Studia Logica (2007) 87: 99128

DOI: 10.1007/s11225-007-9079-0 Springer 2007

Roberto Giuntini
Antonio Ledda
Francesco Paoli
Expanding Quasi-MV
Algebras by a Quantum
Operator
Abstract. We investigate an expansion of quasi-MV algebras ([10]) by a genuine quan-
tum unary operator. The variety

QMV of such

quasi-MV algebras has a subquasi-

variety whose members called cartesian can be obtained in an appropriate way out
of MV algebras. After showing that cartesian

quasi-MV algebras generate

QMV, we
prove a standard completeness theorem for

QMV w.r.t. an algebra over the complex

numbers.
1. Introduction
Algebraic structures arising from quantum computation have stirred consid-
erable interest within the quantum logical community over the last few years
(see e.g. [6], [2]). It was not until very recently, however, that particular
classes of quantum computational algebras became the focus of investiga-
tions carried out from an abstract, universal algebraic viewpoint ([10], [8]).
The present paper intends to press ahead on this tack.
Unlike classical computation, quantum computation [12] allows one to
represent two atomic information bits in parallel. Here, in fact, the appro-
priate counterpart of a classical bit is the qubit, dened as a unit vector
in C
2
:
[) = a [0) + b [1) ,
where a, b are complex numbers s.t. [a[
2
+[b[
2
= 1. Supposing that, in anal-
ogy with the classical case, [0) and [1) represent maximal and precise pieces
of information, the superposition state [) corresponds to an uncertain in-
formation: as dictated by the Born rule, [a[
2
yields the probability of the
information described by the pure state [0), while [b[
2
yields the probability
of the information described by the pure state [1). A system of n qubits,
also called a n-quregister, is represented by a unit vector in the n-fold tensor
product Hilbert space
n
C
2
. Qubits and quregisters, therefore, encode pos-
sibly uncertain, yet maximal information. Non-maximal information pieces
Presented by Heinrich Wansing; Received June 8, 2006
100 R. Giuntini, A. Ledda, F. Paoli
are matched, on a mathematical level, by qumixes, i.e. density operators in
C
2
or in appropriate tensor products
n
C
2
of C
2
.
Similarly to the classical case, we can introduce and study the behaviour
of a number of quantum logical gates operating on such information units.
These gates are mathematically represented by unitary operators on the
appropriate Hilbert spaces. In this way, we end up dening an array of
quantum computational logics ([1], [6], [7]). Here are some signicant ex-
amples of quantum gates, whose behaviour is described in the framework of
quregisters.
Example 1. For any n 1, the negation on
n
C
2
is the unitary operator
Not
(n)
such that, for every element [a
1
, ..., a
n
) of the computational basis
B
(n)
,
1
Not
(n)
([a
1
, ..., a
n
)) = [a
1
, ..., a
n1
) [1 a
n
) .
Example 2. For any n, m 1, the Petri-Tooli gate on
n+m+1
C
2
is the
unitary operator T
(n,m,1)
such that, for every element
[a
1
, ..., a
n
) [b
1
, ..., b
m
) [c) of the computational basis B
(n+m+1)
(shortened
as [

a )

b
_
[c)),
T
(n,m,1)
([

a )

b
_
[c)) = [

a )

b
_

a
n
b
m

+c
_
,
where

+ represents sum modulo 2. The conjunction And([

a ) ,

b
_
) can be
dened as T
(n,m,1)
([

a )

b
_
[0)).
One can easily verify that, when applied to classical bits, Not and And
behave as the standard Boolean truth functions. However, the quantum
computational And is, unlike classical conjunction, reversible one can
retrieve the input values from the output with no loss of information due
to the reversibility of the Petri-Tooli gate.
Example 3. For any n 1, the square root of the negation on
n
C
2
is
the unitary operator

Not
(n)
such that, for every element [a
1
, ..., a
n
) of the
computational basis B
(n)
,

Not
(n)
([a
1
, ..., a
n
)) = [a
1
, ..., a
n1
)
1
2
((1 + i) [a
n
) + (1 i) [1 a
n
)) ,
where i is the imaginary unit.
1
By B
(n)
we denote the set {|a
1
, ..., a
n
: a
i
{0, 1}}, which is an orthonormal basis for
the space
n
C
2
.
Expanding Quasi-MV Algebras by a Quantum Operator 101
The basic property of

Not
(n)
is the following: for any [

a ) in
n
C
2
,

Not
(n)
_

Not
(n)
([

a ))
_
= Not
(n)
([

a )). From a logical point of view, the

square root of the negation can be regarded as a kind of tentative par-
tial negation that transforms precise pieces of information into maximally
uncertain ones. True to form, this gate has no Boolean counterpart.
Although the preceding examples were given in the general framework
of arbitrary n-fold tensor products of C
2
, it can be shown [2] that - from
a logical viewpoint - it is unnecessary to consider information quantities in
Hilbert spaces other than C
2
: in fact, the algebra whose universe is the set of
all qumixes of C
2
and whose operations correspond to appropriate extensions
of the quantum logical gates generates the same logical consequence relation
as the algebra over the set of all qumixes of arbitrary n-fold tensor products
of C
2
. This result smooths things out to a considerable extent, since density
operators of C
2
are amenable to the well-known matrix representation
1
2
_
I + r
1
_
0 1
1 0
_
+ r
2
_
0 i
i 0
_
+ r
3
_
1 0
0 1
__
,
where I is the identity 2 2 matrix, while r
1
, r
2
, r
3
are real numbers s.t.
r
2
1
+ r
2
2
+ r
2
3
1 and, therefore, are in 1-1 correspondence with the points
of the Poincare sphere D
3
. If we restrict ourselves to a given subset of
logical gates, we can even shift down by one dimension and replace qumixes
by points of the closed disc D
2
, or equivalently by points of the disc with
centre

1
2
,
1
2
_
and radius
1
2
and so quantum logical gates become special
operations on the latter set of complex numbers. In this way we can obtain
some standard algebras over the complex numbers, sharing the same universe
but having dierent signatures according to the set of logical gates under
examination ([2], [7]).
In ([10]) we considered, for a start, the standard algebra whose funda-
mental operations included an inverse, a kind of Lukasiewicz-style truncated
sum and a distinguished element

0,
1
2
_
. Such a structure, therefore, has
the similarity type of Changs MV algebras ([5]) and satises all of the MV
algebraic axioms except that

0,
1
2
_
is not a neutral element for truncated
sum. We then introduced the notion of quasi-MV algebra with the following
analogy in mind: the above-mentioned algebra should play w.r.t. quasi-MV
algebras the same role that the standard algebra over the real closed unit in-
terval plays w.r.t. MV algebras. Our choice of axioms was bolstered, indeed,
by a completeness theorem to the eect that an equation in the appropriate
language holds in all quasi-MV algebras i it holds in the standard algebra
over the complex numbers.
102 R. Giuntini, A. Ledda, F. Paoli
So far, so good. The problem with quasi-MV algebras, however, is that
they can be called quantum structures only by courtesy. In fact, consider
the algebra over the complex numbers mentioned in the previous paragraph.
The operations of inverse and truncated sum are mere generalizations of
fuzzy operations on the reals, which allow for no interplay between the real
and the imaginary part of a number. Yet, as a matter of fact, the class of
all qumixes of C
2
gives rise to much richer algebras: we may dene in a
natural way genuine quantum operations whereby the two components of
a complex number are no longer kept separate, but appropriately interact.
Once we had coped with the oversimplied case of quasi-MV algebras, the
next signicant goal was to enrich quasi-MV algebras with such operations.
Thus, in the present paper we consider

quasi-MV algebras, i.e. quasi-

MV algebras expanded by an operation of square root of the inverse, which
is the algebraic counterpart of the logical gate of square root of negation
introduced in Example 3. The square root of the inverse can be seen as
a kind of tentative inversion: by applying it twice to a given element,
we obtain the inverse of the element itself.
2
In the standard algebra, for
example, we have

a, b) = b, 1 a) ;

a, b) = 1 a, 1 b) = a, b)

