REGIMES OF PULSAR PAIR FORMATION AND PARTICLE ENERGETICS

Alice K. Harding,
1
Alexander G. Muslimov,
2
and Bing Zhang
3
Received 2002 April 5; accepted 2002 May 8
ABSTRACT
We investigate the conditions required for the production of electron-positron pairs above a pulsar polar
cap (PC) and the inuence of pair production on the energetics of the primary particle acceleration. Assum-
ing space chargelimited ow acceleration including the inertial frame-dragging eect, we allow both one-
photon and two-photon pair production by either curvature radiation (CR) photons or photons resulting
from inverse Compton scattering of thermal photons from the PC by primary electrons. We nd that while
only the younger pulsars can produce pairs through CR, nearly all known radio pulsars are capable of pro-
ducing pairs through nonresonant inverse Compton scatterings. The eect of the neutron star equations of
state on the pair death lines is explored. We show that pair production is facilitated in more compact stars
and more massive stars. Therefore, accretion of mass by pulsars in binary systems may allow pair production
in most of the millisecond pulsar population. We also nd that two-photon pair production may be impor-
tant in millisecond pulsars if their surface temperatures are above three million kelvins. Pulsars that pro-
duce pairs through CR will have their primary acceleration limited by the eect of screening of the electric
eld. In this regime, the high-energy luminosity should follow an L
HE
/
_
EE
1,2
rot
dependence. The acceleration
voltage drop in pulsars that produce pairs only through inverse Compton emission will not be limited by elec-
tric eld screening. In this regime, the high-energy luminosity should followan L
HE
/
_
EE
rot
dependence. Thus,
older pulsars will have signicantly lower -ray luminosity.
Subject headings: gamma rays: theory pulsars: general radiation mechanisms: nonthermal
relativity stars: neutron
On-line material: color gure
1. INTRODUCTION
The acceleration of particles and the production of elec-
tron-positron pairs are widely considered to be two critical
elements necessary for generating radiation in rotation-
powered pulsars. In polar cap (PC) models (see, e.g., Arons
& Sharlemann 1979; Daugherty & Harding 1996), accelera-
tion occurs in the region of open eld near the magnetic
poles, and -rays fromcurvature and inverse Compton radi-
ation produce pairs primarily by one-photon pair produc-
tion in the strong magnetic eld. These pairs may screen the
accelerating electric eld through the trapping and reversal
of one sign of charge and may be required for the coherent
radio emission process. In outer gap accelerators (see, e.g.,
Cheng, Ho, & Ruderman 1986), a vacuum gap develops
along the null charge surface, and pairs are required to pro-
vide current ow through the gap, which can then operate
as a stable accelerator.
In this paper we discuss plausible regimes of pair forma-
tion above the pulsar PC, including the energetics of relativ-
istic particles and -rays that cause and accompany these
regimes. We treat the acceleration of particles within the
framework of an approach elaborated by Harding &Musli-
mov (1998, 2001, 2002, hereafter HM98, HM01, and
HM02, respectively), which combines rough analytic esti-
mates and simple practical formulae with detailed numerical
calculations. As an underlying PC acceleration model, we
employ the general relativistic version of a space charge
limited ow model developed earlier by Muslimov &
Tsygan (1992) and advanced in a number of important
aspects by HM98, HM01, and HM02. The main focus of
our present study is the physical condition for pair forma-
tion and howthis condition translates into a theoretical pair
death line for the observed radio pulsars. This paper is a log-
ical epilogue of our previous studies (see HM01; HM02),
where we calculated the parameters of pair formation fronts
(PFFs), including the ux of returning positrons, calculated
X-ray luminosities due to PC heating, estimated the lumi-
nosity of the primary beam, and revised derivation of pulsar
death lines. In our calculations of PFFs we employed the
standard mechanism of magnetic pair production by high-
energy photons, generated via curvature radiation (CR)
and/or inverse Compton scattering (ICS). In HM02 we cal-
culated, both analytically and numerically, the theoretical
pair death lines based on the above-mentioned regimes of
pair formation. However, in HM02 we presented the results
of our calculation of pulsar death lines only for a canonical
neutron star (NS) with the mass 1.4 M

and radius 10 km,

even though we pointed out that the eect of deviation of
NS mass and radius from their canonical values might be
important for our calculations.
In the present study we extend our previous analysis
(HM02) of pulsar death lines to explicitly incorporate the
eect of dierent NS mass and radius and, for the short-
period (millisecond) pulsars, to include the possibility of
two-photon pair formation. We must emphasize that the
eect of bigger NS mass is especially worth considering for
the millisecond pulsars, which are believed to be descend-
ants of accreting NSs in low-mass binary systems. It is
remarkable that our present study suggests that the millisec-
1
Laboratory of High Energy Astrophysics, NASA Goddard Space
Flight Center, Code 661, Greenbelt, MD20771.
2
ManTech International Corporation, Lexington Park, MD20653.
3
Department of Astronomy and Astrophysics, Pennsylvania State
University, 525 Davey Laboratory, University Park, PA16802.
The Astrophysical Journal, 576:366375, 2002 September 1
#2002. The American Astronomical Society. All rights reserved. Printed in U.S.A.
E
366
ond pulsars do favor the bigger NS masses, which seems to
be consistent with their standard evolutionary scenario.
This eect is associated with the dominance of the relativis-
tic frame-dragging term in the accelerating voltage drop,
which is a unique feature of the electrodynamic model of
Muslimov & Tsygan (1992). The frame-dragging compo-
nent of the electric potential (eld) is proportional to the
general relativistic parameter i cI,MR
2
, where c r
g
,R,
r
g
is the gravitational radius of an NS of mass M, and I and
R are the stellar moment of inertia and mass, respectively.
Another important aspect of our previous and present stud-
ies is the derivation of a theoretical relationship between the
pulsars -ray luminosity and its spin-down power/lumi-
nosity. In this paper we discuss such a relationship in the
context of the available and forthcoming -ray observations
of pulsars.
The paper is organized as follows: In x 2 we discuss the
determination of pair death lines in pulsars. First, we out-
line the basic denition and main assumptions behind the
death line concept (x 2.1). Second, we discuss the revised
analytic approach in the derivation of theoretical death lines
(x 2.2), and then, in x 2.3, we discuss our numerical calcula-
tion of death lines. In x 2.3.1 we calculate the CR and ICS
death lines for the NS models with a canonical mass 1.4 M

