a

r
X
i
v
:
1
0
0
8
.
0
7
3
3
v
1


[
a
s
t
r
o
-
p
h
.
H
E
]


4

A
u
g

2
0
1
0
Mon. Not. R. Astron. Soc. 000, () Printed 5 August 2010 (MN L
A
T
E
X style le v2.2)
Electron-positron bremsstrahlung and pair creation in very
high magnetic elds
P. B. Jones

Department of Physics, University of Oxford, Denys Wilkinson Building,

Keble Road, Oxford OX1 3RH
ABSTRACT
Cross-sections for Rutherford scattering, Coulomb bremsstrahlung and pair creation,
have been calculated at very high magnetic elds in order to investigate the photo-
production of protons at the polar caps of pulsars whose spin is antiparallel with
the polar magnetic ux density. The Landau-Pomeranchuk-Migdal eect at very high
magnetic elds is included in a simple electron Green function.
Key words: pulsars: general - stars: neutron - stars: magnetic eld
1 INTRODUCTION
Work on quantumelectrodynamic processes at magnetic ux
densities of the order of the critical eld Bc = m
2
c
3
/eh =
4.41 10
13
G has been reviewed recently, and very com-
pletely, by Harding & Lai (2006). The processes concerned
are primarily those relevant to collisions within plasma at
low altitudes above the surface of a neutron star and there-
fore do not include relativistic Coulomb bremsstrahlung and
pair creation in the interaction of electrons and photons
with ions. There appear to be no published cross-sections
for these processes at B Bc apart from a number of pa-
pers on the bremsstrahlung energy-loss of non-relativistic
electrons. But of these, only Lieu (1981) and Lauer et al
(1983) have treated bremsstrahlung as a process of second
order in the electron-photon coupling and the work of the
latter authors is also limited to collisions in which both ini-
tial and nal electrons are in the lowest Landau state. Bus-
sard (1980) and Langer (1981) have considered photon pro-
duction in non-relativistic electron-ion plasma collisions but
have treated it as a sequence of rst-order processes, that
is, Coulomb excitation of a higher Landau state followed by
cyclotron emission.
However, the reverse ux of electrons or positrons pro-
duced by pair creation in the open magnetic ux-line re-
gion above neutron-star polar caps must produce electro-
magnetic showers in the condensed matter at the surface
whose properties are determined by the relativistic second-
order processes. In neutron stars with spin such that
B > 0, the outward particle ux consists of electrons so
that nuclear reactions within the positron-initiated showers
have no obvious observable eect. But in the opposite case,
B < 0, the outward particle ux is positively charged and,
with positrons, also includes protons that are produced in

E-mail: [email protected]
the shower and diuse to the surface. It is likely that the
temporal behaviour of these dierent components does lead
to instabilities that are the basis of observable phenomena
(Jones 2010). In general, there appears to be no reason a
priori why both B cases should not be present in the
neutron-star population and it is of interest to understand
how each might be observed in the electromagnetic spec-
trum. Estimates of shower development in previous work
relied on the correspondence principle, that for large values
of the Landau quantum number the zero-eld cross-sections
are valid, but this is unlikely to be satisfactory at B Bc
where the number of contributing Landau states can be
small. The present paper attempts to give cross-sections that
are rather better than order-of-magnitude estimates but are
not of high accuracy owing to the discrete nature of the
Landau spectrum and to truncations that are made in per-
forming the calculations. Its purpose is to see whether and
in what way the cross-sections dier qualitatively from those
at zero eld. Nucleon production in electromagnetic showers
occurs almost entirely through formation and decay of the
giant dipole resonance. Thus the paper is concerned, par-
ticularly, with factors that inuence the low-energy photon
track length and its depth distribution.
With neglect of radiative corrections and of the anoma-
lous magnetic moment, the energy states of an electron in a
uniform magnetic eld are
E =
_
p
2
+ 1 + 2nB (1)
in which p is the longitudinal momentum component, paral-
lel with B, and n = 0, 1, 2, ... is the Landau quantum num-
ber. Apart from the Appendix and where explicitly stated
otherwise, momentum and energy are here expressed in units
of mc and mc
2
; the magnetic ux density B is in units of
Bc. In electron-photon interactions, the conserved quanti-
ties are energy and the longitudinal momentum component.
Thus the kinematic behaviour of an electron is that of a par-
2 P. B. Jones
ticle conned to one dimension and of mass m
_
1 + p
2

in
which we have dened, purely for ease of notation, a notional
transverse momentum component p

2nB.
We have adopted the solutions of the Dirac equation
found by Johnson & Lippmann (1949) and give a summary
of calculational details in the Appendix. These are unre-
markable except that owing to the one-dimensional kine-
matics of electrons and the small number of Landau states
that contribute signicantly at B Bc, we have found it
convenient, given that numerical computation must be in-
volved, not to proceed with the usual approach of covariant
formalism and Green functions but instead to write down
the transition matrix to second order directly in terms of
time-ordered matrix elements calculated using an explicit
representation of the Dirac matrices. It is also appropriate
to mention here that the states given by equation (1) are
two-fold degenerate for n > 0 only if radiative level shifts,
natural widths, and the anomalous magnetic moment of the
electron are neglected. But for the high-energy spin-averaged
or summed processes considered here, the ne structure is
of no consequence and we use only the completeness of the
Johnson-Lippmann eigenfunctions. References and some fur-
ther details are given in the Appendix.
The most simple problem here is that of nding the re-
placement for Rutherfords scattering formula. Small-angle
Coulomb scattering in the zero-eld case is replaced at
B Bc by the Coulomb excitation of Landau states. This
has been treated by Bussard (1980) and Langer (1981)
but only in the non-relativistic limit. The relativistic cross-
sections that we require appear not to have been published
previously and are given in Section 2. There have been many
calculations of transition rates for cyclotron emission, and
a summary is contained in the review of Harding & Lai
(2006). New transition rates are given in Section 3 to al-
low comparison with previous work and, more specically,
to obtain the distribution of photon transverse momentum
k

