Helicity ip and forward scattering in strong background elds

Victor Dinu,
1,
Tom Heinzl,
2,
Anton Ilderton,
3,
Mattias Marklund,
3,
and Greger Torgrimsson
3,
1
Department of Physics, University of Bucharest, P. O. Box MG-11, Mgurele 077125, Romania
2
School of Computing and Mathematics, Plymouth University, Plymouth PL4 8AA, UK
3
Department of Applied Physics, Chalmers University of Technology, SE-41296 Gothenburg, Sweden
Vacuum birefringence is governed by the probability for a photon to ip helicity (or polarisation
state) in an external eld. Here we calculate the helicity ip and no-ip amplitudes in arbitrary
plane wave backgrounds. Low energy and locally-constant approximations are analysed, along with
the impact of pulse shape and energy on the probabilities. We also provide the rst lightfront-QED
derivation of the coecients in the Euler-Heisenberg eective action.
I. INTRODUCTION
The quantum vacuum, when exposed to intense light,
eectively becomes a birefringent medium [1], because
of the possibility of light-by-light scattering [24]. Pho-
tonic probes of the vacuum then acquire an ellipticity in
their polarisation, in analogy to a probe passing through
a birefringent crystal. Such eects are accessible due to
the advent of high-intensity (optical) and high energy
(X-FEL) laser systems. A agship experiment to mea-
sure vacuum birefringence, using combined optical and
X-ray lasers, has been proposed by the HIBEF consor-
tium [5] following [6]. Continual progress is being made in
meeting the experimental challenges [7], in particular in
developing the required X-ray polarimetry [8]. This pa-
per complements the experimental progress by extending
previous theoretical results [1, 6, 911] obtained for con-
stant (crossed) elds to the case of space-time dependent
plane wave elds.
More generally, one can say that it is becoming a feasi-
ble agenda to study the vacuum as a nonlinear medium.
In this medium, many eects known from classical op-
tics in matter become observable in an entirely photonic
setup, i.e. the quantum vacuum gives us optics without
the optics". Since it is photon scattering which lies be-
hind these eects, they are all captured by the S-matrix,
see Fig. 1 [3]. For example, vacuum birefringence is
part of forward scattering (no momentum change) while
backscattering may be viewed as quantum reection [12].
O-forward scattering in general corresponds to dirac-
tion [1316] or deection [17].
Due to energies in laser-laser scattering being low
(compared to the electron rest mass), one normally con-
siders vacuum birefringence from the perspective of the
low energy eective Euler-Heisenberg action. From this
action one obtains modied Maxwells equations for the
behaviour of a classical electromagnetic eld; solving
these equations in perturbation theory with a plane

dinu@barutu.zica.unibuc.ro

theinzl@plymouth.ac.uk

anton.ilderton@chalmers.se

mattias.marklund@chalmers.se

greger.torgrimsson@chalmers.se

A
ext
A
ext
FIG. 1. Left: the basic photon-photon scattering diagram.
Right: the lowest order contribution to birefringence in a
background eld. Two of the loops legs are attached to the
background, two to a probe photon.
wave ansatz one then nds two refractive indices in
the vacuum, which lead to a quantum-induced elliptic-
ity in an initially linearly polarised plane wave probe.
See [11, 18, 19] for lucid treatments along these lines.
Euler-Heisenberg holds, strictly, only for constant
elds. If one is interested in a particular non-constant
eld geometry, then going beyond Euler-Heisenberg
means calculating the polarisation tensor

for that
eld, and from this extracting the refractive indices.
See [2022] for methods and comprehensive reviews.
The polarisation tensor in plane waves was rst found
in [23, 24]. Most recently, a manifestly symmetric rep-
resentation of this tensor was obtained and the Ward
identity veried in [25], while the tensor in homogeneous
elds has been reexamined in detail in [26].
It is fair to say that such calculations are laborious,
and the expressions obtained involved. What we want to
emphasise here, and in sequel publications, is that it is
not necessary to calculate the whole polarisation tensor
in order to obtain the observables relevant for e.g. bire-
fringence. To understand why, recall that a photon of
denite momentum and helicity propagates freely in vac-
uum. The interaction of the photon with other photons,
or a background eld, leads broadly to three possible ef-
fects. First, the photons momentum is changed. Second,
its helicity is ipped. Third, particle number is changed
via e.g. pair production or photon splitting. For birefrin-
gence, one is interested in the case that photons are ef-
fectively scattered forward with a change in polarisation,
and in the regime where particle production is unlikely
(though see [27]); helicity ip is the physical process un-
derlying the eective birefringence of the vacuum.
a
r
X
i
v
:
1
3
1
2
.
6
4
1
9
v
1


[
h
e
p
-
p
h
]


2
2

D
e
c

2
0
1
3
2
The relevant observables are built from the helicity-
ip and no-ip probabilities, and these can be extracted
directly from quantum scattering amplitudes. These are
the subject of this paper. Before considering the focussed
optical pulses and X-ray probes which will be used in ex-
periments, we consider a basic case for which we have full
analytic control. We therefore restrict attention here to
plane wave backgrounds. (For more complete models of
laser pulses see e.g. [28, 29].) This will allow us to ex-
plicitly check the validity of previous constant (crossed)
eld results [1, 6, 9, 11, 18, 19].
The paper is organised as follows. We begin with a
short review of the eective action approach, and the ob-
servables relevant for birefringence. We then analyse the
helicity ip probability in plane wave backgrounds. For
completeness, we also calculate the no-ip forward scat-
tering amplitude and relate it to the polarisation tensor
and refractive indices. The extension of these results to
realistic elds is discussed in the conclusions.
We use lightfront eld theory throughout, as this is
the easiest and most direct way to perform our calcu-
lations. The details are not needed to understand our
results (which are equivalent to those obtained via the
usual, covariant Feynman diagram approach), and so the
explicit calculations are left to the appendices.
II. PRELIMINARIES AND REVIEW
In this section we briey review the usual approach to
vacuum birefringence in strong elds, based on eective
actions, and then compare with our own approach based
on the calculation of expectation values in QED.
A. Review
The standard derivation of vacuum birefringence [1, 6,
9, 11, 18, 19] starts from the inhomogeneous wave equa-
tion (obtained by varying the quantum eective action)
in a background eld a by linearising in the uctuating
eld b,
b

