 105-118.
H. STETKiER
St[l] D' Alembert's and Wilson's functional equations for vector and 2 x 2 matrix
valued functions. Matematisk Institut, Aarhus Univ. Preprint Ser. 1996, NO 28.
G.!. TARGONSKI
T[I] Seminar on Functional Operators and Equations. Lecture Notes in Math. 33,
Springer Verlag, Berlin, 1967.
L. VERDE-STAR
VS[I] An algebraic approach to convolutions and transform methods. Preprint,
Dept. of Math., Universidad Aut6noma Metropolitana, Mexico City, 1996; Ad-
vances in Applied Mathematics (to appear).
G. VIRSIK
V[I] Right inverses of vector fields. J. Austral. Math. Soc., 58(1995),411-420.
E. WEGERT
Weg[l] Nonlinear Boundary Value Problems for Holomorphic Functions and Sin-
gular Integral Equations. Akademie Verlag, Berlin, 1992.
Weg[2] Nonlinear Riemann-Hilbert problems with unbounded restriction curves.
Math. Nachrichten, 170(1994),307-313.
W. L. WENDLAND
Wen[l] Elliptic Systems in the Plane. Pitman, London-San Francisco-Melbourne,
1979.
J. WILDESHAUS
Wh[l] Polylogarithmic extensions on mixed Shimura varietes. Schriftenreihe des
Math. Inst. der Universitat Munster, 3. Serie, 12, Munster, 1994.
Wh[2] Realization of polylogarithms. Lectures Notes in Math. 1650, Springer
Verlag, Berlin, 1997.
J. WILKOWSKI
Wi[l] On periodic solutions of differential-difference equations with stochastic co-
efficients. Bull. Acad. Polon. Sci., 21(1973), 253-256.
A. WLODARSKA-DYMITRUK
WD[l] Polynomial-periodic solutions of differential-difference equations. Demon-
stratio Math., 8(1975), 44-65.
H. WYSOCKI
Wy[l] A matrix operational calculus. Acta Math. Hungar., 64(3)(1994), 285-303.
References 333
K. YOSIDA
Yl1] Functional Analysis. Springer Verlag, Berlin-Gottingen-Heidelberg, 1965.
Yu SUN TONG
Toll] Higher-order generalized derivations. Chinese Ann. Math., Ser. A, 6(1985),
566-572; An English summary in Chinese Ann. Math., Ser. B, 6(1985), no. 4,
493.
D. ZAGIER
Z[l] The Bloch-Wigner-Ramakrishnan polylogarithm function. Math. Ann.,
286(1990), 613-624.
SUBJECT INDEX
Abelian group 276 Bernoulli

addition 149 numbers 161
algebra 3 operator 161
isomorphism 272 bijective operator 108
Algebraic Analysis viii, 1 bilinear
algebraic form 6
element 32 mapping 7
operation 1 Bittner Operational Calculus 155
operator 118 boundary
of order N 118 of Liapunov type 77
summation formulae 285 value 201
algebraically closed field 3 bounded linear operator 33
almost Bourlet condition 63
averaging initial operator 85 Calculus 3
involution 31 canonical mapping 160,319
Leibniz D-algebra 139 Carleman
quasinilpotent operator 156 condition 243
analytic function 165 function of order N 243
anticommutator 31 operator of order N 242
anticommuting operators 248 shift 243
antilogarithm viii, 13 Cartesian product 35
of higher order 40 Cauchy
of order n 40 functional equation 276
multiplicative with exponent A 83 integral formula for mappings 317
antilogarithmic principal value of integral 201
function vii remainder 5
mapping viii, 13 theorem on integral of mapping 317
of higher order 40 on multiplicaton of series 272
of order n 40 characteristic
Aoki-Rolewicz theorem 262 polynomial 118
argument 16 root
of element 146 of polynomial 32, 118
automorphism 243 of quasipolynomiaI 200
Babbage equation 243 zero 1
backward shift 65 Clairaut equation 183
Banach classical Riemann-Hilbert problem 201
algebra 33 closed
space ix, 122 operator 146
basic equations 11 rectifiable arc 317
basis of exponential function 101 regular arc 201
334
Subject index 335
co dimension 52 coordinatewise
coefficients addition 34
of D-analytic element 196 convergence 65
co-initial operator 3 multiplication 19
common multiple 286 by scalars 34
commutant 78 coset 31
cosine
commutative
element 136
algebra viii, 5
mapping 136
group 149
cotangent Hilbert transform 214
semigroup 149 Cramer system 55
commutator 31 cyclic group 243
complete linear metric space ix, 32
complex D-algebra 9
D-analytic element 160
differentiation 285
D-invariant space 192
extension 146
D-paraanalytic element 187
integration 285
D-polynomial 41
complex of degree n 41
number 1 D-polynomial-exponential element 232
R-shift 264 D-polynomial-exponential-periodic
component of a set 107 element 233
condition (A) 203 D-shift 264
condition (A)l 206 D-R invariant space 197
condition (A)r 206 D-R structure
condition [Cln 136 preserving isomorphism 198
condition [L] 101 deficiency 52
condition [L]# 107 de Moivre formula 49, 136
condition [M] 83 determinant 54
Di Bucchianico theorem 107
condition [M]# 89
diffeomorphism 248
conjugate
difference operator 5
bounded linear operators 108 differential-difference equation 200
harmonic function 215 dimension of kernel 4
imaginary roots 131 direct sum 1
space 54 disjoint projectors 118
constant 1 disk 259
mapping 22 domain
function 4 of multifunction 11
sequence 5 of operator 1
constants variation method 128 with boundary
continuous of Liapunov type 77
functional 287 Duhamel condition 30, 112
right inverse 5 eigenspace 118
convolution 4, 34, 114 eigenvalue 104
336 Subject index
eigenvector 128 fundamental