from the decay of n > 1 Landau states which is of

crucial importance in the development of electromagnetic
showers at B Bc. Cross-sections for Coulomb pair cre-
ation and bremsstrahlung are obtained in Sections 4 and 5.
In the case of pair creation, there appears to be no previ-
ously published work. For bremsstrahlung, the work of Lieu
(1981) and Lauer et al (1983) is in the non-relativistic limit
so that comparison is not possible. There is also the fur-
ther problem that calculations relying on truncation of the
number of Landau states lack external tests of their correct-
ness because the zero-eld limit is not available. But where
minor tests of correctness exist, we nd that they are sat-
ised. Finally, with primary electron energies of the order
of 10
3
GeV and the high condensed matter densities pre-
dicted at the neutron star surface, it is essential to consider
the Landau-Pomeranchuk-Migdal eect and its inuence on
Coulomb pair creation and bremsstrahlung cross-sections.
Its existence and broad properties at zero eld have been
veried through experiments at particle accelerators. We re-
fer, in particular, to the recent work of Hansen et al (2004)
and to the review of Klein (1999). We nd that the eect
exists at B Bc and that the one-dimensional electron kine-
matics makes possible a particularly simple treatment which
is given in Section 6. The cross-section calculations of Sec-
tions 4 & 5 have been repeated with allowance for the LPM
eect which is shown to signicantly change early shower
development. The nal section gives a qualitative summary
of the cross-sections obtained and of the eect of magnetic
elds of the order of the critical eld on shower development.
2 RUTHERFORD SCATTERING
Small-angle Coulomb scattering at zero eld, in which the
transverse momentum component is a continuous variable,
is replaced by a Landau transition n n

with longitudinal
moment transfer q. Free atoms in high magnetic elds have
an axially symmetric but complex electron density distribu-
tion. But the departures from uniform electron density in
condensed matter are not large except for a small number
of inner orbitals and we therefore adopt the potential Ze

V
of a point charge within a spherical Wigner-Seitz unit cell
of uniform electron density. Its radius is
rWS = 2.15 10
10
Z
0.23
B
0.40
cm, (2)
obtained from the ion number density calculated by Medin
& Lai (2006) for atomic number Z. Calculational details are
summarized in the Appendix. The symmetry of

V indicates
that Landau functions in cylindrical polar coordinates are
the appropriate choice. The quantum number l denotes the
spatial degeneracy of these states and the procedure is to
average the transition rate over all l in the range 0 l lm.
Each state has a guiding centre radius

2l + 1rB, where rB
is the cyclotron radius. The incident ux is then c divided by
(2lm + 1)r
2
B
, and the cross-section at incident momentum
p 1, with no spin-ip, is the limit lm of,

nn
= r
2
B
_
Ze
2
hc
_
2 _
2lm + 1
lm + 1
_
lm

l=0

p + q, n

, l

V |p, n, l

2
. (3)
This expression is greatly simplied by the selection rule
(n l) = 0 arising from equation (A3) and from the ax-
ial symmetry of

V . Spin-ip cross-sections are smaller by a
factor of the order of (p

)
2
/4E
2
relative to unity and
as our interest is primarily in electrons with p mc they
have not been calculated. The range of initial-state quan-
tum numbers l should be so large that the cross-section is
independent of lm in the case of an isolated atom. We have
conrmed that, for B = Bc, this condition is satised ad-
equately for lm 30 at which value guiding centres are
at radii approaching the Wigner-Seitz cell radius. Numeri-
cal cross-sections
nn
are given in Table 1, for n, n

4.
They are in units of 1bn = 10
24
cm
2
and are for Z = 26
and longitudinal momentum p = 20, but are almost exactly
independent of p. Calculation to the lowest order in

V is
satisfactory here, as for the zero-eld small-angle Ruther-
ford formula, owing to the pole in the amplitude at zero
momentum transfer. Thus the cross-sections are linear in
Z
2
.
The cross-sections of Table 1 are valid in the relativistic
limit of p 1 so that comparison with the near-threshold
cross-sections found by Bussard (1980) and Langer (1981)
is not possible. But the standard Rutherford cross-section
at small angles can be expressed in terms of a continu-
ous variable n through the correspondence principle relation
p

2nB. It is dR/dn = 0.50Z

2
/n
2
B bn. The numerical
3
Table 1. Cross-sections in units of 10
24
cm
2
for Coulomb-
induced Landau transitions n n

, with n, n

4, are almost
exactly independent of longitudinal momentum at p mc but
have been calculated for electrons of p = 20 mc incident on an
isolated neutral atom of atomic number Z = 26. The magnetic
ux density in the rst column is in units of Bc = 4.41 10
13
G. The signicance of the n n transitions is considered in the
text and later in Section 6.
B n n

= 0 1 2 3 4
0.3 4 91 271 1132 6128 20886
3 182 841 5454 23132 6146
2 525 4483 25901 5467 1138
1 3032 29470 4491 843 272
0 34574 3036 526 182 91
1.0 4 27 80 334 2056 8729
3 54 248 1804 9543 2063
2 156 1460 10521 1810 336
1 968 11746 1463 249 80
0 13439 970 156 54 27
3.0 4 9 25 102 713 3862
3 17 77 621 4179 716
2 49 498 4554 623 103
1 328 5012 500 78 25
0 5628 329 49 17 9
values and dependences on n and B are quite close to those
given in the Table for an initial state with n = 0. The n n
cross-sections are well-dened by equation (3) even though
they are to an evanescent state whose asymptotic nal
form with q = 0 is indistinguishable from the unscattered.
These n = 0 transitions to virtual states of nite q ap-
pear in the calculation of second-order processes, in partic-
ular bremsstrahlung and pair creation cross-sections, and in
condensed matter are relevant to the Landau-Pomeranchuk-
Migdal eect considered in Section 6.
The cross-sections in Table 1 are typically two orders
of magnitude larger than those we obtain for Coulomb
bremsstrahlung in Section 5. Thus Rutherford scattering
excites higher Landau states whose decay is an important
source of photons, particularly in the later stages of shower
development. We shall refer to this in greater detail in Sec-
tion 7.
3 CYCLOTRON EMISSION
Transition rates for relativistic cyclotron emission are obvi-
ously relevant to the low-altitude plasma above polar caps
at B Bc and therefore have been calculated by a number
of authors (Herold, Ruder & Wunner 1982, Latal 1986, Bar-
ing, Gonthier & Harding 2005; see also the review of Harding
& Lai 2006). The principal reason for our calculation here,
which is limited to unpolarized electrons, is to investigate
the sequence of partial transition rates by which an electron
in Landau state n decays to n = 0. These determine the dis-
tribution of the transverse photon momentum k