.
To keep the paper self-contained, we recap in 2 the main results on
quasi-MV algebras from [10]. In 3 we introduce

quasi-MV algebras. In
4 we introduce the subquasivariety of cartesian

quasi-MV algebras, whose

members can be obtained out of MV algebras by means of an appropriate
construction, and in 5 we prove that they generate the variety. In 6 we
briey examine the structure of the congruence and ideal lattices; nally,
in 7 we prove a completeness theorem w.r.t. the standard

quasi-MV
algebra over the complex numbers.
2. Quasi-MV algebras
Definition 4. A quasi-MV algebra is an algebra A = A, ,

, 0, 1) of type
2, 1, 0, 0) satisfying the following equations:
A1. x (y z) (x z) y
A2. x

x
A3. x 1 1
2
Interestingly enough, such operation admits of nice physical models: see e.g. [6].
Expanding Quasi-MV Algebras by a Quantum Operator 103
A4. (x

y)

y (y

x)

x
A5. (x 0)

0
A6. (x y) 0 x y
A7. 0

1
We can think of a quasi-MV algebra as identical to an MV algebra, except
for the fact that 0 need not be a neutral element for the truncated sum .
Of course, a quasi-MV algebra is an MV algebra i it satises the additional
equation x 0 x.
An immediate consequence of Denition 4 is the fact that the class of
quasi-MV algebras is a variety in its signature. Henceforth, such a variety
will be denoted by QMV. The subvariety of MV algebras will be denoted
by MV.
We note in passing that quasi-MV algebras bear some resemblance to
normalizations of MV algebras, considered in [4]; remark, however, that our
A2 is a non-normal condition in the terminology of [3]. If NMV denotes the
variety of normalizations of MV algebras, it turns out that neither NMV
QMV nor QMV NMV.
Definition 5. We introduce the following abbreviations:
x y = (x

x y = x (x

y)
x y = x (x

y)
As already remarked, every MV algebra is an example of quasi-MV al-
gebra. Examples of pure quasi-MV algebras, i.e. quasi-MV algebras that
are not MV algebras, are given by the next two structures over the complex
numbers, S (for square) and D (for disc, already mentioned in the introduc-
tion).
Example 6. (standard quasi-MV algebras). We introduce two standard
quasi-MV algebras. S is the algebra

[0, 1] [0, 1] ,
S
,
S
, 0
S
, 1
S
_
, where:
a, b)
S
c, d) =

min(1, a + c),
1
2
_
;
a, b)
S
= 1 a, 1 b);
0
S
=

0,
1
2
_
;
1
S
=

1,
1
2
_
.
Observe that a, b)
S

0,
1
2
_
,= a, b) whenever b ,=
1
2
. Dis the subalgebra
of S whose universe is the set
a, b) : a, b R and (1 2a)
2
+ (1 2b)
2
1.
104 R. Giuntini, A. Ledda, F. Paoli
We now list some very simple properties of quasi-MV algebras.
Lemma 7. The following equations are satised in every quasi-MV algebra:
(i) x (y z) (x y) z; (v) 0 0 0;
(ii) x y y x; (vi) x 0 x x;
(iii) x x

1; (vii) x y y x;
(iv) x x

0; (viii) x y y x.
It is well-known (see e.g. [5]) that it is possible to introduce a lattice order
on any MV algebra by simply taking a b to hold whenever a(a

b) = a.
This condition is obviously equivalent to a(a

b) = a0 in an MV algebraic
setting, yet it is no longer such in a quasi-MV algebraic one. We dene:
Definition 8. Let A be a quasi-MV algebra. For all a, b A:
a b i a b = a 0.
Lemma 9. Let A be a quasi-MV algebra. (i) For all a, b A, a b i
1 = a

b; (ii) is a preordering, but not necessarily a partial ordering,

of A.
Our preordering relation enjoys some standard properties, including a
few monotonicity properties:
Lemma 10. Let A be a quasi-MV algebra. For all a, b, c, d A:
(i) a 0 b 0, b 0 a 0 a 0 = b 0;
(ii) a b, c d a c b d;
(iii) a b, c d a c b d;
(iv) a b, c d a c b d;
(v) a b, c d a c b d;
(vi) a a 0, a 0 a;
(vii) a b c a b

c;
(viii) a b b

a

;
(ix) 0 a, a 1;
(x) a 0 = b 0 a c = b c om c.
Proof. We show (x), whose proof is not in [10]. If a 0 = b 0, then
a b and b a, whence by (ii) if c A it follows that a c b c and
b c a c. However, since a c = a c 0 and b c = b c 0, in this
case we can apply antisymmetry and thus a c = b c.
Expanding Quasi-MV Algebras by a Quantum Operator 105
If A is a quasi-MV algebra, the term reduct A, , ) turns out to be -
in the terminology of [3], [4] - a bounded distributive q-lattice. In particular,
the operations (pseudo-inf) and (pseudo-sup) are both associative and
commutative but they need not be idempotent; furthermore, the absorption
law fails in general.
Some elements in a quasi-MV algebra (at least one indeed, i.e. 0) are
well-behaved in that they satisfy the equation x0 x; we call them reg-
ular. Of course, MV algebras contain nothing but regulars. Pure quasi-MV
algebras, on the contrary, also have irregular elements that cluster around
the regulars: upon remarking that the relation dened by
ab i a b and b a (i a 0 = b 0)
is a congruence on any quasi-MV algebra, we can view each regular a as
surrounded by the cloud of all the irregulars b s.t. ab.
Definition 11. Let A be a quasi-MV algebra and let a A. We call a
regular just in case a 0 = a. We denote by 1(A) the set of all regular
elements of A.
Lemma 12. Let A be a quasi-MV algebra. The algebra
R
A
=

1(A),
R
,
R
, 0
R
, 1
R
_
where, for any functor f, f
R
is the restriction to 1(A) of f
A
, is a MV-
subalgebra of A, lattice ordered by the restriction to 1(A) of
A
, and iso-
morphic to A/.
Definition 13. Let A be a quasi-MV algebra. We call clouds the elements
of A/.
Every cloud, therefore, contains exactly one regular element; it also turns
out that in a quasi-MV algebra there is at most one cloud whose regular r
is such that r = r

and every element in the cloud is accompanied by its

inverse (possibly coinciding with the element itself):
r = r

b c b

All the remaining clouds come in pairs. If b is in the cloud of r, then
its inverse is in the cloud of r

, and f(b) = b

gives a bijection between such

clouds:
r b c
r

106 R. Giuntini, A. Ledda, F. Paoli
Quasi MV-algebras consisting of just one cloud are called at; they cor-
respond to the subvariety of quasi-MV algebras whose equational basis is
the single equation 0 1.
Definition 14. A quasi-MV algebra F is called at i it satises the equa-
tion 0 1. The subvariety of at quasi-MV algebras will be denoted by
FQMV.
There is a standard technique for extracting a at quasi-MV algebra out
of an arbitrary quasi-MV algebra:
Example 15. Let A =

A,
A
,
A
, 0
A
, 1
A
_
be a quasi-MV algebra, and let
k / A if
A
has no xpoint over 1(A), otherwise let k be such a xpoint.
The k-attening of A is the structure
F(A, k) =

A k ,
F
,
F
, 0
F
, 1
F
_
where:
0
F
= 1
F
= k;
for all a, b A k, a
F
b = k;
for all a Ak, a
F
= a
A
; k
F
= k.
Such an algebra is easily seen to be a at quasi-MV algebra.
It is easy to check that the standard quasi-MV algebra S is just the direct
product of MV
[0,1]
and F(MV
[0,1]
,
1
2
), i.e. the
1
2
-attening of the standard
MV algebra MV
[0,1]
. Interestingly enough, a direct decomposition theorem
holds for all quasi-MV algebras:
Theorem 16. For every quasi-MV algebra Q, there exist an MV algebra M
and a at quasi-MV algebra F such that Q can be embedded into the direct
product MF.
As a corollary to Theorem 16, to Changs completeness theorem for MV alge-
bras and to the completeness of at quasi-MV algebras w.r.t. F(MV
[0,1]
,
1
2
)
(easily proved by means of a syntactical argument), we get the desired com-
pleteness result w.r.t. both the square and the disc:
Theorem 17. If t, s are terms in the language of quasi-MV algebras, the
following are equivalent:
(i) QMV t s;
(ii) S t s;
(iii) D t s.
Expanding Quasi-MV Algebras by a Quantum Operator 107
3.

quasi-MV algebras
We now want to enrich quasi-MV algebras by an additional unary operation.
Definition 18. A

quasi-MV algebra (for short,

QMV algebra) is an
algebra A =
_
A, ,

, 0, 1, k
_
of type 2, 1, 0, 0, 0) such that, upon dening
a

a for all a A, the following conditions are satised:

SQ1. A, ,

, 0, 1) is a quasi-MV algebra;
SQ2. k =

k
SQ3.

