The Gamma Ray Pulsar Population

M. A. McLaughlin1 & J. M. Cordes1

ABSTRACT
arXiv:astro-ph/9912409v2 21 Feb 2000

We apply a likelihood analysis to pulsar detections, pulsar upper limits,

and diffuse background measurements from the OSSE and EGRET instruments
on the Compton Gamma Ray Observatory to constrain the luminosity law for
gamma-ray pulsars and some properties of the gamma-ray pulsar population.
We find that the dependence of luminosity on spin period and dipole magnetic
field is much steeper at OSSE than at EGRET energies (50−200 keV and
>100 MeV, respectively), suggesting that different emission mechanisms are
responsible for low- and high-energy gamma-ray emission. Incorporating a
spin-down model and assuming a pulsar spatial distribution, we estimate the
fraction of the Galactic gamma-ray background due to unidentified pulsars and
find that pulsars may be an important component of the OSSE diffuse flux, but
are most likely not important at EGRET energies. Using measurements of the
diffuse background flux from these instruments, we are able to place constraints
on the braking index, initial spin period, and magnetic field of the Galactic
pulsar population, and are also able to constrain the pulsar birthrate to be
between 1/(25 yr) and 1/(500 yr). Our results are based on a large gamma-ray
beam, but they do not scale in a simple way with beam size. We estimate
that about 20 of the 169 unidentified EGRET sources are probably gamma-ray
pulsars. We use our model to predict the pulsar population that will be seen by
future gamma-ray instruments and estimate that GLAST will detect roughly
750 gamma-ray pulsars as steady sources, only 120 of which are currently known
radio pulsars.

1. Introduction

Since the launch of the Compton Gamma Ray Observatory (CGRO) in 1991, pulsed
gamma-rays have been detected from only 8 spin-driven pulsars. Because of the small
number of detected pulsars, there remain many unanswered questions in pulsar gamma-ray

1
Astronomy Department, Cornell University, Ithaca, NY 14853
 –2–

astronomy. The relationship between pulsar radio and gamma-ray emission is not clear.
Aside from the Crab pulsar, the shapes of radio and gamma-ray profiles are quite
different, with the peaks of the radio and gamma-ray profiles falling at different pulse
phases, suggesting that low- and high-energy emissions come from different regions of the
magnetosphere. In fact, Geminga, an extremely bright gamma-ray pulsar (GRP), is weak
or quiet in the radio band. Furthermore, the gamma-ray profiles are broader than the radio
profiles, suggesting that the beaming solid angle for gamma-rays is larger than that for
radio waves. Models for the GRP emission mechanism, such as the polar cap and outer
gap models, have been invoked to explain the features of GRPs, but no model has been
successful at explaining all of the observations.
Understanding GRPs may be the key to determining the nature of the unidentified
EGRET sources, perhaps the largest mystery remaining from the CGRO mission. Of
the 271 sources listed in the Third EGRET Catalog (Hartman et al. 1999), 169 are not
identified as either pulsars or blazars, the only known EGRET discrete sources. While some
of these sources are distributed isotropically and therefore expected to be AGN, 73 of the
169 unidentified sources are located within 10◦ of the Galactic plane. Since pulsars are the
only known Galactic EGRET sources, it is likely that these unidentified sources are pulsars
or an as-yet-unrecognized source population. Modeling the GRP population is important
for estimating how many of the unidentified sources we expect to be pulsars, which of the
unidentified sources are most likely to be pulsars, and how we should go about detecting
their pulsed emission.
Understanding GRPs may also be important for explaining the intensity, Galactic
distribution, and spectrum of the diffuse Galactic gamma-ray emission, the dominant
feature of the gamma-ray sky. Accurate models for this diffuse emission are essential for
detecting all but the very strongest gamma-ray sources. However, it is not clear whether
this emission is due solely to cosmic ray interactions with interstellar material or partly
due to unidentified point sources. Modeling the GRP population allows us to estimate the
pulsar contribution to the diffuse emission. Furthermore, we can use measurements of the
diffuse gamma-ray background to constrain important properties of GRPs.
As the CGRO mission is ending and future gamma-ray missions are being planned, it
is an appropriate time to assess the population characteristics of GRPs. Suitable modeling
allows us to constrain luminosity laws, luminosity evolution, and spatial distributions of
GRPs and to predict the pulsar population that future high-energy telescopes such as the
Gamma Ray Large Area Space Telescope (GLAST), with an estimated launch date of 2005,
will detect.
In order to address the above issues within the confines of the small number of detected
 –3–