= k sin ,
where is the photon angle with respect to B, and hence
electromagnetic shower development at B Bc.
In the rest frame of the initial electron, the relation
between momentum k and is,
k sin
2
= E0
_
E
2
0
sin
2
(p
2

p
2

)
_
1/2
, (4)
in which E0 =
_
1 + p
2

. The partial transition rate for

unpolarized electrons is

nn
=
1
h
_
1
1
d(cos )
k
2
2h
3
c
E

+ ck cos
2

1
2(lm + 1)

s,s

lm,l

l,l

=0

, n

, s

, k, |e

A|n, l, s

2
, (5)
summed over photon polarization states

and

, respec-
tively parallel with and perpendicular to k B, and over
nal electron spins. The radiation eld vector potential is

A
given by equation (A6) and the nal-state electron energy is
E

. The transition rate is expressed here as an average over

lm + 1 spatially degenerate initial states and we have ver-
ied that the computed rates are lm-independent, as they
should be, and have used the values lm = 0 and l

m
= 12.
The computed values, in units of 10
3
B, where B is the
cyclotron angular frequency, are given in Table 2 for n 4.
The summed transition rates given in the nal row of each
section are the sole results of this paper that can be com-
pared directly with previously published work. Agreement
with the rates shown in Fig. 3 of Harding & Lai (2006) is
satisfactory insofar as accurate comparison is possible. It is
interesting to see that there is eectively no selection rule
on n and that the partial rates display the n-dependence
noted by these authors in that, with increasing magnetic ux
density, there appears an enhanced rate for direct transitions
to the n = 0 ground state. The dependence on photon po-
larization is not given here. It is unremarkable: transitions
to

are approximately twice as strong as those to the

.
Photon angular distributions vary with n but are also un-
remarkable and are not given here. For n = 4 they have
a maximum at = /2; for n = 1 the maxima are near
= 0, .
The maximum photon transverse momentum occurs at
= /2 and is,
k
,max
=
_
1 + p
2

_
1 + p
2

. (6)
It is not possible to make a compact statement about the dis-
tributions of k

because the values of k

,max
are functions of
both n and B. The critical thresholds for single-photon mag-
netic pair creation are those at k

= 2 for electron-positron
Landau numbers n = n = 0 and k

= 1 +

1 + 2B
for n = 0 with n = 1. Rates immediately above the lower
threshold are partially inhibited by a selection rule but above
the higher threshold, transition rates for magnetic pair cre-
ation are so high that photon mean free paths are small
compared with the radiation length, which is the basic unit
of length for shower development. But Table 2 shows that,
except for the highest magnetic eld, the majority of tran-
sitions from n 4 have k

< 2 and so are not followed by

magnetic pair creation. Thus cyclotron emission, following
excitation by Rutherford scattering, is an important source
of photon track length in a shower.
4 P. B. Jones
Table 2. Rates for cyclotron emission with Landau transitions
n n

are given in units of 1.0 10

3

B
for unpolarized elec-
trons. The sums of the partial transition rates for n 4 are in
the nal row of each section.
B n

n = 1 2 3 4
0.3 3 1.33
2 1.33 0.68
1 1.24 0.59 0.38
0 0.76 0.39 0.26 0.20
sum 0.76 1.63 2.18 2.59
1.0 3 0.97
2 1.05 0.59
1 1.14 0.61 0.42
0 1.06 0.67 0.52 0.45
sum 1.06 1.81 2.18 2.43
3.0 3 0.61
2 0.68 0.39
1 0.79 0.43 0.30
0 0.94 0.62 0.50 0.44
sum 0.94 1.41 1.61 1.74
10.0 3 0.35
2 0.39 0.22
1 0.46 0.25 0.17
0 0.62 0.41 0.34 0.29
sum 0.62 0.87 0.98 1.03
4 PAIR CREATION
Coulomb pair creation in a uniform magnetic eld requires
a longitudinal momentum transfer q from the nucleus,
q = k(1 cos )
1
2p+
_
1 + p
2
+
_

1
2p
_
1 + p
2

_
, (7)
an expression valid for nal-state electron and positron mo-
menta p 1, and therefore diers from the zero-eld case
in being a function of both k and k

. The momentum trans-

fer is a minimum for p+ = p and p
+
= p

= 0 for which
conditions it falls to zero at k

= 2, the kinematic threshold

for magnetic pair production. The transition matrix element
to second order for an isolated atom is,
M
P
,
= p, n, l, s| Ze
2

V |p q, n, l, s
(E(p) E(p q) + i)
1
p q, n, l, s|e

A| p+, n+, l+, s+
+p+, n+, l+, s+|Ze
2

V |p+ q, n+, l+, s+

(E+(p+) E+(p+ q) + i)
1
p, n, l, s|e

A| p+ + q, n+, l+, s+ (8)
in which > 0 is innitesimal. The subscripts and refer
to the photon polarization with respect to k B. The spin
and spatial degeneracy states are labelled by s and l (see
equations A7 and A8). There are two further time-ordered
terms in the amplitude but they involve virtual pair creation
in the Coulomb eld. They have energy denominators that
are large, of the order of k rather than k
1
as in the above,
and for k 1 they can be neglected. The total cross-section
is then,

P
,
= Z
2
_
e
2
hc
_
3 _
h
mc
_
2
_
k
_
1+p
2
+
_
1+p
2

dE
2
EE+
pp+

nm

=0
lm

=0

M
P
,

2
(9)
for pair creation by a photon of momentum k, transverse
momentum k

, and of polarization

or . In this expres-
sion, the charges have been factored out from the matrix
element (8). There are a number of approximations here
that merit some consideration. Firstly, truncation of the
Landau quantum numbers to n 4 has been tested by
examination of the computed cross-section as a function of
nm for 0 nm 4. The basis for this truncation is that,
for xed k, the amplitude given by equation (8) is approx-
imately p
1