and with three dierent equations of state to illustrate the

eect of compactness on our death line calculations. In
xx 2.3.2 and 2.3.3 we focus on the death line calculations for
the short-period (millisecond) pulsars; we discuss how the
change in mass and radius of the underlying NS model
aects the pulsar death lines and present the results of our
numerical death line calculation (x 2.3.2); we incorporate
the eect of two-photon pair production for the PC temper-
atures 1 5 10
6
Kand present the corresponding numeri-
cal death lines (x 2.3.3). In xx 2.3.2 and 2.3.3 we illustrate
separately how the mass of an NS and two-photon pair pro-
duction, respectively, may aect the death lines for millisec-
ond pulsars. In x 3 we discuss energetics of the acceleration
of primary electrons and present the theoretical relationship
between the -ray luminosity and spin-down power of a pul-
sar. Finally, in x 4 we summarize our main results and dis-
cuss their most important implications for pulsars.
2. DEATH LINE DETERMINATION IN PULSARS
2.1. General Overview and Denitions
Since the very early attempts to relate the apparent
absence of radio pulsars with long periods in the P-
_
PP dia-
gram with the manifestation of the eect of electron-
positron pair formation as a condition for their operation,
it proved instructive to introduce the term death line to
separate the domain favoring pair formation from the
domain where it would be prohibited (see Sturrock 1971;
Ruderman &Sutherland 1975; Chen &Ruderman 1993 and
references therein). Soon it became almost a common prac-
tice for any theoretical study on radio pulsars to produce
the resulting death lines. Furthermore, some theories devel-
oped the idea that on the P-
_
PP map it is a death valley (see,
e.g., Chen & Ruderman 1993) rather than a death line that
separates radio-active from radio-quiet pulsars. It is impor-
tant that during the past decade the number of new radio
pulsars with a wide range of parameters dramatically
increased, which boosted the various pulsar population
studies. In light of the recent extensive radio pulsar surveys
(e.g., Manchester et al. 2001), pulsar population studies,
and multifrequency (from radio to -ray) pulsar observa-
tions, it seems timely to get back to the basic concept of a
pulsar death line.
The standard denition of a pulsar death line implies
that pulsar radio emission turns o if the energetics of
accelerated particles drops below the minimum required
for electron-positron pair production. Thus, the standard
denition of a death line implicitly assumes that pair for-
mation is a necessary condition for pulsar radio emission
and that pulsars become radio-quiet after crossing the
death lines during their evolution from left to right in the
P-
_
PP diagram. Obviously, this basic condition may not be
sucient (see, e.g., Hibschman & Arons 2001 for a most
recent study in which the sucient condition assumed
was that of pair production with high enough multiplicity
to completely screen the parallel electric eld), and func-
tioning of radio pulsars may imply far more complex
physical conditions (see Usov 2002 for a recent review).
However, numerous theoretical attempts to produce satis-
factory death lines implying even the basic necessary con-
dition meet certain challenges. For example, a number of
observed millisecond radio pulsars fall below most theo-
retical death lines. Also, many normal radio pulsars tend
to be below their death lines based on CR pair forma-
tion. For this reason, and to minimize the underlying
model assumptions, our previous (see HM02) and present
analyses of pulsar death lines are based on the minimal
requirement regarding pair formation. Note that in all
our studies we assume a centered-dipole magnetic eld of
an NS.
In this paper we illustrate how the spread of NS
masses and radii may aect the theoretical death lines for
ordinary and, most importantly, for millisecond radio
pulsars. Needless to say, compactness of an NS is an
important parameter in our calculations of particle accel-
eration (mostly because the accelerating electric eld is of
essentially general relativistic origin) and pair formation
(because of a bigger deection of photon trajectories in
the gravitational eld of a more compact NS). So, the
detailed analysis of the eects of stellar compactness on
the results of such calculations would be quite interesting
by itself. However, this eect is worthy of special consid-
eration in the case of millisecond pulsars. The main rea-
son is that the latter are believed to descend from
accreting NSs in low-mass binary systems and may repre-
sent postaccreting relatively massive NSs. As will be dem-
onstrated in x 2.3.2, the increase in NS mass considerably
facilitates pair formation in short-period pulsars, thus
pushing the corresponding death lines down to or below
the observed (P,
_
PP) values for millisecond pulsars.
We also consider the eect of two-photon pair produc-
tion, where -ray ICS photons interact with thermal X-ray
photons from the NS surface to produce an electron-posi-
tron pair. This process is of primary importance in outer-
gap models (see, e.g., Romani 1996; Zhang & Cheng 1997)
for pulsar high-energy emission. Zhang & Qiao (1998)
investigated the importance of this process above PCs of
normal pulsars, noting that unrealistically high PCtempera-
tures were required. As we will show in this paper, two-pho-
ton pairs can be produced more easily above PCs of
millisecond pulsars because the PC size is larger. We will
compute the death line for two-photon pairs as a function
of PCtemperature.
PULSAR PAIR FORMATION AND PARTICLE ENERGETICS 367
2.2. Analytic Death Lines
In this section we generalize the analytic expressions for
the death lines derived by HM02 to explicitly include the
eect of dierent mass, radius, and moment of inertia of an
NS.
To formulate the analytic death line condition, we need
to know the distribution of voltage drop in the pulsars PC
acceleration region. For a given distribution of voltage, we
can calculate the characteristic Lorentz factor of a primary
electron accelerating above the PC as a function of altitude
z and pulsar parameters B and P:
acc
(z, B, P). The acceler-
ated electron generates (CR and/or ICS) photons that may
pair produce if the condition for the corresponding pair for-
mation process (in most cases magnetic pair production) is
satised. In our previous papers (e.g., HM98; HM01;
HM02) we have demonstrated that use of the pair forma-
tion condition allows us to determine the pair formation
altitude as a function of pulsar parameters B and P alone.
Then the Lorentz factor of a primary electron evaluated at
the pair formation altitude determines a minimum Lorentz
factor an electron should achieve to generate a pair-produc-
ing photon,
min
(B, P). Thus, the death line condition would
require that the Lorentz factor of an accelerating primary
electron is equal to
min
. In our numerical calculation of
death lines we can easily keep track of the fulllment of this
requirement and plot the corresponding points in the P-
_
PP
diagram that constitute the death line (or rather death
curve). However, an analytic derivation of the death line
needs an additional independent relationship between the
pair formation altitude z and pulsar parameters B (or
_
PP)
and P. In our previous paper (HM02) we have demon-
strated that
acc
can be expressed as a function of B, P, and
an additional parameter, the eciency of converting pulsar
spin-down power into the luminosity of the primary beam,
f
prim

