90
DOCS.
76,
77 APRIL
1915
this
follows,
taking
into account
the
theorem of momentum
[conservation],
that
upon magnetic
commutation
of
a
magnetizable body
the
molecules translate into
one
rotation
moment
or
other,
according
to
the
formula
rotation
moment
=
2
electron
masses
-.
electron
charge
d-dt(magnetization)
_
2ßdM
[e]
dt
The
experiments
were
done
by observing
the
oscil-
lation
motions
of
a
little
iron rod
that
was
hanging
on a glass
fiber within
a
coil
through
which
an
al-
ternating
current
was
flowing
(use
of
resonance.)
glass
fiber
coil
little iron
core
77. To Tullio Levi-Civita
[Berlin,]
20
April
[1915]
Highly
esteemed and dear
Colleague,
Just
now
I
was
studying
your very interesting
letter
of
April
15th.
Admittedly
I
do not
concur-as
is
explained
in
more
detail in
the
following-with
your
proof
that the
selection of
the
ôguv's as
“quasi-constants”
was
unworkable;
but
I
gladly
acknowledge
that
you
have
put
your finger
on
the
weakest
point
of
the
proof,
namely,
on
the
independence
of
the
A(uv)’s.
Here
the
proof
lacks
precision;[1]
the
statement
on
page
1072
of
the
paper
“it
will
thus
be
equivalent
to
an
arbitrary
variation
of
the
6guv's"
dispenses
with
rigorous
substantiation; in
the
special
case
of
a
constant
guv
it
is
even
incorrect.[2]
Yet
I
do cherish
the
firm belief
that
in
general
it
applies,
because
the
number
of
freely
selectable variables
that determine
the
10
Sguv,s
is
10,
and because
both
variations
b1
and
b2
are
of
fundamentally
different
kinds,
in
that
a
b2
variation
generally
is not
a
b1
variation.
Now to
your proof. Right
at
the
outset
you say
that
for small
regions
£
A(uv)
=
ƒ
Ôg^dr
=
J
byg^dr,
thus
that the
ƒ S2g^dT
terms
vanish.
This
already
I
contest,[3]
namely
for
exactly
the
same reason
that
I
regard
your
deduction of
the
vanishing
of
the
J
Sadr’s