(a b) 0 = k for all a, b A.

QMV algebras form a variety in their own similarity type, hereafter

named

QMV. We remark in passing that it is impossible to add a square

root of the inverse to a nontrivial MV algebra: letting b be 0 in SQ3, for all
a A we would have

a = k, whence by SQ2 a

a =

k = k and
so a = a

= k

k = k.
We now give some examples of

QMV algebras, starting with the stan-

dard algebras over the complex numbers referred to in the introduction.
Example 19. (standard

QMV algebras). We introduce two standard

QMV algebras. S
r
is the algebra
_
[0, 1] [0, 1] ,
S
r
,

S
r
, 0
S
r
, 1
S
r
, k
S
r
_
,
where:

[0, 1] [0, 1] ,
S
r
,
S
r
, 0
S
r
, 1
S
r
_
is the quasi-MV algebra S of Example 6;

S
r
a, b) = b, 1 a);
k
S
r
=

1
2
,
1
2
_
.
D
r
is the subalgebra of S
r
whose universe is the set
a, b) : a, b R and (1 2a)
2
+ (1 2b)
2
1.
On the other hand, here are two examples of nite

QMV algebras.
Like the standard algebras, they are totally preordered w.r.t. the preorder
of Denition 8. The respective Hasse diagrams are shown in Figure 1.
Example 20. (the Cross). The Cross is the 5-element

QMV algebra
whose operations are given by the following tables:
0 a k a

1
0 0 k k k 1
a k 1 1 1 1
k k 1 1 1 1
a

k 1 1 1 1
1 1 1 1 1 1

0 a
a 1
k k
a

0
1 a

108 R. Giuntini, A. Ledda, F. Paoli

1
k
0
a
a
The Cross
1
k
0
a
a
c b
b
c
The Antenna
Figure 1. Two nite

-QMV algebras.
Example 21. (the Antenna). The Antenna is the 9-element

QMV algebra
whose operations are given by the following tables:
0 b c a k a

1
0 0 0 0 k k k 1 1 1
b 0 0 0 k k k 1 1 1
c 0 0 0 k k k 1 1 1
a k k k 1 1 1 1 1 1
k k k k 1 1 1 1 1 1
a

k k k 1 1 1 1 1 1
b

1 1 1 1 1 1 1 1 1
c

1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1

0 a
b c

c b
a 1
k k
a

0
b

c
c

1 a

Lemma 22. The following equations are satised in every

QMV algebra:
(i) k k

; (iii)

(x

) (

x)

.
(ii) k k 0; (iv)

(x y)

(z w) 1
Proof. (i) By SQ2, k =

k = k

. (ii) By SQ1 and SQ3, k =

(a
b) 0 =

(a b) 0 0 = k 0. (iii)

(a

) =

a = (

a)

; (iv) Let
a, b, c, d A. By SQ3 and (ii),

(ab) 0 = k = k 0, whence by Lemma

10(x),

(ab)

(cd) = k

(cd). However, again by SQ3 and (ii),

Expanding Quasi-MV Algebras by a Quantum Operator 109

(c d) 0 = k = k 0, whence by Lemma 10(x),

(c d) k = k k.
Summing up, by (i)

(a b)

(c d) = k k = k k

= 1.
By Lemma 12, if A is a quasi-MV algebra, the set 1(A) of regular
elements of A is a subuniverse of A. On the other hand, in a

QMV
algebra this set is not, in general, a subuniverse because it is not closed
w.r.t.

. To make it such, we have to add the duals of the regulars

what we call the coregular elements.
Definition 23. Let A be a

QMV algebra and let a A. We call a

coregular just in case

a 0 =

a. We denote by (O1(A) the set of all

coregular elements of A.
Lemma 24. Let A be a

QMV algebra and let a (O1(A). Then: (i) a

0 = k; (ii) a k = a

0 = a

1 = 1.
Proof. (i) Our hypothesis implies

a0) = a

and thus a =

a
0)

0), whence a 0 =

0) 0 = k. (ii) By (i)
a k = a k 0 = k k = k k

= 1. Now, since a 0 =

(0 0) 0 =

00 = k0 = k, then by Lemma 10(x), on the one side

0k = kk = 1,
and on the other side a

0 = k

0 = 1. For a

1 = 1 we argue
similarly.
Observe that in any

QMV algebra A, 1(A) (O1(A) = k. In

fact, by SQ2 and Lemma 22(ii), k is both regular and coregular; the converse
follows from Lemma 24(i). We now have:
Lemma 25. Let A be a

QMV algebra. The algebra

N
A
=
_
1(A) (O1(A),
N
,

N
, 0
N
, 1
N
, k
N
_
where, for any functor f, f
N
is the restriction to 1(A) (O1(A) of f
A
,
is a subalgebra of A.
Proof. 0 and 1 are regular by SQ1 and Lemma 7(v); k is both regular
and coregular as already noticed. Since sums are regular by SQ1, 1(A)
(O1(A) is closed w.r.t. . Lastly, if a 1(A) then a

1(A), i.e.

a (O1(A), and if a (O1(A) then by denition

a 1(A).
The subalgebra of regular and coregular elements of a

QMV algebra A
can always be embedded into a quotient algebra of A modulo a congruence
relation which we are now going to dene.
110 R. Giuntini, A. Ledda, F. Paoli
Definition 26. Let A be a

QMV algebra and let a, b A. We set:

ab i a 0 = b 0 and

a 0 =

b 0
or, equivalently,
ab i a b, b a,

b and

a
It turns out that:
Lemma 27. is a congruence on every

QMV algebra.
Proof. is obviously an equivalence relation. We now prove that it pre-
serves sums. If ab and cd, then in particular a b, b a, c d, d c,
whence by Lemma 10(ii), a c b d and b d a c. To prove

(a c)

(b d) and

(b d)

(a c), it is enough to show

(a c) 0 =

(b d) 0, but both members are equal to k by SQ3.

Thus, (a c)(b d).
Finally, we show that preserves square roots of inverses. Our hypothesis
is that ab, i.e. a b, b a,

b and

a; we have to show
that

b, i.e.

b,

a, a

and b

. But we get the

former two claims for free and the latter two by Lemma 10(viii).
We call the relation the cartesian congruence on a given

QMV al-
gebra. It is easily seen that:
Lemma 28. Let A be a

QMV algebra. The subalgebra N

A
of Lemma 25
is embeddable into A/.
Proof. The canonical mapping sending a to a/ is obviously a homomor-
phism from N
A
to A/. To see that it is one-one, suppose a/ = b/ and
reason by distinction of cases. If a, b are both regular, then a = a 0 =
b 0 = b. If a, b are both coregular, then

a =

a 0 =

b 0 =

b,
whence a

= b

and a = b. W.l.g., suppose a regular and b coregular. Then

a = a 0 = b 0 and

a 0 =

b 0 =

b, whence b

a 0)
and so, by SQ1 and SQ3, a

= (b 0)

= b

0 =

a 0) 0 = k. Thus
a = a

= k

= k and b

k 0) = k, whereby b = b

= k

= k. It
follows that a = b.
Likewise, we introduce a congruence which we call the at congruence
on a

QMV algebra.
Definition 29. Let A be a

QMV algebra and let a, b A. We set:

ab i a = b or a, b 1(A) (O1(A)
Expanding Quasi-MV Algebras by a Quantum Operator 111
If, following common usage, we denote by the identity relation and by
the universal relation, we have:
Lemma 30. (i) is a congruence on any

QMV algebra; (ii) = .