L
prim
_
EE
rot
. 1
where L
prim
is the luminosity of the primary electron beam
and
_
EE
rot
is the pulsar spin-down power [=
4
B
2
0
R
6
,6c
3
f 1
2
,
where B
0
,f 1 is the surface value of the magnetic eld
strength corrected for the general relativistic redshift; all
other quantities have their usual meaning and will be
dened below; see also HM02 for details].
For the typical radio pulsar parameters P and
_
PP and for
most obliquities, excluding the pure orthogonal case, the
dominant termin the expressions for the electrostatic poten-
tial and electric eld in the general relativistic version of the
space chargelimited ow model (Muslimov & Tsygan
1992) is proportional to parameter i. In this paper we shall
use the following general expression for the parameter i to
include its explicit dependence on NS radius and moment of
inertia
i
cI
MR
2
0.15
I
45
R
3
6
. 2
where c is the NS compactness parameter, I
45
I,10
45
g
cm
2
, and R
6
R,10
6
cm; M, R, and I are NS mass, radius,
and moment of inertia, respectively.
Then the explicit expression for
acc
can be written as (see
eq. [51] of HM02)

acc
10
7
f
prim
B
12
P
2
. 3
where
f
prim

"
ff
prim
i
0.15
R
5,2
6
. 4
and the z dependence of
acc
is implicit in f
prim
. Here
"
ff
prim
is
the eciency of converting spin-down luminosity into the
luminosity of the primary beam, calculated for a canonical
NS mass M 1.4 M

, radius R 10
6
cm, and moment of
inertia I 10
45
g cm
2
, and i
0.15
i,0.15, B
12
B
0
,10
12
G,
B
0
is the surface value of the magnetic eld strength, and P
is the pulsar spin period in seconds. Since the eciency f
prim
depends on the NS radius and moment of inertia, we simply
normalize it by
"
ff
prim
to make more illustrative the compari-
son with our calculations performed for a canonical NS
model. Note that in equation (4) the parameter f
prim
is
assumed to be independent of B and P but scales with i and
R
6
in the same way the
acc
(or the corresponding accelera-
tion potential drop; see eq. [49] in HM01) does.
A major advantage of the above formula for
acc
is that it
does not discriminate between unsaturated and saturated
regimes of acceleration of primary electrons (see HM02). It
implies that the Lorentz factor of an accelerating electron is
merely proportional to the maximum voltage drop above
PC, with the coecient of proportionality being the bulk
eciency of the pulsar accelerator f
prim
.
Nowwe can write the pair formation condition as

acc
!
min
. 5
where
acc
is given by equation (3).
The expressions for
min
for dierent underlying mecha-
nisms for pair-producing photons are the same as derived
by HM02 but including the explicit dependence on R and i
and read as follows:
Curvature radiation.

CR
min
10
7
3.4R
5,14
6
i
1,7
0.15
P
1,14
B
1,7
12
. PdP
CR

.
1.2R
3,4
6
i
1,4
0.15
P
1,4
. PeP
CR

.
_
_
_
6
Resonant ICS.

R
min
10
6
0.9R
1,2
6
i
1,3
0.15
P
1,6
B
1
12
. PdP
R

.
0.2 R
5,4
6
i
1,2
0.15
P
3,4
B
1,2
12
. PeP
R

.
_
_
_
7
Nonresonant ICS.

NR
min
10
5
R
1,2
6
i
1,3
0.15
P
1,6
B
1,3
12
. PdP
NR

.
0.6R
5,4
6
i
1,2
0.15
P
3,4
. PeP
NR

.
_
_
_
8
where
P
CR

0.1B
4,9
12
. 9
P
R

0.1B
6,7
12
. 10
P
NR

0.4B
4,7
12
11
are the critical spin periods (see eqs. [4][6] in HM02) in the
criterion dening the unsaturated (upper row) and satu-
rated (lower row) regimes of primary electron acceleration.
368 HARDING, MUSLIMOV, & ZHANG Vol. 576
Now, let us use equations (3) and (5) to get explicit condi-
tions representing the death lines. Note that, following the
reasoning of HM02, in equation (5) we should evaluate
acc
at
"
ff
prim

"
ff
min
prim
, where
"
ff
min
prim
is the minimum pulsar eciency
needed for pair formation. It is this minimum or threshold
value of f
prim
that determines the pulsar death line condi-
tion. We must note that at the pulsar death line, in the case
of ICS, the f
min
prim
denes the voltage drop at the PFF, not the
nal energy of the primary beam (see x 3). For each of the
mechanisms of generation of pair-producing photons we
discuss in this paper, the resultant analytic death line (or
rather parameter space with allowed pair formation) in the
P-
_
PPdiagramreads as follows:
Curvature radiation.
log
_
PP !
21
8
log P
7
4
log
"
ff
min
prim
D
CR
I
R. I 14.6 . PdP
CR

.
5
2
log P 2 log
"
ff
min
prim
D
CR
II
R. I 15.4 . PeP
CR

.
_
_
_
12
where
D
CR
I
R. I 1.5log I
45
6.3 log R
6

and
D
CR
II
R. I 1.5log I
45
7.3 log R
6
.
Resonant ICS.
log
_
PP !
5
6
log P log
2
3
D
ICS
I
R. I 16.6 . PdP
R

.
2
3
log P
4
3
log
"
ff
min
prim

2
3
D
ICS
II
R. I 17.9 . PeP
R

.
_
_
_
13
Nonresonant ICS.
log
_
PP !
7
4
log P
3
2
log
"
ff
min
prim
D
ICS
I
R. I 18.6 . PdP
NR

.
3
2
log P 2 log
"
ff
min
prim
D
ICS
II
R. I 20.0 . PeP
NR

.
_
_
_
14
where
D
ICS
I
log I
45
7.5 log R
6

and
D
ICS
II
log I
45
10.5 log R
6
.
In the above death line conditions we used equation (2) for
i.
Equations (12)(14) dier from the similar expressions
presented in HM02 (see eqs. [52][54]) by the term D, which
takes into account the deviation of NS radius and moment
of inertia from the canonical values of 10
6
cm and 10
45
g
cm
2
, respectively. Thus, for canonical NS parameters, the
above expressions translate into equations (52)(54) of
HM02. One can see from expressions for D that the more
compact the NS is, the lower the death line moves.
The analytic expressions above for the ICS pair death
lines dier signicantly from those derived by Zhang,
Harding, & Muslimov (2000). The reasons for these dier-
ences were discussed in detail in HM02.
2.3. Numerical Death Lines
2.3.1. Eect of NS Equation of State
The details of the method we use to numerically compute
pair death lines can be found in HM02. Briey, we keep
track of the total distance, the sum of the acceleration
length, and the pair production attenuation length of either
CR or ICS-radiated photons since the primary electron is
accelerating. The minimum of this total distance is assumed
to determine the height of the PFF. As the value of surface
magnetic eld decreases for a given pulsar period, the PFF
moves to higher altitude. Performing this calculation for a
range of pulsar periods, we nd the value of surface mag-
netic eld below which a PFF cannot formwithin the pulsar
magnetosphere. This occurs because both the required
acceleration length and the pair attenuation length become
very large. The result is a line in P-B
0
space that we identify
as the death line for pair production by photons of a given
radiation type. In order to compare death lines for dierent
equations of state (EOSs) with the observed pulsar popula-
tion, we must transform the calculated lines to P-
_
PP space
using the magnetodipole spin-down relation
B
0

