/

ON GRAVITATIONAL WAVES.
BY
A. E I N S T E I N

and N. R O S E N .

ABSTRACT.
T h e rigorous solution for cylindrical gravitational waves is given. For the
convenience of the reader t h e theory of gravitational waves and their production,
already known in principle, is given in the first part of this paper. After encountering relationships which cast d o u b t on t h e existence of rigorous solutions for
undulatory gravitational fields, we investigate rigorously t h e case of cylindrical
gravitational waves. It turns out t h a t rigorous solutions exist and t h a t the
problem reduces to t h e usual cylindrical waves in euclidean space.
I. APPROXIMATE SOLUTION OF THE PROBLEM OF PLANE WAVES
AND THE PRODUCTION OF GRAVITATIONAL WAVES.

It is well known t h a t the approximate m e t h o d of integration of the gravitational equations of the general relativity
theory leads to the existence of gravitational waves. The
m e t h o d used is as follows: We start with the equations
-/~. v - - ~1 g , . J ~ =

-- T,..

(I)

We consider t h a t the g., are replaced by the expressions

g., = a., + u.,,

(2)

where
~,v =

=o

if
if

/~ = v.
~v,

provided we take the time co6rdinate imaginary, as was dotLo

by Minkowski. It is assumed t h a t the %. are smal, i.e.
t h a t the gravitational field is weak. In the equatiors the
3`.. and their derivatives will occur in various powers. If the
3`.. are everywhere sufficiently small compared to uaity one
obtains a first-approximation solution of the equations by
neglecting in (I) the higher powers of the 3`.. (~nd their
derivatives) compared with the lower ones. If one introduces
further the ~., instead of the 3`.. by the relations
--

VOL. 223, N0. I333--4

43

44

A.

EINSTEIN

AND N .

ROSEN.

[J. F. I.

then (I) assumes the form

G ....

G ....

~ ....

~ +

v.o, ~ =

2T.,.

(3)

The specialization contained in (2) is conserved if one

performs an infinitesimal transformation on the coordinates:
X ! :

x~ + ~,

(4)

where the ~ are infinitely small but otherwise arbitrary

functions. One can therefore prescribe four of the ~ or
four conditions which the ~ must satisfy besides the equations (3); this amounts to a specialization of the coordinate
system chosen to describe the field. We choose the coordinate system in the usual way by demanding that
~o,

o =

o.

(5)

It is readily verified that these four conditions are compatible

with the approximate gravitational equations provided the
divergence T~, ~ of T~ vanishes, which must be assumed
according to the special theory of relativity.
It turns out however that these conditions do not completely fix the coordinate system. If %, are solutions of (2)
and (5), then the % / a f t e r a transformation of the type (4)

are also solutions, provided the ~ satisfy the conditions

F~

+ ~'~ - } a . ( ~ . o + ;~ o)],~ = o,

or
e,~

o.

(7)

TI a ~:field can be made to vanish by the addition of terms

like chose in (6), i.e., by means of an infinitesimal transformation, then the gravitational field being described is only
an apparent field.
With reference to (2), the gravitational equations for
e m p t y sp~ee can be written in the form

i}
One obtains plane gravitational waves which move in the

Jan., I93~f"

ON

GRAVITATIONAL WAVES.

45

direction of the positive xl-axis b y t a k i n g the ~, of the f o r m

~(xl + ix4)( = ~ ( x l - t)), w h e r e these ~ , m u s t f u r t h e r satisfy
the conditions
"~11 + i'~4 = o,]
%1 + i~44 = 0,['
~:1 Jr-i~24 = O, r

%1 + i%4 = o.