= j

(a)b

. (1)
Thus, the vacuum current j

is dened in terms of the

background dependent polarisation tensor

(a) de-
picted in Fig. 2. The wave equation (1) is solved by
making a geometric optics, or eikonal, ansatz,
b

exp
_
i
_
, (2)
which implies the algebraic transport equation for the
polarisation vector ,
_

2
g

(; a)
_

= 0 , (3)
with the probe wavevector

. (4)
The secular equation then turns into light-cone condi-
tions [11, 30] which may be written as

= 0 , (5)
in which the

are the two nontrivial eigenvalues of the

polarisation tensor.
At low photon energies m, where m is the elec-
tron mass, the appropriate eective Lagrangian is Euler-
Heisenberg, which to leading order in eld strengths is
/
HE,LO
=
1
2

_
c

S
2
+c
+
P
2
_
. (6)
It depends only on the eld invariants
S =
1
4
F

=
1
2
(E
2
B
2
) , (7)
P =
1
4
F

= E B , (8)
and the low energy constants are c

= 4, c
+
= 7 with
:=
4
45

2
m
4
=

45
1
E
2
S
. (9)
In this situation, the eigenvalues of the polarisation ten-
sor may be expressed in terms of the background energy-
momentum tensor, T

0 , (10)
which obey a positive energy theorem [11]. The light-
cone conditions (5) then take on the simple form

= (g

= 0 , (11)
with eective metric g

. Introducing an index
of refraction
n =
[[

0
, (12)
whence n = 1
2
/2
2
, one nds the two solutions
n

= 1 +c

2
2
= 1 +

45
(11 3)
E
2
E
2
S
. (13)
One can obtain (6) and other eective Lagrangians by
simply writing down all possible invariant combinations
of the eld strength, with unknown coecients. The
matching problem is then to calculate these coecients
in the underlying theory. QED, of course, implies the
Heisenberg-Euler-Lagrangian (6), while the low-energy
limit of the U(1) sector in string theories tends to be
Born-Infeld electrodynamics [18, 31, 32]. In this paper
we will give, to the best of our knowledge, the rst calcu-
lation of the matching coecients c

in lightfront QED.
3
B. Observables for birefringence
The proposed setup for measuring birefringence uses
an intense optical laser probed by an X-FEL beam. Both
are linearly polarised. As the probe, with polarisation ,
passes through the optical laser, it acquires an ellipticity
due to light-by-light scattering, and it is this ellipticity
which is to be measured.
The observables of interest are then the polarisation
of the probes elds, or the number of probe photons
detected with a nonzero

-polarisation component, or-

thogonal to that of the initial photons
1
,

= 0. Such
observables can be calculated in quantum eld theory as
the expectation values of (i) the

-projected photon num-

ber operator,

N

, and (ii) the

E component of the laser
eld, both calculated in the nal state. The structure of
such expectation values is, roughly, as described in the
next paragraph. More details will be given elsewhere; we
provide here the basic structure and analyse in detail the
underlying scattering amplitudes in the quantum theory.
A laser probe is best described by an asymptotic co-
herent state. Such closest to classical states give, for
example, non-zero expectation values for electromagnetic
elds, unlike pure photon-number states. The expecta-
tion value for an observable

O in the nal state is

O
out
= in [S

OS[ in , (14)
where S is the S-matrix operator. Working to one-loop,
we nd that the expected number of produced photons
takes the form
N

out
=
_
dl

(l) P
l
(ip) , (15)
in which dl is the invariant on-shell integral over pho-
ton momenta,

(l) is the number density of -polarised

photons in the in-state,
N

in
=
_
dl

(l) , (16)
while P
l
is the probability for that photon to ip to an
orthogonal polarisation in the background eld. We say
ip here because the probability has the same form
whatever orthonormal basis ,

is chosen for the pho-

tons two physical polarisation states; a helicity basis is
natural, and then the probability is for helicity ip. P
l
is
essentially given by
P
l
(ip) = [

[
2
, (17)
where

is the polarisation tensor, but clearly we only

need to calculate the relevant scattering amplitude, not
the whole tensor.
1
This does not mean the beams polarisation is rotated by 90

.
Rather, it is the presence of a small number of orthogonally-
polarised photons which gives the beam a small ellipticity.
l,

l,
FIG. 2. The one-loop contribution to helicity-ip in a back-
ground eld. The tree-level contribution is identically zero.
Consider now the probes electromagnetic elds; we
consider the expectation value of the electromagnetic
eld operator for an initial coherent state. The coherent
state for a monochromatic plane wave probe is dened
by
a

(l

)[ P = i
E
0
2l
0

(l

l)[ P , (18)
in which is the polarisation of the wave, l

is the mo-
mentum and is the invariant delta-function (see the
appendix). The incoming electric eld of this probe is
given by
F
0

in
:= P [F
0
[ P
= Re E
0

e
ilx
= E
0

cos lx .
(19)
After interaction with the background, the expectation
value of the elds becomes
F

out
:= P [S

S[ P . (20)
Take the vector

, with

= 0. From (19), the initial

electric eld has no

component, but one can be induced

by interaction with the background:

E = Re E
0
e
ilx

,= 0 . (21)
This is vacuum birefringence. If we look instead at the
original component, we see that it too is aected by
interaction with the background, becoming
2
E = Re E
0
e
ilx
(1 ) . (22)
Thus there are two objects of interest here.

