.

NATURE [SEPTEMBER 22, 1923

endoparasitic organisms, will determine largely the have formed an important branch of the zoological
extent to which he can use and develop the natural work of Great Britain for forty years, and the studies
resources of the rich tropical and sub-tropical zone of on genetics which made possible an explanation of the
the earth. mode of inheritance of a particular blood-group, and
Other applications of zoology to human well-being of some of the defects (e.g. colour-blindness and
cannot be dealt with here, but mention should be made hremophilia) and malformations which appear in the
of two-the researches on sea-fisheries problems which human race.

The Theory of the Affine Field. 1

By Prof. ALBERT EINSTEIN, For. Mem. R.S.

T HEandtheory of the connexion between gravitation

electromagnetism outlined below is founded
gravitational field , can be obtained from a geometrical
consideration based solely on the law of the affine
on Eddington's idea, published during recent years, of relation given by (z) above. The manner in which the
basing " field physics " mathematically on the theory r:, are expressible in terms of the plays no part
of the affine relation. We shall first briefly consider in this consideration. The behaviour in the case of
the entire development of ideas associated with the differential operations of the absolute differential
names Levi-Civita, Weyl, and Eddington. calculus is analogous.
The general theory of relativity rests formally on These results naturally lead to a generalisation of
the geometry of Riemann, which bases all its concep- Riemann's geometry. Instead of starting off from the
tions on that of the interval ds between points in- metrical relation (r) and deriving from this the co-
definitely near together, in accordance with the efficients r of the affine relation characterised by (z),
2
formula . (r) we proceed from a general affine relation of the type
(z) without postulating (r). The search for the
These magnitudes determine the behaviour of mathematical laws which shall correspond to the laws
measuring-rods and clocks with reference to the co- of Nature then resolves itself into the solution of the
ordinate system, as well as the gravitational field. question : What are the formally most natural con-
Thus far we are able to say that, from its foundations, ditions that can be imposed upon an affine relation?
the general theory of relativity explains the gravita- The first step in this direction was taken by H. Weyl.
tional field. In contrast to this, the conceptual founda- His theory is connected with the fact that light rays
tions of the theory have no relations with the electro- are simpler structures from the physical view-point
magnetic field. than measuring-rods and clocks, and that only the
These facts suggest the following question. Is it not ratios of the g"'" are determined by the law of pro-
possible to generalise the mathematical foundations of pagation of light. Accordingly he ascribes objective
the theory in such a way that we can derive from them significance not to the magnitude ds in (r), i.e. to the
not only the properties of the gravitational field, but length of a vector, but only to the ratio of the lengths
also those of the electromagnetic field ? of two vectors (thus also to the· angles). Those affine
The possibility of a generalisation of the relations are permissible in which the parallel displace-
matical foundations resulted from the fact that Levt- ment is angularly accurate. In this way a theory was
Civita pointed out an element .in the geometry arrived at, in which, along with the determinate (except
Riemann that could be made mdependent of thrs for a factor) other four magnitudes c:p,. occurred,
geometry, to wit, the " affine relation " ; for according which Weyl identified with electromagnetic potentials.
to Riemann's geometry every indefinitely small part of Eddington attacked the problem in a more radical
the manifold can be represented approximately by a manner. He proceeded from an affine relation of the
Euclidean one. Thus in this elemental region there type (z) and sought to characterise this without intro-
exists the idea of parallelism. If we subject a con- ducing into the basis of the theory anything derived
travariant vector A" at the point x, to a parallel from (r), i.e. from the metric. The metric was to
displacement to the indefinitely adjacent point x, +ox,, appear as a deduction from the theory. The tensor
then the resulting vector A"+oA" is determined by an
expression of the form
R p.v =
-r.. .. r/3vo.
or"" .. r11o./l (3)
OXo. p.{3 OXv p.v