GRPs, we have developed a likelihood analysis which employs all available information
about GRPs, including pulsar detections, pulsar upper limits, and diffuse background
measurements, to constrain their luminosity law and population parameters. In this paper,
we describe the included data from the CGRO instruments OSSE (Oriented Scintillation
Spectrometer Experiment) and EGRET (Energetic Gamma Ray Experiment Telescope).
We present our model for pulsar gamma-ray luminosity and luminosity evolution, explain
our likelihood analysis method, and discuss our results and their implications.

2. The Data

2.1. OSSE

Three pulsars (the Crab and Vela pulsars, and PSR B1509−58) have been detected
by OSSE (Ulmer et al. 1994; Matz et al. 1994; Strickman et al. 1996). Table 1 lists
their 50−200 keV fluxes, along with upper and lower distance
√ bounds, spin periods, period
15
derivatives, surface dipole magnetic fields (i.e. B12 ≡ 10 P Ṗ , for P in seconds and B12
in units of 1012 Gauss), and total spin-down energy loss rate (i.e. Ė = IΩΩ̇, where I is
the moment of inertia2 and Ω is the spin frequency). We calculate the distance D to each
pulsar using the Taylor and Cordes (1993) model for Galactic electron density, except for
pulsars with more accurate distance measurements from parallax or a supernova remnant
association. Following Taylor and Cordes, we include a 25% error in our distance estimation.
We also include upper limits to the pulsed OSSE flux for 27 known pulsars (Schroeder et al.
1995; Ulmer et al. 1999). We assume a beaming solid angle Ωγ of 2π, consistent with the
uniformly wide pulse profiles observed for the detected pulsars. A photon spectrum with
a power law index of −2 (i.e. F ∝ ν −2 ) has been assumed to convert all upper limits to
the 50−200 keV energy range. While measured spectral indices for B1509-58 (Wilson et al.
1993) and Vela (Strickman et al. 1996) are greater than −2, adopting a harder spectrum for
the conversion of upper limits to luminosity changes the derived upper limits by only a few
percent and does not affect the results of this analysis. We also make use of measurements
of the OSSE diffuse gamma-ray background at 3 locations in the Galactic plane. Table 2
lists the central longitude l of each of these pointings, the viewing periods used, and the
measured diffuse flux (Skibo et al. 1997; Kinzer et al. 1999). Viewing periods, or time
intervals over which the spacecraft pointing was fixed, are defined in the Third EGRET
Catalog (Hartman et al. 1999). A power law spectrum with index −2 has been assumed to
calculate the integrated 50−200 keV diffuse flux. Choosing a different spectral index does

2
We assume I = 1045 g cm2 throughout.
 –4–

not significantly change results3 .

2.2. EGRET

We include 7 EGRET pulsar detections in our analysis (Nolan et al. 1996; Kaspi et al.
1999). Table 3 lists the E > 100 MeV fluxes for these pulsars, along with distance bounds,
periods, period derivatives, magnetic fields, and spin-down energies. We also include 353
upper limits to the pulsed flux for EGRET pulsars (Nel et al. 1996; Fierro et al. 1995;
Arzoumanian 1997). Again, we assume a beaming solid angle of 2π and a photon power law
index of −2. Table 4 lists the included measurements of the EGRET diffuse gamma-ray
background (Hunter et al. 1997) for E > 100 MeV, calculated by assuming a spectrum with
a power law index of −2. Choosing a different spectral index will change the derived diffuse
fluxes4 ; the impact of this on the analysis is discussed further in Section 5.2. Individual
viewing periods are not listed, as fluxes were calculated from an EGRET composite map.