3c
3
I
_
PPP
2
2
R
6
_ _1,2
. 15
giving
_
PP 2.43 10
16
R
6
6
I
45
_ _
B
2
12
P
_ _
s s
1
. 16
In our numerical death line calculations we employ the
three most representative NS models standardly used in the
calculations of thermal evolution of NSs (see, e.g., Table 1
in Umeda et. al 1993 and references therein) plus a strange
star model (see, e.g., Glendenning 1997). The NS models
correspond to a star with the baryon mass 1.4 M

and dier-
ent radii and moments of inertia: R
6
1.6 and I
45
2.2
(Pandharipande-Pines-Smith 1976 model), R
6
1.1 and
I
45
1.2 (Friedman-Pandharipande 1981 model without
pion condensate), and R
6
0.8 and I
45
0.6 (Baym-
Pethick-Sutherland 1971 model). The strange star model
has a mass 1.4 M

, radius R
6
0.7, and moment of inertia
I
45
0.7. In Figure 1 we show the death line calculations
based on these models. Note that the stellar models were
produced for a nonrotating star. Thus, strictly speaking,
our calculations of death lines shown in Figure 1 are not
very accurate for the millisecond pulsars. However, as it will
be discussed in the next section, our death line calculations
based on a nonrotating NS (and perhaps strange star)
model may still be satisfactory even in the millisecond range.
We should also mention that the only purpose of our inclu-
sion of a strange star model is to demonstrate the eect of
extreme stellar compactness on our death line calculations.
The surface physics of a strange star may be signicantly
dierent from that of an NS, and in this paper we refrain
from any speculation on this issue. It was suggested, how-
ever, that some radio pulsars could well be strange stars
No. 1, 2002 PULSAR PAIR FORMATION AND PARTICLE ENERGETICS 369
rather than NSs (see, e.g., Xu, Qiao, &Zhang 1999; Kapoor
&Shukre 2001).
As was suggested by our analytic expressions in x 2.2,
EOSs with smaller radii will move the death line lower,
allowing a greater number of pulsars to produce pairs. Pair
production is thus facilitated in more compact stars with
bigger i values and having softer EOSs or even having
strange matter EOSs. We demonstrated this eect by
employing the Baym-Pethick-Sutherland (BPS) NS model
and the more or less typical strange star model available in
the literature. The death line corresponding to the strange
star model (having the largest compactness parameter) is
the lowest one of those shown in Figure 1. Note that the ICS
pair death lines are more strongly aected by a change in
NS radius than are the CR pair death lines. In this paper we
have computed ICS death lines for only one PCtemperature
of 10
6
K in order to compare the eect of EOSs. In HM02
we showed that PC temperature has only a small eect on
ICS pair death lines for normal pulsars and is much less sig-
nicant than the eect of dierent NS EOSs. For a PC tem-
perature of 5 10
6
K and a canonical NS model, the ICS
pair death line lies slightly below the BPS model death line
shown in Figure 1.
2.3.2. Eect of NS Mass for Death Lines in Millisecond Pulsars
In our calculations of death lines for pulsars with the
periods in the range of 0.0010.1 s we use rapidly rotat-
ing NS models produced by Friedman, Ipser, & Parker
(1986). In Figure 2 we present our calculated death lines
for NSs with the gravitational masses 1.26, 1.97, and 2.64
M

. This particular sequence of rotating NS models is

calculated by employing the Pandharipande-Smith 1975
EOS and corresponds to the NS spin period of %2 ms.
The NS radii and moments of inertia for this sequence
10
-3
10
-2
10
-1
10
0
10
1
P

(
s

s
-
1
)
10
-24
10
-22
10
-20
10
-18
10
-16
10
-14
10
-12
P (s)
= 0.7
CR, ICS Pair Front
S: ICS - Strange EOS
FP: ICS - R
6
=1.0, I
45
=1.0
PS: ICS - R
6
=1.61, I
45
=2.2
BPS: ICS - R
6
=0.79, I
45
=0.57
FP
PS
BPS
S
PS
Fig. 1.Pair death lines in the pulsar P-
_
PPdiagramfor CR(unlabeled curve) and for inverse Compton radiation for dierent NS EOSs (see text). The param-
eter 0,0
PC
, where 0
PC
R,c
1,2
, indicates the magnetic colatitude of the primary electron acceleration. Also shown are radio pulsars in the ATNF Pul-
sar Catalog (available at http://www.atnf.csiro.au/pulsar).
10
-3
10
-2
10
-1
P

(
s

s
-
1
)
10
-24
10
-23
10
-22
10
-21
10
-20
10
-19
10
-18
10
-17
10
-16
P (s)
ICS Pair Front
M /M = 1.26
1.97
2.64
Fig. 2.Pair death lines for inverse Compton radiation in the pulsar
P-
_
PP diagram of short-period pulsars from the ATNF Pulsar Catalog, for
rotating NS models having dierent masses (see text).
370 HARDING, MUSLIMOV, & ZHANG Vol. 576
are, respectively, R
6
1.59 and I
45
2.28; R
6
1.59 and
I
45
3.9; and R
6
1.47 and I
45
5.18. Note that we
used the same sequence of models (implying the NS spin
period %2 ms) to calculate the death lines for the whole
range of spin periods up to 0.1 s. In fact, for the spin
periods e10 ms, the rotating NS models practically con-
verge with the nonrotating models. Note also that for
these three models, the relative dierences between non-
rotating and rotating sequences in terms of, respectively,
the gravitational mass, radius, and moment of inertia are
3%, 10%, and 8% (for the 1.26 M

model), 1.5%, 5%,

and 5% (for the 1.97 M

model), and 6%, 8%, and 4%

(for the 2.64 M

model). The dierences between rotating

and nonrotating models of this magnitude are more or
less typical for other sequences of models based on a rea-
sonable EOS. Thus, the eect of rotation, by itself, is not
very important for the death line calculations, and we
can justify using this particular sequence of models for
our death line calculations for the period range under
consideration. What may actually be important for the
death line calculations in millisecond pulsars is the mass
of an NS. We nd that our numerical calculations of
death lines in millisecond pulsars favor more massive NS
models, which is consistent with the hypothesis that the
millisecond pulsars descend from accreting NSs in low-
mass binaries. Figure 2 illustrates the eect of NS mass
on the death line calculations for the millisecond pulsars.
It shows that the increase in NS (gravitational) mass by
0.60.7 M

