425

Foundations of the New Field .

By M. B orn and L. lNFELD,t Cambridge.

(Communicated by R. H. Fowler, F.R .S.—Received Jan u ary 26, 1934.)

§ 1. Introduction.
The relation of m a tte r and th e electrom agnetic field can be in te rp re te d
from tw o opposite standpoints :—
The first which m ay be called th e Unitarian standpoint^. assum es only one
physical entity, th e electrom agnetic field. The particles of m a tte r are con-
iidered as singularities of th e field and m ass is a derived notion to be expressed
by field energy (electrom agnetic mass).
The second or dualistic standpoint takes field a n d particle as tw o essentially
lifferent agencies. The particles are th e sources o f th e field, are acted on b y
the field b u t are n o t a p a rt of th e field ; th e ir characteristic p ro p erty is in e rtia
measured by a specific constant, th e mass.
A t the present tim e nearly all physicists have adopted th e dualistic view ,
which is supported by three facts.
1. The failure of any attempt to develop a Unitarian theory.— Such a tte m p ts
aave been m ade w ith tw o essentially different tendencies: (a) The th e o rie s
started by H eaviside, Searle and J . J . Thomson, a n d com pleted b y A braham ,
Lorentz, and others, m ake geometrical assum ptions a b o u t th e “ shape ” a n d
Sdnematic behaviour of th e electron and distribution of charge density (rigid
electron of A braham , contracting electron of L o re n tz ); th e y break dow n
because they are compelled to introduce cohesive forces of non-electrom agnetie
origin ; (b) th e theory of Mie§ form ally avoids this difficulty b y a generalization
of Maxwell’s equations m aking them n o n -lin ear; th is a tte m p t breaks dow n

t Research Fellow of the Rockefeller Foundation. I should like to thank the Rocke­
feller Foundation for giving me the opportunity to work in Cambridge.
t This expression has nothing to do with “ unitary ” field theory due to Einstein, W eyl,
Eddington, and others where the problem consists of uniting the theories of gravitational
and eleetro-magnetio fields into a kind of non-Riemannian geometry. Specially some of
Eddington’s formulae, developed in § 101 of his book “ The M athematical Theory of
Relativity ” (Cambridge), have a remarkable formal analogy to those of this paper, in
3pite of the entirely different physical interpretation.
§ ‘ Ann. Physik,’ vol. 37, p. 611 (1912); vol. 39, p. 1 (1912); vol. 40, p. 1 (1913). A lso
Born, ‘ Gottinger Nachr,’ p. 23 (1914).2

VOL. CXLIV.— A. 2 F
 426 M. Born and L. Infeld.
because M ie’s field e q u atio n s h a v e th e un accep tab le p ro p erty , th a t their
so lutions d ep en d on th e a b s o lu te valu e o f th e po ten tials.
2. The result o f relativity ,th
eory t h a t th e
o n v elocity is in no w a y c h a ra c te ristic o f electrom agnetic m ass, b u t can be
d eriv ed fro m th e tra n sfo rm a tio n law .
3. L a st, b u t n o t least, the great success o f quantum, mechanics which in its
p re se n t fo rm is essentially b a sed on th e d u a listic view. I t s ta rte d from the
co n sid eratio n o f oscillators a n d p a rticle s m oving in a Coulom b fie ld ; the
m eth o d s developed in th e se cases h a v e th e n b een a p p lie d even to th e electro­
m ag n etic field, th e F o u rie r coefficients o f w hich beh av e like harmonic
oscillators.
B u t th e re a re in d ic a tio n s t h a t th is q u a n tu m electrodynam ics m eets con­
sid erab le difficulties a n d is q u ite insufficient to ex p lain several facts.
T he difficulties a re chiefly c o n n ected w ith th e fa c t t h a t th e self-energy of a
p o in t charge is infinite, f T he fa c ts u n e x p la in ed concern th e existence of
e le m e n ta ry p a rticle s, th e c o n stru c tio n o f th e nuclei, th e conversion of these
p a rtic le s in to o th e r p a rtic le s o r in to p h o to n s, etc.
I n all th e se cases th e re is sufficient evidence t h a t th e p rese n t th e o ry (formu­
la te d b y D ira c ’s w ave e q u a tio n ) h o ld s as long a s th e w ave-lengths (of the
M axw ell or o f th e de B roglie w aves) a re long co m p ared w ith th e “ radius of
th e electro n ” e2/mc2, b u t b rea k s do w n fo r a field co n tain in g sh o rter waves.
T h e n o n -ap p e a ran c e o f P la n c k ’s c o n s ta n t in th is expression fo r th e radius
in d ic a te s t h a t in th e first p lace th e electro m ag n etic law s are to be modified;
th e q u a n tu m law s m a y th e n b e a d a p te d to th e new field equations.
C onsiderations o f th is s o rt to g e th e r w ith th e conviction o f th e g reat philo­
so p h ical su p e rio rity o f th e U n itarian id ea h a v e led to th e rec e n t attem pt^ to
c o n stru c t a new electrodynam ics, b a sed on tw o r a th e r d ifferent lines of th o u g h t:
a new th e o ry o f th e electro m ag n etic field a n d a new m eth o d of quantum
m ech an ical tre a tm e n t.
I t seem s desirable to keep th ese tw o lines se p a ra te in th e fu rth e r development.
T he p urpose o f th is p a p e r is to give a deeper fo u n d atio n o f th e new field
e q u a tio n s on classical lines, w ith o u t to u ch in g th e question o f th e quantum
th e o ry .