3. Luminosity and Population Model

3
For a spectral index of −1.5, total diffuse flux over the 50− 200 keV band is 1.1 times that in Table 2.
For a spectral index of −2.5, total diffuse flux is 0.9 times that given in Table 2.
4
For a spectral index of −1.5, diffuse fluxes are 3 times that listed in Table 4. For a spectral index of
−2.5, total diffuse fluxes are 0.5 times that listed in the table.

Table 1. OSSE Pulsar Detections

Name F Dl Du P Ṗ B12 Ė
(ergs kpc−2 s−1 ) (kpc) (kpc) (ms) (s/s) (1012 G) (ergs/s)

B0531+21 3.33 ± 0.4 × 1034 1.50 2.50 33.4 4.21 × 10−13 3.8 4.47 × 1038
B0833−45 8.34 ± 5.6 × 1031 0.40 0.60 89.3 1.25 × 10−13 3.4 6.92 × 1036
B1509−58 1.24 ± 0.2 × 1033 3.50 5.30 150.6 1.54 × 10−12 15.5 1.78 × 1037
 –5–

Table 2. OSSE Diffuse Measurements at b = 0◦

l VPs F
−2
(degrees) (ergs kpc s−1 sr−1 )

0 5 + 16 1.41 ± 0.2 × 1037

25 7 + 13 4.67 ± 1.3 × 1036
95 515 + 519 3.53 ± 1.2 × 1036

Table 3. EGRET Pulsar Detections

Name F Dl Du P Ṗ B12 Ė
(ergs kpc−2 s−1 ) (kpc) (kpc) (ms) (s/s) (1012 G) (ergs/s)

B0531+21 9.55 ± 0.96 × 1033 1.50 2.50 33.4 4.21 × 10−13 3.80 4.47 × 1038
J0633+1746 3.53 ± 0.35 × 1034 0.12 0.22 237.1 1.10 × 10−14 1.62 3.24 × 1034
B0833−45 6.78 ± 0.68 × 1034 0.40 0.60 89.3 1.25 × 10−13 3.39 6.92 × 1036
B1046−58 2.39 ± 0.57 × 1033 2.24 3.73 123.7 9.59 × 10−14 3.47 2.00 × 1036
B1055−52 4.01 ± 0.80 × 1033 1.15 1.91 197.1 5.83 × 10−15 1.10 3.02 × 1034
B1706−44 7.92 ± 1.59 × 1033 1.37 2.28 102.4 9.30 × 10−14 3.09 3.39 × 1036
B1951+32 2.29 ± 0.46 × 1033 1.78 2.96 39.5 5.84 × 10−15 0.49 3.72 × 1036

Table 4. EGRET Diffuse Measurements at b = 0◦

l F
(degrees) (ergs kpc−2 s−1 sr−1 )

20 1.70 ± 0.30 × 1041

40 1.44 ± 0.25 × 1041
60 7.85 ± 0.40 × 1040
 –6–

3.1. Luminosity and Period Model

We model a pulsar’s gamma-ray luminosity L as a power law in period and magnetic

field,
L = γP −αB12 β , (1)
where P is the period in seconds and B12 is the surface dipole magnetic field in units of 1012
Gauss. If the luminosity predicted by Eq. 1 is greater than some fraction ǫγ of the pulsar’s
total spin-down energy Ė (i.e. Ė = IΩΩ̇ ∝ Ṗ P −3 ∝ P −4B 2 ), the model luminosity is set
equal to ǫγ Ė (see Section 5).
Given a luminosity law and assuming a spindown law Ω̇ ∝ Ωn , where Ω is the spin
frequency and n is the braking index, we may calculate a population-averaged gamma-ray
luminosity, from which the total diffuse flux from GRPs may be derived.
The spin period at age t of a pulsar born with initial spin period P0 is
P (t) = P0 (1 + t/τ0 )1/(n−1) , (2)
where τ0 ≡ P0 /(n − 1)P˙0 is the initial spindown time. We assume
√ the magnetic field remains
constant throughout the pulsar’s lifetime. Therefore, as B12 ∝ P Ṗ for n = 3, P0 P˙0 = P Ṗ .
However, while we expect n = 3 for magnetic dipole radiation, all measured pulsar braking
indices are less than 3 (Lyne et al. 1993; Kaspi et al. 1994; Gouiffes et al. 1992). If n does
not equal 3, then the magnetic field estimate B12 must be viewed as only an effective field
strength that may differ substantially from the actual dipole field strength.
For a given magnetic field strength B12 , the probability density function (PDF) for
period P over a population of objects born at constant birth rate is
 −1
fP (P |B12 ) = Tg Ṗ , P0 ≤ P ≤ Pg , (3)