The magnitudes r are symmetrical in the lower indices, is symmetrical in the special case of Riemann's
and are expressed in accordance with Riemann geometry. In the general case is split up into a
geometry by the and their first derivatives symmetrical and an " anti-symmetrical " part :
(Christoffel symbols of the second kind). We obtain
these expressions by formulating the condition that Rp.v=/'p.v+</>p.v • (4)
the length of a contravariant vector formed in One is confronted with the possibility of identifying
accordance with (r) does not change as a result of the /'p.v with the symmetrical tensor of the metrical or
parallel displacement. gravitational field, and </>p.v with the antisymmetrical
Levi-Civita has shown that the Riemann tensor of tensor of the electromagnetic field. This was the
curvature, which is fundameNtal for the theory of the course taken by Eddington. But his theory remained
• Translated by Dr. R. W. Lawson.
incomplete, because at first no course possessed of the
2 In accordance with custom, the signs of summation are omitted. advantages of simplicity and naturalness presented
NO. 2812, VOL. I 12]
©1923 Nature Publishing Group
 SEPTEMBER 22, 1923] NATURE 449
itself, for the determination of the 40 unknown functions
r }.<:. The following brief statement will serve to show <P}Lv = t(ar }.<':.. _ ol';,';,.) . (u)
OXv ox}L
how I have endeavoured to fill in this gap. 3
If the German capital SJ be a scalar density that ;v
In these equations the r on the right-hand side are
. . only on the functions r }Lav, then Hamilton's
depends to be expressed by means of (8) in tPrms of the
pnnCiple gJLv and Moreover, if SJ is known, then on the
. (s) basis of (7) YfLv and i.e. the left-hand sides of (ro)
and (n), can also be expressed in terms of and
supplies us with 40 differential equations for the
This latter calculation can be simplified by means
functions r' when we stipulate that during the variation
of the following artifice. Equation (6) is equivalent to
the functions r are to be treated as magnitudes in-
dependent of each other. Further we assume that SJ the statement that
depends only on the magnitudes Y}Lv and <f>J.<v, and oSJ*=yfLVOgfLV+</>fLVOfJLV , , (6a)
thus write
is also a complete differential, so that if SJ* is an
(6) unknown function of the and ffLV, the following
where we have relations will hold :
oSJ
= g}LV I - o&J*)
YfLv-ogJLv
. (?a)
= fJLV J.
. (7) _a&J*
o</>J.<v 1

At this point it should be noticed that in the theory We now have only to assume SJ*. The simplest
developed here, the small German letters respectively possibility is obviously
represent the contravariant density (gJ.<v) of the metrical
tensor, and the contravariant tensor density (f1Lv) of the SJ*= _@JfLvf}Lv • • (12)
2
electromagnetic field. Thus in a well-known manner
is given the transition from tensor densities (expressed In this connexion it is interesting that this function
by German letters) to contravariant and covariant does not consist of several summation terms which are
tensors by the corresponding italic letters), logically independent of each other, as was the case
and a metnc 1s mtroduced which rests exclusively on with the theories hitherto proposed.
the affine relatiDn. In this way we arrive at the field equations
By performing the variation we obtain after some R}Lv= · (13)
amount of calculation
whereby RfLv is the Riemann tensor of curvature. K
and yare constants, ffL is the electromagnetic potential,
which is connected with the field strength by the
relation
f = 'iJjJL- Oj,_ . (r4)
fLV OXv oxfL .
and with the electrical current density by the relation
Equation (8) shows that o:Jr extension of the theory
which appears to be so general, leads to a structure of • . (rs)
the affine relation that does not deviate more strongly In order that these equations may be in accord with
from that of the geometry of Riemann than is required experience, the constant y must be practically in-
by the actual structure of the physical field. definitely small, for otherwise no fields would be possible
We now obtain the field equations in the following without noticeable electrical densities.
manner. From (3) and (4) we first derive the relations The theory supplies us, in a natural manner, with
the hitherto known laws of the gravitational field and
ar a
YJ.<v= - __!LV+l'
1ol' a ar a)
t( ___e+----"" - r a rllll (ro) of the electromagnetic field, as well as with a connexion
OXa }L va \ OXv cx}L }LV a as regards their nature of the two kinds of field ; but
3 it brings us no enlightenment on the structure of
Herr J?roste of Leyden hit upon the same idea independently of the
present writer. electrons.

Further Determinations of the Constitution of the Elements by the Method

of Accelerated Anode Rays. 1
By Dr. F. W. AsToN, F.R.S.
BY rays,
further use of the method of accelerated anode
results have been obtained with a number
method will be published later. Most of the following
results were obtained by the use of fluorine compounds
of elements since the publication of the isotopes of of the elements investigated.
copper (NATURE, Aug. 4, p. r62). Details of the The mass-spectrum of strontium shows one line only;
1 at 88. This was obtained in considerable intensity.
A paper on September r8 before Section A of the British Association
Meetmg at Liverpool. If any other constituents exist they must be present
NO. 2812, VOL. 112]
©1923 Nature Publishing Group