■f The attem pt to avoid this difficulty by a new definition o f electric force acting on a
particle in a given field, made b y W entzel (‘ Z. Physik,’ vol. 86, pp. 479, 636 (1933),vol.
87, p. 726 (1934)), is very ingenious, but rather artificial and leads to new difficulties.
% B om , * Nature,’ vol. 132, p. 282 (1933); ‘ Proc. Roy. Soo.,’ A, vol. 143, p. 410 (1934),
■cited here as I.
 Foundations of New Field Theory. 427
In th e papers cited above, th e new field th eo ry has been introduced ra th e r
cgmatically, b y assum ing t h a t th e L agrangian underlying M axwell’s th eo ry

L = £ (H 2 — E 2) (1.1)

(I and E are space-vectors of th e electric a n d m agnetic field) has to be replaced

lr the expression!
L = ( a / 1 + i ( H 2- E * ) - l ) . (1.2)

The obvious physical idea of th is m odification is th e following :—

The failure in th e present th e o ry m ay be expressed b y th e sta te m e n t th a t
violates th e principle of finiteness which postulates th a t a satisfactory th eo ry
lould avoid letting physical quantities become infinite. A pplying th is
inciple to th e velocity one is led to th e assum ption of a n upper lim it of
alocity c and to replace th e N ew tonian action function \ mv2 of a free particle
y th e relativity expression me2 (1 — V l — A pplying th e sam e condition
a th e space itself one is lead to th e idea of closed space as introduced b y
linstein’s cosmological theory. %A pplying i t to th e electrom agnetic fie
i lead im m ediately to th e assum ption of a n upper lim it of th e field stren g th
and to th e modification of th e action function (1.1) into (1.2).
This argum ent seems to be quite convincing. B u t we believe th a t a deeper
oundation of such an im p o rta n t law is necessary, ju s t as in E in ste in ’s m echanics
he deeper foundation is provided b y th e p o stulate of relativ ity . Assum ing
hat th e expression me2(1 —\ / 1 — v2jc2) has been found b y th e idea of a veloc
imit it is seen th a t it can be w ritten in th e form

me2 (1 —
vhere
c2 d r2— c2 dt2— dx2 — dy2 — dz2,

tnd therefore it has th e p ro p erty th a t th e tim e integral of is in­

variant for all transform ations for which dx* is invariant. This four-dim en-
lional group of transform ations is larger th a n th e three-dim ensional group of
transformations for which th e tim e integral of th e N ew tonian function

\mo2 — (ds/dt)2 ; ds2 — + dy2 -J- dz2,

is invariant.

t See Bom and Infeld, ‘ Nature,* vol. 132, p. 1004 (1933).

t See Eddington, “ The Expanding Universe,” Cambridge, 1933.
2 f 2
428 M. Born and L. Infeld.
So we believe t h a t we ought to search for a group of transform ations for
which th e new L agrangian expression has a n in v a ria n t space-time integral
a n d which is larger th a n t h a t for th e old expression (1.1). This la tte r group is
th e know n group of special re la tiv ity b u t n o t th e group of general space-time
transform ations, f Now it is v ery satisfying th a t th e new Lagrangian belongs
to th is group of general re la tiv ity ; we shall show th a t it can be derived from
th e p o stulate of general invariance w ith a few obvious additional assumptions.
Therefore th e new field th e o ry seems to be a consequence of th is very general
principle, a n d th e old one n o t m ore th a n a useful p ractical approximation,
ju s t in th e sam e w ay as for th e m echanics o f N ew ton and E instein.
I n th is p ap er we develop th e whole th e o ry from th is general standpoint.
W e shall be obliged to re p e a t som e of th e formulae published in the previous
paper. The connection w ith th e problem s of grav itatio n and of quantum
th eo ry will be tre a te d later.

§ 2. Postulate o f Invariant Action.

W e s ta r t from th e general principle t h a t all laws of n a tu re have to be expressed
b y equations co v arian t for all space-tim e transform ations. This, however,
should n o t be ta k e n to m ean t h a t th e g rav itatio n a l forces play an essential
p a rt in th e c o n stitu tio n of th e physical w orld ; therefore we neglect the
g rav itatio n al field so t h a t th e re exist co-ordinate system s in which th e metrical
tensor gkl has th e value assum ed in special re la tiv ity even in th e centre of an
electron. B u t we p o stu la te t h a t th e n a tu ra l law s are independent of the
choice of th e space-tim e co-ordinate system .
W e denote space-tim e co-ordinates b y

x 1, x 2, z?, cc4 =■■x, y, , ct.

The differential dxk is, as usual, considered to be a con trav arian t vector.
One can pull th e indices up a n d dow n w ith help of th e m etrical tensor which
in a n y cartesian co-ordinate system (as used in special relativity) has the form.

f The adaptation o f the function L (1.1) to the general relativity by multiplication wi

• / — g is quite formal. Any expression can be made generally invariant in this way.
 Foundations of New Field Theory. 429
t is n o t th e u n it m atrix, because of th e different signs in th e diagonal. There-
i)re we have to distinguish betw een cov arian t a n d c o n tra v a ria n t tensors even
a th e co-ordinate system s of special relativ ity . I n th is case, however, th e
ule of pulling up and down of indices is very simple. This operation on th e
ndex 4 does n o t change th e value of th e ten so r com ponent, th a t on one of th e
ndices 1, 2, 3 changes only th e sign.
We use th e well-known convention th a t one has to sum over a n y index
vhich appears twice.
To obtain th e laws of n a tu re we use a v ariational principle of least action
)f th e form
S f dx —0,

We 'postulate: the action integral has to he an invariant. W e have to find

she form of satisfying th is condition.
W e consider a covariant tensor field a k l; we do n o t assum e a n y sym m etry
property of akl. The question is to define $£ to be such a function of akX
th a t (2.2) is invariant. The well-known answ er is th a t S£ m u st have th e
formf ___
& — a/|«*. i | > (| a** | = d eterm inant of (2.3)

I f th e field is determ ined by several tensors of th e second order, can be

any homogeneous function of th e determ inants o f th e co v ariant tensors of
the order
Each a rb itrary tensor akl can be split up into a sym m etrical and anti-
symm etrical p a r t :

a kl — 9kl ~*rf hi> 9kl 9Ik » fkl = — flk- (2.4)

The simplest sim ultaneous description of th e m etrical and electrom agnetic

field is the introduction of one a rb itra ry (unsym m etrical) tensor a k l; we
identify its sym m etrical p a rt gkXw ith th e m etrical field, its an
p a rt with the electromagnetic field. J

t See Eddington, “ The Mathematical Theory of Relativity,” Cambridge, § 48 and 101

(1923).
The proof is simple : by a transformation with the Jacobian I = d (xx ... x*) dx is changed
d (x1 ... rc4)
into dx = I dx and \aki\ into \aki\ = \aki\ I-2 ; for the dxk are contravariant, aki covariant,
t assumption has already been considered by Einstein, 6Berl. Ber.,’ pp. 75/37
1923) and p. 414 (1925), from the standpoint of the affine field theory.
430 M. Bom and L. Infeld.
W e have th e n th re e expressions w hich m ultiplied by d r are invariant

V— \akl\= V—| gki+ / m| ; V—

where th e m inus sign is added in order to get real values of th e square roots;
for (2.1) shows t h a t = — 1, therefore alw ays \gki\ < 0.
The sim plest assum ption for ££ is a n y linear function of (2.5):

^ —V —J gia+ /* ;i| + A a/ — | +
B u t th e la s t te rm can be om itted. F o r if is th e ro ta tio n of a potential vector,
as we shall assum e, its space-tim e in te g ral can be changed in to a surface
in tegral a n d h as no influence on th e v ariatio n al equation of th e field, f There­
fore we can ta k e
B = 0. (2.7)

W e need a n o th er condition for th e d eterm ination o f A. I ts choice is obvious.