where Tg is the age of the Galaxy and Pg ≡ P0 (1 + Tg /τ0 )1/(n−1) is the period of the oldest
pulsar. We note that for fixed B12 , Ṗ is uniquely determined by P . Throughout the
analysis, we make the simplistic assumption that the PDF for B12 is a delta function to
keep the number of fitting parameters small.
The total number of pulsars in the Galaxy is
Npsr = Ṅpsr Tg = 108 Ṅ0.01 Tg,10 , (4)
for a birth rate Ṅpsr = 0.01 yr−1 Ṅ0.01 and Galactic age Tg ≡ 1010 Tg,10 yr. The differential
distribution of their spin periods is
dNpsr
= Npsr fP (P |B12 ) = 2Ṅpsr Tg P Pg−2 = 108.3 s−1 Ṅ0.01 Tg,10 P Pg−2. (5)
dP
 –7–

Note that, because Pg depends on n, this expression is dependent upon n.

The luminosity PDF for a given magnetic field strength B12 , calculated from Eqs. 1
& 3 is
fP (P |B12 ) 2γ 2/α
fL (L|B12 ) = = B12 2β/α L−(1+2/α) , (6)
|dL/dP | αPg2
and the population-averaged GRP luminosity is therefore
n−1 )
( 2−α
 
2γB12 P02−α
β
Tg

hL|B12 i = 1 − 1+  for α 6= 2 and n 6= 1 (7)
Pg2 (α − 2) τ0

1015 γB12 β−2 P02−α

≈ for α > 2 and n = 3. (8)
Tg (α − 2)

3.2. Spatial Distribution

We adopt a spatial distribution in order to calculate the total gamma-ray flux from
pulsars in some direction through the Galaxy. We assume that pulsars are distributed in a
Gaussian disk of radius rg with exponential scale height h. We also include a molecular ring
of width wr at Galactic radius rr . We can express the pulsar number density at position x
as a function of radius from the Galactic center r and height above the Galactic plane z as
1 2 2 2 2
ρ(x) = ρ(r, z) = √ e−z/h (e−r /rg + ηe−(r−rr ) /wr )ρ0 (9)
4πh π(rg + ηwr )
where η is a dimensionless ring strength parameter and ρ0 ≡ Npsr /Vg is the average Galactic
pulsar number density, with Npsr the number of pulsars in the Galaxy and Vg the total
Galactic volume. This expression is normalized according to d3 xρ(x) = Npsr . We also
R

define a dimensionless spatial distribution G(x) ≡ ρ(x)Vg /Npsr such that d3 xG(x) = Vg .
R

To convert the PDF in r and z (Eq. 9) to a PDF fD (D) in distance from the Sun D, we use
the relation
D = [r 2 + z 2 − 2R⊙ r sin φ + R⊙
2 1/2
] (10)
where r 2 = x2 + y 2 for x = r cos φ and y = r sin φ, and R⊙ is the distance of the Sun from
the Galactic center. Assuming cylindrical symmetry, we integrate over φ to find
 !2 −1/2
D
Z Z
1  r 2 + z 2 + R⊙2
− D2
fD (D) = dr dz ρ(r, z) 1−  (11)
2πR⊙ r 2R⊙ r
where |r 2 + z 2 + R⊙
2
− D 2 | ≤ 2R⊙ r for all r and z. The integrals over r and z are done
numerically.
 –8–

3.3. Flux Distribution

The PDF for the flux F for a given gamma-ray luminosity L in terms of the distance
PDF fD (D) is
!1/2
fD (D) 1 L
fF (F |L) = = fD (D)F −3/2 (12)
|dF/dD| 2 Ωγ
for beaming solid angle Ωγ . The flux PDF after integrating over a luminosity PDF, for
fixed B12 , is
Z
fF (F |B12 ) = dL fL (L|B12 )fF (F |L). (13)

