VOL.

38, 1952

MA THEMA TICS: V. HLA VA T

243

Acknowledgments.-The authors wish to thank Creighton Buck for suggesting the geometrical interpretation of the forward interpolation scheme; John Tukey for giving the authors the benefit of his experience with the solution of "stiff" equations; Albert Schaeffer, Kenneth Arnold and J. Barkley Rosser for their interest in this problem. * This work was carried out under Contract NOrd-9938 with the Navy Bureau of
Ordnance. Assuming the value of a(x, y) to be independent of y.

THE ELEMENTARY BASIC PRINCIPLES OF THE UNIFIED THEORY OF RELATIVITY*

By VACLAV HLAVAT.Y
DEPARTMENT OF MATHEMATICS, INDIANA UNIVERSITY

Communicated by T. Y. Thomas, January 14, 1952

1. Introduction.-The unified theory of relativity exposed recently by Einstein' is based on three principles: (A) The introduction of a nonsymmetric tensor gAy in the space X4 of the relativity; (B) the introduction of a non-symmetric connection rF, by means of

axy

gw,

= r ,g,

+ rJ'g@

(1)

and finally (C) the introduction of a (seemingly) overdetermined system of conditions imposed on the rF, which yields gA,. In the subsequent sections we will deal with each of these basic principles. However, we confine ourselves to results only. The corresponding detailed proofs will be given in a subsequent series of three papers in the Journal of Rational Mechanics and Analysis. 2. Principle A.-Denote by hx,,(k],,) the symmetric (the skew symmetric) part of gAF and by g, h, k the corresponding determinants. Throughout this paper we assume h # 0. If n = 4 and if hx,, is of the signature +++ - then there are in general two sets of bivectors B1, B2 totally perpendicular, which are privileged in the sense that they are polar conjugate with respect to the cone hx,, as well as with respect to the linear complex k),,, of bivectors. Closely connected with them are four sets of (imaginary) bivectors each of them being polar self conjugate with respect to h),,, and k),,,. Projecting this configuration from an arbitrary point P of our space X4 into the ideal space H3 of the tangent space T4(P) of X4 at P, one obtains a linear line complex K containing a linear line congruence C (whose axes

244

MA THEMA TICS: V. HLA VA TY'

PROC. N. A. S.

are the ideal lines of B1(P) and B2(P)). Both line manifolds K and C are defined intrinsically, by means of g,,, only. In another paper2 I have shown that the projective three-dimensional spinor space, S3, associated with the (centered) four-dimensional vector space of the (special) theory of relativity is fully characterized by a linear line complex and a linear line congruence contained in it. Also these manifolds are defined intrinsically. Identifying these manifolds with K and C we may identify H3 and S3. This identification is not affected by the coordinate transformations (which are different in Hs and S3) because in both cases the line manifolds are defined intrinsically. Hence the introduction of a non-symmetric tensor gx,, in X4 leads in the most natural way to the theory of the spinor space of the relativity theory.

3. Principle B.-In the subsequent sections we denote by{>} the Christoffel symbols of h),,,, by H.,, its curvature tensor, by V,, the symbol }. Moreover we put A of covariant derivative with respect to

a2~ ~~X
h xJaVV,. and OA = Oh aXP where Oh)' = hxIL(P). The curvature tensor of r winl be denoted by P.' . If n = 4 then ew"" denotes the four vector Ricci density (with components + 1, -1, 0). The equations (1) split in two systems

(a)
(b)

{= } +

S)

+ ha'(Sak, S;,,k, )
(2)

= 2S 8,6X1W,,,a, ,

where X,X,,, is a tensor, function of h],,, and k,,. Computing S.tY from (2(b)) and substituting it in (2(a)) one obtains the solution r, of (1). A necessary condition for (2(b)) to admit exactly one solution is g 0 0. A sufficient condition for (2(b)) to admit more than one solution SY is g = 0. However, there are cases where (2(b)) does not admit any solution Sayp (and consequently (1) does not admit any solution rF). 4. Almost Riemannian Space.-Our space X. will be termed an almost Riemannian space Rn if kx, = e K),,, S), = e s,, where e -O 0 is a constant and Kx,, s , are tensors different from zero. In R. we obtain from (2(b))3

(a)
and

s,-= h&(VaKW,; +

-V;Kac.) (mod e)

(3)

(b)

r} + E [S>,- hl'(sc? k, + sAykx)] (mod e2).

(3)

VOL. 38, 1952

MA THEMA TICS: V. HLA VA TY

245

Substituting from (3(a)) in (3(b)) we obtain the solution rT, of (1). 5. Principle C.-If n = 4 the conditions = 0 ew'4'P4X = 0 Pj),J = 0

(4)

represent 4 + 6 + 6 = 16 differential equations for 16 unknowns h.o and kx,. These conditions reduce in 1? to

(a) (b) (c)

= 0 (mod e), (Vx = h-V,,) AKx,, + ha$(H^X:Kp, + 2Hp,,x,K,A,O) = 0 (mod e) V [PIPI + - Ka,pHx,',g hal) =_ 0 (mod e) (5)

2ePxf(hP]aH`a,Kp

where IP = e`#PV"K,,, is (up 'to a numerical factor) the Einstein electric current vector density. In order to find a solution of (5) we introduce a neighborhood NE(P) of a general point P(x) N0(P) is the set of all points whose coordinates x' satisfy the condition I(x' - x') 1 < e. If hAr = Oh/F + 'e h' in Ne(P) (6) then (5) reduces to

0AK

0 (mod e),

= 0 (mod e).