, also that matrix elements tend rapidly to

zero as they are constructed between states with increas-
ingly dierent quantum numbers and hence nodal surfaces.
There is evidence of reasonably satisfactory convergence, but
cross-sections obtained here should be seen as lower limits.
Truncation of the sum over the spatial degeneracy quantum
numbers to l 12 again produces fair convergence. But in
this case, there is a natural upper limit determined by the
radius of the Wigner-Seitz cell. Finally, the matrix elements
of the Coulomb potential Ze

V have been limited to those

with n = 0. This is satisfactory, for the purposes of the
present work, because these are typically much larger than
o-diagonal matrix elements, as shown by the cross-sections
of Table 1. But errors arising from this latter truncation are
not positive-denite.
Numerical values of the isolated-atom total cross-
sections are given in Table 3 for ranges of values of B, k and
k

and for both photon polarizations. They are for Z = 26

but being of second order in electron-photon coupling, are
linear in Z
2
. The dierential cross-sections are unremark-
able symmetric functions of the nal-state energies E, not
strongly peaked within the kinematic limits present in the
integral of equation (9). They are not given here primar-
ily because in the relevant case of condensed matter they
may be unpredictably distorted. The explanation for this is
a consequence of the partial order which may be present in
neutron-star surface condensed matter, and is deferred until
Section 7.
A problem in this Section is that there are no external
tests of the validity of expressions derived from equations (8)
and (9) or of their numerical evaluation. But we note that
the dierential cross-sections are symmetric in E and that
under the restriction nm = 0, the cross-section is
P

= 0
for photons polarized parallel with kB. This is consistent
with the fact that for this polarization, the matrix element
(A7) should be exactly zero for n+ = n = 0 (see Semionova
& Leahy 2001).
The cross-sections in Table 3 are quite slowly varying
functions of k and can be compared with the asymptotic
k 1 Bethe-Heitler value of
P
= 5.0 bn for unpolar-
ized photons incident on a screened Z = 26 nucleus in zero
magnetic eld. The correspondence principle indicates that
the Bethe-Heitler expression with modied screening should
be a fair approximation at moderate elds, of the order of
10
12
G, for which the Landau quantum number associated
with p

1 is large. It is dicult to obtain cross-sections

for elds between 10
12
G and those of the Table because
the truncations we have used are unlikely to be satisfac-
tory approximations. But as a function of decreasing B, the
5
Table 3. Numerical values of the total cross-section (in units
of 1bn = 10
24
cm
2
) for pair creation in the Coulomb eld of
an isolated atom with nuclear charge Z = 26 are given for both
photon polarizations in columns 3-5 of the upper sector of the
Table. The photon momentum k and transverse momentum k

are in units of mc. The magnetic ux density is in units of the

critical eld Bc = 4.4110
13
G. The three-row lower sector gives
cross-sections re-calculated in condensed matter with inclusion
of the Landau-Pomeranchuk-Migdal eect. We refer to Section 6
for the details of this latter calculation, which assumes a static
structure function S(q) = 1.
B k
P

,
P

= 0.5 1.0 1.5

0.3 40 2.09, 2.03 4.21, 4.20 4.39, 4.60
400 2.28, 2.21 4.55, 4.53 4.70, 4.91
4000 2.30, 2.23 4.59, 4.57 4.73, 4.95
1.0 40 0.61, 0.59 1.22, 1.23 2.00, 2.52
400 0.71, 0.69 1.41, 1.43 2.30, 2.87
4000 0.72, 0.70 1.44, 1.49 2.33, 2.90
40000 0.73, 0.70 1.44, 1.45 2.34, 2.91
3.0 40 0.19, 0.19 0.29, 0.31 0.44, 0.87
400 0.24, 0.24 0.37, 0.40 0.57, 1.11
4000 0.25, 0.25 0.38, 0.41 0.59, 1.13
10.0 40 0.06, 0.06 0.07, 0.08 0.08, 0.23
400 0.07, 0.07 0.08, 0.10 0.11, 0.35
4000 0.07, 0.07 0.09, 0.11 0.11, 0.36
1.0 40 0.61, 0.59 1.22, 1.23 1.99, 2.47
4000 0.21, 0.21 0.51, 0.50 0.78, 0.79
40000 0.02, 0.02 0.05, 0.05 0.08, 0.08
cross-sections in the Table are not inconsistent with smooth
convergence at intermediate elds toward the Bethe-Heitler
value. The most noticeable features of the Table are: (i)
the rapid increase of cross-section as k

approaches the n-
dependent thresholds for magnetic pair production at which
q and hence the energy denominators in equation (8) be-
come zero; and (ii), the rapid decrease of cross-sections at
B > Bc.
5 BREMSSTRAHLUNG
The emission of a photon of momentum k at an angle
with B, and a change of Landau quantum number n n

,
requires a longitudinal momentum transfer from the atom
of
q = k(1 cos ) +
1 + p
2

2p

1 + p
2

2p

, (10)
where the initial electron longitudinal momentum is p mc
and the nal momentum is p

= p + q k cos .
The transition matrix element, to second order for an
isolated atom, is
M
B
,
= p

, k, n

, l

, s

|e

A|p + q, n, l, s
(E(p) E(p + q) + i)
1
p + q, n, l, s| Ze
2

V |p, n, l, s +
p

, n

, l

, s

| Ze
2

V |p

q, n

, l

, s

(E(p

) E(p

q) + i)
1
p

q, k, n

, l

, s

|e

A|p, n, l, s, (11)
in which the subscripts and again refer to the pho-
ton polarization with respect to k B. As in the previous
Section, and for the same reason, we have neglected two
time-ordered terms in the amplitude that involve virtual
pair creation in the Coulomb eld. The existence of single-
photon magnetic pair creation thresholds implies that the
bremsstrahlung cross-section in a high magnetic eld is best
expressed in terms of the independent variables k and k

.
The cross-section dened by the energy-loss rate (equivalent
to the radiation length) is then given by an integration over
photon energy,