elementary operator 33 solution 154
entire function 63 system 132
Euler function 256
Euler-Maclaurin formula 162 Gauss kernel 165
exponential 96, 156 general solution 115
element 128 generalized
function 15, 48, 101 complex differentiation 285
type 14 integration 281
exponential-periodic Floquet theorem 222
element 218 Riccati equation 179
function 218 Riemann-Hilbert problem 203
extended functional shifts 305 Schwarzian 184
generating function 165
F-norm 122 geometric series 173
factorization 129 graph of multifunction 11
field group 149
of characteristic zero 1
of scalars 1 harmonic function 215
final point 316 Hermite derivative 69
finite Hilbert
deficiency 52 space 33
dimensional operator 52 cotangent transform 214
index 52 transform 214
nullity 52 homeomorphism 243
first formal Holder condition 201
derivative 304 with exponent p 201
integral 304 holomorphic function 214, 304
fixed-point 243 homographic function 184
Floquet theorem 218 hyperbolic
formal cosine 296
derivative 304 sine 296
integral 304 tangent 298
forward shift 64
Fourier ideal of finite dimesional operators 52
expansion 148 idempotent 51
series 148 identity operator 1
fractional indefinite integral of mapping 317
calculus 256 index
integral 256 of operator 52
functional of complex point 317
equation 15, 96 of Riemann-Hilbert problem 211
shift 165, 264 infinitesimal generator
functional-differential equation 4 of operators family 277
of Carleman type 251 of semigroup 160
Subject index 337
initial left
condition 16 almost exponential type 96
operator 2 antilogarithm 13
point 316 of higher order 40
of order n 40
value problem 4
antilogarithmic mapping 13
integer 1 of higher order 40
integral of order n 40
equation 255 basic equation 11
over r 316 inverse of operator 3
remainder 5 invertible operator 3
integrable over r function 316 logarithm 13
intermediate value of higher order 40
of order n 40
of function 5
logarithmic
of shift 165 mapping 13
invariant metric 263 of higher order 40
inverse of order n 40
element 10 Leibniz
function 37 algebra 6
mapping 13 condition 5, 6
multifunction 11 D-algebra 9
formula 8
operator 56
Liapunov type boundary 77
invertible limit
element 10 point of sequence 277
mapping 13 property of operator family 277
operator 3 linear
selector viii, 13 equation with scalar coefficients 4
involution 31, 201 functional 54
metric space 32
of order N 216
operator 1
isomorphic algebras 271
ordinary differential equation 4
isomorphism 198 with periodic coefficients 218
ring 3
kernel of operator 1
space 1
Kronecker symbol 10
linearly independent
LAE-type 96 elements 54
functionals 54
Lagrange remainder 5
Liouville formula 132
Lagrange-Poisson formula 161 locally
Laguerre differential operator bounded space ix, 122
of the second order 155 convex space ix, 122
Laplace operator 77 pseudoconvex space ix, 122
338 Subject index
logarithm vii, 13 nullity 52

of higher order 40
of order n 40 w-almost quasinilpotent operator 156
logarithmic w-periodic function 218
derivative viii Operational Calculus 4
function vii operator
mapping viii, 139 of complex differentiation 285
of higher order 40
integration 285
of order n 40
of multiplication by element 9
m-normalized of rotation 249
left logarithm 18
order
logarithm 18
right logarithm 18 of algebraic operator 118
m-pseudoconvex algebra 122 of involution 216
mapping of power type 168 of nilpotent element 33
with exponent.x 168 ordinary differential equation 4
matrix function 109 oriented arc 201
Mikusinski field of operators 4
minor determinant 240 p-homogeneous F-norm 122
mixed boundary value conditions 6 partial differential equation 4
modulus of concavity 262 of the first order 75
monomial 105
partition of unit 118
multifunction n viii, 11
multiple characteristic roots 228 path 107
multiplication 9 pathwise connected set 107
multiplicity of root 232 periodic
multiplicative element 216
initial operator 85 function 215
involution of order N 217
mapping 146
mapping 9
linear mapping 10 operator 221
R-shift 232 permuting operators 243
piecewise analytic function 201
nth formal
point derivation 29
derivative 304
integral 304 pointwise multiplication 5
nth root of element 10 Poisson-d' Alembert formula 163
natural logarithm 104 polylogarithmic function 170
neutral element 272 polynomial-periodic function 232
nilpotent element 33 polynomial-exponential-periodic
of order n 33
function 233
noncommutative algebra 6
Pommiez
non-Leibniz component 6
nonlinear equation 175 operator 68
nonnegative integer 1 type operator 296
normalized polynomial 118 positive integer 1
Subject index 339
power right
function 170 almost exponential type 95
mapping 168 antilogarithm 13
type 168 antilogarithmic mapping 13
primitive 1 basic equation 11
principal space 118 inverse 1
invertible operator 17
product rule 6
logarithm 13
projection 2
of order n 40
projector 118
logarithmic mapping 13
proper
of order n 40
non-Leibniz component 7 Rolewicz theorem 262
two-sided ideal 31 Rolle theorem 5
quasinilpotent operator 156 root
of element 10
quasipolynomial 200
of quasipolynomial 218
quasi-Leibniz
rotation 249
algebra 39
D-algebra 9 *-modulus 146
quotient SN -periodic element 205
algebra 31 Sm-periodic operator 221
space 52 scalar
coefficient 4
RAE-type 95 matrix 109
R-shift 218 Schwarz derivative 184
range of operator 1 Schwarzian 184
rational selector viii, 13
function 3 semigroup 149
number 1 separation of points 160
real number 1 shift 160
rectifiable arc 317 similarity property 72
recurrence 7 simple Duhamel algebra 64
sine
reflection of argument 243
element 136
regular value of operator 27
mapping 136
remainder 5
single characteristic root 118
resolving
single-valued multifunction 38
equation 175 singular
operator 72 element 187
restriction of operator 274 integral
Riccati equation 179 equation 199
Riemann-Hilbert problem 201 operator 6, 201
Riemann-Liouville fractional Smerl theorem 108
integral 258 smooth element 159
340 Subject index
space symmetric
of linear functionals 54 domain of multifunction 11, 136
of all linear continuous form 8
functionals 287 matrix 109
of analytic functions 165 Targonski theorem 63
of coefficients of Taylor
D-analytic elements 196 expansion 161, 312
of constants 1 formula 2
of D-analytic elements 160 Taylor-Gontcharov formula 2
of D-paraanalytic elements 187 total space 54
of D-polynomials 159 transformation of argument 37
of D-polynomial-periodic triangle matrix 110
elements 232 trigonometric
of D-polynomial-exponential- element 136
periodic elements 233 identity 138
of exponentials 160 mapping 1364
of exponential-periodic true shift 160
elements 218 two-sided ideal 31
functions 218
uniform convergence
of periodic elements 216 on compact sets 170
of singular elements 187 on closed intervals 170
of smooth elements 159 uniformly continuous operators 317
special generalized unit 1
Clair aut equation 183 disk 259
Riccati equation 180 of algebra 15
square matrix 109
standard Vandermonde determinant 238
subalgebra 108 Volterra
operator sub algebra 108 operator 4
right inverse 4
stochastic differential-difference
continuous right inverse 5
equation 226
vulgar fractions 3
structure operation 160
strong differentiation 318 well-posed initial value problem 4
strongly continuous Wronski theorems 132
function 322 Wronskian 132
group 160 zero 1
semigroup 160 divisor 16
holomorphic mapping 314 element 1
sub algebra 38 of quasipolynomial 235
submultiplicative F-norm 122
summation
formulae 285
operator 162
superposition 8
symbol function 199
AUTHORS INDEX
Anosov 201, 324 Jahnke 1, 327

