(Knebelman M. S.) Conformal Geometry of Generalize (BookFi) PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

376 MA THEMA TICS: M. S. KNEBELMAN PRoS. N. A. S.

ti, . . ., t ranging over H. Hence A has rank tr with respect to H and


its rank equation has the symmetric group with respect to K = HQ1,
... .) and is irreducible in K.
By applying theorem 1 we have
THnOREM 4. Let A be a simple algebra over F expressed by A = B 0 M
where M is a total matric algebra of order r2 and B is a normal division
algebra in t2 units over its central field R(v) of order s. Then the rank of
A is str and the rank function of A with respect to F is irreducible in the
field of its coefficients.
1 Dickson, L. E., Algebren und ihre Zahlentheorie, p. 260.
2 Cf. Noether, E., Mathematische Annalen, vol. 78.
3 Wedderburn, Bulletin of the American Mathematical Society, 31 (1925), 11-13.
4 Cf. the author's paper, "On the Group of the Rank Equation, etc.," these PRo-
COUDINGS 14, (1928), No. 12, pp. 906-7.

CONFORMAL GEOMETRY OF GENERALIZED METRIC SPACES


BY M. S. KNEBsLMAN*
DEPARTMENT OF MATHUMATICS, PRINCETON UNIVERSITY
Communicated February 25, 1929
1. Let V,, be an n-dimensional generalized metric space-one to which
there is assigned an absolute scalar differential invariant f(x, dx), f being
positively homogeneous of degree two in dxl, . . ., dxT. If f.i =flbdx'
and gij =/2fij.J the length of a vector t relative to the element dx is by
definition
t2= g(xdx)t1, (1.1)
the summation convention for every repeated index being used.
I define two metrics f(x, dx) and f'(x, dx) as conformal if the length of
an arbitrary vector in the one is proportional to the length of this vector
in the other. Let 'p(x, dx) be the factor of proportionality; then from
(1.1) we obtain
gij = 'pgij
and therefore
gij.k = Vgqj.k + (o.kgVj = Vgjk.i + '.igjk
Hence
X'P.k = 5k4.jX
from which we get by contraction V.k = 0. Hence the factor of propor-
tionality is at most a point fulrction.
For convenience we write 1 2a,
gij e
= (1.2)
VOL. 15, 1929 MATHEMATICS: M. S. KNEBELMAN 377

then if gt' is the normalized cofactor of gij in g,,e which is assumed


not to vanish, we have
g I ij e- 2ai
g. (1.3)
Let r and r1 be the fundamental affine connections of the two spaces,
their components being given' by expressions of the form
rik = {jk} + {ik}1dx
+ ti}.kdX+/2
1
{a }j.k dxadoXa ; (1.4)
{jk} being the Christoffel symbols of the second kind formed out of the
g's. When these components are compared we find that
rk= rk- A'0k o (1.5)
where oac = bal/?)x and A"a = l/2fgia -dxidxa.
If Kekz are the components of the affine curvature tensor, we find by
means of (1.5) that
Kjkl (As l~ha i#
Kjl= Kk-
I

(A Jk 0,al -Ail1 0 ak) + 0a0


af (A.k A.lh
AIh.k).i - a,a(AT,i.j - A-[,kj), (1.6)
where a subscript preceded by a comma indicates covariant differentiation
with respect to the r's.
The conditions of integrability of (1.5) and a., = 0 furnish the com-
plete set of conformal invariants, for the two metrics f and f'. Before
outlining the method for obtaining these invariants it may be worth
noticing the essential difference between our problem and the correspond-
ing one for a Riemann space. If f is a Riemann metric, the coefficient
of 0,a in (1.5) vanishes identically and the algebraic elimination of the
( + ) quantities of ma, carries with it the elimination of 0,a C,, so
that the result of elimination is a tensor, -the well-known conformal
curvature tensor of Weyl. In the generalized metric space the elimina-
tion of ar as gives a conformal invariant which in general is not a tensor
while the elimination of the n2 + 2n quantities a a a,o-aa,, and oma if it
can be performed, will give a conformal tensor.
2. An invariant which we call a conformal connection may be defined
in two ways, each of which involves an additional assumption about the
metric f. .The first connection is analogous to the one developed by
J. M. Thomas2 and the second one by T. Y. Thomas.3 For Riemann
spaces and in fact for generalized metric spaces whose Hessian with respect
to dx is a point function these connections are identical, both assumptions
being valid for spaces of this sort.
Let f be any generalized metric. With T. Y. Thomas we let F
f gI - 1/", F being a relative scalar of weight - 2/n. We assume that
378 MA THEMA TICS: M. S. KNEBELMAN PROC. N. A. S.