Proof. We conne ourselves to (ii). Suppose ab. If in addition ab, then
either a = b (and we are done) or a, b 1(A) (O1(A), whence applying
Lemma 28, it likewise follows that a = b.
4. Cartesian and at

QMV algebras
We have just seen that the cartesian relation on a

QMV algebra A
is always a congruence on A. We now investigate two special classes of

QMV algebras: cartesian algebras, where is the identity, and at algebras,

where is the universal relation.
Definition 31. A

QMV algebra A is called cartesian i = , i.e. i it

satises the quasiequation
x 0 y 0

x 0

y 0 x y
A

QMV algebra A is called at i = . We denote by F the class

of at

QMV algebras, and by C the class of cartesian

QMV algebras.
As a consequence of the denition, the only

QMV algebra which is both

cartesian and at is the trivial one-element algebra. It is readily seen that
the algebras of Example 19 (i.e. the standard

QMV algebras), Example

20, and Example 21 are all cartesian. The algebras in the next example, on
the other hand, are at.
Example 32. F
100
is the algebra
3
whose universe is the 2-element set 0, b,
s.t. all truncated sums equal 0, while

0 = 0 and

b = b. F
020
is the
algebra whose universe is the 3-element set 0, a, b and whose operations
are given by the tables
0 a b
0 0 0 0
a 0 0 0
b 0 0 0

0 0
a b
b a
3
In general, we denote by F
nmp
the nite at algebra which contains n xpoints for

beside 0, m xpoints for the inverse which are not xpoints for

, and p elements which

are not xpoints under either operation.
112 R. Giuntini, A. Ledda, F. Paoli
Both F
100
and F
020
are at; moreover, F
100
is simple, while F
020
is a
nonsimple subdirectly irreducible algebra having three congruences: , =
and the monolith whose cosets are a, b and 0.
C and F do not exhaust the variety of

QMV algebras: for example,

the direct product of the Cross and F
100
is neither cartesian nor at. It is
also worth noticing that F is a variety, while C is a quasivariety which is not
a variety.
Lemma 33. (i) F is a variety; (ii) C is a quasivariety but not a variety.
Proof. (i) We show that F is axiomatized by the equation 0 1. In
fact, if A is at then, for every a, b A, we have that a 0 = b 0,
whence 0 = 0 0 = 1 0 = 1. Conversely, if such equation holds, then
a 0 = a 1 = 1 = b 1 = b 0, and similarly for

a 0.
(ii) It is sucient to nd a quotient of a cartesian

QMV algebra which

is not cartesian. Consider the Antenna, which is cartesian. Take the con-
gruence whose congruence blocks are 0, k, 1, a, a

and b, c, b

, c

. Quo-
tienting the Antenna by we get the nontrivial at algebra F
100
, which is
outside C.
Cartesian

QMV algebras are amenable to a clean representation in

terms of algebras of pairs. Consider the standard

QMV algebra S
r
. One
may think of it as obtained out of the standard MV algebra MV
[0,1]
by
taking the cartesian square of its universe and dening the operations in
such a way that each component of the result may be extracted out of the
components of the argument(s) simply by means of polynomial MV
[0,1]
-
operations. It turns out that this construction can be carried out not just
for MV
[0,1]
, but for an arbitrary MV algebra provided that inverse has a
xpoint as the next denition shows.
Definition 34. Let A =

A,
A
,
A
, 0
A
, 1
A
_
be an MV algebra and let
k A be such that k = k

. The pair algebra over A is the algebra

T(A) =
_
A
2
,
P(A)
,

P(A)
, 0
P(A)
, 1
P(A)
, k
P(A)
_
where:
a, b)
P(A)
c, d) =

a
A
c, k
_
;

P(A)
a, b) =

b, a
A
_
;
0
P(A)
=

0
A
, k
_
;
Expanding Quasi-MV Algebras by a Quantum Operator 113
1
P(A)
=

1
A
, k
_
;
k
P(A)
= k, k).
For a start, we show that pair algebras are cartesian

QMV algebras.
Lemma 35. Every pair algebra T(A) over an MV algebra A is a cartesian

QMV algebra.
Proof. It is immediate to see that T(A) is a

QMV algebra. Now, sup-

pose that a, b)
P(A)

0
A
, k
_
= c, d)
P(A)

0
A
, k
_
, i.e. a, k) = c, k), and
that

P(A)
a, b)
P(A)

0
A
, k
_
=

P(A)
c, d)
P(A)

0
A
, k
_
, i.e. b, k) =
d, k); it follows that a = c and b = d, viz. a, b) = c, d).
Conversely, every cartesian

QMV algebra is embeddable into a pair

algebra:
Theorem 36. Every cartesian

QMV algebra A is embeddable into the

pair algebra T(R
A
) over its MV term subreduct R
A
of regular elements.
Proof. For the sake of notational irredundancy, throughout this proof it
will be understood that operations without superscripts are operations of
A. Let f : A 1(A)
2
be dened by
f(a) =
_
a 0,

a 0
_
.
The quasiequation of Denition 31 guarantees that f is one-one. Now
we must check that it preserves the operations. First, remark that f(0) =
_
0 0,

0 0
_
. However, since 0 = 0 0, by SQ3 we get f(0) = 0, k) =
0
P(R
A
)
, and similarly for f(1) and f(k). As for the sum,
f(a)
P(R
A
)
f(b) =
_
a 0,

a 0
_

P(R
A
)
_
b 0,

b 0
_
;
= a b, k) ;
=
_
a b 0,

(a b) 0
_
;
= f(a b).
Finally, as regards the square root of the inverse,

P(R
A
)
f(a) =

P(R
A
)
_
a 0,

a 0
_
;
=
_

a 0, (a 0)

_
;
=
_

a 0, a

0
_
;
= f(

a).
114 R. Giuntini, A. Ledda, F. Paoli
5. Cartesian algebras generate

QMV
Cartesian algebras are a quasivariety; hence, it makes sense to try and nd
out the variety of

QMV algebras they generate. In the light of Lemma

33(ii), what we can claim so far is the following:
C V(C)

QMV
The aim of the present section is to prove that V(C) =

QMV.
As a rst step towards this goal, we show that a variant of the direct
decomposition for quasi-MV algebras provided by Theorem 16 carries over
to our enriched structures.
Theorem 37. For every

QMV algebra Q, there exist a cartesian algebra

C and a at algebra F such that Q can be embedded into the direct product
CF.
Proof. Let Q =
_
Q,
Q
,

Q
, 0
Q
, 1
Q
, k
Q
_
be a

QMV algebra; through-

out this proof, it will be understood that operations without superscripts
are operations of Q. The ingredients of our representation are the following:
C = T(R
Q
), the pair

QMV algebra over the MV algebra R

Q
of regular
elements of Q;
F = Q/.
Now, let h : Q 1(Q)
2
F be given by:
h(a) =

__
a 0,

a 0
_
, k/
_
, if a 1(Q) (O1(Q);
__
a 0,

a 0
_
, a/
_
, otherwise.
The function h is clearly well-dened. It is also injective; in fact, suppose
h(a) = h(b). If a and b are either both regular or both coregular, then we
have
_
a 0,

a 0
_
=
_
b 0,

b 0
_
, whence either a = a0 = b0 = b
or

a =

a 0 =

b 0 =

b, i.e. a

= b

and again a = b. If w.l.g.