Antonevich 324 Kaplansky 327
Antosik 69, 324 Karayannakis 327
Apostol 32, 324 Karwowski 327
Araslanov 324 Kaufmann vii, 328
Arscott 218, 324 Kierat 155, 327
Babbage 243,324 Kiryakova 256, 327
Bart 108, 324 Kolodner 327
Bernoulli vi Kornacki 77, 327
Binderman viii, 132, 147, 165, Kuczma 243, 276, 327
189, 215, 264, 265, 277, Lagrange 1, 328
278, 279, 280, 281, 283, Lausch 66, 90, 155, 328
284, 285, 290, 294, 296, Lee 328
304, 323, 324, 325 Leibniz vi, 328
Bolibruch 201, 324 Levin 328
Bourlet 63 Linchuk 68, 287,328
Brichkov 329 Loeb 328
Carleman 243, 325 Lumer 32, 328
Cauchy 1,325 MacPherson 328
Conway 325 Mal'cev 329,
Daletskii 324 Meister 201, 328
Davis 325 Mercator vii, ix, 328
Di Bucchianico 98, 107, 108, 326 Merchiston, Lord of vii
Dimovski 326 Michalski 329
Ditkin 326 Michlin 199, 201, 329
Dong 326 Mikusinski 4, 69, 324
Dudek 9,32,90,326 Miller 296, 32
Erhardt 108, 324 Morel 325
Euler vi Nagnibida 285,290,326
Fage 285, 290, 326 Napier vii, 329
Faber 329 Nashed 328
Fialkov 32, 324 Neper vii, 329
Fraser 1, 326 Nguyen Van Mau 6, 32, 72, 182, 329
Gangl 326 Pallaschke ix, 328
Gelfand 326 Phillips 277, 278, 327
Gradstein 298, 326 Pingping 326
Haack 326 Pogorzelski 201, 329
Hanamura 326 Poisson 329
Haplanov 285, 327 Poulsen 329
Hille 277, 278, 327 Pringsheim 329
Ince 215, 327 Prudnikov 4, 265, 298, 329
341
342 Authors index
Przeworska-Rolewicz viii, 1, 5, 6, 8,
9, 16, 29, 31, 32, 34, 52, 66
72, 74, 77, 90, 92, 106, 112,
118, 122, 128, 132, 138, 143,
144, 147, 150, 154, 155, 156,
158, 160, 161, 162, 163, 165,
181, 187, 189, 190, 191, 196,
199, 200, 201, 202, 206, 210,
214, 216, 218, 224, 225, 226,
227, 228, 235, 243, 244, 245,
246, 247, 251, 255, 262, 264,
265, 269, 278, 279, 280, 281,
283, 284, 290, 296, 298, 301,
319,325,327,328,329,330,
331
Rizhik 298, 326
Rogulski 187, 331
Rolewicz vii, ix, 5, 52, 122, 163,
218, 262, 328, 330, 331
Roman 331
Rosenblum 32, 328
Rota 328, 331
Silbermann 108, 324
Sk6rnik 129, 155, 181, 327, 331
Smer! 108, 332
Steenhrink 332
Stetkrer 148, 329, 332
Targonski 63, 332
Tong 32,333
von Trotha 106, 122, 198, 331
Ulam vi
Verde-Star 332
Virsik 75, 332
Wegert 201, 215, 332
Wendland 201,326,332
Wildeshaus 332
Wilkowski 226, 332
Wloka 130, 181, 331
Wlodarska-Dymitruk 228, 332
Wysocki 155, 332
Yosida 165, 333
Yu Sun Tong 32, 333
Zagier 173, 333
Zelazko 163
LIST OF SYMBOLS
d .
<IT IX,
1 A(X) 6
C 1 IV(x, y) 6
IF ix, 1 CD 6
N 1 Ibn) 7
Q 1 Aa 9
IR 1 M(X) 10
/l 1 ML(X) 10
No 1 leX) 10
1141 In(Y) 10
L(X) ix, 1 x l / n 10
dom A 1 a+- 10
ker A 1 IF[tJ 10
Lo(X) 1 Oij 10
R(X) 1 n1 11
RD 1 nr 11
EB 1 n viii, 11
FD 1 n-I 1 11
A(X) 3 n;1 11
iT 3 n- 1 11
QT 3 graph n1 11
L(X) 3 graph nr 11
Q(D) 3 graph n 11
Q(I,R) 3 n1(y) 12
VeX) 4 nr(y) 12
C[O,IJ 4 Ll 13
J: 4 Lr 13
dim ker D 4 El 13
C(n) 5 Er 13
CN[O, 1] 5 L 13
C([O, I] X [0,1]) 5 E 13
C'(n) 5 G[nd 13
ft 5 G[nr] 13
(s)JF 5 G[n] 13
A(X) 6 cent 15,48
343
344 List of symbols
LgI(D) 15 En 40
Lgr(D) 15 G[Or,n] 40
Lg(D) 15 G[OI,n] 40
C(lR+) 16 G[On] 40
lnt 16
lin 40
GR,m[Od 18
GRn ,rrOr,n] 41
GR,m[Or] 18
GRn,rrOI,n] 41
GR,m[O] 18
gD,I(U) 22 Ar,n 45
gD,r(U) 22 A1,n 45
gD(U) 22 An 45
Vl(X) 27 C(lR) 48
Vl(X) 27 e 48
WA(X) 27 C(O;lF) 48
Aa+- 27 C([e,7r - e; lR) 49
vlFA 27
sint 49
C[a,b] 29
arcsinu 49
HD(X,y) 29
cot t 49
GD,r(X,y) 30
C[-7r/2+e,7r/2-e;lR) 49
G D,I(X, y) 30
[a, b] 31 cost 49
la, b[ 31 arccosv 49
[a] 31 tant 49
D+ 31 C([e,7r - e]; C) 49
MA,B(X) 33 C([-7r/2 + e,7r/2 - e];C) 49
X(a,b) 33 X 2 (a,b) 49
x *y 34 K(X) 52
Or,n 39 co dim 52
Ol,n 39
<lA 52
0-r,n1 39
(3A 52
0-
l,n 39
1
II:A 52
On 39
Lgr # 53
Lr,n 40
Er,n 40 LgI# 53
Ll,n 40 Lg# 53
El,n 40 X' 54
Ln 40 6.(R) 54
List of symbols 345
C'[O,T] 56 XD 101
8(S) 57 [L] 103
Lg# 57 [L]# 107
G(L) 58 exp(X) 107
AD(a,c,d) 63
{h 107
Hu 64
B(H) 107
h = ad - c? + c 64
B(X) 108
P(D) 69
An(OjlF) 109
P(I,R) 70
Q(D) 70 detu 109
Q(I,R) 71 R a 114
Q(D) 71 R a +- 114
Q(R, J) 71 II· lin 122
Q[D] 72 R~n 123
IP
8t8s 74 R~,n 123
2:j=o aj(t)%t 75 €>.z 127
8(u,v) 76 WD 132
C(O, JR) 77
WD(Xl, ... ,XN) 132
80 77
det(Dj-1Xk)j,k=1, ... ,N 132
A 77
flA,r 133
ND(a,b,c,d) 77
yc 78 DA 133
ND(a, c, d) 78 RA 133
BD(a, c, d) 78 FA 133
B!];(a, c, d) 78 LA,r 133
[M] 83 EA,r 133
[M]# 89 OA,1 134
Lr(D) 92 LA,1 134
Ll(D) 92 EA,l 134
L(D) 92
[Cln 136
In(x,y) 99
En 136
In(x) 99
Cx 136
rn(x, y) 99
Sx 136
rn(x) 99
s(x) 100 c>. 144
eX 101 s>. 144

