338

DOCS.

349,

350

JUNE

1917

Einstein discusses

Felix

Ehrenhaft’s

theory

of

“subelectrons,”

providing

equations

for

its

testing.

The last

section

is

devoted to

the

question

of

the

compatibility

with

the

wave

theory

of

directed

momentum

transfer:

...

Although Pointing’s

theory

of momentum

is

compatible

with

Maxwell’s

equations,

it

is

not

a

consequence

of them. Our

inability

until

now

to

discover

a

detailed

energy-momentum

localization

analogous

to

quanta,

should not imme-

diately

lead to

the

interpretation

that

it

is

an impossibility

...

350. To

Paul

Ehrenfest

[Berlin,]

3

June

1917

Dear

Ehrenfest,

I

am

corresponding quite busily

with Adler

because

he

has written

something

about

relativity theory

that

is

unfortunately

only very flimsily supported.[1]

But

otherwise,

he

is

a

terrific

fellow.

It

would

be

best

to

send

offprints

to

his

wife

(Mrs.

Adler,

junior,

for Dr. Fritz

Adler,

1

Blumel

Alley,

Vienna

VI).

I

optimistically

hope

that

nothing

will

happen

to him.

He

has

the

sympathy

of

all discrimi-

nating

persons,

also here

more

than

would be

expected.

Slowly

but

surely,

a

change

is

generally

taking place

toward

a

moderate

and

natural

mentality.

But

a

tremendous

amount

of fertilizer

is

needed before

the little

plant

can grow!

The

generalization

by

Sommerfeld

is

as

follows.[2]

Let

there

be

a

problem

in

which

as

many integrals

L(qv, pv)

=

const

exist

as

degrees

of

freedom. Then

the

momenta

pv

can

be

expressed

as (multiple)

functions of

qv.

On

the other

hand,

the

path’s

curve

fills

a

certain

portion

of

the

qv-space entirely,

so

that it

comes

arbitrarily close to

every

point

in it.

Then the

system’s

path

in

the

qv-space

generates

a

vector field

for

the

pv’s.

In

a

“Riemann

foliated”

qv-space,

the

pv’s

can

be

interpreted

as

unique

and

always

constant

functions of

the

qv’s.

Now

we

regard

the

sum

da

=

Y^Vudqv,

V

formed for

a

random line element

of

the

qv-space.

This

sum

is

invariant

under

coordinate transformations and

is

in

addition

a

complete differential.

The latter

can

be inferred from

Jacobi’s

law.

The

integral

fda