is just
the amplitude for helicity ip (or polarisation ip), while
is the no-ip amplitude, which relates anomalous
dispersion to pair production and the refractive indices.
Note, though, that to calculate the observables (15) and
(21) one needs only the helicity ip amplitude. This sin-
gle amplitude clearly corresponds to a particular compo-
nent of the polarisation tensor, but the tensor as a whole
is not needed.
The remainder of this paper is given over to an inves-
tigation of the amplitudes and probabilities for helicity
ip and no-ip.
2
Here we have neglected terms of higher order in the probe eld
strength E
0
. These vanish for helicity ip, where

= 0.
4
III. HELICITY AND POLARISATION FLIP
Consider a photon probing a background plane wave.
The plane wave is a transverse function of nx where n

is
a lightlike vector, n
2
= 0. We take n

= (1, 0, 0, 1) so that
nx = x
0
+ x
3
x
+
, lightfront time (see the appendix).
In order to make approximations and estimations, we
will sometimes use := kx where k = n, with a
typical frequency scale associated with the background.
This is useful for making variables dimensionless. The
background plane wave is
eF
ext

() = k

() a

()k

, a

() := eE

()/ .
(23)
The normalised photon state of denite helicity and
momentum wavepacket
1
(l

) peaked around l

= l, is
[ , =
_
dl

(l

(l

) a

(l

)[ 0 ,
_
dl

[(l

)[
2
= 1 .
(24)
The two possible helicity states of a photon with momen-
tum l are described by polarisation vectors

(l) which,
in the lightfront form, can be chosen to be orthogonal to
both the photons momentum and to the lightfront/laser
direction n

, so l

(l) = n

(l) = 0 . The two helicity

state vectors are (vector components in order +, , )

(l) = (0,
l

=
1

2
(1, i) , = 1 ,
(25)
and they obey the completeness relation

(l)

(l) = g

+
n

+l

nl
. (26)
In the presence of a background, there is a nonzero
probability that the photons helicity will ip ( )
as it propagates. In general backgrounds, this can of
course be accompanied by scattering; when the back-
ground is a plane wave, though, conservation of three
lightfront momentum components means that asymp-
totic scattering is automatically forward. (See [19] for
non-forward, non-asymptotic contributions using the ef-
fective action approach.) Hence the probability for he-
licity ip and forward scattering is given by the total
probability of helicity ip,
P(ip) =
_
dl

[S[

2
, (27)
where the outgoing sum is taken over states l

[ with
helicity opposite to that of the incoming state. The con-
tributing one-loop Feynman diagram is given in Fig. 2,
and the double lines are dressed fermion (Volkov) prop-
agators. While the method of calculation, starting from
this diagram, is by now standard, it is more ecient and
elegant to calculate in lightfront eld theory. This is
discussed in detail in the appendix, but here we jump
straight to the nal result and its analysis.
We drop the subscripts on the polarisation vectors,
since our results hold for more general polarisation states,
see below. The helicity-ip probability depends on the
background eld through the moving average
a

:=
1

+
1
2

1
2

dx a

(x) , (28)
where and are lightfront times which arise in the
S-matrix calculation as centre-of-mass and relative co-
ordinates, respectively. The probability also depends on
the projection of this average onto the polarisation vec-
tors, A := a and

A :=

a. We write a subscript or
for a derivative, so A

and so on. The ip

probability is
P
l
ip) =

kl

_
0
d
1
_
0
ds e

i
2kl
M
2
(,)
s(1s)
_

A

1
4

+
_
1
2

1
4s(1s)
__

A

2
, (29)
where we have taken the wavepacket to be peaked around
momentum l

, and where
M
2
(, ) = m
2
+a
2
a
2
(30)
is Kibbles eective mass [33]. The probability vanishes
when the eld vanishes, as it should. There is no UV di-
vergence here. The variable s is a lightfront momentum
fraction, s = p

/l

for loop momentum p [3436]. The

s-integral in (29) can be performed exactly in terms of
Mejers G-function, but the explicit expression is not re-
vealing so we do not present it. (Alternatively, M
2
m
2
for high-intensity backgrounds and kl/m
2
1 for X-FEL
probes, in which case the asymptotic expansion of the s-
integral can be used.) The remaining , integrals can
be performed numerically.
5
A. Low energy expansion
For laser-based investigations of birefringence, we are
mainly interested in low-energy (compared to the elec-
tron rest mass) probes in low-energy backgrounds, which
means that the invariant b
0
:= kl/m
2
1 and we can
simplify the probability. The most naive way of doing so
is to change variables b
0
and then expand in pow-
ers of b
0
. Both the and s-integrals then become simple,
and we are left with
P
l
(ip)
b01
=

30
kl
m
4
_
d a

()a

()

2
. (31)
In Fig. 3 we compare (31) with the exact polarisation
ip probability (29). We use here, and in subsequent
examples, the following background eld prole:
_
a
1
a
2
_
= a
0
me

2
/T
2
_
cos sin(
01
)
sin sin(
02
)
_
, (32)
which has a Gaussian envelope and a polarisation deter-
mined by and the carrier phases
0i
. For Fig. 3 we
have chosen a circularly polarised background, = /4,

01
= 0,
02
= /2. We see rst that the approximation
(31) is excellent in the low energy regime of interest. As
the probe energy rises, (31) underestimates the full re-
sult, over a range of around one order of magnitude in
b
0
. Thus quantum eects boost the ip probability. How-
ever, for even higher energies, (31) greatly overestimates
the true probability, which begins to drop. The reason is
that at high energies, there is a greater probability that
the photon will decay via e.g. stimulated pair produc-
tion [3742]), and it follows from unitarity that photon
persistence probabilities must drop.
From the gures, we can read o when the approxi-
mation (31) holds. We nd that this can be phrased in
terms of a single parameter, which is the photons quan-
tum eciency parameter [37, 38, 43],