(9)

One Can a c c o r d i n g l y s u b d i v i d e the m o s t general (progressing)

plane g r a v i t a t i o n a l w a v e s into t h r e e t y p e s :
(a) pure longitudinal waves,
o n l y 7n, 7~, ~y44tim t r e n t from zero,
(b) half longitudinal, half t r a n s v e r s e w a v e s ,
o n l y ~21 and ~24, or o n l y ~al a n d ~34 different f r o m zero,
(c) pure t r a n s v e r s e waves,
only ~22, ~'23, ~33 are different from zero.
On the basis of the p r e v i o u s r e m a r k s it can n e x t be s h o w n
t h a t e v e r y w a v e of t y p e (a) or of t y p e (b) is an a p p a r e n t field,
t h a t is, it can be o b t a i n e d b y an infinitesimal t r a n s f o r m a t i o n
f r o m the e u c l i d e a n field (~,~ = v , , = 0).
W e c a r r y o u t the p r o o f in the e x a m p l e of a w a v e of t y p e
(a). A c c o r d i n g to (9) one m u s t set, if ~ is a s u i t a b l e f u n c t i o n
of the a r g u m e n t x~ + iX4,
hence also
"Yll ----- (P,

"}/14 = i(~,

"Y44 =

-- (~.

If one n o w chooses ~' a n d ~4 (with ~'~ = ~3 = o) so t h a t

~1 = X(Xl _[_ iX4),

~4 = i x ( x l + i:~4),

then one has

~1.1 + }',1 = 2x',

~1,4 + }4,1 = 2ix',

~4,4 + }4,4 = - 2x'.

T h e s e agree w i t h the v a l u e s given a b o v e for 7 . , 714, ~44 if one

chooses x' -- -~. H e n c e it is s h o w n t h a t these w a v e s are

46

A.

EINSTEIN

AND

N.

ROSEN.

[J. F. I.

apparent. An analogous proof can be carried o u t for the

w a v e s of t y p e (b).
F u r t h e r m o r e we wish to show t h a t also t y p e (c) contains
a p p a r e n t fields, namely, those in which ~22 = ~aa ~ o, ~2a = O .
The corresponding %~ are y u -- 744 # o, all others vanishing.
Such a w a v e can be obtained b y taking }' = x, ~4 = _ ix,
i.e. b y an infinitesimal t r a n s f o r m a t i o n from the euclidean
space. Accordingly there remain as real w a v e s only the two
pure transverse types, the non-vanishing c o m p o n e n t s of
which are
722 ----- - - "~/33,

(Cl)

7 3.

(c2)

or

It follows however from the transformation law for tensors

t h a t these two t y p e s can be t r a n s f o r m e d into each other b y
a spatial rotation of the co6rdinate s y s t e m a b o u t the xl-axis
through the angle rr/4. T h e y represent merely the decomposition into c o m p o n e n t s of the pure transverse w a v e (the only
one which has a real significance). T y p e ct is characterized
b y the fact t h a t its c o m p o n e n t s do not change under the
transformations
X2 !

- -

X2,

~'1 ! ~

~'1,

A~3/

: -

X3,

X4 ! ~

X4,

X3 f

: -

- -

~3,

Xl ! =

W1,

~3~2f

~2~

K4 !

0C4,

OF
=

in c o n t r a s t to c2, i.e. cl is s y m m e t r i c a l with respect to the

x,-x2-plane and the &-xa-plane.
W e now investigate the generation of waves, as it follows
from the a p p r o x i m a t e (linearized) gravitational equations,
T h e s y s t e m of the equations to be integrated is
....

(m)

L e t us suppose t h a t a physical s y s t e m described b y T ~ is

found in the neighborhood of the origin of co6rdinates. T h e
~,-field is then determined m a t h e m a t i c a l l y in a similar w a y to
t h a t in which an electromagnetic field is determined through
an electrical current system. T h e usual solution is the one

Jan., I937.]

ON GRAVITATIONAL WAVES.

47

given by retarded potentials

- =__'
~'"P

271-J

(Ii)

Here r signifies the spatial distance of the point in question

from a volume-element, t = x4/i, the time in question.
If one considers the material system as being in a volume
having dimensions small compared to r0, the distance of our
point from the origin, and also small compared to the wavelengths of the radiation produced, then r can be replaced by
r0, and one obtains

t" ~T""-](t-")dv'

Or

271"10

I-fT,,dye(t-r).