I n th e lim iting case o f th e cartesian co-ordinate system and of small values
o f f kl, ££ has to give th e classical expression

L = i/« /“. (2-8)

W e now leave th e general co-ordinate system w hich has guided us to th e expres­
sion (2.6) fo r §£ a n d calculate in cartesian co-ordinates. Then we have
w ith gkl = 8kl (see (2.1) )

—1 /1 2 /l3 fu
= 1 + (/2a2 “h / a i 2 + /1 2 2 fu
/2 1 — 1 f z 3fzA
— I Skl + / h| = — /242 / 34s)
/3 1 fz2 — 1 /34
— (/23/14 + /3 1 /2 4 +/12/34)2
/4 1 fi 2 fi3 1
= 1 + ( / . 32 + /3 1 2 + / l 22 ~ / l 4 2 42 ~ / 3 4 2) - IA ll•
F o r sm all values of f ktth e la st d e te rm in a n t can be neglected
equal to (2.8) only if
A = — 1. (2-9)
W e have therefore th e r e s u l t :—
The action function of th e electrom agnetic field is in general co-ordinates
(2. 10)
se = V
a n d in cartesian co-ordinates
L = V l + F — G2 — 1, ( 2. 11)

t See Eddington, loc. tit., § 101.

 Foundations of New Field . 431
where
F — f z 32+ / 312 _t~ /l 22 /l4 2 / 24s / 34s
( 2 . 12)

G = / 2s /l 4 + /31/24 4 *f iz f u ' (2.13)

L et us go back to th e expression fo r in a general co-ordinate system .

W e denote as usual
l&tl = 9 ,
and we develop th e d e te rm in a n t \gkX + / * i | in to a pow er series in f kl. We
have th en
| 9kl+ /f c l| — 9+ ®
The transform ation properties of | gki + f ki |, | a n d therefore also
of <J) (gkl, f kl) are th e same. T hey tra n sfo rm in th e sam e w ay as g. I f we
w rite
j + ® + l / « l = ? ( i + | + (2.14)

we see, th a t all expressions in th e b rac k e t on th e rig h t side of (2.14) are in ­

varian t. W e have calculated th e ir value in a geodetic co-ordinate system a n d
have f o u n d :
I = ! / » ,/ “ = F = 1 y / j " .
<E> igsi an invariant. W e have therefore in a n a rb itra ry co-ordinate system :
I0 m + / h I =** 91( + F — G2)
(2.15)
Se = \ f — ~g (VI + F — G2 — 1 )
F == 'G2 = I f klI = ( / 2 3 /1 4 H~ / 3 1 /2 4 4 ~ / 12 / 3 4 )2
~ —9 —9
B oth F and G are in variant. W e shall bring G in to such a form, t h a t its
invariance will be evident. F o r th is purpose let us define an antisym m etrical
tensor j tklm for an y p air of indices, t h a t isf
1
if skim is an even perm u tatio n of 1, 2, 3, 4
2V —0
:sklm
3 — 1 if skim is an odd perm u tatio n of 1, 2, 3, 4 ►. (2.17)
2V — g
0 in any other case
We can write now G in th e following form :

6 = i (2.18)
t Einstein and Mayer, ‘ Berl. Ber.,’ p. 3 (1932).
 432 M. Born and L. Infeld.
JTrom tlie last equation we can deduce th e tensor character of j sklm, We can
also w rite G in th e form
Cl = if2.19)
(
w h e r e /* sfc is th e dual tensor defined b y
f%sk ~8klm f
J J Jim? ( 2. 20)
t h a t is
/*2S — f f ^*31 _ f
J — / ----- /14? J — /----- J 24? / * 12 =
V —g V —g V —g
, (2.21)
-F*14 — F
*24 1 y*34 _ 1
J / J 23’ J
V —g V —g V —g'
o r also

*
f*2* = - V - g f “ , / * 3i = — V — g f 24.

II
11
rH
<N
( 2. 22)
/ » , . = ------V - g f ™ , f* * = - V - g Jf* 34 —
—_
■h
because
f*ak ~jsklmflm J

W e shall need la te r th e following formulae :

f * klf * u = ~ f klfk i (2.24)

j Uabfksfab.= f * lsfks = < W
/ * * fe = — f u. (2.26)

(2.24)-(2.26) follow from th e definition o f f * kl, f * kl an d G given above.

T he function represented b y (2.15) is th e sim plest L agrangian satisfying
th e principle of general invariance. B u t i t differs from th a t considered in I
b y th e te rm G2. This is of th e fo u rth order in th e f kl a n d can, therefore, be
neglected except in th e im m ediate neighbourhood of singularities (i.e., electrons,
see § 6). B u t th e L agrangian used in I can also be expressed in a general
c o v a rian t form ; for G2 is a d eterm in an t, nam ely, | f H | , therefore

j (V — I 9ki + fk i | - f I

is also in v a r ia n t; in cartesian co-ordinates it has exactly th e form

J (V 'l + F — 1) dr. (2-28>

W hich of these action principles is th e rig h t one can only be decided by their
consequences. W e ta k e th e expression given b y (2.15) and can then easily
 Foundations of New Field Theory. 433
return to th e other (2.27) or (2.28) b y p u ttin g G = 0. I n a n y case th e solution
of th e statical problem is identical for b o th action functions because one has
G = 0 in th is special case.

§ 3. Action Principle, Field Equation and Conservation Law.

We w rite (2.15) in th e general form

& = ^ /~ 9 L = ^L (9n, F , G).