= a
0
b
0

E
E
S

l
m
, (33)
which is just (in this case the peak value of) the contrac-
tion of the probe momentum with the energy momentum
tensor in a plane wave [6, 11],

2
=
e
2
m
6
l

. (34)
In terms of , we nd that dierence between (29) and
(31) is within 5% for < 0.4, and within 10% for
< 0.54 . (35)
Even for an optical eldstrength of E = 10
4
E
S
, which
corresponds to an intensity of around 10
23
W/cm
2
, the
approximation (31) therefore remains valid for photon
energies reaching
l
1 GeV. The worst underestimate,
(31)/(29) 0.72, occurs at 1.38. The probability
10
4
0.001 0.01 0.1 1
b
0
10
5
10
4
0.001
0.01
0.1
1
P, Pc
10
4
0.001 0.01 0.1 1
b
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
P
Pc
FIG. 3. Comparison of the exact polarisation helicity ip
probability (29) and its approximation (31) as a function
of invariant b0 = kl/m
2
(log-log scale). The background is
circularly polarised with a Gaussian prole, see (32), width
T = 10 and a0 = 200, 100, 50 (left to right). In the regime
of interest, the approximation (31) is excellent.
for helicity ip is maximal for 17. This is all consis-
tent with expectations, which are that quantum eects
become important for 1, although we have been able
to give more specic values.
Interestingly, and perhaps surprisingly, Fig. 3 shows
that (31) again becomes a good approximation at even
higher energies, due to the behaviour of the probability.
Because the probability turns and begins to drop, the
low-energy expansion changes from an under- to an over-
estimate, and there is then a (small) kinematic regime
around 3 in which the approximation becomes close
to the true probability again.
B. Locally constant approximation
More about the low-energy approximation, and what
one loses when making it, is revealed by considering the
locally constant approximation to (29). This can be ob-
tained either by expanding (29) in the relative coordinate
6
, or by writing down (29) for a constant crossed eld
and then replacing E E() everywhere. For the latter
method, take = m, and dene c
j
= E
j
/E
S
(c
0
= 0),
the ratio of the electric eld to the Schwinger eld. A
vector potential for the crossed eld is a

() = mc

.
The probability becomes
P
CF
=

m
2
kl

_
0
d
1
_
0
ds e
i
m
2
kl
E
2

3
/12
2s(1s)
1
4
(c)( c)

2
.
(36)
This diverges quadratically with the volume of lightfront
time . However if we replace c c() then we ob-
tain the locally constant approximation, P
LCFA
. The -
integrals can be performed exactly in this case, giving
Airy (Ai) and Scorer (Gi) functions (both of which are
typical of one-loop results in constant elds, see [11, 44]),
P
LCFA
=

kl
m
2
_
d c()c()
_
ds[s(1 s)]
2

2
d
d
_
Ai() iGi()

2
(37)
where
=
_
m
2
kl
1
s(1 s)
_
2/3
(2c
2
)
1/3
. (38)
The kinematic regime of interest is
kl
m
2
1 1;
we therefore turn to the asymptotic expansions of the
Airy and Scorer functions, which for 1 are
Ai

() =
1
2

1/4
exp
2
3

3/2
, Gi

() =
1

2
.
(39)
The Airy function term is nonperturbatively small in the
invariant b
0
= kl/m
2
, relative to the Scorer function.
The locally-constant approximation reveals that there is
a nonperturbative (in b
0
) part to the helicity ip prob-
ability which is, naturally, missed when one simply ex-
pands in powers of b
0
. Nevertheless, since this nonpertur-
bative part is exponentially small in the regime of inter-
est, it should be safe to neglect it; doing so, the remain-
ing s-integral is trivial and we recover precisely (31), so
P
LCFA
= P(ip) for b
0
1. The validity of this ap-
proximation in the regime of interest is conrmed by the
excellent agreement between the approximation (31) and
the exact result (29) show in Fig. 3 and Fig. 4.
We have therefore seen, either by direct low-energy
expansion or by going via the locally-constant approx-
imation, that the ip probability collapses to a simple
expression which is quadratic in the background, i.e. the
same as would be obtained to lowest order by treating the
background perturbatively. This is an explicit example
of the general statement given in [45], that higher-order
terms (in the background) are hard to observe in loop
processes, when probes have low-energy.
C. Dependence on geometry and pulse shape
The result (29) is not restricted to purely the helic-
ity states; we can choose any two orthogonal combina-
tions ,

of the helicity states as a basis of photon

polarisations, and in fact (29) holds for the probability
of polarisation ip between these states. This is relevant
because the experimental case of interest, birefringence,
takes both the background and probe to be linearly po-
larised. Let us then make the connection between the
above results and those in the literature on birefringence.
We take the incoming photon to have polarisation
= (
+1
+
1
)/

2, and we look at the probability that

this photon will have ipped to the opposite linear polar-
isation

= (
+1

1
)/(i

2) after interaction with the

background. (Both these vectors are real.) The prob-
ability (29) and its approximation (31) are plotted in
Fig. 4 for this case, with a linearly polarised background
given by (32) with = /4 and both
0i
= 0. In this
case the approximation (31) is within 10% of the full
probability for < 0.6, and the worst underestimate,
(31)/(29) 0.74, occurs at around = 1.1.
Considering (31) in the low energy regime, we can iden-
tify three parameters on which the probability depends;
the collision angle, the angle between the polarisations of
the background and probe, and (somewhat broadly) the
background eld prole. Since the polarisation vectors
are real in this linear setup, we can look directly at the
amplitude o such that P = [o[
2
in (31). Let be the
angle between the beam directions k and l, and let be
the angle between

and a

. For probe frequency

l
we
have,
o =

30

l
sin
2

2
sin 2
_
dx
+
E
2
(x
+
)
E
2
S
. (40)
We have a straightforward factorisation of dependencies,
and the probability is clearly maximal for head-on colli-
sions, = , and a 45

angle between the probe polarisa-

tion and the background, = /4. (This is well known in
the case of birefringence.) This leaves us with the pulse
shape. For these values, consider rst a constant eld of
duration x
+
in lightfront time. Introducing the probe
wavelength
l
= 2/
l
, we nd
(40)