CD(X) 101 w* 144
346 List of symbols
qX) 144 6. h 161

Iw I 144 J(h) 161
]R(X) 144 A(hD) 162
arg w 146 B(hD) 162
[C]# 147 AR(D; h) 162
Xa 149
S(h) 162
Xo(D) 149
SR(h) 162
XeD) 149
LA(X) 162
Xo(D;F) 149
X(D, F) 149 EA(X) 163
F'Xo(D) 150 eAx 163
F'X(D) 150 B(6. h ) 164

F'D 150 D(a; c) 164
8>.(p) 150 H(n) 165
8 a (p) 150 E r ,>. 167
8>.(p,q) 150 E 1,>. 167
8 a (p,q) 150 E>. 167
nill(R) 152
EI 167
n-I(R) 152
t'b(X) 170
Eo(R) 152
u>' 170
E(R) 152
Yen) 171
FEo(R) 152
FE(R) 152 xy 171
()(t,p) 155 Lipx 173
B(X) 156 di}CX) 173
QN(X) 156 f(aEx) 182
AQN(E;w) 156 Ef(x) 183
AQN(E) 156 S(f) 184
Sm,n(t) 156
Sn 184
Doo 159
QR(D) 187
S = peR) 159
PAR(D) 187
E(R) 160
AR(D) 160 r~(u) 194
{ShhEA(IR) 160 'Yn(x) 196

xA(t) 160 XO 196
/l, 160 DO 196
ehD 161 RO 196
List of symbols 347
po 196 R~ 219
ZD,R 197 E~l) 219
.p 198
Q 222
ax 199
Q(t,s) 222
W(D) 199
WA(t) 199 Q(t) 222
an 201 Q(D,S) 224
n 201 h(n)
k,m
229
n+ 201 Xpp(hjno, ... ,nM) 232
n- 201 XPEP(hj Ajj Kjj Njj M) 233
.p+ 201
x y 245
.p- 201 0
Di3 256
fr ~<:) dr 201
HIL(f) 201 Ri3 256
p± 201 f(x) 256
X± 202 P(g) 257
x± 202 Po (g) 227
(A) 203 gilL 258
(A)l 206 OC(ajr) 259
(A)r 206
OCr 259
"'a211
OC 259
Aij 212
S(k) (D) 259
aij 212
Aj 213 S,(D) 259
1
2tr
ttr ()scot
0 X
s-t
- 2 ds 214 S,(D) 259
He 214 f(hD) 260
H 214 E)..(R) 261
if 214 E 261
XSN 216 X1(D) 262
Pj 216
p,(x,y) 263
e 216
T,,'K. = {T"hhE'K. 264
Xj 216
Enr 217
Tg,G = {Tg,hhEG 264
A(D) 268
E(j) 217
XEP(hj Al, .... , AM) 218 T(OC) 271
Ph 219 Ts(OC) 271
Dh 219 O(h) 272
348 List of symbols
l(h) 272
Soc (D) 273
TC(IK) 274
Tf,ociy 274
Xl = qt] 285
X* 287
x* 288
sinh t 296
cosht 296
P 296
I.l3 296
tanht 298
ih 304
6- 1
MI 304
6n
6h n 304
6- n
6h n 304
fen) 304
f(-n) 304
T;;i = {T;~)hEOC (p E Z) 305

€~) 311
c(p) 311
oc
ch~) 311
s(p) 311
oc
sh~) 311
Tf(x; h) 312
T;m) (x; h) 312
Xn 315
Ir g( ()d( 316
Indr(h) 317
2;i Ir ~d( 317
I\;j~~) (p, q E Z) 319
I\;f(x;h) 320
I\;~m) (x; h) 320
Other Mathematics and Its Applications titles of interest:
J. Chaillou: Hyperbolic Differential Polynomials and their Singular Perturbations.

1979, 184pp. ISBN 90-277-1032-5
S. Fucik: Solvability o/Nonlinear Equations and Boundary Value Problems. 1981,
404pp. ISBN 90-271-1077-5
V.1. Istratescu: Fixed Point Theory. An Introduction. 1981, 488 pp.
out o/print, ISBN 90-271-1224-7
F. Langouche, D. Roekaerts and E. Tirapegui: Functional Integration and Semiclas-
sical Expansions. 1982,328 pp. ISBN 90-277-1472-X
N.E. Hurt: Geometric Quantization in Action. Applications 0/ Harmonic Analysis in
Quantum Statistical Mechanics and Quantum Field Theory. 1982,352 pp.
ISBN 90-271-1426-6
F.H. Vasilescu: Analytic Functional Calculus and Spectral Decompositions. 1983,
392 pp. ISBN 90-271-1376-6
W. Kecs: The Convolution Product and Some Applications. 1983,352 pp.
ISBN 90-271-1409-6
C.P. Bruter, A. Aragnol and A. Lichnerowicz (eds.): Bifurcation Theory,
Mechanics and Physics. Mathematical Developments and Applications. 1983,400
pp. out o/print, ISBN 90-271-1631-5
J. Aczel (ed.) : Functional Equations: History, Applications and Theory. 1984,256
pp. out o/print, ISBN 90-271-1706-0
P.E.T. Jorgensen and R.T. Moore: Operator Commutation Relations. 1984,512 pp.
ISBN 90-271-1710-9
D.S. Mitrinovic and J.D. Keckic: The Cauchy Method 0/ Residues. Theory and
Applications. 1984,376 pp. ISBN 90-277-1623-4
R.A. Askey, T.H. Koomwinder and W. Schempp (eds.): Special Functions: Group
Theoretical Aspects and Applications. 1984,352 pp. ISBN 90-271-1822-9
R. Bellman and G. Adomian: Partial Differential Equations. New Methods/or their
Treatment and Solution. 1984,308 pp. ISBN 90-271-1681-1
S. Rolewicz: Metric Linear Spaces. 1985,472 pp. ISBN 90-271-1480-0
Y. Cherruault: Mathematical Modelling in Biomedicine. 1986,276 pp.
ISBN 90-271-2149-1
R.E. Bellman and R.S. Roth: Methods in Approximation. Techniques/or Mathemati-
cal Modelling. 1986,240 pp. ISBN 90-271-2188-2
R. Bellman and R. Vasudevan: Wave Propagation. An Invariant Imbedding
Approach. 1986,384 pp. ISBN 90-271-1766-4
A.G. Ramm: Scattering by Obstacles. 1986,440 pp. ISBN 90-277-2103-3
C.W. Kilmister (ed.): Disequilibrium and Self-Organisation. 1986,320 pp.