(I2)

The ~,, are more and more closely approximated by a plane

wave the greater one takes r0. If one chooses the point in
question in the neighborhood of the xl-axis, the wave normal
is parallel to the xt direction and only the components ~22,
N23, 733 correspond to an actual gravitational wave according
to the preceding. The corresponding integrals (I2) for a
system producing the wave and consisting of masses in motion
relative to one another have directly no simple significance.
We notice however that /'44 denotes the (negatively taken)
energy density which in the case of slow motion is practically
equal to the mass density in the sense of ordinary mechanics.
As will be shown, the above integrals can be expressed
through this quantity. This can be done because of the
existence of the energy-momentum equations of the physical
system :
T,~, ~ = o.
(13)
If one multiplies the second of these with x2 and the fourth
with x22 and integrates over the whole system, one obtains
two integral relations, which on being combined yield

fr ,&

l O ' f x2~T44dv"
20x42

(13a)

48

A. EINSTEIN AND N. ROSEN.

[J. F. I.

Analogously one obtains

f T33d.-I a2f
20x4 2
I

x32T44dv'

02

f T2 d -2o f xmT4d..
One sees from this t h a t the time-derivatives of the moments
of inertia determine the emission of the gravitational waves,
provided the whole method of application of the approximation-equations is really justified. In particular one also
sees that the case of waves symmetrical with respect to the
x l - x 2 and x1-x~ planes could be realized by means of elastic
oscillations of a material system which has the same symm e t r y properties. For example, one might have two equal
masses which are joined by an elastic spring and oscillate
toward each other in a direction parallel to the x3-axis.
From consideration of energy relationship it has been
concluded t h a t such a system, in sending out gravitational
waves, must send out energy which reacts by damping the
motion. Nevertheless, one can think of the case of vibration
free from damping if one imagines that, besides t h e waves
emitted by the system, there is present a second Concentric
wave-field which is propagated inward and brings to the
system as much energy as the outgoing waves remove. This
leads to an undamped mechanical process which is imbedded
in a system of standing waves.
Mathematically this is connected with the following
considerations, clearly pointed out in past years by Ritz and
Tetrode. The integration of the wave-equation
[B~ = - 4~rp
by the retarded potential
q~ =

is mathematically not the only possibility.

it with
g~ =

One can also do

Jan., 1937.]

ON

GRAVITATIONAL WAVES.

49

i.e. b y means of the " a d v a n c e d " potential, or by a mixture

of the two, for example,
I ~" EP](t+r) -Ir- EP](t--r)
dr.
g~ -= 2 J
r

T h e last possibility corresponds to the case w i t h o u t damping,

in which a s t a n d i n g wave is present.
It is to be r e m a r k e d t h a t one can t h i n k of waves generated
as described above which approximate plane waves as closely
as desired. One can obtain them, for example, t h r o u g h a
limit-process b y considering the wave-source to be removed
f u r t h e r a n d f u r t h e r from the point in question and at the
same time the oscillating m o m e n t of inertia of the former
increased in proportion.
II. RIGOROUS SOLUTION FOR CYLINDRICAL WAVES.

We choose the coordinates Xl, x2 in the meridian plane in

such a w a y t h a t Xx = o is the axis of rotation and x2 runs from
o to infinity. Let xa be an angle coordinate specifying the
position of the meridian plane. Also, let the field be symmetrical a b o u t every plane x2 = const, and a b o u t every
meridian plane. The required s y m m e t r y leads to the vanishing of all components g,, which contain one and only one
index z; the same holds for the index 3- In such a gravitational field only
gll,

g 2 2 , g33, g44,

g14

can be different from zero. For convenience we now take all

the cobrdinates real. One can f u r t h e r t r a n s f o r m the coOrdinates xl, x4 so t h a t two conditions are satisfied. As such we
take
g14 = O,
[
g l l = - - g44. [

(I4)

It can be easily shown t h a t this can be done w i t h o u t introducing a n y singularities.

We now write
- - gll = g44 = A , ]
-

g:,~ =

C,

5o

A. EINSTEIN AND N. ROSEN.

where A, B, C > o.
lates t h a t

(
2 .Rll

In t e r m s of these quantities one calcu-

I
)
B44 C44
2 gl,R = ~- -[- ~

I[B42
C42
2 ~ -[-

-- B--~+ A k B
B1C1

+NK
2A(R22

[J. F. I.