W e shall see th a t all considerations hold if L is an in v arian t function of these

argum ents. As usual we assum e th e existence of a p o tential vector so
th a t
f d<f>k /o jv
fkl ~ da?dad'
Then we have th e id en tity
| Qfmk l — A (3.2)
dxk dxl dxm ’

which can w ith th e help of (2.20) be w ritten :

^ = ^ ‘ = 0. (3.2 a )

W e introduce a second kind of antisym m etrical field tensor p ki, which has to
f kt a relation sim ilar to th a t which, in Maxwell’s theory of m acrospic bodies,
th e dielectric displacem ent and m agnetic induction have to th e field strengths :

V— a v kl— — a/ _ (o f*ki\ (.fkl- G f * klW - 9

(3.3)
V 9P dfkl V gVdEJ + dGJ ) Vl + F — G8
The variation principle (2.2) gives th e E ulerian equations

d -y/—
(3.4)
dxl

The equation (3.2) (or (3.2a ) ) and (3.4) are th e complete set of field equations.
We prove the validity of th e conservation law as in Maxwell’s theory.
Assuming a geodetic co-ordinate system , we m ultiply (3.2) by p lm :
434 M. Born and L. Infeld.
I n th e second a n d th ird te rm we can ta k e p 1™ u n d er th e differentiation symbol
because o f (3.4); in th e first te rm we use th e definition (3.3) of p 1™:

2 + £ |? = < > > 1

or

I f we introd uce th e ten so r

(3.6)
w here

(3.7)
l o if

\
we have 1 &

o
ii (3.8)

I n a n a rb itra ry co-ordinate sy stem we h a v e

^ I l _ i V zr-g
T ^ = 0 , (3.9)

or, w ith th e udual n o ta tio n o f c o v a ria n t differentiation

T'*,, = 0. (3.9a)

I t follows from (3.3) a n d (2.25) t h a t we can w rite also T*.1in th e form

T kl = L B kl — ~ . (3.6a)
V i + F — G2

§ 4. Lagrangian Hamiltonian.
££ can be considered as a function o f gkl a n d f ki. W e shall show that
__2 dS£
.. . *8 th e energy-im pulse tensor. W e find

dV: (4.1)
i V — 9 9ki

0F (4.2)
C*~kl
dg & JksJlr

0G (4.3)
a / ~l — $®9kl-
 Foundations of New Field . 435

Therefore
/0 l 0 F , a t- 0 G \ \
— 2 V — g { \ - g kl —
d fl - 2 \0 F 0 / 1 ^ 0G 0 / 7 J

V ^g — (4 .4 )
V l + F - G2 / *

I t follows from (3.6 a ) a n d (3.6)

3 C£ ___ ___
- 2 -^ET = V - ? T » = V - 5 ( L f e <4 6 >

Now i t is v ery easy to generalize o u r actio n principle in such a w ay t h a t i t

contains E in ste in ’s g ra v ita tio n laws ; one h as only to a d d to th e a c tio n in te g ral

tiie te rm I R V — g d r , w here R is th e scalar o f c u rv atu re. B u t we do n o t

discuss problem s connected w ith g ra v ita tio n in th is paper.

S£ was regarded as a function o f gkl a n d W e can, how ever, express
also as a function o f gkl a n d p kl. I t can be show n t h a t it is possible to solve
th e equations
* » _ /** ~ Q f *** (3 .3 )
V l -b F — G2

w ith respect to f kl.For th is purpose we have to calculate

±P**lP*H = P , (4.6)
I p MP \ i = Q, (4.7)

i.e., P and Q corresponding to F a n d G. U sing th e formulae (3.3) a n d (2.24)-

(2.26), we obtain
p — F + G2F + 4G2
(4.8)
1 4- F — G2
Q = G. (4.9)

The last equations can be w ritten in a m ore sym m etrical form :

1 + F — G2 _ l + Q2
(4.8a )
1 + G2 1 + P — Q2
G = Q. (4 .9 a )

W e are now able to solve th e equations (3.3). I t follows from (3.3) a n d (2.26)
th a t
r **»- / * ” + (¥** (3.3 a )
V l + F — G2’
436 M. Born and L. Infeld.
Solving (3.3) a n d (3.3 a ) we o btain (taking in to account (4.8 a ) and (4.9a ) )

fki _ Pkl + Qp*kl . f*kt __ — Qpfct

(4.10)
J V i + P - Q 2’ V i + P - Q 2'

The tensors f kl and p kt can now be tre a te d com pletely symmetrically.

In ste a d o f th e L a g ran g ia n L we can use in th e principle of action th e Hamil­
to n ia n function H :
H = L — lspklfk(4.11)

where H has to be regarded as a function of gkl a n d F ro m (4.8), (4.9),

a n d (4.10) it follows for H as a function of gkl an d :

= H V — g= V —

a n d th is can be expressed in th e form

& = V — | gki + p * H | — V —| |. (4.12a)

The function H leads us to ex actly th e sam e equations of th e field as th e function

L. W e see t h a t th e equations

v* = _ 8 ^ * fc (^ * fc = anti-potential-vector) (4.13)
F kl dxk dxl
8JT (4.14)
\ / — g f * kl es
dp*u
d 's / — (4.15)
0,
dxl

are entirely equivalent to the equations (3.4), (4.10), (3.2 a ).

The energy-im pulse ten so r (3.6a ) can also be expressed with help of H
instead of L. One has

ay = H6Y = <l - ip<*fab) a.1 (*•»«)

T he id e n tity o f th is expression w ith (3.6) is e v id e n t,t if we appeal to th e following
form ula w hich can be deduced from (2.21), (2.22)

= p m,f m» - i p Mf<« v
t In I it has been stated that the two expressions for T kl, obtained with help of L an
H, are different; this has turned out to be a mistake.
 Foundations of New Field 437
Generally, from each equation containing f kl, one o b tain s
another correct equation changing these q u a n titie s correspondingly in to
4***, p\ * f*M-
§ 5. Field Equations in Space-vector Form.
W e now introduce th e conventional u n its in ste a d of th e n a tu ra l u nits. W e
denote by B , E a n d D, E , th e space-vectors w hich characterize th e electro­
m agnetic field in th e conventional u nits. W e h a v e in a c a rte s ia n co-ordinate
system :
(a?1, a;2, sc8, cc4) -> ( , y, z, ct) (5.1)
(^l> <
f>2> ^8> ^ 4 ) (^> (^*2)
( / 28 . / 31 . A 2 -B
(5.3)
/l4> fz&> 4) ®
(*>23> Psi’ H
(5.4)
Pu> P m >P m ) ?+ D
The quotient of th e field stren g th expressed in th e conventional u n its divided
b y th e field stren g th in th e n a tu ra l u n its m ay be denoted b y This c o n sta n t
of a dimension o f a field stren g th m ay be called th e absolute field ; la te r we sh all
determ ine th e value o f b, w hich tu rn s o u t to be v ery great, i.e., of th e order o f
m agnitude 1016 e.s.u.
W e have
L = V l + F — G2 — 1, ( 2 . 11 )