15
x
+

_
E
E
S
_
2
=
2
15
d

_
E
E
S
_
2
, (41)
where, in the last step, we have introduced the distance
d traveled by the probe in the birefringent medium; for
the case of a head-on collision between a photon and a
plane wave, this distance is d = x
+
/2, as shown in
Fig. 5. We recognise the structure of the birefringence
signal from [6], which has come directly from the ampli-
tude for helicity ip in a background eld. Note that the
observable is (41)-squared; in the quantum theory, this
is the probability of no-ip, and in the eective theory it
is the phase shift of the probe beam, see [6].
7
10
4
0.001 0.01 0.1
b
0
10
5
10
4
0.001
0.01
0.1
1
P, Pc
10
4
0.001 0.01 0.1
b
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
P
Pc
FIG. 4. Comparison of the exact polarisation ip probability
(29) and its approximation (31) as a function of invariant
b0 = kl/m
2
(log-log scale). The background has a Gaussian
prole with width T = 10 and a0 = 200, 100, 50 (left to
right). Top: individual values. Bottom: the ratio of the two.
In the regime of interest, the approximation is excellent.
x
+
= 2d
z = d
t
z
Probe
x
+
= x
+
x
+
= 0
z =
1
2
x
+
t
z
Probe
FIG. 5. Head-on collision between the probe and the back-
ground eld. The background eld has lightfront duration
x
+
, and the probe emerges after a distance d = x
+
/2.
It is clear that the constant eld probability can be in-
creased by raising the elds amplitude or extending its
duration, so let us now consider a nonconstant wave of
arbitrary prole. Returning to (40), we could calculate
the x
+
-integral for dierent intensities and eld congu-
rations and then compare. However, the physically real-
istic scenario is that a laser will have a xed amount of
energy which can be manipulated (e.g. via focussing) into
dierent pulse shapes. The parameter we should there-
fore keep constant when comparing pulses is the energy.
The best we can do in a plane wave is to hold constant
the total energy per transverse area (or, integrated in-
tensity), which is
1
2
_
dx
+
E
2
(x
+
) +B
2
(x
+
) =
_
dx
+
E
2
(x
+
) . (42)
We see immediately that if the energy is xed, (40) is
also xed, and the ip probability becomes independent
of the pulse shape, under the constraint of xed energy.
This is a potentially positive result, as it suggests that
birefringence signals, while weak, are robust. The im-
portant question is of course how well this result extends
to more realistic, focussed background elds. This will
be investigated elsewhere.
IV. FORWARD SCATTERING, PAIR
PRODUCTION AND THE REFRACTIVE
INDICES
For completeness we discuss here the no-ip forward
scattering amplitude and its relation to the refractive in-
dices and pair production.
A. Structure and renormalisation
For a one-photon state [ , of the form (24) the no-
ip forward scattering amplitude is
T = , [S[ , = 1 +T
1
+. . . , (43)
where S is the S-matrix and a subscript 1 means the one-
loop term, on which we concentrate. The relevant Feyn-
man diagram is as in Fig. 2 but with the same incoming
and outgoing states. T
1
is related to the polarisation
tensor by
T
1
=
_
dl [(l)[
2
(l, l) , (44)
where is here given in Fourier space. T
1
is calculated
in the appendix, again using lightfront quantisation. A
photon is associated with two independent no-ip am-
plitudes, because the photon has two possible polarisa-
tion/helicity states. We present the simplest case here,
which is when a = 0. Taking the wavepacket to be
peaked around l

, and writing a
2
a( + /2), and
a
1
a(/2), the no-ip forward scattering amplitude
for a photon of polarisation obeying a = 0 is
8
T
1
=
i
2kl

_
0
d
1
_
0
ds
_
e

i
2kl
M
2
s(1s)
(a
2
a
1
)
2
4s(1 s)
+
M
2
m
2

1
_
0
d e

i
2kl
M
2

s(1s)
_
, (45)
where M
2

:= m
2
+(M
2
m
2
). This is the renormalised
result; the nal term, with the integral, ensures that T
1
is UV nite and vanishes when the eld vanishes (because
the photon is stable), so that T 1. In this form, with
the nal state integrals performed, the UV divergence
appears as a contact term [21] in position space, at =
0. We can make this explicit by performing the integral,
which is exact. (We left the integral unevaluated in (45)
only for the sake of presentation.) The nal term in T
1
becomes,
M
2
m
2

1
_
0
d e

i
2kl
M
2

s(1s)

i
2kl
M
2
s(1s)
(a = 0)

2
,
(46)
where we see that taking the dierence of amplitudes in
the theories with and without backgrounds (a structure
which extends trivially to the whole of T
1
) regulates a
1/
2
divergence at = 0.
At this stage we can verify our results by compari-
son with the well-known constant crossed eld case. The
crossed-eld polarisation tensor is well known, and con-
tracting this with , where a = 0, should yield our the
no-ip amplitude in a crossed eld.
In a constant crossed eld we have a

= mc

, taking
= m for convenience, and M
2
= m
2
(1c
2

2
/12). (Re-
call that c
2
= E
2
/E
S
> 0.) In this case (45) becomes
independent of , and the d-integral contributes the in-
nite volume of lightfront time, which we can divide out.
Now take the rst term in (45), and change variables as
follows:
t :=

2b
0
s(1 s)
, s :=
1
2
(1 +v) . (47)
Then the rst term becomes
ib
0
16
_
E
E
S
_
2
1
_
1
dv (1 v
2
)