AI(B,

clh]
+ A ,B +-~]J'

C44

.A 11
Cll
A
C
i [ C12 C42

A44

2 g22R = - -t- C
- -

__

+ ~ [ -~
c - c-;
2A19" , 2A4 2]
+

2A(
-C-

Raa - 2 gaaR

-A44
-+
A

A2

B44

~ ; ]'

An

Bn

B
A
i [ 2A12
+ 2 [ A2

B
2A4 2
A2

BI 2
+ Bz
"~ Bn
Cn
2 (.R44 -- - e . R } = ~ +
I

B42 ]
B2 '
I [ B12

(I6)

C12

2[-~-+ ~

AI( I

B1C~
- - -BC
+

A ,B +-d]

<]!

B4G

+ N K + A k B + C].]
B,4

G4

I[ BIB,

2R,4 = B - + C - -

CiC4

- - ~ + c--~

A4(B1

AI(B4 c & l
+ A ,B +-g]]'
where subscripts in the r i g h t - h a n d m e m b e r s denote (lifter-

Jan., 1937.]

ON

GRAVITATIONAL

51

WAVES.

eutiation.
If we t a k e as field equations these expressions set
equal to zero, replace the second and third b y their s u m and
difference, and introduce as new variables
a = log A,

log (B/C),p
1og(BC),J

3"

(15a)

we get
23"44 AI- ~1 [ ~ 4

-~- 33"42 "@ ~t 2

3"12 - - 20L13"1

2o

3" -1

o,

(17)

2(O~11 - - O!44) -IT. 23"11 - - 23"44 -[- [~12 -~- 3"12 - - ~42 - - 3"42] = O, ( 1 8 )
~11 - - ~44 -t- [~13"1 - - ~4'~4-] = O, ( I 9 )
23"ll -[- 1[~12 -t- 33"12 -t- ~42 - - 3"42 - - 2G~l'Y1 - - 20/43"4] = O, ( 2 0 )
23"14 -t- [ ~ i ~ 4 -t- 3"13"4 - - 2o~13"4 - - 20~43"1] = O. ( 2 1 )

T h e first and fourth equations of this group give

3"1I - - 3"44 -Jr- (3"12 - - 3"42) = O.

(22)

(BC)~,

(23)

The substitution
3" = log 0",

0" =

leads to the w a v e equation

0"11 - - 0"44 = O,

(24)

which has the solution

0" = f ( X l

At- X4) -[- g(Xl - - X4),

where f and g are a r b i t r a r y functions.

(25)

Eq. (18) reduces to

Ogll __ O~44 .71- 1(~12 __ ~42 ...[- 3"42 __ 3"12) = O.

(I8a)

E q u a t i o n (I 7) then shows t h a t 3" cannot vanish everywhere.

W e m u s t now see w h e t h e r there exist u n d u l a t o r y processes
for which 3" does n o t vanish. W e note t h a t such an undulat o r y process is represented, in the first approximation, b y an
u n d u l a t o r y fl, t h a t is b y a 5-function which, so far as its
dependence on x, and also its dependence on x4 is concerned,
possesses m a x i m a and minima; we m u s t expect this also for
a rigorous solution. W e know a b o u t 3" t h a t e~ = ~ satisfies
the w a v e equation (24) and therefore takes the form (25).
F r o m this, however, the u n d u l a t o r y nature of this q u a n t i t y

52

A . EINSTEIN AND N . ROSEN.

[J. F. I.

does n o t necessarily follow. W e shall in fact show t h a t q,

can have no minima.
Such a m i n i m u m would i m p l y t h a t the functions f and g
in (25) have minima. A t a point (Xl, x4) where this were the
case we should have ~'! = ~'4 = o, ~'n =>o, ~'44 ~ o .
But
b y (I7) and (2o) this is impossible. Therefore ~, has no
minima, t h a t is it is not u n d u l a t o r y b u t behaves, at least in a
region of space arbitrarily extended in one direction, monotonically. W e shall now consider such a region of space.
It is useful to see w h a t sort of t r a n s f o r m a t i o n s of xl and x4
leave o u r s y s t e m of e q u a t i o n s (I4) invariant.
F o r this
invariance it is necessary and sufficient t h a t the transformation satisfy the e q u a t i o n s
021