F = ~ (B2 — E 2) ; G = 1(B .E ) (2.12 a ); (2.13 a )

w __ 7,2 _ B — GE
0B V l + F — G2’
► (3.3 a )
~ ,2 0L E — GB
0E V l + F — G2

B = ro t A ; E = — — grad <b (3.1a )

c
ro t E -f- - ^ — 0 ; div B = 0 (3.2b )
c ot

ro t H — -^ = 0 ; div D = 0. (3.4a )
c ot
Our field equations (3.2 b ) and (3.4a ) are formally identical with Maxwell’s
equations for a substance which has a dielectric Constance and a susceptibility,
438 M. Born and L. Infeld.
being c e rtain functions of th e field strength, b u t w ithout a spatial distribution
of charge an d current.
F o r th e energy-im pulse tensor we find :
X, Xv X, cGx
Y* Y, Y,
z* zv z, cG* (3.6a)

-s* -sv U
c * c
47CX, = H yB y + H 2B 2 - D^E* - 62L "
47rYa. = 47 z X v= —
(3.6b)
— S* = 4:7zcG x = B yB z-
c
47cU = DgMx -f- D vEj, + B zE 2 -f- 62L

One gets an o th er set of expressions for these q uantities b y changing L, B, E,

H , D in to H, H , D, B , E.
The conservation laws are :

dx dy dz c2 dt
(3.8a)
dSx. dSe dU
+ %■ + S z = - - ¥
The function H is given b y
H = V l -f- P — Q2 — 1 (4.12a)

P = i (D 2 - H 2 ) ; Q = p(D.H). (4.6 a ); (4.7a)

Solving (3.3 a ) we o b tain :

B 6,8 H H + QD
0H V l + P — Q2
►. (3.10a)
E==f e23H = = _ D + Q H =a
0D V l 4- p — Q2

§ 6. Static Solution o f the Field Equations.

W e consider (in th e c a rtesia n co-ordinate system ) th e electrostatic case
where B = H = 0 an d all other field com ponents are independent of t. Then
th e field equations reduce t o :
ro t E = 0 (®d 1
div D = 0. (6-2)
 Foundations of New Field Theory. 439
fe solve this equation for th e case of central sym m etry. T hen (6.2) is
imply
f (r3D r) = 0, (6.3)
dr
nd (6.3) has th e solution
D r = e/r2. (6.4)

n this case th e field D is exactly th e sam e as in M axwell’s th eo ry : th e sources

f D are point charges given b y th e surface integral

4ne = 1 D r da. (6-5)

H e equation (6.1) gives
Er = — ~j ~ — — 4>' W
dr
md from (3.3a )
D„ .----------- i ------- ---------------- / (r) . (6.7)

The com bination of (6.4) an d (6.7) gives a differential equation for <f>(r) of th e
irs t order, w ith th e solution

* M - ; f W = f" - ; = a /«
ro Vo/ J*Vi + V
This is the elementary 'potential o f a point charge e, w hich has to replace
Coulomb’s l a w ; th e la tte r is an approxim ation for ^ 1, as is seen im ­
mediately, b u t th e new p o ten tial is finite everyw here.
W ith help of th e su b stitu tio n x — ta n one obtains

f (x ) — \ = / ( 0) (6-9)
P(*)Vl — £ sin2 p
where
(3 = 2 arc ta n x, (6 . 10)

1
and F (&,p)is the Jacobian elliptic integral of th e first kind for k — = sin J 7c
V2
(tabulated in m any books)f

( 6 . 11 )

t ^.0., Jahnfce-Emde, “ Tables of functions ” (Teubner 1933), p. 127.

440 M . B o m a n d L . In fe ld .

F o r x — 0 one has / •< \

(6.12)
^ (0) = F v i ’ ) = 1’8541'

The p o ten tial has its m axim um in th e centre a n d its value is

<f>(0) = 1,8541 ejr0. (6.13)

1-0 _

F ig. 1.

The function / (x )is p lo tte d in fig. 1. I t has very sim ilar properties to the
function arc cot x. F o r exam ple, one has
P (1/a?) = 2 arc ta n 1/a; = 2 — arc ta n x) = n — P (x ) >

on th e other h and
 Foundations of New Field Theory. 441
Therefore one has
/(*)+/(!/*)=/(<>)• (6-14)
It is sufficient, therefore, to calculate / (xrfom 0 to
k> (3 = J tc.
One sees, th a t th e D field is infinite for r = 0 ; E an d <
inite. One has
e/r2, (6.4)
E„ (6.15)
r0*V 1 + ( r /r 0)4
The com ponents E e, E„, E z are finite a t th e centre, b u t h ave th ere a d is­
continuity.
§ 7. Sources o f the Field.
In th e older theories, w hich we have called dualistic, because th e y considered
n a tte r a n d field as essentially different, th e ideal w ould be to assum e th e
particles to be p o in t charges ; th is w as im possible because o f th e infinite self-
energy. Therefore i t w as necessary to assum e th e electron h av in g a finite
iia m e te r a n d to m ake a rb itra ry assum ptions a b o u t its in n er stru c tu re , w hich
lead to th e difficulties po in ted o u t in th e in tro d u ctio n . I n our th e o ry th ese
difficulties do n o t appear. W e have seen t h a t th e field (or D-field) h as
a singularity which corresponds to a p o in t charge as th e source of th e field.
D and E are identical only a t large distances (r r 0) from th e p o in t charge,
bu t differ in its neighbourhood, a n d one can call th e ir q u o tien t (which is function
of E) “ dielectric co n stan t ” of th e space. B u t we shall now show t h a t
another in te rp re ta tio n is also possible w hich corresponds to th e old idea of a
spatial distribution of charge in th e electron. I t consists in ta k in g div E
(instead of div D = 0) as definition of charge d en sity p, w hich we propose to
call “ free charge d ensity.”
L et us now w rite our set of field equations in th e following form :
0 a/ — g p
dxl 0, (3.4)
dflcl l dflm
'Jfl _i_ v lm |I ____
0,
sV-gf (3.2)
dxm dxk dxl dxl
Pkl is a given function of/** a n d if we p u t in (3.4) for p kl th e expression (3.3),
in which L is no t specified, we obtain :