Rad
=
_
km
0
dk
k
p
d
B
dk
, (12)
in which the dierential bremsstrahlung cross section for
an electron in Landau state n, summed over both photon
polarization states, is,
d
B
dk
= (2lm + 1)r
2
B
Z
2
_
e
2
hc
_
3 _
1 + n0
4
_
_
k

k
dk

lm

l,l

=0

s,s

nm

=0

M
B

2
, (13)
The summations have been truncated to nm = 4 and
lm = 12, as in the case of pair creation, but this approxi-
mation is much less satisfactory for bremsstrahlung and has
lead us to omit values for B < Bc. Bearing in mind the
importance of the variable k

, the cross-section
Rad
given
by equations (12) and (13) has been divided into two com-
ponents:
Rad
1
is for the interval 0 < k

< 2.5 and

Rad
2
for 2.5 < k

< 5.0 mc. The computed values in units of

10
24
cm
2
, for electrons in an initial n = 0 Landau state
incident on isolated atoms are given in the upper sector of
Table 4. They are summed over both photon polarizations,
but we note that the

cross-section is typically an order

of magnitude larger than that for

. The computed values

are almost completely independent of p and there is little
cross-section outside the intervals of k

considered. The to-

tal cross-section is simply dened here as the sum of
Rad
1,2
.
Cross-sections for initial Landau states n > 0 are of similar
magnitude but are not given here. These initial states are,
of course, naturally unstable against photon decay, which
appears as a zero in the denominators of equation (11) and
is discussed further in Section 7.
6 THE LANDAU-POMERANCHUK-MIGDAL
EFFECT
In both this and previous work, we have adopted expressions
for the zero-pressure ion number density,
N = 2.5 10
28
Z
0.7
B
1.2
cm
3
, (14)
and for the interatomic separation in a linear chain, parallel
with B,
as 1.5 10
10
Z
1/2
B
1/2
cm, (15)
conveniently summarizing the calculated values of Medin &
Lai (2006). We anticipate that the bremsstrahlung and pair
creation cross-sections obtained in previous Sections for iso-
lated atoms will be considerably modied in the condensed
6 P. B. Jones
Table 4. Numerical values of the cross-section equivalent to the
radiation length, dened by equation (12), are given in the up-
per sector of the Table for electrons in the n = 0 Landau state
incident on isolated atoms of Z = 26. The values of electron mo-
mentum p are in units of mc, and B is in units of the critical
eld Bc = 4.41 10
13
G. Columns 3 and 4 are partial cross-
sections dened for the photon transverse momentum intervals
0 < k

< 2.5 and 2.5 < k

< 5.0 mc, respectively. The total

cross-section in column 5 is dened here as the sum of these par-
tial cross-sections. The 15 rows in the lower sector of the table
give the above cross-sections re-calculated in condensed matter
assuming a static structure function S(q) = 1 but with inclusion
of the Landau-Pomeranchuk-Migdal eect. We refer to Section 6
for further details.
p B
Rad
1

Rad
2

Rad
200 1.0 4.83 1.85 6.68
3.0 0.77 1.44 2.21
10.0 0.08 0.25 0.33
2000 1.0 4.82 1.85 6.67
3.0 0.77 1.44 2.21
10.0 0.08 0.25 0.33
20000 1.0 4.82 1.85 6.67
3.0 0.77 1.44 2.21
10.0 0.08 0.25 0.33
200 1.0 22.83 1.87 24.70
3.0 9.94 1.63 11.54
10.0 3.92 0.76 4.68
2000 1.0 20.04 2.41 22.45
3.0 6.95 3.59 10.54
10.0 1.75 1.88 3.63
20000 1.0 5.35 2.15 7.50
3.0 1.32 2.66 3.98
10.0 0.22 0.78 1.00
60000 1.0 1.52 0.95 2.47
3.0 0.32 1.05 1.37
10.0 0.05 0.25 0.30
200000 1.0 0.23 0.21 0.44
3.0 0.04 0.21 0.25
10.0 0.01 0.04 0.05
matter of the neutron-star surface. In particular, it is essen-
tial to consider the Landau-Pomeranchuk-Migdal (LPM) ef-
fect, whose existence at zero eld has been established with-
out doubt by experiments with laboratory electron beams of
several hundred GeV (Hansen et al 2004). The essence of the
eect is that, for example in bremsstrahlung and at high en-
ergies, the isolated-atom longitudinal coherence length (of
the order of q
1
) can be so long that the coherence that
would be present in the case of an isolated atom is par-
tially removed by multiple scattering of the initial or nal-
state electron within the medium. We refer to the review by
Klein (1999) for further details, and for a discussion of other
processes in condensed matter, such as the dielectric eect,
that are not relevant here. In the present Section, we exam-
ine the LPM eect at high magnetic elds and nd that the
one-dimensional nature of kinematics at B Bc allows a
particularly simple description.
We begin by considering the eect of scattering of an
electron with initial longitudinal momentum p mc by a
(dimensionless) potential U
i
= Ze
2

V /hc whose origin is

at z = zi. We consider transitions with no change of Landau
quantum number so that the nal state is given by the Green
function,
G = i(t t

>0
n(r

n
(r

)e
ip

(zz

)iE

(tt

)
, (16)
constructed from the Landau functions given by equation
(A3). The truncation to n = 0 is possible because, as we
found in Section 2, the n = 0 transition rates are at least
an order of magnitude smaller. It also excludes spin-ip and
back-scattering (p

< 0) which are both negligible in the

present context. The scattered state is
_
d
3
r

dt

G(r, t; r

, t

)U
i
n(r

)e
ipz

iEt

=
_

dq
2
U
i
q
q + i
e
iq(zz
i
)
_
n(r

)e
ipziEt
_
, (17)
in which U
i
q
is the matrix element of U
i
between states p and
p