_
0
dt t exp
_
it i
1
48

2
t
3

,
(48)
where = b
0
(1 v
2
)E/E
S
. For the second term in
(45), it is easiest to keep the integral unevaluated for
now. Change variables t as above, then from
:= s
2
(1 s)
2
. The integrand becomes independent
of s, and that integral can be performed. After this,
one can change variables again, v with = (1
v
2
)
2
/16. (Note that these changes of variables must be
dened piecewise in order to be invertible and properly
cover the integration volume.) One obtains (48) but with
an additional factor of v
2
/3 in the integrand. Finally,
adding the two terms together gives us
T
1
_
d
=
ib
0
16
_
E
E
S
_
2
1
_
1
dv (1 v
2
)(1
1
3
v
2
)

_
0
dt t exp
_
it i
1
48

2
t
3

, (49)
in which we have the same integral as in equation 4.26 in
[11] and in equation (4) in [9]) (contracted with where
a = 0). A further check is provided below.
B. The imaginary part and refractive indices
Take an incoming monochromatic laser with momen-
tum l

. As stated in the introduction, the no-ip ampli-

tude appears in the expectation value of the nal states
elds as
E
out
(1 + Re T
1
) cos x

+ (ImT
1
) sin x

Re e
ix

+F1
. (50)
from which we identify T
1
as a (complex) phase shift in-
troduced into the wave by its interaction with the back-
ground. Let us again begin with constant crossed elds.
From (50), we can identify the phase shift in the crossed
eld using (2),
= lx nx
T
1
_
d
, (51)
in which the volume of lightfront time is rewritten as nx,
and then we can dene the wavevector as in (4),

= l

T
1
_
d
. (52)
From this the refractive index is extracted as before, n =
[[/
0
, and in the low energy limit we correctly recover
the result (28) in [9],
n = 1 + 7

90
m
2

2
l
, (53)
9
where
l
is the incoming probe frequency. Had we taken
| a, we would have obtained the second refractive index
with coecient 2/45 instead of 7/90, see the appendix.
Returning to non-constant elds, the low-energy ex-
pansion of the no-ip amplitudes imaginary part is
Im T
1
nl
7
90
_
dx
+
E
2
(x
+
)
E
2
S
. (54)
Clearly, the indices will be functions of spacetime. In or-
der to be able to give the local refractive indices, though,
a non-asymptotic calculation must be performed. The
reason is that there can be nonzero contributions to am-
plitudes which vanish asymptotically, and so it is strictly
not allowed to extract the local indices from T
1
in (45).
We emphasise, though, that it is T
1
which enters observ-
ables. One does not need the indices, which are an eec-
tive description of a scattering process. Note also that
while the indices will be dependent on pulse shape, only
integrated quantities appear in observables. We there-
fore see once again that the only relevant property of the
pulse is its total energy.
C. The real part
Turning now to the real part no-ip forward scattering
amplitude (essentially the imaginary part of the refrac-
tive index) , we recover the well-known optical theorem
result that, for all energies,
2Re T
1
= P(pair) , (55)
where P(pair) is the total probability of stimulated pair
production [4042], written here in the form (45), not
previously found in the literature, in which the nal state
momentum integrals have been performed
3
. One nds
that the low energy (small b
0
) expansion of Re T
1
is zero.
Naturally, it is zero to all orders because pair produc-
tion is nonperturbative in the kinematic invariant b
0
. In
vacuum, pair production by two photons k

and l

is a
threshold process forbidden for kl < 2m
2
b
0
< 2.
For a photon l

in a background eld, this threshold is

removed because of the (in principle) arbitrarily high fre-
quency components of the background; if k

is the central
frequency and s parameterises the frequency range, the
S-matrix element will contain the structure
_
ds
_
skl/m
2
2
_
. (56)
While this does indeed have support for arbitrarily low
l

, it remains identically zero to all orders in a low energy

expansion.
3
That P(pair) is here related to the real, rather than imaginary,
part of F is just a question of denitions.
V. CONCLUSIONS
We have presented the helicity-ip and non-ip am-
plitudes for a photon in an arbitrary plane wave back-
ground. In the kinematic regime to be employed by up-
coming birefringence experiments, the relevant probabili-
ties have simple and accurate low-energy approximations.
Interestingly, these extend to fairly high probe energies
as currently available laser frequencies (/m) and eld
strengths (E/E
S
) remain low. All the amplitudes we
have considered contain parts which are nonperturba-
tively small in the kinematic invariant kl/m
2
.
We have concentrated on helicity/polarisation eects
without changes in momentum. As we saw, a single pho-
ton is always scattered forward in a plane wave back-
ground. In a general background, though, it may be that
forward scattering is more likely than non-forward scat-
tering, or vice versa, and this may provide clues for iden-
tifying the most promising experiment to detect eects
due to light-by-light scattering. Alternatively, it is easy
to imagine that the eld structure could be tuned to em-
phasise particular eects. These are interesting questions
for the future.
Appendix A: Strong eld QED in lightfront
quantisation
All the amplitudes in this paper can be calculated
from the standard Feynman diagrams, such as that in
Fig. 2. Expressions for the dressed propagator can be
found throughout the literature [43, 46], state normalisa-
tions and wave-packet manipulation are given in [47], and
nal state integrals are calculated as in [48]. Calculations
are most easily performed using lightfront coordinates.
However, there is a more convenient way to calculate
scattering amplitudes in plane wave backgrounds. That
is, as rst advocated in [46], to employ lightfront quanti-
sation, in which one quantises on lightlike hypersurfaces,
taking x
+
= x
0
+ x
3
as the time direction (see [35, 36]
for reviews). There are several advantages to using this
front form over the usual instant form of quantisa-
tion. First, we can rotate the background plane wave so
that its phase is aligned with (it depends only on) x
+
,
which makes calculations simple because the background
enters only through time-dependent factors. Second, be-
cause lightfront eld theory is an on-shell formalism, all
of the p
+
integrals in Feynman diagrams are done for free.
Third, many terms vanish immediately because p

