2 0 D O C . 2 5 A P R I L 1 9 1 9

nothing higher than 1st derivatives of

.)[3]

The reasons are set forth precisely in

the proof of the paper I sent

you.[4]

Now it seems to me, though, that another un-

solved puzzle is hiding behind energy conservation, because in the current solution

there is nothing as comparably compelling as the case without gravitation.

Furthermore, I would like to relay to you an argument that allows the spherical

option to appear preferable to the elliptic

one.[5]

In the spherical world every closed

line can be contracted continuously into one point, but not in the elliptic one; i.e.,

only the spherical, not the elliptic, world is simply connected. (Because a line that

in the spherical world links one point with its antipodal point is, in the respective

elliptic one, a noncontractable closed line.)

In addition to the Euclidean line element, there are, of course, also finite spaces

of arbitrary size that can be obtained from infinity by postulating a threefold peri-

odicity, if it is further postulated that periodically lying points are identical. These

possibilities, which incidentally do not apply in the case of general relativity, suffer

from the characteristic that these spaces are connected in multiple ways.

With cordial greetings, yours very truly,

A. Einstein.

25. From Leonhard Grebe[1]

Bonn, 17 April 1919

Highly esteemed Professor,

Pursuant to your request during our discussion in Berlin, I take the liberty today

of presenting you with the following:

Together with my colleague Dr. A.

Bachem,[2]

I have been working, since my

return from the battlefield, on the redshift of spectrum lines in the solar spectrum

predicted in your investigations and, specifically, as did Schwarzschild, on the

soc[alled] cyanogen

band.[3]

In the process, as yet unexplained displacement

anomalies have emerged for the various lines of this band. While some lines exhibit

the effect to sufficient approximation, others yield deviations far above the obser-

vational error, which are all the more remarkable since they belong to lines that,

owing to their association within a single series of band lines, would lead one to

expect the same behavior. The assumption suggests itself that these deviations are

elicited by closely lying attendant lines or by asymmetries in the intensity distribu-

tion of the lines, and that Schwarzschild’s result was substantially affected by the

inclusion of these lines to obtain a

mean.[4]

gμν