ISBN 90-277-2300-1
A.M. Krall: Applied Analysis. 1986, 576 pp.
ISBN 90-277-2328-1 (hb), ISBN 90-277-2342-7 (pb)
J.A. Dubinskij: Sobolev Spaces of Infinite Order and Differential Equations. 1986,
164 pp. ISBN 90-277-2147-5
H. Triebel: Analysis and Mathematical Physics. 1987,484 pp.
ISBN 90-277-2077-0
B.A. Kupershmidt: Elements of Superintegrable Systems. Basic Techniques and
Results. 1987,206 pp. ISBN 90-277-2434-2
M. Gregus: Third Order Linear Differential Equations. 1987,288 pp.
ISBN 90-277-2193-9
M.S. Birman and M.Z. Solomjak: Spectral Theory of Self-Adjoint Operators in
Hilbert Space. 1987,320 pp. ISBN 90-277-2179-3
V.1. Istratescu: Inner Product Structures. Theory and Applications. 1987,912 pp.
ISBN 90-277-2182-3
R. Vich: Z Transform Theory and Applications. 1987,260 pp.
ISBN 90-277-1917-9
N.V. Krylov: Nonlinear Elliptic and Parabolic Equations of the Second -Order.
1987,480 pp. ISBN 90-277-2289-7
W.I. Fushchich and A.G. Nikitin: Symmetries of Maxwell's Equations. 1987,228
pp. ISBN 90-277-2320-6
P.S. Bullen, D.S. Mitrinovic and P.M. Vasic (eds.): Means and their Inequalities.
1987,480 pp. ISBN 90-277-2629-9
V.A. Marchenko: Nonlinear Equations and Operator Algebras. 1987, 176 pp.
ISBN 90-277-2654-X
Yu.L. Rodin: The Riemann Boundary Problem on Riemann Surfaces. 1988,216 pp.
ISBN 90-277-2653-1
A. Cuyt (ed.): Nonlinear Numerical Methods and Rational Approximation. 1988,
480 pp. ISBN 90-277-2669-8
D. Przeworska-Rolewicz: Algebraic Analysis. 1988,640 pp. ISBN 90-277-2443-1
V.S. Vladimirov, YU.N. Drozzinov and B.1. Zavialov: Tauberian Theorems for
Generalized Functions. 1988,312 pp. ISBN 90-277-2383-4
G. Morosanu: Nonlinear Evolution Equations and Applications. 1988, 352 pp.
ISBN 90-277-2486-5
A.F. Filippov: Differential Equations with Discontinuous Righthand Sides. 1988,

320 pp. ISBN 90-277-2699-X
A.T. Fomenko: Integrability and Nonintegrability in Geometry and Mechanics.
1988,360 pp. ISBN 90-277-2818-6
G. Adomian: Nonlinear Stochastic Systems Theory and Applications to Physics.
1988, 244pp. ISBN 90-277-2525-X
A. Tesar and Ludovt Fillo: Transfer Matrix Method. 1988,260 pp.
ISBN 9O-277-2590-X
A. Kaneko: Introduction to the Theory of Hyperfunctions. 1989,472 pp.
ISBN 90-277-2837-2
D.S. Mitrinovic, J.E. Pecaric and V. Volenec: Recent Advances in Geometric
Inequalities. 1989,734 pp. ISBN 90-277-2565-9
A.W. Leung: Systems of Nonlinear PDEs: Applications to Biology and Engineer-
ing. 1989,424 pp. ISBN 0-7923-0138-2
N.E. Hurt: Phase Retrieval and Zero Crossings: Mathematical Methods in Image
Reconstruction. 1989,320 pp. ISBN 0-7923-0210-9
V.1. Fabrikant: Applications of Potential Theory in Mechanics. A Selection of New
Results. 1989,484 pp. ISBN 0-7923-0173-0
R. Feistel and W. Ebeling: Evolution of Complex Systems. Seiforganization,
Entropy and Development. 1989,248 pp. ISBN 90-277-2666-3
S.M. Ermakov, V.V. Nekrutkin and A.S. Sipin: Random Processes for Classical
Equations of Mathematical Physics. 1989, 304 pp. ISBN 0-7923-0036-X
B.A. Plamenevskii: Algebras of Pseudodifferential Operators. 1989, 304 pp.
ISBN 0-7923-0231-1
N. Bakhvalov and G. Panasenko: Homogenisation: Averaging Processes in
Periodic Media. Mathematical Problems in the Mechanics of Composite Materials.
1989, 404 pp. ISBN 0-7923-0049-1
A.Ya. Helemskii: The Homology of Banach and Topological Algebras. 1989,356
pp. ISBN 0-7923-0217-6
M. Toda: Nonlinear Waves and Solitons. 1989,386 pp. ISBN 0-7923-0442-X
M.1. Rabinovich and D.1. Trubetskov: Oscillations and Waves in Linear and
Nonlinear Systems. 1989,600 pp. ISBN 0-7923-0445-4
A. Crumeyrolle: Orthogonal and Symplectic Clifford Algebras. Spinor Structures.
1990, 364 pp. ISBN 0-7923-0541-8
V. Goldshtein and Yu. Reshetnyak: Quasiconformal Mappings and Sobolev
Spaces. 1990,392 pp. ISBN 0-7923-0543-4
I.H. Dimovski: Convolutional Calculus. 1990,208 pp. ISBN 0-7923-0623-6