It is not difficult to find a solution KX, of this system. Thus for instance K --A (z -t) Ku -B(x + iy), KXp O for XA 6 12, 34 (mod e) in N (P) is a solution, provided B is an analytic (regular) function of x + iy (and = 'hu = -.hO4 = 1, the remaining = 0). 'hi 6. The First Einstein Conditions.-If we put

=hr

0hx,,

r(,) RxS= ri; - rX$r - 2 (8x~~~~~ +8x r(a) axa

rS(

then Einstein's conditions imposed on the r',p are (7) S,tA= 0, RXp = O and represent in this form n + n2 conditions for n2 unknowns hx and k,".

sltin(n 2 However, in RX they spltin n( +l) + n

(a) H,, = 0 (mod e2),

.(n -i) = n2 conditions namely 2

haz'H4Kpp = 0
(mod e).

(b) AKx,, +

(8) This system admits a solution h=,,ohX,, (mod e2) in N.(P) where o' is an explicitly defined function of position. Moreover we have in this

246

MA.THEMATICS: V. HLA VAT

P PROC. N. A. S.

case H =,,x 0 (mod _2) in N.(P). Hence a coordinate system may be found for which hx, are constant in NE(P) so that (8(b)) reduces to 'AK), 0 (mod e) in N. (P). Moreover one obtains the same condition if (6) holds, or if one assumes that the set of bivectors B1(B2) generates a non-holonomic surface which differs only mod e in N.(P) from a set of holonomic surfaces. However, in this case (8(a)) leads to hk,, = h,, (mod e2) in N.(P) while if (6) holds the equation (8(a)) reduces to

h axab hL + axabX) h'Pp - ?xPI)xX hal) (mod e). (9) It is not difficult to find a non-trivial solution of (9). In particular if 1 a oh )a-haOXa a h (mod e)' kX
=-.

oAh)

then (9) reduces to

(10) (mod e) If Ohl, = 0hkn = 0h33 = -h44 = 1 (and the remaining Oh), = 0), then (10) admits a solution h= p,, (x + iy, z - t) (mod e) in NE(P) = where p p,,, are arbitrary (regular) functions of x + iy. 7. The Second Einstein Conditions.-Any solution of (8) is also a solution of the second Einstein system
0 Sx,,= R(Ix) =
0

oAh I -=0

&X [R1,X] =

(11)

which (for n = 4) represents 4 + 10 + 4 = 18 conditions for 16 unknowns. This system reduces in R4 to

(a) H,

(mod E2)

(12) (b) AIT + ewI (H:,VKep - 4H,VpK,,p)h" 3 0 (mod e). w In all cases mentioned in section (6) we obtain the same solution for h),... while (in these cases) the equation (12(b)) reduces to 'AI" = 0 (mod e). If Ohl, = h22 = 0h33 = - 0h44 = 1 the remaining 0h,,, = 0, then IF -f"(x + iy, z-t) (mod e) in Ne(P) is a solution of this equation, provided that fv is analytic (regular) in x + iy.
* Prepared under Army Contract DA-33-008-Ord-224. 1 Einstein, A., The Meaning of Relativity, Princeton University Press, 1950, Appendix II. 2 HlavatO, V., "Spinor Space and Line Geometry," Canadian J. Math., 3, 442-459 (1950), and "Spinor Space and Line Geometry II" (to be published in the Journal of

X OL. 38, 1952

MA THEMA TICS: M. MORSE

247

Rational Mechanics and Analysis) where the Dirac equations are interpreted by means of the complex and the congruence mentioned above. 3A B (mod em) is the analytical description of the statement that the difference A - B is of the magnitude order of em" m' > m. 4 It is worth while to stress that (9) and (10) are covariant equations [thanks to (6)]. In particular (10) cannot be enforced by a proper choice of coordinates!

HOMOLOG Y RELA TIONS ON REG ULAR ORIENTABLE MANIFOLDS

By MARSTON MORSE
INSTITUTE FOR ADVANCED STUDY

Communicated January 25, 1952

1. Introduction.-The theorems stated in 1927 by Morse in' reference 1 are becoming increasingly significant in questions of homology, homotopy and category. Studies related to the unsolved Schoenflies problem by Morse and Baiada are based on these theorems. It is for this reason that the unpublished proofs of these theorems as written in 1927 will be here given with relatively obvious extensions whereby chains over the field of integers mod 2 are replaced by chains over any field cL. The connectivities used in reference 1 will here be replaced by the dimensions R1 of the respective homology groups, always using cycles over (D. The complexes involved will be similar to those used in reference 2. They are broken up into cells with the properties of "Block-ketten" of reference 5, page 78. The ranks of the incidence matrices will be determined "relative" to cI, replacing the elements a of a matrix by ea, where e is the unit in (D, and evaluating the subdeterminants of the resultant matrix as elements in (D. 2. Regular Manifolds and Regions.-A regular n-manifold Ej. in a Euclidean (n + r)-space E+,, n > 0, r > 0 is a closed, compact, connected Hausdorff subspace of E.+, specially representable as follows. Let U be an arbitrary open set in a Euclidean space En with rectangular coordinates u = (ul, . . ., us). Let y = (yi, . ., Yn+r) similarly represent a point in En+r. Let P be an arbitrary point of En. Some neighborhood N(P) of P relative to En shall be the homeomorph of an open set U e En under a mapping of u e U onto [a/1(u), . .., lIn+r(u)] e En+,r where the functions y6i are of class Cm, m > 0, and possess a functional matrix of rank n. By a regular subregion of En will be meant a closed non-empty subset of En bounded by a finite set of disjoint regular (n - 1)-manifolds. Unless otherwise stated all manifolds including boundaries of regions are to be regular and given by representations of class C"'. Other regular representations of a lower class may sometimes be used.