> 0,
which is related to the properties of the ligthfront vac-
uum.
One small disadvantage of the formalism is that a
Feynman diagram with n vertices leads to (roughly) n!
lightfront-time-ordered diagrams. Even though this in-
crease is minimal for the low-order processes we consider
here, it is nevertheless simplest to drop the diagrammatic
approach entirely, and calculate directly with the states
and Hamiltonian. This may be viewed as a particular
10
incarnation of old-fashioned perturbation theory [49].
Coordinates are x

= x
0
x
3
, x

= x
1
, x
2
, p

=
(p
0
p
3
)/2, p

= p
1
, p
2
. The invariant, on-shell mo-
mentum space measure dp is, in lightfront coordinates,
_
dp
_
d
2
p

(2)
2

_
0
dp

(2)2p

, (A1)
and p
+
is always given by the mass-shell condition, p
+
=
(p
2

+m
2
)/(4p

), m 0 for photons. We dene (p) by

_
dp (p) = 1 . (A2)
The mode expansions for the photon is
A

=
_
dl a

(l)

(l)e
ilx
+ c.c., (A3)
with A
+
= 0 (lightfront gauge) and A

the
constrained eld. We dene a

(l) = a

(l)

(l). For the

fermion elds we use the four-component, rather than
two-component, spinor formalism,
(x) =
_
dp b

(p)
+
K
p
u
sp
+

K
p
v
sp
d

(p)

, (A4)
in which the spinor and scalar components are, respec-
tively [46, 50],
K
p
() = 1 +
/ k/ a()
2kp
,

(x) = exp
_
ipx +

_
0
2pa a
2
2kp
_
.
(A5)
The commutators of the fermions modes are,
b

(p), b

(p

) = d

(p), d

(p

) = (pp

, (A6)
while the photon modes obey
[a

(l), a

(l

)] = (l l

)
_
g

+l

nl
_
. (A7)
Of interest here are the one-loop amplitudes for single
photon scattering, momentum l

and polarisation

to
momentum l

and polarisation

. The general one-loop

amplitude between one-photon states [ l, is the follow-
ing lightfront time-ordered product,
S
fi
= l

[
_
dy
+
y
+
_
dz
+
H
1
(y
+
)H
1
(z
+
)[ l, , (A8)
in which H
1
is the order e term in the lightfront QED
Hamiltonian,
H
1
=
e
2
_
dx

A

. (A9)
The order e
2
terms in the Hamiltonian, the seagull and
four-fermion terms, both particular to lightfront quanti-
sation, do not ultimately contribute to the processes of
interest here, and so we discard them. (In terms of Feyn-
man diagrams, one nds that the instantaneous parts of
the fermion propagators do not contribute.)
We need rst to evaluate the action of the Hamiltonian
on the incoming state,
H
1
(z
+
)[ l, = H
1
(z
+
)(l)a

(l)[ 0 . (A10)
The integral over x in H
1
yields a delta function in three
momentum components. Because all momenta have pos-
itive minus component, q

> 0, and because there are no

fermions in the incoming state, only terms like
H
1
(z
+
)(l)a

(l)[ 0 b

aa

(l)[ 0 (A11)
can contribute. The photon operators can then be com-
mutated away. It follows that the only contributing term
of the second factor of H
1
can be bda

, and all the

commutators can be performed immediately. One nds,
after a standard, tedious, calculation of the trace that
S
fi
= (l

l)R, where
R =
e
2
kl
_
d
2
2
_
d
1
_
dp
(kl kp)
kl kp
e
i
2

1
l
klkp
_

(2

1)
+
_
1
1
2
kl
2
kpk(l p)
_

[2

1]
+

kl
2
4kpk(l p)
(a
2
a
1
)
2
_
,
(A12)
with 1, 2 indicating arguments
1
,
2
and bracketed sub-
scripts indicating symmetrisation and antisymmetrisa-
tion. The Lorentz momentum for a particle in a plane
wave is
= p a +
2ap a
2
2kp
k . (A13)
The integrals in (A12) are regulated in transverse dim
reg. The eld-dependent terms are UV nite, and renor-
11
malisation is performed by subtracting the divergent free-
eld contribution from the integrand of R. After this,
one can return to 3+1 dimensions, and perform the p

integrals, which are Gaussian. We also change variables

from p

to the lightfront momentum fraction s := p

/l

,
and from the ordered times to = (
2
+
1
)/2 and
=
2

1
. This leaves us with the expressions (29)
and (45) in the text, for appropriate choices of polari-
sation vectors. We note that the s(1 s) factor in the
exponents of our nal expressions is typical of lightfront
wavefunctions [3436].
For small kl we can expand (A12) to nd
R(a) R(0) =
i
4
kl
m
2
1
45
_
14

a
2
6

. (A14)
There are two cases discussed in the text. First, no-ip,

= , for which R(a) R(0) = T

1
. There are two inde-
pendent subcases of here; if a = 0 we pick up the term
with coecient 14, and if | a we pick up both terms,
146 = 8. These are the well known coecients of the re-
fractive indices. When