— ((2 — 4- — f* kl) V — a l = 0. (7.1)

dx1^ + dG J' V
We can now write th e equation (7.1) in th e form

dV-fff* 4tc p* V - (7.2)

dxl
VOL. CXL1V.--- A. 2 G
442 M. Born and L. Tnfeld.
where

~ 47Cp* = 2 SL/0B ( o t ) + ^ * fcZ 0»l ( s s ) } * (7-3)

The equations (7.2) and (3.2) are form ally identical w ith th e equations of
th e Lorentz theory. B u t th e im p o rta n t difference consists in this, that p*
is not a given function o f th e space-tim e co-ordinates, b u t is a function of the
unknow n field strength. I f we have a solution of our set of equations, we are
able to find th e density of th e “ free charge ” or th e “ free current ” with help
of (7.2) or (7.3).
W e see im m ediately th a t p7c satisfies th e conservation law :

This follows from (7.2), th a t is from th e antisym m etrical character o f / and can
also be checked from (7.3).
I n L orentz’s th eo ry th ere exists th e energy-im pulse tensor of the electro­
m agnetic field, defined b y
4 w S** = i 8kl F — f lsf ks> (7-5)
b u t its divergence does n o t vanish, where th e density of charge is not zero.
Therefore to preserve th e conservation principle in th e L orentz’s theory it was
necessary to introduce an energy-im pulse tensor of m atter, MK, th e meaning of
which is obscure. The tensor M / h ad to fulfil th e condition th a t the diver­
gence of Sfcl -f- M / vanishes. This difficulty does n o t appear in our theory.
W e do n o t need to introduce th e m a tte r tensor Mfcl because th e conservation
laws are alw ays satisfied b y our energy-im pulse tensor T / .
W e shall, however, show th a t it is possible b y introducing th e free charges to
bring our conservation law
T \ , fe = 0

into th e form used in th e L orentz theory, nam ely,

The calculations are sim ilar to those used in § 3. The sim plest way is to
choose a geodetic co-ordinate system . We have th en :
df^ (7.2a)
= 4tc p*
dxl
dfki , dfi™ _i_ dfmk = 0. (3.2)
dxm dxk ”t" dxk
 Foundations of New Field Theory. 443

M tiplying (3.2) b y / * 1 we find :

i 3 / ffki\ o ^ ( f ) — 2 (7.7)
* a ? (/h / }“ 2 a**( / /sfc) ~ zu
sd therefore
i |a ( /« /“) - ^ ( /'% * ) = / « p *. <7 -8 )

ud taking account of (7.5):

H = /.* ? * • p -»>

me can derive the same equation directly from the conservation formula
i.8) writing it, in a geodetic co-ordinate system, in the form

(3.8)

ad introducing the expression (3.6) for Ts*. The two m ethods are equivalent.
L et us now specialize our equations for th e case in which L has the form given
i (2.15). We obtain then for p* in (7.3)

__ (
| f ki -JL __________
1 \ ^ vk*fM / _______ jl
- to p = v r + T r = T p dxl VV l + F — G *) J dxl VV I + F — G '
(7.10)
a the space-vector notation, where

(Pi> Pa> P3> P4) 4

\c
re have

' 4,1 c V l + F - G2 ( B x grad (•V l + F — Q*

— E x grad I — )\
V 1 H—F — G2/J
1
+ c V l + F —G2 \ \ V l + F — G'2
d ►. (7.10a )
B r-
& \V l + F — G2/J

4,tp “ ~ v T+f - g4 e •grad(y i + y - 4

B. gKMl ( V J 4 - F - 11

2a 2
444 M. Born and L. Infeld.
W e shall now a p p ly th e results here obtained to th e case of the statical
field. I n th is case E is alw ays finite a n d has a n o n -v a n ish in g divergence
w hich represents th e free charge. W e can, therefore, regard an electron
eith er as a p o in t charge, i.e., as a source of th e D (pki) field, or as a continuous
d istrib u tio n of th e space charge w hich is a source o f th e E ( f kl) field. It can
easily be show n t h a t th e whole charge is in b o th cases th e sam e (as is to be
expected). B o th
| d iv D dv an d 1 di

have th e sam e value, i.e., 4ne.F o r th e first in teg

§ 6. F o r th e second we have

E e
T-ir ~ — as o •.
r2

everyw here else E r is finite. The disco n tin u ity of E^, E v, E a a t the origin
is also finite a n d gives no c o n trib u tio n to th e integral. Therefore

I div E dv — 4tc I p dv =

L e t us now calculate th e d istrib u tio n o f th e free charge in th e statical case.

W e could calculate i t from th e eq u atio n (7.10 a ), b u t it is easier to do it from
th e equation
div E = \ E r ) = 4np,
r2dr
where
IT! — _iL _____ l -------- f6.151
r V V l + (r/r0)* *1
The result is

p = 2^ Z T T + M T * •
r0 X/

For r r0, p oc r 7, therefore dim inishing very rapidly as r increases. For

r r0, p oc 1 jr ,therefore p -> oo , b u t r 2p -> 0 for -^0. I t is eas
t h a t th e space in teg ral of p is equal to e. F o r one has, p u ttin g rjr0 V ta n <f>

—e cos <f>dy = e.
Jo
 Foundations of New Field Theory. 445
Our theory combines th e tw o possible aspects o f th e field ; tru e p o in t charges
ad free spatial densities are entirely equivalent. The question w hether th e
ae or the other picture of th e electron is rig h t has no m eaning. This confirms
lie idea which has proved so fruitful in q u a n tu m m echanics, th a t one
as to be careful in applying notions from th e m acroscopic w orld to th e world of
tom s : i t m ay happen th a t tw o notions contradictory in m acroscopic use are
iuite com patible in m icrophysics.

§ 8. Lorentz’s Equations o f Point Motion and Mass.