a 1(Q) and b (O1(Q), then a = a0 = b0 = k =

a0 =

b0 =

b, whence a = b = k. If a and b are neither regular nor coregular, then

a/ = b/, whence a = b since the -cosets are all singletons except for
1(Q) (O1(Q). The remaining cases cannot arise; suppose w.l.o.g. a
is either regular or coregular and b is neither regular nor coregular: then
b/ = k/, a contradiction.
Expanding Quasi-MV Algebras by a Quantum Operator 115
By SQ1, SQ3 and Lemma 22(ii) h(0) = 0, k) , k/) , h(1) = 1, k) , k/)
and h(k) = k, k) , k/).
We nally show that basic operations are preserved by the embedding.
By SQ1 and SQ3,
h(a)
CF
h(b) =
__
a 0,

a 0
_
,
2
(h(a))
_

CF
__
b 0,

b 0
_
,
2
(h(b))
_
;
= a b 0, k) , k/) ;
= a b, k) , k/) ;
= h(a b).
Now, suppose that a is either regular or coregular. Then by SQ1

CF
h(a) =

CF
(
__
a 0,

a 0
_
, k/
_
);
=
__

a 0, (a 0)

_
, k/
_
;
=
__

a 0, a

0
_
, k/
_
;
= h(

a).
Finally, if a is neither regular nor coregular.

CF
h(a) =

CF
(
__
a 0,

a 0
_
, a/
_
);
=
__

a 0, (a 0)

_
,

a/
_
;
=
__

a 0, a

0
_
,

a/
_
;
= h(

a).
By virtue of the preceding result, it will be enough to show that every
at algebra is a homomorphic image of a cartesian algebra. With this goal
in mind, we now proceed to prove that the subvariety of at algebras is
generated by its nite members.
Lemma 38. If F
fin
is the class of nite at algebras, then V(F
fin
) = F.
Proof. Let t s be an equation of type 2, 1, 0, 0, 0) which has a counterex-
ample in some at algebra. If either t or s contain at least an occurrence
of , then either the equation has no counterexample (contrary to the hy-
pothesis) or we are in position to falsify t s in F
100
. If neither t nor s
contain any occurrence of , t and s are terms in at most one variable. The
following cases may arise:
116 R. Giuntini, A. Ledda, F. Paoli
t is a constant preceded by k square roots of the inverse (0 k), s is
a constant preceded by j square roots of the inverse (0 j). Such an
equation can have no counterexample in the variety.
t (w.l.g.) is the variable x preceded by k square roots of the inverse
(0 k), s is a constant preceded by j square roots of the inverse (0 j).
To falsify the equation in in F
100
, simply assign x the value b.
t is the variable x preceded by k square roots of the inverse (0 k), s
is the variable y preceded by j square roots of the inverse (0 j), and
x ,= y. To falsify the equation in F
100
, simply assign x the value 0, and
y the value b.
t is the variable x preceded by k square roots of the inverse (0 k),
s is the variable x preceded by j square roots of the inverse (0 j).
Since x x

is always satised, it is sucient to conne ourselves to the

following six equation types: (i) x

x; (ii) x x

; (iii) x

; (iv)
x

x; (v)

x; (vi) x

. We are in a position to falsify

all these equations in F
004
.
In the previous lemma we actually proved a stronger claim. Not only is
F generated as a variety by its nite members, but it has a single nite gen-
erator: we may take either the algebra F
104
, which contains as subalgebras
both algebras mentioned in our proof, or even F
004
, which suces to provide
the required counterexamples. However, we will not need to use this fact in
what follows. We now show that every nite at algebra is a quotient of a
cartesian algebra.
Theorem 39. For every nite at algebra F, there exist a cartesian algebra
C and a congruence on C such that F is isomorphic to C/.
Proof. We dene by simultaneous induction a denumerable family
C
n

nN
of algebras of type 2, 1, 0, 0, 0) and a family of associated subsets
1(C
n
)
nN
of their universes.
C
0
is the Cross (cfr. Example 20); 1(C
0
) = 0, k, 1.
Let C
n
=
_
C
n
,
C
n
,

C
n
, 0
C
n
, 1
C
n
, k
C
n
_
, and let
Z
n+1
=
_
a
n+1
,

a
n+1
, a

n+1
,

n+1
, b
n+1
,

b
n+1
, b

n+1
,

n+1
_
where Z
n+1
C
n
= . We put
C
n+1
=
_
C
n
Z
n+1
,
C
n+1
,

C
n+1
, 0
C
n
, 1
C
n
, k
C
n
_
,
Expanding Quasi-MV Algebras by a Quantum Operator 117
where for every a C
n
,

C
n+1
a =

C
n
a, and the behaviour of

C
n+1
over Z
n+1
is built into the names of the elements. (Since the interpre-
tation of the symbols 0, 1, k remains constant throughout C
n

nN
, we
will drop the relative superscripts from now on.) As regards
C
n+1
, it
is dened as follows. Let 1(C
n+1
) = 1(C
n
)
_
a
n+1
, a

n+1
_
; by induc-
tion, 1(C
n
) is totally ordered by
C
n
1(C
n
) and we extend such an
order by assuming that k covers a
n+1
and is covered by a

n+1
. Obviously,
there exists an order-preserving bijection f between 1(C
n+1
) and the
set of all fractions of the nite MV algebra MV
2n+5
; thus, for every
a, b 1(C
n+1
), we dene
a
C
n+1
b = f
1
(f(a)
MV
2n+5
f(b)).
We also stipulate:
a
C
n+1
b = c
C
n+1
d if c, d 1(C
n+1
), a, b C
n
,
a
C
n
0 = c
C
n
0
and b
C
n
0 = d
C
n
0;

a
n+1

C
n+1
x =

n+1

C
n+1
x = k
C
n+1
x, for every x C
n+1
;
b
n+1

C
n+1
x =

b
n+1

C
n+1
x = a
n+1

C
n+1
x, for every x C
n+1
;
b

n+1

C
n+1
x =

n+1

C
n+1
x = a

n+1

C
n+1
x, for every x C
n+1
.
All the preceding stipulations are mutually consistent and exhaust all the
possible cases, so each C
n
is well-dened. The next gure gives a graphical
illustration of the whole construction.
1
.
.
.
b

n
a

n
[
b

n+1
a

n+1

n+1
.
.
.
a

a
n

a
n+1
k

n+1

n
a

.
.
.

b
n+1
a
n+1
b
n+1
[

b
n
a
n
b
n
.
.
.
0
118 R. Giuntini, A. Ledda, F. Paoli
We now show that each C
n
is a cartesian

QMV algebra.
We already know that C
0
is such. Thus, suppose inductively that C
n
is
a cartesian

QMV algebra and consider C

n+1
. Let
(O1(C
n+1
) =
_

x : x 1(C
n+1
)
_
,
and let F =

(O1(C
n+1
),
F
,
F
, 0
F
, 1
F
_
, where 0
F
= 1
F
= k, and for every
a, b (O1(C
n+1
), a
F
b = k and a
F
= a
C
n+1
. It is easy to see that F
is a at quasi-MV algebra and that the appropriate term reduct of C
n+1
is
embeddable into MV
2n+5
F, which guarantees that SQ1 holds for C
n+1
. It
is nearly immediate to check that also SQ2 and SQ3 hold. Moreover, by our
construction, if two distinct elements of C
n+1
are in the same cloud, their
square roots of the inverse are in dierent clouds i.e., C
n+1
is cartesian.
We are now left with the task of showing that each nite at

QMV
algebra is a quotient of some C
n
modulo an appropriate congruence relation.
So, let F
nmr
be a nite at algebra which can be assumed to have the
universe
0 g
i