moves the death line down by a factor of a

few or more (see also eq. [14], saturated case). Even
though the mass of 2.64 M

we use in our calculations

may seem to be rather large, the main result of our calcu-
lations shown in Figure 2 is that the increase in the NS
baryon mass by $30%60% may signicantly facilitate
the process of pair formation above the PC in a millisec-
ond pulsar. This eect may account for the fact that
many millisecond pulsars tend to scatter below the theo-
retical death lines. The fact that the millisecond pulsars
might be more massive NSs processed in binary systems
could naturally explain this eect.
2.3.3. Eect of Two-Photon Pair Formation
We also investigate the eect of two-photon pair produc-
tion on the pair death line. The process we consider is that
of ICS photons interacting with soft X-ray photons from
the hot PC, drawing from the same pool of thermal photons
that are responsible for creating the ICS photon spectrum.
In PCmodels, two-photon pair production has traditionally
not been considered important in comparison to one-
photon pair production. Zhang & Qiao (1998) noted that
two-photon pair production could be important in PCmod-
els if the temperature of the PC was high enough (>4 10
6
Kfor P 0.1 s). Zhang (2001) estimated the photon attenu-
ation length for two-photon pair production above a hot
PCof radius R
t
to be
l
2
4.7 10
5
T
3
6
gzc
1
cm . 17
where
gz 0.27 0.507l
c
0.237l
2
c
.
l
c

z

z
2
z
2
t
_ . 18
where z
t
R
t
,R, T
6
T,10
6
K, c
2
,3
ln0.117c,c, T,mc
2
, and c is the photon energy
in units of mc
2
. Near the NS surface l
c
$ 0, and at
threshold
c
th

2
1 l
c

. 19
where c $ 1,, and c $ 1 so that l
2
1.7 10
6
T
3
6
cm. For surface temperatures T
6
$ 1, the photon attenua-
tion length l
2
is much larger than the acceleration length
required for the electron to radiate an ICS photon above
threshold. Therefore, l
2
sets the distance to the two-photon
PFF. Since the soft photon density declines with height
above the surface on a scale roughly equal to R
t
, a reason-
able criterion for two-photon pair production is then
l
2
< R
t
. Since R
t
r
PC
R,c
1,2
R for a heated PC, this
condition becomes
T
6
e1.6
P
1 ms
_ _
1,6
R
1,2
6
. 20
For normal pulsars, T
6
4 5 is required, which is unrealis-
tically high, but for millisecond pulsars, the temperature
required for signicant two-photon pair production is in the
range detected for some millisecond pulsars. It is clear that
the advantage millisecond pulsars hold over normal pulsars
in the facilitation of two-photon pair production is a large
PC size, which allows both larger angles between the ICS
-rays and the thermal PC photons thus lowering the
threshold energy for producing a pair, and an increase in the
scale length over which the soft photon density decays. The
primary electrons therefore can reach the energies needed to
radiate photons at threshold in a shorter distance.
In order to compute the two-photon pair death lines
numerically, we need an expression for the rate of pair pro-
duction of a high-energy photon of energy c:
R
2
c. 0 c
_
dc
_
dl
ti
_
dc
s
o
2
wn
s
c
s
. l
s
1 l
ti
.
21
where l
ti
is the cosine of the polar angle between the propa-
gation direction of the two photons, c
s
and n
s
are the energy
and cosine of the polar angle of the soft photon, and o
2
(w)
is the cross section,
o
2
w
r
2
e
w
6
_
2w
4
2w
2
1
_ _
ln w

w
2
1
p
_ _
w w
2
1
_ _

w
2
1
p
_
. 22
in the center of momentumframe in terms of the variable,
w
cc
s
1 l
ti

2
_ _
1,2
. 23
The above cross section does not take into account the eect
of the strong magnetic eld near the NS surface. Although
these eects may be signicant in the highest pulsar elds,
the magnetic two-photon pair production cross section is
very complicated (Kozlenkov & Mitrofanov 1986), and
since the process will only be important for millisecond pul-
sars having low elds, we will not consider these eects here.
The above (eld-free) cross section may be simplied in two
No. 1, 2002 PULSAR PAIR FORMATION AND PARTICLE ENERGETICS 371
limits, near threshold and for large w (Svensson 1982):
o
2
w
r
2
e

w
2
1
p
. w 1 .
r
2
e
,w
6
2 ln2w 1 . w41 .
_
24
where r
e
is the classical electron radius. We choose the coor-
dinate system so that the z-axis is along the magnetic pole.
To simplify the geometry of the calculation, we assume that
the -ray travels along the positive z-axis and that the soft
photons are uniformly radiated over the PC. There is thus
azimuthal symmetry about the magnetic pole and the polar
angle l
ti
ranges from 0 to l
c
, where l
c
is given in equation
(18). The thermal photons from the PC are described by the
blackbody distribution,
n
s
c
s
1 l
c

8
`
3
C
c
2
s
expc
s
, 1
. 25
where `
C
is the electron Compton wavelength. Changing
variables from l
ti
to w and using the expressions for o
2
(w)
dened by equation (24), the expression for the rate in equa-
tion (21) becomes
R
2
c 16
2
r
2
e
c
c
2
_
1
0
dc
s
n
s
c
s

c
2
s
w
s
. 26
where
w
s

_
1
5
w
2
s
1
5,2

1
3
w
2
s
1
3,2

. w
s
1 .
2.39
2 ln 2w
s
w
s

1
w
s
_ _
. w
s
41 .
_

_
27
and
w
s
max 1.
cc
s
1 l
c

2
_ _
. 28
Equation (26) is then integrated numerically to obtain the
two-photon pair production and attenuation length.
As in the case of the one-photon PFF calculation, we
minimize the sum of the acceleration length and the pair
production attenuation length of ICS-radiated photons as
the primary electron is accelerating. Performing this cal-
culation for a range of pulsar periods, we nd the value
of surface magnetic eld below which a PFF cannot form
within the pulsar magnetosphere. Figure 3 shows the
computed pair death lines in P-
_
PP space that include the
possibility of two-photon pair production for dierent
values of the PC surface temperature. We display three
cases for illustration: (1) death lines for one-photon pairs
only, (2) death lines for two-photon pairs only, and (3)
death lines for one-photon and two-photon pairs. All
cases assume a canonical NS model with M 1.4 M