Y.M. Svirezhev and V.P. Pasekov: Fundamentals of Mathematical Evolutionary
Genetics. 1990,384 pp. ISBN 90-277-2772-4
S. Levendorskii: Asymptotic Distribution of Eigenvalues of Differential Operators.
1991,297 pp. ISBN 0-7923-0539-6
V.G. Makhankov: Soliton Phenomenology. 1990, 461 pp. ISBN 90-277-2830-5
I. Cioranescu: Geometry of Banach Spaces, Duality Mappings and Nonlinear
Problems. 1990,274 pp. ISBN 0-7923-0910-3
B.I. Sendov: Hausdorff Approximation. 1990, 384 pp. ISBN 0-7923-0901-4
A.B. Venkov: Spectral Theory of Automorphic Functions and Its Applications.
1991,280 pp. ISBN 0-7923-0487-X
V.1. Arnold: Singularities of Caustics and Wave Fronts. 1990,274 pp.
ISBN 0-7923-1038-1
A.A. Pankov: Bounded and Almost Periodic Solutions of Nonlinear Operator
Differential Equations. 1990,232 pp. ISBN 0-7923-0585-X
A.S. Davydov: Solitons in Molecular Systems. Second Edition. 1991,428 pp.
ISBN 0-7923-1029-2
B.M. Levitan and I.S. Sargsjan: Sturm-Liouville and Dirac Operators. 1991, 362
pp. ISBN 0-7923-0992-8
V.1. Gorbachuk and M.L. Gorbachuk: Boundary Value Problems for Operator
Differential Equations. 1991,376 pp. ISBN 0-7923-0381-4
Y.S. Samoilenko: Spectral Theory of Families of Self-Adjoint Operators. 1991, 309
pp. ISBN 0-7923-0703-8
B.I. Golubov A.V. Efimov and V.A. Scvortsov: Walsh Series and Transforms.
1991,382 pp. ISBN 0-7923-1100-0
V. Laksmikantham, V.M. Matrosov and S. Sivasundaram: Vector Lyapunov
Functions and Stability Analysis of Nonlinear Systems. 1991,250 pp.
ISBN 0-7923-1152-3
F.A. Berezin and M.A. Shubin: The Schrodinger Equation. 1991, 556 pp.
ISBN 0-7923-1218-X
D.S. Mitrinovic, J.E. Pecaric and A.M. Fink: Inequalities Involving Functions and
their Integrals and Derivatives. 1991, 588 pp. ISBN 0-7923-1330-5
Julii A. Dubinskii: Analytic Pseudo-Differential Operators and their Applications.
1991,252 pp. ISBN 0-7923-1296-1
V.I. Fabrikant: Mixed Boundary Value Problems in Potential Theory and their
A.M. Samoilenko: Elements of the Mathematical Theory of Multi-Frequency

Oscillations. 1991,314 pp. ISBN 0-7923-1438-7
Yu.L. Dalecky and S.V. Fomin: Measures and Differential Equations in Infinite-
Dimensional Space. 1991,338 pp. ISBN 0-7923-1517-0
W. Mlak: Hilbert Space and Operator Theory. 1991,296 pp. ISBN 0-7923-1042-X
NJa. Vilenkin and A.V. Klimyk: Representation of Lie Groups and Special
Functions. Volume 1: Simplest Lie Groups, Special Functions, and Integral
Transforms. 1991,608 pp. ISBN 0-7923-1466-2
N.la. Vilenkin and A.V. Klimyk: Representation of Lie Groups and Special
Functions. Volume 2: Class I Representations, Special Functions, and Integral
Transforms. 1992,630 pp. ISBN 0-7923-1492-1
N.la. Vilenkin and A.V. Klimyk: Representation of Lie Groups and Special
Functions. Volume 3: Classical and Quantum Groups and Special Functions. 1992,
650 pp. ISBN 0-7923-1493-X
(Set ISBN for Vols. 1,2 and 3: 0-7923-1494-8)
K. Gopalsamy: Stability and Oscillations in Delay Differential Equations of
Population Dynamics. 1992, 502 pp. ISBN 0-7923-1594-4
N.M. Korobov: Exponential Sums and their Applications. 1992,210 pp.
ISBN 0-7923-1647-9
Chuang-Gan Hu and Chung-Chun Yang: Vector-Valued Functions and their
Z. Szmydt and B. Ziemian: The Mellin Transformation and Fuchsian Type Partial
Differential Equations. 1992,224 pp. ISBN 0-7923-1683-5
L.I. Ronkin: Functions of Completely Regular Growth. 1992,394 pp.
ISBN 0-7923-1677-0
R. Delanghe, F. Sommen and V. Soucek: Clifford Algebra and Spinor-valued
Functions. A Function Theory of the Dirac Operator. 1992,486 pp.
ISBN 0-7923-0229-X
A. Tempelman: Ergodic Theorems for Group Actions. 1992, 400 pp.
ISBN 0-7923-1717-3
D. Bainov and P. Simenov: Integral Inequalities and Applications. 1992,426 pp.
ISBN 0-7923-1714-9
I. Imai: Applied Hyperfunction Theory. 1992, 460 pp. ISBN 0-7923-1507-3
Yu.I. Neimark and P.S. Landa: Stochastic and Chaotic Oscillations. 1992,502 pp.
ISBN 0-7923-1530-8
H.M. Srivastava and R.G. Buschman: Theory and Applications of Convolution
Integral Equations. 1992,240 pp. ISBN 0-7923-1891-9
A. van der Burgh and J. Simonis (eds.): Topics in Engineering Mathematics. 1992,
266 pp. ISBN 0-7923-2005-3
F. Neuman: Global Properties of Linear Ordinary Differential Equations. 1992,
320 pp. ISBN 0-7923-1269-4
A. Dvurecenskij: Gleason's Theorem and its Applications. 1992,334 pp.
ISBN 0-7923-1990-7
D.S. Mitrinovic, J.E. Pecaric and A.M. Fink: Classical and New Inequalities in
Analysis. 1992,740 pp. ISBN 0-7923-2064-6
H.M. Hapaev: Averaging in Stability Theory. 1992,280 pp. ISBN 0-7923-1581-2
S. Gindinkin and L.R. Volevich: The Method of Newton's Polyhedron in the
Theory ofPDE's. 1992,276 pp. ISBN 0-7923-2037-9
Yu.A. Mitropolsky, A.M. Samoilenko and D.I. Martinyuk: Systems of Evolution
Equations with Periodic and Quasiperiodic Coefficients. 1992,280 pp.
ISBN 0-7923-2054-9
I.T. Kiguradze and T.A. Chanturia: Asymptotic Properties of Solutions of Non-
autonomous Ordinary Differential Equations. 1992,332 pp. ISBN 0-7923-2059-X
V.L. Kocic and G. Ladas: Global Behavior of Nonlinear Difference Equations of
Higher Order with Applications. 1993, 228 pp. ISBN 0-7923-2286-X
S. Levendorskii: Degenerate Elliptic Equations. 1993,445 pp.
ISBN 0-7923-2305-X
D. Mitrinovic and J.D. Keckic: The Cauchy Method of Residues, Volume 2. Theory
and Applications. 1993, 202 pp. ISBN 0-7923-2311-8
R.P. Agarwal and P.J.Y Wong: Error Inequalities in Polynomial Interpolation and
Their Applications. 1993,376 pp. ISBN 0-7923-2337-8
A.G. Butkovskiy and L.M. Pustyl'nikov (eds.): Characteristics of Distributed-
Parameter Systems. 1993,386 pp. ISBN 0-7923-2499-4
B. Stemin and V. Shatalov: Differential Equations on Complex Manifolds. 1994,
504 pp. ISBN 0-7923-2710-1
S.B. Yakubovich and Y.F. Luchko: The Hypergeometric Approach to Integral
Transforms and Convolutions. 1994,324 pp. ISBN 0-7923-2856-6
C. Gu, X. Ding and C.-c. Yang: Partial Differential Equations in China. 1994, 181
pp. ISBN 0-7923-2857-4
V.G. Kravchenko and G.S. Litvinchuk: Introduction to the Theory of Singular
Integral Operators with Shift. 1994,288 pp. ISBN 0-7923-2864-7
A. Cuyt (ed.): Nonlinear Numerical Methods and Rational Approximation II. 1994,
446 pp. ISBN 0-7923-2967-8
G. Gaeta: Nonlinear Symmetries and Nonlinear Equations. 1994,258 pp.