the Hessian of F with respect to dx does not vanish identically. Let x' =
(p'(2) be any analytic transformation of coordinates whose jacobian
(x,x) 0 0; then if Gi1= 2F~ij,
G-y = (x, )2/n Gaj ugaU7 (2.1)
where uE' = ?Xa/)ti and if GCV is the normalized cofactor of G(y in I Gas I
- (x, )2/n Gaf Vi (2.2)
where v ia= 6xa.
When we form the affine connection out of the G's (cf. (1.4)) we find
that its components Kjk transform according to the law
K~kUs = K,8Sy~t~k + U,,k + A4,klisUi' (2.3)
where
=2xa/8jaxk
- , 4 = 1/n a log (x,2)/bs
and (2.4)
AiS - 1/2 F USS - d~idf.
The invariant whose components are Kk we call the conformal con-
nection and if P~ki is the analogue of the curvature tensor formed out of
the K's, its law of transformation is
piki P5U-ukUl vt
pa
(Aiakl-',a/l - A.j l+1/A/k) -
+
ll'/aol/-+
(A.k - Ah +k(A-. .

-A-i/k j) (2.5)
a subscript preceded by a solidus denoting covariant differentiation with
respect to the K's. When the expressions for Pjkl + 1/2 Pijk are evaluated
we get n(n + 1) 2 equations that may be solved for the n(n + 1) 2 quanti-
ties lI'//,a giving expressions of the form
= S'azl [Piki'/2PXt)z4kuZ1
vJ/a/ + 1/2Ptjk - (P;.i + +
TVa,6/j~lk + Uj ,I//j. (2.6)
By means of (2.6) we are able to eliminate higher derivatives of 4i and
thus obtain an infinite sequence of conformal invariants which are not
tensors in general. These invariants together with Kjk.lh.12. .. /ml/m2... and
Gij.ki.k.... constitute a complete set of conformal invariants, the proof of
this fact being analogous to that used in the metric equivalence of
generalized spaces.
3. By making a further assumption about f(x, dx) we can form a con-
formal connection which leads to conformal tensors. This assumption is
false for spaces whose Hessian is a point function and consequently false
for Riemann spaces. That it is not always false can be shown by a simple
example.
VoL. 15, 1929 MATHEMATICS: M. S. KNEBELMAN 379

From (2.3) we obtain by contraction


khk = Kafki + As(,na + Ahk)
or (3.1)
Khh = K au + #s (1/2Fp Z).k-
We denote the absolute conformal tensor ('/2FG%).k by Tk. If we
I
assume that Tkj 0, we may define Sk, by
Tk St = Mp. (3.2)
Equations (3.1) may then be solved for h giving
Ek=-hrr Khrgua -

Hence if we define an invariant whose components are


Ck = '/2(K',Oix.dxax - A KfSa).Jk, (3.3)
we shall find their law of transformation to be
CGkUi = Cla U4U+ (3.4)
Thus, if this invariant exists the complete set of conformal tensors is
obtained by replacing r~k in the complete set of metric tensors by the
corresponding C0.*
* NATIONAL RSARCH FoLLow IN MATrsmAnIcs.
1 L. Berwald, Jahr. der Deut. Math. Ver., 1925, 34, pp. 213-20.
2 These PROCOUDINGS, 12, (1926), 389-393.
3 Ibid., 12, (1926), 352-359.

You might also like