,
I
45
1, and R 10 km. It is apparent that two-photon
pair production is not important at all for any of the
known radio pulsar population unless the PC tempera-
ture T
6
e3. The position of the two-photon death line is
sensitively dependent on PC temperature for 3.0dT
6
d
5.0 and then saturates at about T
6
$ 5.0, reecting the
eect of the two-photon pair threshold. For T
6
d3.0, the
ICS photons never reach the pair threshold during the
particle acceleration. For 3.0dT
6
d5.0, the photons are
pair producing near the threshold where the cross section
10
-3
10
-2
10
-1
10
0
10
1
10
-25
10
-23
10
-21
10
-19
10
-17
10
-15
10
-13
10
-11
P (s)
ICS 1 and 2 Pair Fronts
1 and 2 pairs
P

(
s
/
s
)
4.0
T
6
= 3.0
5.0
2 pairs only
1 pairs only
Fig. 3.One-photon and two-photon pair death lines for inverse Compton radiation in the pulsar P-
_
PP diagram for dierent PC surface temperatures,
T
6
T,10
6
K, as labeled. Also shown are radio pulsars in the ATNFPulsar Catalog.
372 HARDING, MUSLIMOV, & ZHANG Vol. 576
is sharply rising, and for T
6
e5.0, the photons are pair
producing above the threshold where the cross section is
decreasing. The two-photon death lines curve upward to
become almost vertical with increasing P because for lon-
ger periods particles must accelerate to high altitudes to
reach the pair threshold where the thermal photon den-
sity is declining. Thus, as we had noted previously, two-
photon pair production is only important in short-period
and millisecond pulsars. Since young, short-period pul-
sars with high magnetic elds do not have detected
PC surface temperatures as high as T
6
$ 3, two-photon
pairs are eectively not important for any but millisecond
pulsars. The combined one-photon plus two-photon
death lines blend into the one-photon death lines since
one-photon pairs dominate at higher elds and longer
periods.
HM02 found that substantial PC heating by trapped
positrons returning from an ICS pair front can occur in
millisecond pulsars if PC temperatures exceed T
6
$ 1. In
order for millisecond pulsars to sustain these high tempera-
tures through PC heating, the heated area must be much
smaller than the area of the standard PC, which is
A
PC
R
2
R,c. This is in fact consistent with the non-
uniformheating distribution found by HM02. However, for
the PC temperatures T
6
3 needed for two-photon pair
production, the question of the stability of two-photon
PFFs must be addressed. Positrons returning from the PFF
will radiate ICS photons that can produce two-photon pairs
in a relatively small distance because the pair production
threshold (see eq. [19]) is much lower for head-on collisions.
Creation of enough two-photon pairs by the returning posi-
trons at high altitudes could disrupt the acceleration of the
primary electrons. Investigation of this eect will require
inclusion of full angular dependence of the pair production
rate and will be considered in a future paper.
3. ACCELERATION AND GAMMA-RAY LUMINOSITY
Establishing the regimes of pair formation above pulsar
PCs not only is important for understanding the behavior of
the radio emission but also allows us to predict regimes of
particle acceleration and thus high-energy emission since
the acceleration of the primary particles may be limited by
screening at a PFF. HM02 found that CR pairs are very
eective in screening the electric eld at the PFF, whereas
ICS pairs are less eective and may only screen the electric
eld above the PFF in some cases. They found that when
ICS screening does occur, it only screens the eld locally but
will not screen at higher altitudes. Thus, ICS pairs may
retard but do not ultimately limit acceleration of the pri-
mary electrons, which may then also produce CR pairs at
higher altitude. In fact, the luminosity of the thermal X-rays
from a hot PC detected from PSR B1929+10, a 3 Myr old
pulsar where a detectable cooling component is not
expected, would require and is consistent with heating by
positrons produced at a CR pair front (HM01) since the
heating by positrons produced only at an ICS pair front
would not be detectable (HM02).
The luminosity of the primary electron beam in the PC
pulsar model can be calculated as
L
prim
cc
_
j,
e
jdS . 29
where
j,
e
j
B
0
2ccj
3
f j
f 1
1 i 30
is the value of an electron charge density calculated at
cos % 1 (where is the pulsar obliquity), (z, , c) is the
electric potential, and
dS
R
3
cf j
j
3
d dc 31
is the element of a spherical surface cut by the last open eld
lines at the radial distance r (=Rj). Here c 1 r
g
,R
_ _
1,2
,
r
g
is the gravitational radius of an NS, c is the velocity of
light, z is the altitude above the PC in units of stellar radius,
is the magnetic colatitude of a eld line scaled by the
magnetic colatitude of the last open eld line, and c is the
magnetic azimuthal angle.
In our previous papers (e.g., HM02) we calculated L
prim
using in equation (29) the expression for the electric poten-
tial evaluated at the relatively smaller altitudes (both for the
unsaturated and saturated regimes of acceleration of pri-
maries) where the bulk of the CRpair formation and electric
eld screening occur. In the regime where CR pairs are cre-
ated, i.e., above the CR death line, the luminosity of the pri-
mary beam is therefore set by the CR pair front. HM02
derived the following expressions for the luminosity of the
primary electron beam based on the altitude of the CR pair
front:
L
CR
prim
10
16
ergs s
1
_ _
1,2
_
EE
1,2
rot
P
1,14
B
1,7
12
. PdP
CR

.
0.3P
1,4
. PeP
CR

.
_
_
_
32
In the case in which there are no pairs produced by CR (and
therefore no electric eld screening) the most appropriate
expression for the electric potential is (see eq. [24] in HM01
and eq. [13] in HM98)

j
cos %1

0
R
f 1c
1
2
_ _
i
3
2
z .
r
PC
R
5z < 1 .
1
2
1
1
j
3
_ _
. j 1 ! 1 .
_

_
33
where j r,R. This formula applies for the altitudes much
greater than the PC radius and corresponds to the saturated
and unscreened regime of acceleration of primaries. How-
ever, in some cases the acceleration of primary electrons
may be limited by CR losses, where the general equation
(29), which does not take into account the radiation reaction
of accelerating particles, may not be applicable.
After substituting equation (30) for |,
e
| and the above
expression for into equation (29) and performing integra-
tion over and c, we get
L
prim
j
j
cos %1

3
4
i1 i 1
1
j
3
_ _
_
EE
rot
. 34
Formally, in the above formula we should put j !1 to
calculate the maximum power in the primary beam. Thus,
for the maximum power of the primary beam, we can write
No. 1, 2002 PULSAR PAIR FORMATION AND PARTICLE ENERGETICS 373
(see eq. [76] in Mulsimov &Harding 1997)
L
prim. max

3
4
i1 i
_
EE
rot
. 35
To estimate the pulsar bolometric photon luminosity, it
is reasonable to assume that L

% 0.5L
prim. max
, where
L
prim, max
is given by equation (35). Then, using equation (2)
for i, we can write
L