We consider once m ore th e problem of th e electron a t rest. W e in te n d to

alculate th e m ass and to determ ine th e absolute field co n stan t b in term s of
bservable quantities. I t is convenient to use th e space vector n o tation.
The im pulse-energy tensor is according to (3 .6 b )

E *
ItcX . = — D x E B — L = — -7
V I — 1/62E 2

-* * ( V 1- p " - 1,

i " X ,= - D» E » = -
. (3.6c)
s . = s„ = s» = o

btU = D . E + b2 L = b 2H =

, E2 1 \
V I — 1/62 E a l V l — 1/62 E 2 l ).

We calculate th e space integrals of these quantities. Obviously one has w ith

lv = dx dy d z :

M X * dv — 4tc( y dv = 47t [z z dv — — i f — = £ = = dv
* J J 3J V I — 1/62 E 2

— 62j ( \ / l - p E 2 - l (8.1)
J X v dv = J X 2 dv = I Zv dv = 0. (8.2)

Using (6.15) and (6.8) we find :

j x* dv = b2 (Ix — I a), (8.3)

446 M. Born and L. Infeld.
w here

Ii x 2dx
Vl +
dx (8.4)
12 ~ 4 lo V i T * ~ i/(0 >
The in te g ra l I x can be tra n sfo rm e d b y p a rtia l in te g ra tio n
r° / \ dx3
t “ l f. (] V l -f- ;W dx
dx =
V l ~h
xxdx
5 Jo (1 + affi*

T he first te rm vanishes ; th e second can be tra n sfo rm e d b y another partial

in te g ra tio n :

ii _ i j „ 21 s U + J 1
dx = i 1d xv\ Vr + > xt'
= 1 1 4 /(0 ).

The re su lt is th e so-called “ th e o re m o f L a u e ” t

JXa,dv — j*
d Z2 = 0.

I n th e s ta tic a l case a n d in a co -o rd in ate sy stem in w hich th e electron is a t rest

th e in te g rals o f all co m p o n en ts o f th e te n s o r T kl v a n ish e x cep t th e to ta l energy

E = Ju< fo = j ^ j H (8.5)

W e find fro m (3.6c), (6.15), a n d (8.4)

E = = 8 V (312 - I ,) = - 21, = § —/(0 ) = 1-2361 f . (8.6)

ro r0 To
W e h a v e o b tain e d a fin ite value o f the energy or the mass o f the electron with a
definite nu m erical facto r. T his re la tio n enables us to com plete our theory
concerning th e v alu e of th e ab so lu te field b in th e conventional runts.
(8.6) gives th e “ ra d iu s ” o f th e electron expressed in te rm s of its charge and
m ass :
r 0 = 1-2361 — = 2, 28 X 10 18 cm. (8.7)

and
( 8. 8)
9, 1 8 . 1015 e.s.u.

The enorm ous m ag n itu d e of th is field justifies th e application of th e Maxwefif

f Mie, ‘ Ann. P h ysik,’ vol. 40, p. 1 (1913).

 Foundations of New Field Theory. 447
quations in th eir classical form in all cases, except those w here th e in n er
tru c tu re of th e electron is concerned (field of th e order b, distance or w ave­
length of th e order r 0).
I t can be shown t h a t th e m otion of a n elem entary charge, on w hich an
external field is acting, satisfies an equation w hich is an obvious generalization
th e classical equation of L orentz. To find th is equation we shall use here a
3artesian co-ordinate system .
We assum e th a t th e stren g th of th e ex tern al field in a region surrounding
the electron is very sm all com pared w ith th e proper field of th e p o in t charge.
We denote th e proper field of th e electron b y

P h {0), (8-9)
and th e external field b y
V e) (8 .1 0 )

we do not take into consideration the sources of the external field. From the
assumption, that
p * r > / « • • ’ ; / h 10’ > / , ( 8 . 11)

inside th e sphere surrounding th e electron, it follows evidently tfia t th e real

solution of th e field equations cannot be very different from th a t obtained b y
adding th e unperturbed proper field a n d th e ex tern al field. W e co nstruct
therefore a sphere $ <0) w ith its centre a t th e singularity of H an d w ith a radius
r <0), which is so small, th a t inside th e sphere (8.11) is alw ays satisfied. B u t th e
radius r !0) of th e sphere has to be great com pared w ith th e radius of th e electron,
so th a t we can assum e th e valid ity o f M axwell’s equations on th e surface of
th e sphere ju s t as outside th e sphere.
We m ake th e fu rth er assum ption th a t th e acceleration (curvature o f th e
world fine) is n o t too large, i.e., one can choose th e radius in such a w ay t h a t
the field p H(0) inside S(0> is essentially identical w ith th a t of th e charge e in
uniform m otion and can be derived from th e form ula of § 7 b y a L orentz
transform ation. Now we sp lit th e integral

H dx ( 8 . 12)

into a p art corresponding to th e sphere St0) and th e rest of space R. In S<0) we

have
H = V l — it P k iP 1
1 (8.13)
= V i — W 0) p(0>kl— Pki0)f e)— i
Q= 0 J
448 M. Born and L. Infeld.
Corresponding to (8.11) one can consider th e term s f ki 0>f (e)kl as small of the
first order (com pared w ith f k i 0)f lom)>th e term s f
second order, a n d these la tte r will be neglected. T hen we have by developing
(8.13) a n d using (4.14) :

H = V l - i W ° ’« - l - i /„«»’/ <«*>, (8.14)

w hich holds inside th e sphere S(0). W e can w rite (8.14) in a n o th er fo rm :

H = H (0) — % fki ° f (e)kl

(8.15)
H (0) = Vl — 1 J

(8.15) differs from (8.14) only in th e term s of th e second order. B ut (8.15)

holds n o t only inside b u t also outside th e sphere. F o r in R th e equation (8.15)
takes, according to our assum ptions a b o u t r (0), th e following form :

H = - W V *» - - i / « l' 7 “,*‘ ij
= - ~ (8.16)
This is, how ever, th e know n expression for H in M axw ell’s th e o ry ; (L = — H).
Therefore (8.15) holds as well in th e sphere S(0) as in R . One has

j H dx = J H (0) d x — \ | / fcl(0) f ie)kl (8.17)

£J

W e introduce th e n o ta tio n

4ttA = j H (0) dv - £ \ f i a (0)f {e)kld v - i (8.18)