= 0, R(a) R(0) = R(a) o,

the helicity ip amplitude, and we pick up only the -
nal term with coecient 6, which is the dierence of the
indices.
[1] J.S. Toll, PhD thesis, Princeton, 1952 (unpublished).
[2] O. Halpern, Phys. Rev. 44, 855 (1934).
[3] H. Euler and B. Kockel, Naturwiss. 23, 246 (1935).
[4] W. Heisenberg and H. Euler, Z. Phys. 98, 714 (1936)
[physics/0605038].
[5] See the HIBEF website at
http://www.hzdr.de/db/Cms?pNid=427&pOid=35325
[6] T. Heinzl, B. Liesfeld, K. -U. Amthor, H. Schwoerer,
R. Sauerbrey and A. Wipf, Opt. Commun. 267 (2006)
318 [hep-ph/0601076].
[7] H.-P. Schlenvoigt, T. E. Cowan, T. Heinzl, R. Sauerbrey,
U. Schramm, to appear.
[8] B. Marx, K.S. Schulze, I. Uschmann, T. Kmpfer,
R. Ltzsch, O. Wehrhan, W. Wagner, C. Detlefs,
T. Roth, J. Hrtwig, E. Frster, T. Sthlker,
G.G. Paulus, Phys. Rev. Lett. 110, 254801 (2013).
[9] N.B. Narozhnyi, Zh. Eksp. Teor. Fiz. 55, 714 (1968) [Sov.
Phys. JETP 28, (1969)].
[10] V. I. Ritus, Annals of Physics 69 (1972) 555.
[11] G. M. Shore, Nucl. Phys. B 778, 219 (2007) [hep-
th/0701185].
[12] H. Gies, F. Karbstein and N. Seegert, New J. Phys. 15
(2013) 083002 [arXiv:1305.2320 [hep-ph]].
[13] E. Lundstrom, G. Brodin, J. Lundin, M. Marklund,
R. Bingham, J. Collier, J. T. Mendonca and P. Norreys,
Phys. Rev. Lett. 96 (2006) 083602 [hep-ph/0510076].
[14] A. Di Piazza, K. Z. Hatsagortsyan and C. H. Keitel, Phys.
Rev. Lett. 97, 083603 (2006) [hep-ph/0602039].
[15] B. King, A. Di Piazza and C. H. Keitel, Nature Photonics
4 (2010) 92.
[16] J. M. Davila, C. Schubert and M. A. Trejo,
arXiv:1310.8410 [hep-ph].
[17] J. Y. Kim and T. Lee, JCAP 1111, 017 (2011)
[arXiv:1101.3433 [hep-ph]].
[18] I. Bialynicki-Birula, In Quantum Theory Of Particles
and Fields" (1984) 31, B. Jancewicz and J. Lukierski
(Eds.).
[19] I. Aeck, J. Phys. A 21 (1988) 693.
[20] C. Schubert, Nucl. Phys. B 585 (2000) 407 [hep-
ph/0001288].
[21] W. Dittrich and H. Gies, Springer Tracts Mod. Phys. 166
(2000) 1.
[22] G. V. Dunne, In From elds to strings, vol. 1" 445-522
M. Shifman (ed.) et al. [hep-th/0406216].
[23] W. Becker and H. Mitter, J. Phys. A 8, 1638 (1975).
[24] V. N. Baier, A. I. Milshtein and V. M. Strakhovenko, Zh.
Eksp. Teor. Fiz. 69, 1893 (1975) [Sov. Phys. JETP 42,
961 (1976)].
[25] S. Meuren, C. H. Keitel and A. Di Piazza, Phys. Rev. D
88 (2013) 013007 [arXiv:1304.7672 [hep-ph]].
[26] F. Karbstein, arXiv:1308.6184 [hep-th].
[27] T. N. Wistisen and U. I. Uggerhj, Phys. Rev. D 88
(2013) 053009.
[28] N. B. Narozhny, M. S. Fofanov, JETP 117 (2000) 867.
[29] A. M. Fedotov, Laser Physics 19 (2009) 214.
[30] H. Gies and W. Dittrich, Phys. Lett. B 431 (1998) 420;
Phys. Rev. D 58 (1998) 025004.
[31] M. Born and L. Infeld, Proc. Roy. Soc., Series A 144
(1934) 425.
[32] B. Zwiebach, A rst course in string theory (2004) 2nd
ed., Cambridge Univ. Press 2009.
[33] T. W. B. Kibble, A. Salam and J. A. Strathdee, Nucl.
Phys. B 96 (1975) 255.
[34] G. Lepage, S. Brodsky, T. Huang, and P. Mackenzie,
Hadronic wave functions in QCD", Proceedings of the
Ban Summer Institute (1981).
[35] S. J. Brodsky, H.-C. Pauli and S. S. Pinsky, Phys. Rept.
301 (1998) 299 [hep-ph/9705477].
[36] T. Heinzl, Lect. Notes Phys. 572 (2001) 55 [hep-
th/0008096].
[37] A. I. Nikishov and V. I. Ritus, Zh. Eksp. Teor. Fiz. 46 ,
776 (1963).
[38] A. I. Nikishov and V. I. Ritus, Zh. Eksp. Teor. Fiz. 46,
1768 (1964).
[39] R. Schutzhold, H. Gies and G. Dunne, Phys. Rev. Lett.
101 (2008) 130404
[40] T. Heinzl, A. Ilderton and M. Marklund, Phys. Lett. B
692 (2010) 250 [arXiv:1002.4018 [hep-ph]].
[41] T. Nousch, D. Seipt, B. Kampfer and A. I. Titov, Phys.
Lett. B 715 (2012) 246.
[42] A. I. Titov, H. Takabe, B. Kampfer and A. Hosaka, Phys.
Rev. Lett. 108 (2012) 240406 [arXiv:1205.3880 [hep-ph]].
[43] A. Di Piazza, C. Muller, K. Z. Hatsagortsyan and
C. H. Keitel, Rev. Mod. Phys. 84 (2012) 1177
[arXiv:1111.3886 [hep-ph]].
[44] T. Heinzl and O. Schroeder, J. Phys. A 39 (2006) 11623
[hep-th/0605130].
[45] A. Di Piazza, Annals Phys. 338 (2013) 302,
12
[arXiv:1303.5353 [hep-ph]].
[46] R. A. Neville and F. Rohrlich, Phys. Rev. D 3 (1971)
1692.
[47] A. Ilderton and G. Torgrimsson, Phys. Rev. D 87 (2013)
085040.
[48] V. Dinu, Phys. Rev. A 87 (2012) 052101.
[49] S. Weinberg, The Quantum theory of elds. Vol. 1: Foun-
dations, Cambridge University Press (1995).
[50] D. Volkov, Z. Phys. 94 (1953) 250.