= p +q, and > 0 is innitesimal. It is reached by intro-

ducing the integral representation of the Heaviside function
, followed by integration over r

and t

. Equation (17) shows

that the electron has components of momentum p + q and
that these are signicant at distances zzi < q
1
. Either by
performing the integral, or taking the limit z , we can
see that the prefactor outside the brackets evolves to an am-
plitude U
i
0
, as expected. Apart from the forward-scattered
component of the wave-function given by equation (17), ux
conservation shows that there is also an undisturbed compo-
nent of amplitude (1 |U
i
0
|
2
)
1/2
. Equation (17), giving the
forward-scattered state, is no more than a concrete expres-
sion of the uncertainty principle, but it shows that forward
scattering by a sequence of atoms, though unobservable in
the motion of the electron itself, has the eect of broadening
the distribution of q and so reducing the small-q amplitudes
that mediate most of the bremsstrahlung cross-section at
high energies.
The simplicity of equation (17) enables us to nd the
Green function for an electron undergoing a sequence of
forward scatters and so provides an easy way to calculate
bremsstrahlung cross sections including the LPM eect. We
begin by noting that the distribution in the variable q in
equation (17) is unchanged in form by multiple forward scat-
tering as is shown by the amplitude for a second and suc-
cessive scattering at zj,
_

du
2
_

dq

2
e
iu(zz
j
)
u + i
U
j
u
e
iq

(zz
i
)
q

+ i
U
i
q
=
i
_

dq
2
U
j
q
U
i
0
e
iq(zz
j
)
q + i
, (18)
in which u = q q

. We assume that the properties of U

allow the limits of integration to be extended to with
completion of the contour in the upper half of the complex-q

plane. We see that the amplitude has the spatial dependence

z zj of the last scattering.
Consequently, the amplitude for an electron propagat-
ing without back-scattering or n = 0 transitions in a long
interval of z must be a superposition of q-distributions, each
of the form of equation (17), with components which are
necessarily identical except that their individual amplitudes
decay exponentially as functions of z zj, reecting a re-
duction factor (1|U0|
2
)
1/2
for each atom traversed. This
exponential representation is satisfactory provided the mean
7
square value of |U
j
0
|
2
, satises the condition |U0|
2
1, as
it does here. Therefore, we can nd the Green function that
couples the electron to a nal state of total longitudinal mo-
mentum K

. It is given by,

j
_

z
j
dze
iK

z
_

dq
2
U
j
q
q + i
e
ipz+iq(zz
j
)(zz
j
)/2
=

j
_
1
p K

+ i/2
_
U
j
0
e
i(pK

)z
j
, (19)
in which = a/|U0|
2
, where a is the mean interatomic
separation zj zj1 and we have integrated rst over z and
then over q by completion of the contour in the semi-innite
upper half of the complex q-plane. The right-hand side of
equation (19) contains, within brackets, the momentum-
space Green function for an electron propagating without
back-scattering, spin-ip or n = 0 transitions. Forming the
square modulus of the remaining term (and neglecting any
inhomogeneity in the U
j
0
) yields the (one-dimensional) static
structure function for a linear chain of Na atoms,
S(q) =
1
Na

Na

j=1
e
iqz
j

, (20)
which contains, as a function of q = K

p, the dependence
of the transition rate on the condensed-matter structure.
Allowance for transitions with change of Landau quantum
number should not be dicult but is not essential because,
as can be seen from the Table 1 cross-sections, their mean
free path is considerably longer than .
Pair production and bremsstrahlung cross-sections in-
cluding the LPM eect are obtained immediately, assuming
a random distribution of atoms dened by a structure func-
tion S(q) = 1, simply by substituting this Green function
for the energy denominators contained in equations (8) or
(11). We have done so and have repeated the calculations
by which the Table 3 and 4 cross-sections were obtained.
The revised cross-sections given in the lower sectors of these
Tables are for Z = 26 and require some comment.
In the case of pair production, the re-calculated cross-
sections are given in Table 3 just for B = Bc and show
that, owing to the very high matter density ( 2 10
5
g
cm
3
) the LPM eect becomes noticeable at k = 4000 and
reduces cross-sections by more than an order of magnitude
at k = 40000. However, its eect on bremsstrahlung is less
simple. The re-calculated cross-sections in Table 4 show that
there is an initial increase at low values of p. The explanation
for this is as follows. For xed values of the kinematic vari-
ables and quantum numbers, there is a partial cancellation
of the two terms in equation (11) which arises because the
energy denominators are almost exactly equal in magnitude
( q) but have opposite signs. The presence of
1
in the
modied Green function creates a further amplitude term
in which this cancellation is not present, so producing the
cross-section increase. But with increasing p, values of q de-
cline and eventually the
1
component of the denominator
becomes dominant. Its eect becomes marked for p > 20000
mc.
The presence of the LPM eect removes the simple Z
2
scaling of cross-sections, but equations (14) and (15) allow
us to see that Z

, with 1.77 < < 1.5. Thus for

a xed magnetic eld, cross-sections become asymptotically
dependent on Z
2+2
in the high-energy limit.
But the possibility that zero-pressure neutron-star mat-
ter has some degree of one-dimensional order must also be
considered. Formation of one-dimensional chains of atoms of
homogeneous Z, oriented parallel with the magnetic ux, is
energetically favoured at B Bc (see Medin & Lai 2006)
and such structure may be present at the surface to the
extent that it survives the formation of the approximately
Poisson distribution of nuclear charge produced by photodis-
integration reactions in the reverse-electron electromagnetic
showers. Its eect can be studied simply by reference to the
structure factor given by equation (21) neglecting, for the
moment, any inhomogeneity in atomic number. Consider,
for example, the structure function for a nite linear chain
of Na homogeneous atoms with spacing zj zj1 = as,
S(q) =
1
Na
_
sin(Naqas/2)
sin(qas/2)
_
2
. (21)
This has maxima, S(q) = Na, when q coincides with an in-
tegral multiple of the basic reciprocal-lattice wavenumber
g = 2/as, but has zeros at |q g| = g/Na. Its eect is
primarily to signicantly distort spectral shapes given by
isolated-atom dierential cross-sections. Unfortunately, it is
not easy to be specic about changes in total cross-sections
because they are dependent on the (unknown) structure
function and on all other parameters, but with a change
of variable in a particular dierential cross-section to q we
can note that the modied total cross-section is,

S
=
_
S(q)
d
dq
dq, (22)
taken over the complete interval of q for the process, and
that this diers from the S = 1 cross-section by a factor of
order unity, not Na.
7 CONCLUSIONS
The cross-sections obtained here are those that, with Comp-
ton scattering, determine electromagnetic shower develop-
ment. For magnetic ux densities of the order of 10
12
G,
Landau quantum numbers n 10
2
are associated with
transverse momenta p

mc. Thus, on the basis of the cor-

respondence principle, zero-eld expressions are used with
some degree of condence. But in many neutron stars, elds
more than two orders of magnitude larger have been in-
ferred from the observed spin-down rate. For this reason,
it is necessary to be aware of any eld-dependence of these
cross-sections that would qualitatively change the nature of
shower development. In this respect, the processes that ap-
pear not to have been considered previously are those of
Sections 2, 4 and 5. Of these, Rutherford scattering mer-
its little comment except to note that its cross-section de-
creases with increasing magnetic eld in the region B Bc.
Pair creation is a little more complex because the ampli-
tude (equation 8) becomes singular at the threshold values
of k

for magnetic pair creation, beyond which it becomes

no more than a Coulomb eld-dependent correction to that
process. Thus we are concerned with the cross-section for k

below the rst two thresholds discussed following equation

8 P. B. Jones
(6) in Section 3. This is the explanation for the intervals of
k

adopted for columns 3-5 of Table 3.