09;1
02,

OX4 '
_

Ox4

T h u s we m a y arbitrarily
equation

024 ]

(26)

024
Oxl '

choose 21(xl, x4) to satisfy the

0221

0221

Oxl 2

Ox4~-

-- o

(26a)

and then (26) will d e t e r m i n e the corresponding 24. Since ev is

i n v a r i a n t u n d e r this t r a n s f o r m a t i o n and also satisfies the w a v e
equation, there exists a t r a n s f o r m a t i o n where 21 is respectively
equal or proportional to e~. In the n e w co6rdinate s y s t e m we
have
e~ = a x l

or

~, = log a -}- log

(27)

Xl-

If we insert this expression for ~, in (I7)-(27) the equations

reduce to the e q u i v a l e n t s y s t e m
~11 -

~44 + i
Xl

1 = o,

(2s)
I

and

-3
al = x1(tl2 + /342) B -2xl

(29)

~4 = x1~1~4.

(30)

Jan., I937.1

O N GRAVITATIONAL WAVES.

53

E q u a t i o n (28) is the equation for cylindrical w a v e s in a threedimensional space, if xl denotes the distance from the axis
of rotation. T h e equations (29) and (3o) determine, for
given/3, the function a up to an (arbitrary) additive constant,
while, b y (27), "r is already determined.
In order t h a t the w a v e s m a y be regarded as w a v e s in a
euclidean space these e q u a t i o n s m u s t be satisfied b y the
euclidean space when the field is i n d e p e n d e n t of x4. This
field is represented b y
A = I;

B =

I;

C = xl 2,

if we denote the angle a b o u t the axis of rotation b y x3.

relations correspond to
a = o,

3 = -- logxl,

These

~, = log xl,

and from this we see t h a t the equations (27)-(3o) are in fact

satisfied.
W e have still to investigate w h e t h e r stationary w a v e s
exist, t h a t is w a v e s which are purely periodic in the time.
F o r 3 it is at once clear t h a t such solutions exist. Although it is n o t essential, we shall now consider the case where
the variation of 3 with time is sinusoidal. Here 3 has the
form
3 = X0 + XI sin ~x4 + X2 cos ~x4,
where X0, X1, X2 are functions of xl alone. F r o m (30) it then
follows t h a t a is periodic if and only if the integral
f/31/~4dx4

taken over a whole n u m b e r of periods vanishes.

In the case of a s t a t i o n a r y oscillation, which is represented
by
= X 0 + X I sin ~x4,
this condition is actually fulfilled since
f31~4dx4

= f(Xo

t Av X I '

sin ox4)coX 1 c o s wx4dx 4 ~-- o.

On the other hand, in the general case, which includes the case
of progressive waves, we obtain for this integral the value
3l r\ X - ~1-~-2p - - X 2 X I ' ) o ~ T ,

54

A.

EINSTEIN

AND N .

ROSEN.

[J. F. I.

where T is the interval of time over which the integral i s

taken. This does n o t vanish, in general. A t distances xl
from xi = o great compared with the wave-lengths, a progressive wave can be represented with good a p p r o x i m a t i o n in
a d o m a i n c o n t a i n i n g m a n y waves by
= X0 -t- a sin ~0(x4 - -

Xl),

where a is a c o n s t a n t (which, to be sure, is a s u b s t i t u t e for a

function depending w e a k l y on xl). In this case XI = a cos ~0xi,
X2 = -- a sin ~OXl,so t h a t the integral can be (approximately)
represented b y -- a~o2T, and t h u s c a n n o t vanish and always
has the same sign. Progressive waves therefore produce a
secular change in the metric.
This is related to the fact t h a t the waves t r a n s p o r t energy,
which is b o u n d up with a s y s t e m a t i c change in time of a
g r a v i t a t i n g mass localized in the axis x = o.
Note.--The second part of this paper was considerably altered by me after
the departure of Mr. Rosen for Russia since we had originally interpreted our
formula results erroneously. I wish to t h a n k m y colleague Professor Robertson
for his friendly assistance in the clarification of the original error. I t h a n k also
Mr. Hoffmann for kind assistance in translation.
A. EINSTEIN.