% 0.05
I
45
R
3
6
1 0.15
I
45
R
3
6
_ _
_
EE
rot
. 36
Thus, the energetics of CR photons generated by accelerat-
ing electrons above the pulsar PC is proportional to the pul-
sar spin-down luminosity, and according to equation (36),
the maximum eciency of conversion of pulsar spin-down
power into the high-energy quanta may amount to 10%.
Note that this estimate of eciency implies that only half of
the power of accelerating electrons gets consumed by -ray
photons.
In Figure 4 we plot the predicted high-energy luminosity
as a function of spin-down luminosity. In the CR pair
regime, where equation (36) applies, we have plotted the
luminosity for the screened unsaturated case (see top of eq.
[32]). Belowthe CR death line,
_
EE
rot
d10
34
ergs s
1
, the lumi-
nosity for the unscreened case of equation (36) applies. Also
plotted are the luminosities of the pulsars with detected
high-energy emission and their predicted luminosities. The
detected high-energy pulsars are all above the CR death
line, although Geminga and PSR B105552 are just above
the line. The detected high-energy pulsars seem to follow
the predicted relationship L

/
_
EE
1,2
rot
. We predict that
this relationship will break to L

/
_
EE
rot
at E
rot
d10
34
ergs
s
1
, so that older pulsars will have lower predicted luminosi-
ties than what would be predicted by an extrapolation of the
trend seen in the younger pulsars. Nonthermal high-energy
emission has in fact not been detected from older nearby
pulsars such as PSR B1929+10 and PSR B0950+08,
although a thermal emission component has been detected
from PSR B1929+10 (Wang & Halpern 1997), which may
be due to PC heating (see HM01). EGRET upper limits for
pulsed emission from these pulsars (Thompson et al. 1994)
are plotted in Figure 4 and lie above the predicted luminosi-
ties, but not by much. GLAST should be able to detect -ray
emission from these older pulsars and test the predicted
L

/
_
EE
rot
dependence and its location. Note that the break
value of
_
EE
rot
in the L

_
EE
rot
dependence depicted in Figure 4
can be estimated by equating the screened expression for
L

% 0.5L
CR
prim

_
EE
rot
given by equation (32) and L
. max
%
0.5L
prim. max
, where L
prim, max
is given by equation (35),
5 10
15
_
EE
1,2
rot
P
1,4
0.04
_
EE
rot
. 37
which gives
_
EE
rot. break
1.4 10
34
P
1,7
B
2,7
12
ergs s
1
. 38
For P % 0.1 s and B
12
% 4, which is shown in Figure 4,
_
EE
rot. break
% 6.6 10
33
ergs s
1
. Because of the B
2,7
12
depend-
ence,
_
EE
rot. break
should generally occur at higher
_
EE
rot
values
for millisecond pulsars.
There are a number of millisecond pulsars that, according
to equation (32), should have observable high-energy emis-
sion. Several of these, including PSR J04374715, PSR
J0030+0451, PSRJ18242452, and PSRJ0218+4232, have
nonthermal pulsed X-ray emission components but have
not, with the possible exception of PSR J0218+4232 (see
Kuiper et al. 2000), been detected in the -ray band. It is
important to remember, however, that the predicted lumi-
nosity L

is a bolometric luminosity. The luminosity in a

particular band is very dependent on the actual emission
spectrum. The iF
i
spectra of the bright -ray pulsars all
peak in the -ray or hard X-ray bands with typical high-
energy turnovers around several GeV. The high-energy
spectrum of the millisecond pulsars may be quite dierent.
The bulk of the millisecond pulsars are near or below the
CR death line, so that their accelerating electric eld is
unscreened. The accelerating primary electrons will reach
an energy where CRlosses are compensated by the accelera-
tion energy gain (see also Luo, Shibata, &Melrose 2000),

crr
1.8 10
7
B
1,4
12
i
1,4
0.15
P
1,4
. 39
The CR emission spectrum of these electrons will be quite
hard (photon index
2
3
) and will not be cut o by magnetic
pair production at an energy of a fewGeVbut will extend to
the natural cuto of the CRspectrumat
c
cr

3
2
"h
mc
2

3
crr
,
c
2.3 10
5
B
12
i
0.15
P
_ _
3,4
. 40
where ,
c
is the radius of curvature of a dipole eld line. The
CR iF
i
spectrum of the millisecond pulsars is therefore
expected to peak at 50100 GeV energies. The numerical
model spectra for millisecond pulsars of Bulik, Dyks, &
Rudak (2000) have also shown this result. The high-energy
CR from millisecond pulsars therefore falls in an energy
10
30
10
31
10
32
10
33
10
34
10
35
10
36
10
37
10
38
10
30
10
31
10
32
10
33
10
34
10
35
10
36
H
i
g
h

E
n
e
r
g
y

L
u
m
i
n
o
s
i
t
y

(
e
r
g
/
s
)
L ~ E
rot
Spin-Down Luminosity (erg/s)
1
0
5
5
-
5
2
G
e
m
i
n
g
a
0
6
5
6
+
1
4
1
9
5
1
+
3
2
V
e
l
a
1
7
0
6
-
4
4
1
5
0
9
-
5
8
C
r
a
b
L
CR
1
0
4
6
-
5
8
L ~ 0.5 L
,max
J
0
2
1
8
+
4
2
3
2
B
1
9
2
9
+
1
0
B
0
9
5
0
+
0
8
J
1
4
5
6
-
6
8
Fig. 4.Predicted and observed high-energy luminosity vs. spin-down
luminosity. The solid curve is the theoretical prediction from rst expres-
sion in eqs. (32) and (36). The lled circles are the luminosities of the
detected -ray pulsars Thompson (2001) derived fromdetected uxes above
1 eV assuming a 1 sr solid angle. The upper limits are for more than
100 MeV from Thompson et al. (1994). The open triangles are predicted
luminosities for the detected pulsars. [See the electronic edition of the
Journal for a color version of this gure.]
374 HARDING, MUSLIMOV, & ZHANG Vol. 576
band that has been above that of satellite detectors like
EGRET and below that of air Cerenkov detectors. There
could be a second spectral component of almost comparable
total luminosity due to the synchrotron pair cascade from
ICS pairs, but the iF
i
spectrum of this component would
peak at much lower energies. There is also a question of
whether two-photon pairs could screen the accelerating eld
and limit the voltage to very low values. Our preliminary
investigations indicate that screening by two-photon pairs
occurs only for NS surface temperatures high enough
(T
6
e4) that two-photon pairs from returning positrons are
likely to disrupt the acceleration (see discussion at end of
x 2.3.3). This issue will be a subject for future, more detailed
investigation.
4. SUMMARY AND CONCLUSIONS
n this paper we have outlined the status of two basic
regimes of primary particle acceleration above a pulsar PC
in the general relativistic version of the space chargelimited
owmodel: the regime of acceleration with subsequent elec-
tron-positron pair formation by CR and screening of the
electric eld and the regime of unscreened acceleration with
ICS pairs and the emission of energetic CR photons. For
the pulsar physical parameters we discussed in this paper
such as the pulsar spin period, NS surface magnetic eld
strength, and PC temperature, the pair production may
involve both the one-photon (magnetic) and two-photon
(mostly -ray ICS photons on thermal X-ray photons)
mechanisms, with high-energy photons generated via CR
and ICS processes. In HM02 we began analyzing the onset
of pair formation and addressed how the corresponding
pair formation criterion transforms into the pulsar theoreti-
cal death lines. HM02 presented the death line calculations
for the canonical NS model (mass of 1.4 M