\ f kl(e

a n d have for th e actio n principle

8 J A d t = 0. (8.19)
The in teg ral
( 8. 20)

in (8.17) gives zero, because

= 0; (P*(e)= 0). (8.21)
dx1
I f we bear in m ind th a t in th e co-ordinate system , where th e point charge is at

r e s t,jH (0) dv is p roportional to th e m ass, we have :

J*H(0) dv — m 0c2j 1 — \ 2/c?dv, (&f|jl

 Foundations of New Field Theory. 449
here v is the velocity of th e centre of th e electron. In th e second in teg ral
f (8.17) we have
, <e> _ _
(8.23)
11 dxk dxl
0r(O)W
id by p artial integration we find, using ■ . —- = 4rcpfc :

i j f 0klf klle) — 4tu| pl dv dt. (8.24)

rhe additional surface integral over th e infinitely large surface can be om itted,
because it gives no co n tribution to th e v ariatio n (8.19). The resu lt i s :

A — m 0c2 a/ 1 — v2/c2 — | dv. (8.25)

We can w rite (8.25) in th e space-vector form :

A = w 0c2 \ / l —v 2/c2 — j" F e)? -j- - | A I dv.(8.25 a )

An electron behaves therefore like a mechanical system f with the rest mass m0,
acted on by the external field f ke(e).%
If the external potential is essentially constant in a region surrounding the
electron considered, the diameter of which is large compared with r0, one gets
instead of (8.25):

| A dt= | m 0c2 's/ 1 — v2/c2 -j- e (<

and this is entirely equivalent to Lorentz’s equations of motion. B ut our

t Born, 6Ann. Physik/ vol. 28, p. 571 (1909); Pauli, “ Relativitatstheorie,” p. 642
(Teubner).
+ The method used in I for deriving the equation of motion is not correct. It started
from the action principle in the form

8 j L dx — 0 (instead 8 J H dx = 0 );

then in the development instead of the coefficients f ki(°) the ^>^(°) appear, which become
infinite at the centre of the electron. Therefore the transformation of the space integral
is not allowed. In the first approximation we have
Pki — P k l ^ + P k l ^
and not
fkl =/fcZ(0)
The mistake in the former derivation is also shown by the wrong result for the mass (the
numerical factor was half of that given here).
450 M. Born and L. Infeld.
form ula (8.25) holds also for fields w hich are n o t constant. A ny field can be
sp lit u p in F o u rier com ponents or elem entary w a v e s ; we m ay consider each
of those separately, a n d choosing th e Z-axis parallel to th e propagation of the
wave, we can assum e t h a t <f>s(e) is pro p o rtio n al to e2ni z/A. Then we see that
th is F o u rier com ponent gives a co n trib u tio n to th e in teg ral (8.25) of the form
(8.26), where <f>s{e) is now th e am p litu d e of th is com ponent a n d e has to be
replaced b y a n “ effective ” charge e, given b y

pezlxdv.

U sing th e expression of pgiven b y (7.12), a n d p u ttin

dv = r2 sin d& d(f> dr,

one has
2 rrir
r2dr
rp ( n/*^ \3/2 C0Rv sin & . 'HHH
o , %
jo f i 1 + 7i>

The -h in te g ratio n can be perform ed, a n d one can w rite

/2nrn\ , .2 f
e“ 6 9 !fi-[ ; 9*
<> ( T T ^

F o r w aves long com pared w ith r0 one has e — e, because (0) = 1. But for
decreasing w ave-lengths th e effective charge dim inishes, as th e little table for
g ( x s) hows :—
Table o f g (cc).f

X. 9 (* )• X• 9 (*)•
1

0 1 1-25 0-796
01 0*988 1-50 0-7 30
0-2 0*984 1*75 0*659
0-3 0*968 2*00 0*588
0-4 0*959 2*25 0*526
0-5 0*949 2*50 0*457
0-6 0 -9 2 9 3*00 0*347
0-7 0 -9 1 7 3-50 0*252
0-8 0*901 4*00 0*186
0-9 0 -8 8 0 5*00 0*094
10 0 -8 5 6

The decrease begins to becom e rem arkable where — 1, or X 2tvr0. For

large x one has g {x) = 2 jx .2
f Calculated by Mr. Devonshire.
 Foundations of New Field Theory. 451

I f we introduce th e qu an tu m energy corresponding to th e w ave-length X

}y E = hcj'k, then using (8.6) one has

= 1 - 2 3 6 - J L =SU 5 ® _ ® _ = J _ J L .
X hem<t? 137-1 111 m„c2
x — 1 corresponds to a q u an tu m energy of a b o u t 100 m Qc2 = 5 . 107 e. volt.
For energies larger th a n th is th e interaction of electrons w ith o th er electrons
(or light waves excited b y those) should become sm aller th a n th a t calculated
by th e accepted theories. This consequence seems to be confirm ed b y th e
astonishingly high penetrating power of th e cosmic ra y s .f

Summary.
The new field theo ry can be considered as a revival of th e old idea of th e
electrom agnetic origin of mass. The field equations can be derived from th e
postulate th a t there exists an “ absolute field ” b which is th e n a tu ra l u n it for
all field com ponents and th e upper lim it of a purely electric field. F rom th e
standpoint of relativ ity transform ations th e th eo ry can be founded on th e
assum ption th a t th e field is represented by a non-sym m etrical tensor akl, an d
th a t th e Lagrangian is th e square ro o t of its d e te rm in a n t; th e sym m etrical
p a rt gkl of akt represents th e m etric field, th e antisym m etrical p a rt th e
electrom agnetic field. The field equations have th e form of M axwell’s equa­
tions for a polarizable m edium for which th e dielectric c onstant and th e m agnetic
susceptibility are special functions of th e field com ponents. The conservation
laws of energy and m om entum can be derived. The static solution w ith
spherical sym m etry corresponds to an electron w ith finite energy (or mass) ;
the true charge can be considered as concentrated in a point, b u t i t is also
possible to introduce a free charge w ith a spatial distribution law. The m otion
of the electron in an external field obeys a law of th e Lorentz ty p e where th e
force is the integral of th e product of th e field and th e free charge density.
From this follows a decrease of the force for alternating fields of short wave-
lengths (of the order of th e electronic radius), in agreem ent w ith th e observa­
tions of the penetrating power of high frequency (cosmic) rays.

f Bom, ‘ Nature,’ vol. 133, p. 63 (1934).