1in

_
g
j
,

g
j
_
n+1jn+
m
2

_
g
k
,

g
k
, g

k
,

k
_
n+
m
2
+1kn+
m
2
+
r
4
,
where for i n g
i
=

g
i
= g

i
, while for n+1 j n+
m
2
, g
j
=

g
j
= g

j
.
Consider C
n+
m
2
+
r
4
+1
, and dene on its universe an equivalence having the
following blocks:
1(C
n+
m
2
+
r
4
+1
) (O1(C
n+
m
2
+
r
4
+1
);
for every i n,
_
b
i+1
,

b
i+1
, b

i+1
,

i+1
_
;
for every n + 1 j n +
m
2
,
_
b
j+1
, b

j+1
_
and
_

b
j+1
,

j+1
_
;
for every n+m+1 k n+
m
2
+
r
4
, b
k+1
,
_

b
k+1
_
,
_
b

k+1
_
,
_

k+1
_
is a congruence on C
n+
m
2
+
r
4
+1
, because all regular and coregular el-
ements are in the same coset (hence sums are preserved) and it is readily
seen from the denition of that if ab then

b. Moreover, it is
F
nmr
C
n+
m
2
+
r
4
+1
/ via the following isomorphism:
(0) = 1(C
n+
m
2
+
r
4
+1
) (O1(C
n+
m
2
+
r
4
+1
);
for i n, (g
i
) =
_
b
i+1
,

b
i+1
, b

i+1
,

i+1
_
;
for n + 1 j n +
m
2
, (g
j
) =
_
b
j+1
, b

j+1
_
and
(

g
j
) =
_

b
j+1
,

j+1
_
;
Expanding Quasi-MV Algebras by a Quantum Operator 119
for n + m + 1 k n +
m
2
+
r
4
, (g
k
) = b
k+1
,
_

g
k
_
=
_

b
k+1
_
, (g

k
) =
_
b

k+1
_
and
_

k
_
=
_

k+1
_
.
This concludes the proof of our theorem.
Theorem 40. F H(C).
Proof. By Theorem 39, F
fin
H(C). Thus, by well-known properties of
class operators ([13]), HSP(F
fin
) HSPH(C) HHSP(C) = HSP(C).
Since C is a quasivariety, we get HSP(F
fin
) H(C). However, by Lemma
38, F = V(F
fin
) = HSP(F
fin
) H(C).
We now have all we need to prove the result we were after.
Theorem 41. V(C) =

QMV.
Proof. By Theorem 40, F C H(C), whereby SP(F C) SPH(C)
HSP(C). However, by Theorem 37 A SP(F C) for an arbitrary

QMV
algebra, whence our conclusion.
6. The congruence and ideal lattices
In [10] we proved that QMV is rather ill-behaved as a variety. Essentially,
the same negative results carry over to

QMV. We begin with the failure

of congruence modularity.
Lemma 42.

QMV is not congruence modular.

Proof. We provide an example of an algebra which fails to be congruence
modular. Consider the direct product of the Cross and F
100
(cp. Examples
20, 32). Let be the congruence whose blocks are
0, 0) , k, 0) , 1, 0) , a, 0) , a

, 0)
0, b) , k, b) , 1, b) , a, b) , a

, b)
It can be checked that the congruences , , , , are pairwise distinct
and such that , , is incomparable with either
or . Moreover, = as identies distinct pairs just in case their
second coordinates are equal, while never does it. Finally, =
as the following argument shows. Take two distinct pairs; if their second
coordinates are both 0, they are identied by ; if the second coordinate of
at least one pair is b, they are -related to two -related pairs. In sum, the
congruences , , , , form a sublattice of the congruence lattice isomorphic
to N
5
, which suces to establish our conclusion.
120 R. Giuntini, A. Ledda, F. Paoli
As a consequence,

QMV is neither congruence distributive, nor con-

gruence e-regular, nor congruence permutable. As regards permutability of
congruences, a stronger result in complete analogy with quasi-MV alge-
bras holds.
Lemma 43. For no nullary operation e in its type is

QMV congruence
e-permutable.
Proof. This time we consider the direct product of the Antenna and F
100
(cp. Examples 21, 32). Since 1, b) 1, 0) and 1, 0) 0, 0), we have that
1, b) 0, 0); however, it cannot be the case that 1, b) 0, 0), as
1, b) /0, 0) / = . Analogously, 0, b) 1, 0) yet not 0, b) 1, 0);
and b, b) k, 0) yet not b, b) k, 0).
Applying well-known results from [14], it follows that:
Corollary 44.

QMV is not subtractive.

As a consequence of the previous results, we can also conclude that

QMV is not ideal-determined: just like quasi-MV algebras,

QMV alge-
bras fail to admit a reasonable (in the sense of [9]) notion of ideal that can
stand in for the notion of congruence. However, we remarked in [10] that
there is a quasi-MV algebraic concept of ideal, directly borrowed from the
theory of MV algebras, for which a one-one correspondence can be estab-
lished w.r.t. an important subclass of congruences - namely, those congru-
ences which are kernels of homomorphisms onto MV algebras. Something
similar happens in the present framework.
Definition 45. Let A be a

QMV algebra. An ideal of A is an ideal of

its quasi-MV algebraic term reduct, i.e. a subset J A s.t. for all a, b A
the following conditions are satised:
I1 0 J;
I2 a, b J a b J;
I3 a J, b a b J.
Henceforth we will denote by ((A), respectively 1(A), the lattice of
congruences, respectively ideals, of the

QMV algebra A and sometimes,

by a notational abuse, their universes. We now proceed by singling out a
notable subclass of the class of congruences.
Definition 46. Let A be a

QMV algebra. A congruence on A is called

a

QMVC congruence i A/ is cartesian; in other words, i for any

a, b A, a 0/ = b 0/ and

a 0/ =

b 0/ implies a/ = b/.
Expanding Quasi-MV Algebras by a Quantum Operator 121
It is readily seen that

QMV C congruences on a given

QMV al-
gebra A form a sublattice of ((A), hereafter denoted by (
I
(A). From the
previous denition it follows that:
in any

QMV algebra, is the smallest

QMV C congruence;
in a at

QMV algebra, = is the unique

QMV C congruence;
in a cartesian

QMV algebra,

QMV C congruences are exactly the

relative congruences.
Theorem 47. Let A be a

QMV algebra. The following lattices are mutu-

ally isomorphic:
(
I
(A);
1(A);
((R
A
);
1(R
A
).
Proof. The isomorphism between ((R
A
) and 1(R
A
) follows from well-
known results about MV algebras. Therefore, it suces to show that (
I
(A)
is isomorphic to 1(A) and that 1(A) is isomorphic to 1(R
A
). We will conne
ourselves to establishing the existence of appropriate bijections, leaving up
to the reader the task of checking order preservation.
The correspondence between ideals and

QMV C congruences is
shown essentially as in [10], Theorem 45; as a consequence, we will not go
into much detail. The general idea is this. Given an arbitrary ideal J and
an arbitrary

QMVC congruence , we dene correspondences f and g

as follows:
f(J) =
_
a, b) A
2
: a b

, b a

J
_
;
g() = x A : (x 0)0 .
It turns out that f(J) is a

QMV C congruence, and that g() is an

ideal. Moreover, g (f(J)) = J since
g (f(J)) = a : (a 0)f(J)0
=
_
a : (a 0) 1, 0 (a 0)

(a 0)

1,

(a 0)

J
_
=
_
a : a 0,
_

(a

0)

0
_

,
_

(a 0)
_

J
_
= a : a 0 J = J.
122 R. Giuntini, A. Ledda, F. Paoli
(Notice the application of Lemma 22(iv).) Finally, the fact that
f(g()) = is established as in [10], Theorem 45, using the properties of .
The bijection between 1(A) and 1(R
A
) is provided as expected: given
an arbitrary ideal I of Aand an arbitrary ideal J of R
A
, we dene mappings
f

and g

as follows:
f

(I) = I 1(A);
g

(J) = (J] (the smallest ideal of A containing J).