Bremsstrahlung at B Bc is an even more complex
process. It is of second order in the coupling of the electron
with the electromagnetic eld, as at zero eld. But there
are two signicant dierences. The Table 4 cross-sections
are for an electron initially in the n = 0 Landau ground
state, but in the development of a shower in condensed
matter, higher-n states are certainly populated owing to
the very large Rutherford cross-sections listed in Table 1.
Secondly, a pole in the amplitude (equation 11) for such
initial states represents the possibility of real intermediate
states of n

1. The process we refer to as bremsstrahlung

then consists of two separate rst-order processes; Ruther-
ford scattering followed by cyclotron emission with the nat-
ural widths given in Table 2. The initial electron momentum
determines whether or not this is more important than the
second-order process. An estimate of the critical value is
easily made from the cyclotron emission rates in Table 2
which are all of the order of 10
3
B. Although the mean
free path lRu for Rutherford scattering is always small com-
pared with the radiation length l
Rad
which is characteristic
of the second-order bremsstrahlung process, time dilation
can extend the mean ight path lce for natural decay by
cyclotron emission to lengths greater than l
Rad
. Then the
rst-order processes are the dominant source of photon pro-
duction provided the initial-state momentum is,
p
mc
<
l
Rad
c
n1

=0

nn
. (23)
From the transition rates given in Table 2 and the values
of
Rad
given in Table 4, we can see that its value is not
strongly B-dependent. Specically, for B = Bc, Z = 26,

Rad
= 22 bn and l
Rad
= 1.8 10
5
cm with inclusion
of the LPM eect, so that for n = 4, the inequality (23)
becomes p < 1100 mc. At lower momenta, the process of
bremsstrahlung becomes simply a sequence of excitation
(or de-excitation) of higher-n Landau states by Ruther-
ford scattering and de-excitation by cyclotron emission. At
even lower momenta, the balance between excitation and
de-excitation changes and the typical values of the Landau
quantum number decrease until for lRu lce, that is, for
p
mc

lRu
c
10 (24)
they are n = 1. With the cross-section 01 from Table 1,
we nd lRu = 4.0 10
7
cm. The transition rate 10 =
1.06 10
3
B gives p 11 mc.
The above summary of the properties of the
bremsstrahlung and pair creation processes at high magnetic
elds serves to show qualitatively how the characteristics of
electromagnetic showers dier from the zero-eld case. In
the initial stages of an electron-initiated shower of about
10
3
GeV, the Landau-Pomeranchuk-Migdal (LPM) eect is
far more eective in reducing cross-sections owing to the
very high density of the condensed matter at the neutron
star surface. Therefore, shower development is displaced in-
ward until typical electron or photon energies are reduced
to the extent that the eect becomes unimportant. At this
stage, electron energy loss becomes rapid, occurring within
lengths much shorter than l
Rad
. Photon mean free paths are
dependent on k

, but below the lowest threshold for mag-

netic pair creation at k

< 2 they are a little larger than the

value of given by the zero-eld Bethe-Heitler cross-section.
The principal interest in shower development here is in
proton production by decay of the nuclear giant dipole res-
onance (GDR). Cross-sections for its electro-production are
small so that the important shower property is the distribu-
tion of total photon track length in the neighbourhood of the
resonant momentum, k 41 mc. An estimate of this, made
under the assumptions that bremsstrahlung occurs through
the sequence of rst-order processes decribed above and
that the Bethe-Heitler pair creation cross-section is valid,
was given by Jones (2010). We have found that shower de-
velopment at B Bc is more complex than was assumed
in that work, the important question being the fraction of
photon production in the GDR region that has transverse
momentum below the magnetic pair creation thresholds at
k

= 2 for Landau quantum numbers n+ = n = 0 and

k

= 1 +

1 + 2B for n = 0 with n = 1.
It is signicant that cyclotron decay, which is dominant
in the later stages of shower development, always leads to a
decrease in Landau quantum number, so bringing about a
convergence of the shower toward small k

. The only quali-

cation here is the tendency, noted by Harding & Lai (2006)
and conrmed in Table 2, for partial transition rates direct
to the ground state n = 0 to be large at B Bc. Complete
and quantitative calculations of shower development would
require an expansion of Tables 1, 2 and 4 to much larger
initial-electron values of the Landau quantum number, and
a Monte Carlo calculation using them would be a substan-
tial undertaking which has not been attempted here. Qual-
itatively, we can conclude that the photon track length dis-
tribution assumed in previous work is valid at B Bc. For
very high elds at B Bc, the pair creation cross-sections
become small and the photon mean free path at transverse
momenta below magnetic pair creation thresholds are then
determined by Compton scattering. Therefore, we expect
that the eect of the large cyclotron partial transition rates
to n = 0 will begin to reduce total photon track length in
the GDR region of momentum at B Bc.
REFERENCES
Baring M. G., Gonthier P. L., Harding A. K., 2005, ApJ,
630, 430
Bussard R. W., 1980, ApJ, 237, 970
Hansen H. D., Uggerhoj U. I., Biino C., Bellestrero S., Man-
giarotti A., Sona P., Ketel T. J., Vilakazi Z. Z., 2004, Phys.
Rev. D, 69, 032001
Harding A. K., Lai D., 2006, Rep. Progr. Phys., 69, 2631
Herold H., Ruder H., Wunner G., 1982, A&A, 115, 90
Johnson M. H., Lippmann B. A., 1949, Phys. Rev., 76, 828
Jones P. B., 2010, MNRAS, 401, 513
Klein S., 1999, Rev. Modern Phys., 71, 1501
Langer S. H., 1981, Phys. Rev. D, 23, 328
Latal H. G., 1986, ApJ, 309, 372
Lauer J., Herold H., Ruder H., Wunner G., 1983, J. Phys.
B, At. Mol. Phys., 16, 3673
Lieu R., 1981, Ap. & Sp. Sci., 80, 157
Medin Z., Lai D., 2006, Phys. Rev. A, 74, 062508
Semionova L., Leahy D., 2001, A&A, 373, 272
9
APPENDIX A: SOLUTIONS OF THE DIRAC
EQUATION
The solutions used here are the spinors obtained by Johnson
& Lippmann (1949) for a free electron in a uniform magnetic
eld with the gauge A = (Br)/2. As these authors noted,
they can be easily expressed in terms of any complete set of
eigenfunctions of the non-relativistic Schrodinger Hamilto-
nian, whether in cartesian or polar coordinates. The Dirac
equation is,
( + m) = E, (A1)
in which = p + eA with e > 0 and c = 1. The solutions
for E > 0 and for Landau quantum number n 0 are,
1,n =
_