, radius of 10
km, and moment of inertia I 10
45
g cm
2
) and one-photon
pair formation mechanism only. Here we extended our
study of pulsar theoretical death lines to include the eect of
NS mass and radius and, for the millisecond pulsars with
relatively hot PCs [T $ 3 4 10
6
K], to incorporate the
mechanism of two-photon pair formation. In the present
study we demonstrate that the eect of NS mass is impor-
tant for the death line calculations in millisecond pulsars. In
fact, if the millisecond pulsars are more massive postaccret-
ing NSs spun up in low-mass binary systems, then the phase
space they occupy in the P-
_
PP diagram would be consistent
with our theoretical death lines calculated for NSs with
masses more than 1.4 M

.
For the regime of acceleration without CR pairs, we cal-
culate the -ray luminosity of the primary beam as a func-
tion of pulsar spin-down power. It is important that the
form of the theoretical L

-
_
EE
rot
dependence is determined by
the regime of primary acceleration. For example, for the
regime of acceleration accompanied by pair formation
capable of screening the accelerating eld, L

/
_
EE
1,2
rot
,
whereas for the unscreened regime of acceleration,
L

/
_
EE
rot
. We discuss our theoretical L

-
_
EE
rot
plots in the
light of currently available pulsar -ray data and predict
that for pulsars in the regime of unscreened acceleration
(with relatively low values of
_
EE
rot
), L

should turn down

fromthe
_
EE
1,2
rot
dependence (see Fig. 4).
The main conclusions of our study can be summarized in
the following way:
1. We revised the prescription for the derivation of pulsar
theoretical death lines to include the eect of variation (by
more than an order of magnitude) of pair formation altitude
with pulsar parameters Pand
_
PP.
2. Pulsar theoretical death lines are strongly aected by
the EOS of an NS, with the onset of pair formation facili-
tated in more compact NSs.
3. Theoretical death lines for millisecond pulsars pro-
duced for massive NS models are in a good agreement with
the empirical death line and with the hypothesis that milli-
second pulsars are descendants of accreting NS in low-mass
binaries.
4. The millisecond pulsars with relatively hot PCs [tem-
perature $3 4 10
6
K] may allow the occurrence of two-
photon pair production, which signicantly facilitates pair
formation and may also move down the theoretical death
lines in the millisecond range.
5. We derive a theoretical expression for L

and illustrate
that it scales simply as
_
EE
rot
or
_
EE
1,2
rot
, respectively, for the
unscreened regime of acceleration of the primary beam or
for the acceleration implying pair formation and screening
of the accelerating electric eld. We predict that the break in
L

-
_
EE
rot
dependence, attributed to the transition from the
regime of acceleration with pair screening to the unscreened
regime, might be seen in observational data.
We acknowledge support from the NASA Astrophysics
Theory Program.
REFERENCES
Arons, J., &Scharlemann, E. T. 1979, ApJ, 231, 854
Bulik, T., Dyks, J., &Rudak, B. 2000, MNRAS, 317, 97
Chen, K., &Ruderman, M. 1993, ApJ, 402, 264
Cheng, K. S., Ho, C., &Ruderman, M. A. 1986, ApJ, 300, 500
Daugherty, J. K., &Harding, A. K. 1996, ApJ, 458, 278
Friedman, J. L., Ipser, J. R., &Parker, L. 1986, ApJ, 304, 115
Glendenning, N. 1997, Compact Stars: Nuclear Physics, Particle Physics,
and General Relativity (Berlin: Springer)
Harding, A. K., &Muslimov, A. G. 1998, ApJ, 508, 328 (HM98)
. 2001, ApJ, 556, 987 (HM01)
. 2002, ApJ, 568, 862 (HM02)
Hibschman, J. A., &Arons, J. 2001, ApJ, 554, 624
Kapoor, R. S., &Shukre, C. S. 2001, A&A, 375, 405
Kozlenkov, A. A., & Mitrofanov, I. G. 1986, Soviet Phys.-JETP Lett., 64,
1173
Kuiper, L., Hermsen, W., Verbunt, F., Thompson, D. J., Stairs, I. H., Lyne,
A. G., Stickman, M. S., &Cusumano, G. 2000, A&A, 359, 615
Luo, Q., Shibata, S., &Melrose, D. B. 2000, MNRAS, 318, 943
Manchester, R. N., et al. 2001, MNRAS, 328, 17
Muslimov, A. G., &Harding, A. K. 1997, ApJ, 485, 735
Muslimov, A. G., &Tsygan, A. I. 1992, MNRAS, 255, 61
Romani, R. W. 1996, ApJ, 470, 469
Ruderman, M., &Sutherland, P. G. 1975, ApJ, 196, 51
Sturrock, P. 1971, ApJ, 164, 529
Svensson, R. 1982, ApJ, 258, 335
Thompson, D. J. 2001, in AIP Conf. Proc. 558, High-Energy Gamma-Ray
Astronomy, ed. F. A. Aharonian &H. J. Volk (NewYork: AIP), 103
Thompson, D. J., et al. 1994, ApJ, 436, 229
Umeda, H., Shibazaki, N., Nomoto, K., &Tsuruta, S. 1993, ApJ, 408, 186
Usov, V. V., 2002, in Neutron Stars, Pulsars, and Supernova Remnants, ed.
W. Becker, et al. (Munich: MPI), in press
Wang, F. Y.-H., &Halpern, J. P. 1997, ApJ, 482, L159
Xu, R. X., Qiao, G. J., &Zhang, B. 1999, ApJ, 522, L109
Zhang, B. 2001, ApJ, 562, L59
Zhang, B., Harding, A. K., &Muslimov, A. G. 2000, ApJ, 531, L135
Zhang, B., &Qiao, G. J. 1998, A&A, 338, 62
Zhang, L., &Cheng, K. S. 1997, ApJ, 487, 370
No. 1, 2002 PULSAR PAIR FORMATION AND PARTICLE ENERGETICS 375