Now, f

(I) 1(R
A
) and obviously g

(J) 1(A). Moreover g

(J) =
a A : a 0 J. The latter set, in fact, is an ideal: to check that I3
holds, just suppose that a 0 J and that b a, whence b 0 a 0. It
follows that b 0 J and so b g

(J). Clearly, it is also the smallest ideal

of A containing J. Taking this into account, we can compute
f

(g

(J)) = f

(a A : a 0 J) = a A : a 0 J 1(A) = J;
g

(f

(I)) = g

(I 1(A)) = a A : a 0 I 1(A) = I.
Using known properties of MV, the previous theorem implies that:
Corollary 48. C is a relatively congruence distributive, relatively congru-
ence permutable, relatively point regular quasivariety; the assertional logic of
C is regularly algebraizable with C as equivalent quasivariety semantics.
7. Standard completeness
The aim of this section is showing that the conditions spelled out in De-
nition 18, provide an appropriate axiomatization of the equational theory
of the standard algebra S
r
over the complex numbers. We believe that a
standard completeness result w.r.t. the subalgebra D
r
, which is even more
signicant from a motivational viewpoint, could be attained by adapting
Theorem 62 in [10]. However, we refrain from tackling the issue here. We
observe that a completeness result for a related class of algebras has recently
been given along similar lines, although quite independently in [8].
First, some notational preliminaries. For the sake of readability, we
use vectorial notation for n-tuples of variables or elements, leaving their
dimension implicit. By Term() we mean the set of all terms of type , and
by Subterm(t) the set of all subterms of the term t. If Ais a

QMV algebra
and t(

x ) Term(2, 1, 0, 0, 0)), then for every variable v (whether it occurs

in t or not) and for every

c A we let v
A
t
(

c ) be the value assigned to v in

a xed assignment whereby each x
i
is assigned the value c
i
; hence, if v x
k
,
Expanding Quasi-MV Algebras by a Quantum Operator 123
then as usual v
A
t
(

c ) =
k
(

c ) = c
k
. By square-rooted variable in a term t,
we mean a variable which has occurrences in t preceded by

. Apices will
be omitted whenever no danger of confusion is impending.
Lemma 49. Let t(

x ) Term(2, 1, 0, 0, 0)) contain a subterm of the form

s
1
s
2
, and let A be a

QMV algebra. Then, for any

a A, if t
A
(

a )0 =
1, then t
A
(

a ) = 1.
Proof. If t contains a subterm of the form s
1
s
2
, then for any

a A,
t
A
(

a ) is either regular or coregular. If the former then, upon supposing

that t
A
(

a ) 0 = 1, it follows immediately that t

A
(

a ) = 1. If the latter,
t
A
(

a ) 0 = k = 1 implies that A is at and, as t contains a subterm of the

form s
1
s
2
, t
A
(

a ) = 1.
Definition 50. Let t(

x ,

y ) Term(2, 1, 0, 0, 0)) contain the variables

x
and the square-rooted variables

y . Moreover, let Z(t) =

z be a sequence of
variables of length l(

x ) + l(

y ) and disjoint from

x ,

y . The mapping
t
:
Subterm(t)
_

t : t Subterm(t)
_
Term(2, 1, 0, 0, 0)) is inductively
dened as follows:

t
(x
i
) = x
i
;

t
(

x
i
) = z
i
;

t
(

y
j
) = z
l(

x )+j
, unless there is a k s.t. y
j
= x
k
;

t
(c) = c for any constant c occurring in t;

t
(s
1
s
2
) =
t
(s
1
)
t
(s
2
);

t
(

s
1
) =
_

t
(s
2
)

if s
1

s
2
;
k if s
1
s
2
s
3
.
Lemma 51. Let A be a

QMV algebra, let

a ,

b A, and let t(

x ,

y )
Term(2, 1, 0, 0, 0)) contain the variables

x and the square-rooted variables

y . Then there exist

c ,

d A s.t., for every subterm s of t,

t
(s)
A
(

c ,

d ) = s
A
(

a ,

b ) 0.
Proof. For any variable v, choose

c ,

d such that:
v
A
t
(

c ,

d ) =
_
v
A
t
(

a ,

b ) 0 if v / Z(t);

1
t
(v)
A
(

a ,

b ) 0 if v = z
i
Z(t).
We now prove, by induction on the construction of s, that

t
(s)
A
(

c ,

d ) = s
A
(

a ,

b ) 0.
124 R. Giuntini, A. Ledda, F. Paoli
(1. s x
i
). Since x
i
/ Z(t), and since
t
(x
i
) = x
i
, we have by denition
that

t
(x
i
)
A
(

c ,

d ) = x
A
i
(

a ,

b ) 0.
(2. s s
1
s
2
). In this case
t
(s) =
t
(s
1
s
2
) =
t
(s
1
)
t
(s
2
). Then,
using the inductive hypothesis:

t
(s)
A
(

c ,

d ) =
t
(s
1
)
A
(

c ,

d )
t
(s
2
)
A
(

c ,

d )
= s
A
1
(

a ,

b ) 0 s
A
2
(

a ,

b ) 0
= (s
1
s
2
)
A
(

a ,

b ) 0
= s
A
(

a ,

b ) 0
(3. s

s
1
). We distinguish three subcases.
(3.1. s
1
is the variable y
j
). Then, if y
j
is distinct from every x
i
,

t
(s)
A
(

c ,

d ) =
t
(

y
j
)
A
(

c ,

d )
= z
l(

x )+j
A
(

c ,

d )
=

y
j
A
(

a ,

b ) 0
= s
A
(

a ,

b ) 0
If y
j
coincides with some x
i
, the proof is analogous - only the subscript
changes.
(3.2. s
1

s
2
). Then:

t
(s)
A
(

c ,

d ) =
t
(

s
2
)
A
(

c ,

d )
=
_

t
(s
2
)
A
(

c ,

d )
_

=
_
s
A
2
(

a ,

b ) 0
_

= s
A
2
(

a ,

b )

0
= s
A
(

a ,

b ) 0
(3.3. s
1
s
2
s
3
). Then:

t
(s)
A
(

c ,

d ) =
t
(

(s
2
s
3
))
A
(

c ,

d )
= k
A
(

c ,

d ) = k
=

(s
2
s
3
)
A
(

a ,

b ) 0
= s
A
(

a ,

b ) 0
Theorem 52. Let t(

x ,

y ) Term(2, 1, 0, 0, 0)) contain the variables

x
and the square-rooted variables

y . If S
r
t 1, then for any

QMV
algebra A, A t 1.
Expanding Quasi-MV Algebras by a Quantum Operator 125
Proof. Suppose ex absurdo that S
r
t 1 but that there are a

QMV
algebra A and

a ,

b A s.t. t
A
(

a ,

b ) ,= 1. Since S
r
t 1, t must
contain at least an occurrence of . In fact, if S
r
t 1 and t contains
no variables, it readily follows that A t 1, against the hypothesis. If t
is a variable preceded by some occurrences of

, t 1 is falsiable in S
r
,
against the hypothesis. It follows, by Lemma 49, that t
A
(

a ,

b ) 0 ,= 1.
Applying Lemma 51, for some

c ,

d A we have that
t
(t)
A
(

c ,

d ) ,= 1.
Now,
t
(t) is a term containing just the functors ,

, 0, 1, k, and the
values of the term function
t
(t)
A
for the arguments

c ,

d are regular el-

ements of A. Thus the equation
t
(t)(

x ,

y ) 1 has a counterexample
in the MV term subreduct R
A
of A, expanded by the constant k - hence
also in MV
[0,1]
, by a slight generalization of Changs completeness theorem
due to Lewin et al. ([11]). It follows that there exist

r
1
,

r
2
[0, 1] s.t.