ISBN 0-7923-3048-X
V.A. Vassiliev: Ramified Integrals, Singularities and Lacunas. 1995,289 pp.
ISBN 0-7923-3193-1
NJa. Vilenkin and A.U. Klimyk: Representation of Lie Groups and Special
Functions. Recent Advances. 1995,497 pp. ISBN 0-7923-3210-5
Yu. A. Mitropolsky and A.K. Lopatin: Nonlinear Mechanics, Groups and Sym-
metry. 1995,388 pp. ISBN 0-7923-3339-X
R.P. Agarwal and P.Y.H. Pang: Opial Inequalities with Applications in Differential
and Difference Equations. 1995,393 pp. ISBN 0-7923-3365-9
A.G. Kusraev and S.S. Kutateladze: Subdifferentials: Theory and Applications.
1995,408 pp. ISBN 0-7923-3389-6
M. Cheng, D.-G. Deng, S. Gong and c.-c. Yang (eds.): Harmonic Analysis in
China. 1995,318 pp. ISBN 0-7923-3566-X
M.S. Livsic, N. Kravitsky, A.S. Markus and V. Vinnikov: Theory of Commuting
Nonselfadjoint Operators. 1995, 314 pp. ISBN 0-7923-3588-0
A.I. Stepanets: Classification and Approximation of Periodic Functions. 1995,360
pp. ISBN 0-7923-3603-8
C.-G. Ambrozie and F.-H. Vasilescu: Banach Space Complexes. 1995,205 pp.
ISBN 0-7923-3630-5
E. Pap: Null-Additive Set Functions. 1995,312 pp. ISBN 0-7923-3658-5
CJ. Colboum and E.S. Mahmoodian (eds.): Combinatorics Advances. 1995, 338
pp. ISBN 0-7923-3574-0
V.G. Danilov, V.P. Maslov and K.A. Volosov: Mathematical Modelling of Heat
and Mass Transfer Processes. 1995, 330 pp. ISBN 0-7923-3789-1
A. Laurincikas: Limit Theorems for the Riemann Zeta-Function. 1996, 312 pp.
ISBN 0-7923-3824-3
A. Kuzhel: Characteristic Functions and Models of Nonself-Adjoint Operators.
1996,283 pp. ISBN 0-7923-3879-0
G.A. Leonov, I.M. Burkin and A.I. Shepeljavyi: Frequency Methods in Oscillation
Theory. 1996,415 pp. ISBN 0-7923-3896-0
B. Li, S. Wang, S. Yan and c.-c. Yang (eds.): Functional Analysis in China. 1996,
390 pp. ISBN 0-7923-3880-4
P.S. Landa: Nonlinear Oscillations and Waves in Dynamical Systems. 1996, 554
pp. ISBN 0-7923-3931-2
A.J. Jerri: Linear Difference Equations with Discrete Transform Methods. 1996,
462 pp. ISBN 0-7923-3940-1
I. Novikov and E. Semenov: Haar Series and Linear Operators. 1997,234 pp.
ISBN 0-7923-4006-X
L. Zhizhiashvili: Trigonometric Fourier Series and Their Conjugates. 1996,
312 pp. ISBN 0-7923-4088-4
R.G. Buschman: Integral Transformation, Operational Calculus, and Generalized
Functions. 1996,246 pp. ISBN 0-7923-4183-X
V. Lakshmikantham, S. Sivasundaram and B. Kaymakcalan: Dynamic Systems on
Measure Chains. 1996,296 pp. ISBN 0-7923-4116-3
D. Guo, V. Lakshmikantham and X. Liu: Nonlinear Integral Equations in Abstract
Spaces. 1996,350 pp. ISBN 0-7923-4144-9
Y. Roitberg: Elliptic Boundary Value Problems in the Spaces of Distributions.
1996,427 pp. ISBN 0-7923-4303-4
Y. Komatu: Distortion Theorems in Relation to Linear Integral Operators. 1996,
313 pp. ISBN 0-7923-4304-2
A.G. Chentsov: Asymptotic Attainability. 1997,336 pp. ISBN 0-7923-4302-6
S.T. Zavalishchin and A.N. Sesekin: Dynamic Impulse Systems. Theory and
U. Elias: Oscillation Theory of Two-Term Differential Equations. 1997,226 pp.
ISBN 0-7923-4447-2
D. O'Regan: Existence Theory for Nonlinear Ordinary Differential Equations.
1997,204 pp. ISBN 0-7923-4511-8
Yu. Mitropolskii, G. Khoma and M. Gromyak: Asymptotic Methods for Investigat-
ing Quasiwave Equations of Hyperbolic Type. 1997,418 pp. ISBN 0-7923-4529-0
R.P. Agarwal and P.J.Y. Wong: Advanced Topics in Difference Equations. 1997,
518 pp. ISBN 0-7923-4521-5
N.N. Tarkhanov: The Analysis of Solutions of Elliptic Equations. 1997,406 pp.
ISBN 0-7923-4531-2
B. Rieean and T. Neubrunn: Integral, Measure, and Ordering. 1997,376 pp.
ISBN 0-7923-4566-5
N.L. Gol'dman: Inverse Stefan Problems. 1997,258 pp. ISBN 0-7923-4588-6
S. Singh, B. Watson and P. Srivastava: Fixed Point Theory and Best Approxima-
tion: The KKM-map Principle. 1997,230 pp. ISBN 0-7923-4758-7
A. Pankov: G-Convergence and Homogenization of Nonlinear Partial Differential
Operators. 1997,263 pp. ISBN 0-7923-4720-X
S. Hu and N.S. Papageorgiou: Handbook of Multivalued Analysis. Volume I:

