14 DOC.

8

MAY

1914

8. To Paul Ehrenfest

[Charlottenburg, 18

May

1914]

Dear

Ehrenfest,

I

would have

had the

devil to

answer

to

long ago,

had

he done his

duty,

for not

answering

such

a

dear

person

as

you

for

so

long.

The

photographs

are charming.

I

was delighted

with

them.

The business

about

rotating

bodies

in

a

magnetic

field is

not

quite

clear

to

me.[1]

Mutual

energy certainly

does exist between

two

electrical

circuits.

But

there

is

no

mutual

energy

between

a

current

and

a

magnet;

should

this

fiction be

unpermissible

in

principle?

Visualize the state of

affairs

clearly!

magnet

fixed plane

rotating

electic

charge

rod.

Depending

on

the

position

of

the

magnet,

the

probabilities

of

the

various

rotation

velocities should be different

a

priori,

even

though

the

electrodynamic

forces

are

constantly

being compensated by

the

rod. This

necessarily follows

if

you

assume

that the

“probabilities

a

priori”

are

not

altered

by

adiabatic

change

(infinitely

slow

movement

of

the

magnet).

I

have considered

your

last letter

thoroughly.

You

are

inclined

to

the

view

that

lg

W deviates from

thermodynamic

entropy.

You

prove

that this

is

really

the

case

if

the

probability

function

G(q, p,

a) depended

on

the

a

parameters.[2]

I

believe, however,

that the

assumption

of such

a

dependency

is

not

permissible,

that

it

is

even

completely

contrary to

Boltzmann’s

conception.

For if

you

just

have

a

look

at

the

states

the

system

assumes

of

its

own

accord in

the

course

of

time,

it

is

evident

that then

variable

parameters

a are

altogether

inconceivable

as

independent

entities. Therefore

I

am

of the

opinion

that

G

can

only

be

dependent

on

p

and

q,

but

not

on a.

The scale

by

which

the

probabilities

of the

system

states

are

evaluated

may

not itself be made

dependent

on

the

system

state. Otherwise

a

comparison

of

probabilities

calculated

under such

varying

states

would be futile.