t
(t)
MV
[0,1]
(

r
1
,

r
2
) ,= 1.
For any variable v occurring in t or in Z(t), choose now complex numbers

m
1
,

m
2
[0, 1]
2
s.t.:
v
S
r
t
(

m
1
,

m
2
) =

_
v
MV
[0,1]
t
(

r
1
,

r
2
),
t
(

v)
MV
[0,1]
(

r
1
,

r
2
)
_
if v / Z(t);
_
v
MV
[0,1]
t
(

r
1
,

r
2
),
_
x
MV
[0,1]
(

r
1
,

r
2
)
_

_
if v Z(t)
and
t
(

x) = v
We now prove that:
t
S
r
(

m
1
,

m
2
) =
_

t
(t)
MV
[0,1]
(

r
1
,

r
2
),
t
(

t)
MV
[0,1]
(

r
1
,

r
2
)
_
.
We already remarked that t cannot be a variable, a square-rooted variable
or a constant. Thus we distinguish two cases.
(1. t s
1
s
2
). Applying the induction hypothesis, we have that:
(s
1
s
2
)
S
r
(

m
1
,

m
2
)= s
S
r
1
(

m
1
,

m
2
) s
S
r
2
(

m
1
,

m
2
)
=
_

t
(s
1
)
MV
[0,1]
(

r
1
,

r
2
),
t
(

s
1
)
MV
[0,1]
(

r
1
,

r
2
)
_

t
(s
2
)
MV
[0,1]
(

r
1
,

r
2
),
t
(

s
2
)
MV
[0,1]
(

r
1
,

r
2
)
_
=

t
(s
1
)
MV
[0,1]
(

r
1
,

r
2
)
t
(s
2
)
MV
[0,1]
(

r
1
,

r
2
),
1
2
_
=
_

t
(s
1
s
2
)
MV
[0,1]
(

r
1
,

r
2
),
t
(

(s
1
s
2
))
MV
[0,1]
(

r
1
,

r
2
)
_
(2. t

s). We already ruled out that s be a variable. Since S

r
t 1,
it cannot be a sum or a constant, because square-rooted sums,

0,

1,

k
are all coregular elements of [0, 1]
2
, hence distinct from

1,
1
2
_
Thus s

s
1
.
126 R. Giuntini, A. Ledda, F. Paoli
It follows that:
_

s
1
_
S
r
(

m
1
,

m
2
)= s
S
r
1
(

m
1
,

m
2
)

=
__

t
(s
1
)
MV
[0,1]
(

r
1
,

r
2
),
t
(

s
1
)
MV
[0,1]
(

r
1
,

r
2
)
__

=
_

t
(s
1
)
MV
[0,1]
(

r
1
,

r
2
)

,
t
(

s
1
)
MV
[0,1]
(

r
1
,

r
2
)

_
=
_

t
(s

1
)
MV
[0,1]
(

r
1
,

r
2
),
t
(

1
)
MV
[0,1]
(

r
1
,

r
2
)
_
=
_

t
(

s
1
)
MV
[0,1]
(

r
1
,

r
2
),
t
(

s
1
)
MV
[0,1]
(

r
1
,

r
2
)
_
Now, since by hypothesis S
r
t 1, we have that
t
S
r
(

m
1
,

m
2
) =
_

t
(t)
MV
[0,1]
(

r
1
,

r
2
),
t
(

t)
MV
[0,1]
(

r
1
,

r
2
)
_
=
_
1,
1
2
_
whence
t
(t)
MV
[0,1]
(

r
1
,

r
2
) = 1, a contradiction.
Corollary 53. Let t(

x ,

y ), s(

z ,

w) Term(2, 1, 0, 0, 0)). Then S

r

t s i

QMV t s.
Proof. Right to left: trivial.
Left to right. Suppose there exist a

QMV algebra A and

a ,

b ,

c ,

d
A s.t. t
A
(

a ,

b ) ,= s
A
(

c ,

d ). Since cartesian

QMV algebras generate

QMV, there is a cartesian

QMV algebra B and

e ,

f ,

g ,

h B s.t.
t
B
(

e ,

f ) ,= s
B
(

g ,

h ).
Since B is cartesian, either t
B
(

e ,

f ) s
B
(

g ,

h ) or s
B
(

g ,

h )
t
B
(

e ,

f ) or

(t
B
(

e ,

f ))

(s
B
(

g ,

h )) or

(s
B
(

g ,

h ))

(t
B
(

e ,

f )). Let w.l.g. t

B
(

e ,

f ) s
B
(

g ,

h ) (for the other cases we

argue analogously). Then 1 ,= t
B
(

e ,

f )

s
B
(

g ,

h ) and thus, by Theorem

52, there exist

m,

n ,

r ,

o [0, 1]
2
s.t. 1 ,= t
S
r
(

m,

n )

s
S
r
(

r ,

o ), whence
t
S
r
(

m,

n ) ,= s
S
r
(

r ,

o ).
8. Conclusion
There is a whole stack of open problems concerning quasi-MV algebras with
a square root of the inverse, and more generally the algebras and logics
arising from quantum computation, which are currently the focus of intensive
investigations. Among these, we mention a couple which may deserve special
attention:
What can be said about the assertional logic of

QMV? Which rela-

tionship does it bear to the logics of [7]?
Expanding Quasi-MV Algebras by a Quantum Operator 127
While some fundamental operations of

QMV algebras correspond to

reversible quantum gates, other ones notably, truncated sum are
the algebraic counterparts of irreversible quantum operations. From the
viewpoint of quantum computational signicance, it may be worth to
leave aside irreversible operations and focus on structures endowed only
with product, inverse and the square root of the inverse. Of course, in
this case it would be impossible to rely on parasitic strategies in order
to prove standard completeness; some sort of self-contained proof would
be needed. What form could such a proof take?
Acknowledgements. We thank Daniele Mundici and Felix Bou for the
stimulating conversations on the topics covered by this paper, as well as two
referees of Studia Logica who helped to improve our presentation.
References
[1] Cattaneo G., M. L. Dalla Chiara, R. Giuntini, and R. Leporini, An unsharp
logic from quantum computation, International Journal of Theoretical Physics 43,
7-8 (2001), 18031817.
[2] Cattaneo G., M. L. Dalla Chiara, R. Giuntini, and R. Leporini, Quantum
computational structures, Mathematica Slovaca 54 (2004), 87108.
[3] Chajda, I., Normally presented varieties, Algebra Universalis 34 (1995), 327335.
[4] Chajda, I., R. Halas, J. Kuehr, and A. Vanzurova, Normalization of MV alge-
bras, Mathematica Bohemica 130, 3 (2005), 283300.
[5] Cignoli, R., I.M.L. DOttaviano, and D. Mundici, Algebraic Foundations of
Many-Valued Reasoning, Kluwer, Dordrecht, 1999.
[6] Dalla Chiara, M.L., R. Giuntini, and R. Greechie, Reasoning in Quantum
Theory, Kluwer, Dordrecht, 2004.
[7] Dalla Chiara, M. L., R. Giuntini, and R. Leporini, Logics from quantum com-
putation, International Journal of Quantum Information 3, 2 (2005), 293337.
[8] Domenech, G., and Freytes H., Fuzzy propositional logic associated with quantum
computational gates, International Journal of Theoretical Physics 45, 1 (2006), 228
261.
[9] Gumm, H.P., and A. Ursini, Ideals in universal algebra, Algebra Universalis 19
(1984), 4554.
[10] Ledda, A., M. Konig, F. Paoli, and R. Giuntini, MV algebras and quantum
computation, Studia Logica 82, 2 (2006), 245270.
[11] Lewin, R., M. Sagastume, and P. Massey, MV* algebras, Logic Journal of the
IGPL 12, 6 (2004), 461483.
[12] Nielsen, M., and I. Chuang, Quantum Computation and Quantum Information,
Cambridge University Press, Cambridge, 2000.
128 R. Giuntini, A. Ledda, F. Paoli
[13] Pigozzi, D., On some operations on classes of algebras, Algebra Universalis 2 (1972),
346353.
[14] Ursini, A., On subtractive varieties I, Algebra Universalis 31 (1994), 204222.
Francesco Paoli
Department of Education
University of Cagliari
Cagliari, Italy
corresponding author:
paoli@unica.it
Roberto Giuntini
Department of Education
University of Cagliari
Cagliari, Italy
giuntini@unica.it
Antonio Ledda
Department of Education
University of Cagliari
Cagliari, Italy
antonio.ledda@inwind.it