Theory. 1997,980 pp. ISBN 0-7923-4682-3 (Set of 2 volumes: 0-7923-4683-1)
L.A. Sakhnovich: Interpolation Theory and Its Applications. 1997,216 pp.
ISBN 0-7923-4830-0
G.V. Milovanovic: Recent Progress in Inequalities. 1998,531 pp.
ISBN 0-7923-4845-1
V.V. Filippov: Basic Topological Structures of Ordinary Differential Equations.
1998,530 pp. ISBN 0-7293-4951-2
S. Gong: Convex and Starlike Mappings in Several Complex Variables. 1998,
208 pp. ISBN 0-7923-4964-4
A.B. Kharazishvili: Applications of Point Set Theory in Real Analysis. 1998,
244 pp. ISBN 0-7923-4979-2
R.P. Agarwal: Focal Boundary Value Problems for Differential and Difference
Equations. 1998,300 pp. ISBN 0-7923-4978-4
D. Przeworska-Rolewicz: Logarithms and Antilogarithms. An Algebraic Analysis
Approach. 1998,358 pp. ISBN 0-7923-4974-1

Logarithms and Antilogarithms (Danuta Przeworska-Rolewicz (Auth.) )

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Logarithms and Antilogarithms (Danuta Przeworska-Rolewicz (Auth.) )

Hochgeladen von

Copyright:

Verfügbare Formate

Logarithms and Antilogarithms

Mathematics and Its Applications

SPRINGER SCIENCE+BUSINESS MEDIA, B.V.

ISBN 978-94-010-6194-0 ISBN 978-94-011-5212-9 (eBook)

Printed on acid-free paper

AII Rights Reserved

Beyond this (and extremely important), there is the

logarithm - the power to which a fixed number, called the base,

Another old name of antilogaritm is numerus logarithmi. Logarithms and their

Recently, I had a feeling of a need of some" generalized" logarithmic and antilog-

nu = {x Edam D: Du = uDx} far u E dom D.

This multifunction is well-defined. Any invertible selector L of n is said to be a

complete linear metric spaces

" Methodus construendi

Warszawa, May 25, 1997

Van uta Przeworska-Rolewicz

PRELIMINARIES. INTRODUCTION TO ALGEBRAIC ANALYSIS

Any F E F D is said to be an initial operator for D corresponding to R. One can

R' = R - F'R independently of the choice of an R E 'RD.

(1.2) F-y = I - R-yD on dom D (-y E r).

Formula (1.2) yields (by a two-lines induction) the Taylor-Gontcharov Formula,

Let F be an initial operator for D corresponding to an R E 'RD. Putting in (1.3)

Let D E R(X). Then

(1.5) 'RD = {R + FA: A E Lo(X)}, FD = {F(I - AD): A E Lo(X)},

where F is an arbitrary (but fixed) initial operator for D corresponding to a right

Then D E R(X), R E 'RD and F is an initial operator for D corresponding to R.

If there is an R E RD such that the operator

is invertible, then Q(D) E R(X) and R N [Q(I,R)]-1 E 'RQ(D). Thus an initial

FDkx = Yk, where Yk E ker D are given (k = 0,1, ... , N -1),

i.e. this problem is well-posed.

of the equality: (Rax)(t, s) =

ker D = {x E dom D = G 1 [0, 1] : x(t, s) = <p(s) far all t, s E [0, In,

L Pk(n)xn+k = Yn, where Pk(n) E IF (n EN),

with either initial conditions: xm+i = y:" (m E N is fixed, j = 0, ... , N - 1) or

(1.6) D(xy) = xDy + (Dx)y for x,y E dom D.

(1.7) fD(X,y) = D(xy) - cD[xDy + (Dx)y] for X,y E dom D,

where CD is a scalar dependent on D only. Clearly, fD is a bilinear (i.e. linear

D(xy) = cD[xDy + (Dx)y] + fD(X,y) for X,y E dom D.

This representation is not uniquely determined, as it is shown by

D(xy) = cv[xDy + (Dx)y] + fD(X, y), and D(xy) = fb(x, y)

(1 - cD)fb(x, e) = (1 - cD)Dx = cDxDe + fD(X, e),

(a., (3 E IF \ {O}). If it is the case, we say that fD is a proper non-Leibniz com-

(1.8) Dk(xy) = ct[xDky + (Dkx)y] + fi;)(x,y),

(1.9) fi;)(x,y) = ct[(Dx)Dk-ly + (Dk-1x)Dy]+

+ct-1[JD(X, Dk-1y) + fD(Dk-1x, y)] + Dfi;-l) (x, y)

(1.9' ) fi;)(x,y) = ct[(Dx)Dk-ly + (Dk-1x)Dy]+

+cD[J1k - 1) (x, Dy) + fi;-l) (Dx, y)] + D k- l fD(X, y),

D[DN-1(xy)] = D{C~-I[xDN-Iy + (DN-1x)y] + f1N - 1)(x,y)}

A similar proof for Formula (1.9').

In a similar manner as Proposition 1.1, we prove

Proof. Use the equality yx = xy.

D(xy) = xDy + yDx + (Dx)(Dy) for all x = {xn},y = {Yn} EX.

D(xy) = xDy + yDx + d(Dx)(Dy) for x, y E dom D, where dE IF \ {a},

Dp(xy) = p(xDpy + yDpx) + (Dpx)(DpY) + (p2 - p)xy

for x = {xn},y = {Yn} E X, i.e. here CDp = P and fDp(x,y) = (Dpx)(DpY) +

Aax = A(ax) for A E L(X), a,x E X, ax E dom A.

Let D E A(X). We shall consider the following sets:

• the set of all multiplicative linear operators belonging to L(X):

M L(X) = M(X) n L(X);

• the set I(X) of all invertible elements belonging to X;

If x E In(Y) and yn =x then we write y = x 1 / n , (n EN).

(1.15) at-x = xa where x E X.

Then, by definition, for D E A(X)