_
0
(m + E)n
p

n1
pn
_

_
1,n1 =
_

_
(m + E)n1
0
pn1
p

n
_

_
each multiplied by the common normalization factor G =
1/
_
2E(E + m). The longitudinal momentum component
is p. These solutions have eigenvalues given by equation (1)
and are two-fold degenerate for n > 0, but they are not
identical with the true physical states owing to radiative cor-
rections that have been neglected. Natural line-widths, and
level shifts arising from the electron anomalous magnetic
moment and from radiative corrections, are all of similar
magnitude and give physical states that are quite dierent
from the above solutions (see Herold, Ruder & Wunner 1982;
also Baring, Gonthier & Harding 2005). The distinction
would be important if spin-polarized electron or positron
states were being considered. However, for the cross-sections
obtained here, which are for relativistic energies, lepton spin-
polarization is not of interest and the ne-structure in energy
levels can be safely neglected. In this paper, we always aver-
age or sum over lepton spin states so that we require only the
completeness of the Johnson-Lippmann states. These, with
the negative energy solutions given by the transformation
E E, have been used explicitly in Sections 2 - 5. The
corresponding representation of the Dirac matrices is easily
found from equations (47)-(54) of Johnson & Lippmann and
is
=
_
0
0
_
=
_
I 0
0 I
_
.
The subscript s = 1 labelling the spinors indicates that
for the case n = 0, or in the limit B 0, they are also
eigenstates of z with eigenvalues as indicated. The func-
tions n are a complete orthonormal set of eigenfunctions
of
2
labelled by the Landau quantum number n. Their rel-
ative phases are not arbitrary because they are required to
satisfy the conditions,
(x + iy)n1 = p

(n)n,
(x iy)n = p

(n)n1, (A2)
where p

(n) =

2nB and B is the magnetic ux density in

units of Bc. The atomic Coulomb potential has axial sym-
metry, reecting the shape of the electron orbitals in the
magnetic eld. (We neglect any small component with a lat-
tice symmetry.) Thus cylindrical polar coordinates r

, , z,
with B parallel with the z axis appear to be the optimum.
The functions are then,

nl
=
(i)
n
rB

n!l!

(ln)/2
e
/2
fn(l, )
e
i(ln)

2
e
ipz
, (A3)
conveniently expressed in terms of the variable = r
2

/2r
2
B
,
where rB = mc
2

B/e is the cyclotron radius. The Landau

quantum number can be any positive integer n = 0, 1, 2, ..
and all except the n = 0 state are two-fold degenerate. In
addition, the quantum number l represents the innite spa-
tial degeneracy of all Landau states whose guiding centres
can be at radii

2l + 1rB from the z-axis. The polynomials

fn are,
f0 = 1
f1 = l
f2 = l(l 1) 2l +
2
f3 = l(l 1)(l 2) 3l(l 1) + 3l
2

3
f4 = l(l 1)(l 2)(l 3) 4l(l 1)(l 2)
+ 6l(l 1)
2
4l
3
+
4
(A4)
in obvious sequence and they satisfy,
fn(l, ) = ()
nl
f
l
(n, ), (A5)
which is worth noting in relation to states with l < n.
We adopt gaussian cgs units so that the ne struc-
ture constant is e
2
/hc. Perturbations of the correct sign
are introduced into equation (A1) by modifying the energy-
momentum four-vector, A A+

A and E EZe
2

V , to
include the radiation eld

A and the atomic Coulomb po-
tential for nuclear charge Z. Thus the absorption of a photon
with wave-vector k and angular frequency is represented
by the matrix element of
e

A =
_
2hc
2

_
1/2
e e
ikr
, (A6)
in which is the polarization vector either parallel with, or
perpendicular to, the vector k B.
The creation of a pair in states specied by
|p, n, l, s, as in lines three and six of equation (8),
is given by the matrix element
_
r

dr

ddz

,n

(p, E, l)e

A
s
+
,n
+
(p+, E+, l+), (A7)
in which the positron energy is E+ > 0. Evaluation of this
proceeds by matrix multiplication and immediately yields,
for each set of values of s and of photon polarization, a
linear combination of terms each of which, apart from kine-
matic factors, contains an integral of the form,
I(k

, n, l, n+, l+) =
_
r

dr

de
ik

cos

,l

(r

)
n
+
,l
+
(r

).
These expressions, and the integrals contained in them, have
all been evaluated numerically. Therefore, we are unable
to give useful analytical expressions for any of the cross-
sections calculated in this work. Evaluation is the more te-
dious because the transverse dipole approximation is not ad-
equate so that there is no radiative transition selection rule
that is eective for physically signicant values of k

. But
10 P. B. Jones
matrix elements of the Coulomb potential satisfy (nl) = 0
strictly for isolated atoms; also in condensed matter to the
extent that components with the symmetry of the crystalline
eld are small.
This paper has been typeset from a T
E
X/ L
A
T
E
X le prepared
by the author.