Substituting Eq. 6, we do this integral numerically, with integration limits determined

by the luminosities corresponding to periods P0 and Pg . If the luminosity predicted for
period P0 is greater than ǫγ Ė, the maximum luminosity is set equal to ǫγ Ė. In this regime,
fL (L|B12 ) scales as L−3/2 instead of the L−(1+2/α) of Eq. 6. Including a maximum allowed
gamma-ray efficiency is therefore important for all models with α < 4.
The differential number of pulsars vs. flux, with beaming solid angle Ωγ , is
dNpsr Ωγ
= Npsr fF (F |B12 ), (14)
dF 4π
which can be expressed as a histogram of F with logarithmic bins of width ∆ log F
dNpsr
∆Npsr (F ) = 2F sinh(1.15∆ log F ) . (15)
dF

Eq. 15 gives the expected flux distribution for all Galactic pulsars. To calculate the
expected flux from all pulsars in one beam (i.e. a cone of directions encompassing a small
solid angle) centered on a particular direction, we take the average luminosity from Eq. 7
and integrate across the solid angle of the beam and our dimensionless spatial distribution.
The total flux due to pulsars for a beam centered on Galactic longitude l and latitude b and
of solid angle Ωb is
Npsr Z Z
F (l, b, Ωb ) = hL|B12 i dΩb (l, b) dD G(x), (16)
12πVg
where distance D from the Sun is related to r and z of Eq. 9 through

x = D cos b sin l (17)

y = R⊙ − D cos b cos l (18)
z = D sin b. (19)
 –9–

4. Likelihood Analysis

We use a likelihood analysis which exploits the detected GRPs, GRP upper limits,
and GRP diffuse background measurements to find the best model to describe the GRP
population. The total likelihood for model Θ (i.e. one combination of parameters α, β, γ,
n, P0 , and B12 ) is factorable into the likelihoods for detections, upper limits, and diffuse
background measurements and can be expressed as

Ltot (Θ) = Ldet (Θ)Lup (Θ)Ldif (Θ). (20)

The total likelihoods for the detections, upper limits, and diffuse measurements are the
products over the individual measurements such that
N
Y det

Ldet (Θ) = Ldet,i (Θ), (21)

i=1
Nup
Y
Lup (Θ) = Lup,i(Θ), and (22)
i=1
Ndif
Y
Ldif (Θ) = Ldif,i (Θ), (23)
i=1

where Ndet , Nup , and Ndif are the number of included detections, upper limits, and diffuse
pointings, respectively.

4.1. Detected Pulsars

To calculate Ldet,i (Θ), the likelihood of model Θ given the ith pulsar detection, we use
Eq. 1 to calculate the predicted model luminosity Li (Θ) for the pulsar’s measured P and
Ṗ and for model parameters α, β, and γ. For an assumed pulsar distance D, we calculate
a predicted flux F̂i (Θ, D) = Li (Θ)/Ωγ D 2 and compare this to the detected flux Fi . The
form of the likelihood function is chosen to be a Gaussian with width determined by the
flux measurement error σF i . Integrating over a distance PDF f (D) (n.b. different from the
distance PDF of Eq. 11) to account for uncertainties in distance estimation, we may express
the PDF for model Θ given detection i as
Z Du
Ldet,i (Θ) = dD f (D) g((F̂i (Θ, D) − Fi ); σF i ), (24)
Dl

where g(x, σ) = (2πσ 2 )−1/2 exp(−x2 /2σ 2 ). The likelihood for a particular model maximizes
when the pulsar’s predicted flux F̂i (Θ, D) equals the detected flux Fi . We assume a uniform
distance distribution (i.e. f (D) = 1/(Du − Dl ) for Dl ≤ D ≤ Du and 0 otherwise).
 – 10 –

4.2. Upper Limits

For each pulsar with a measured upper limit, we calculate the predicted luminosity
given its measured P and Ṗ and for model parameters α, β, and γ. For an assumed pulsar
distance D, we calculate a model flux F̂i (Θ, D) and compare this to the measured upper
limit Fi . The form of the upper limit likelihood function is chosen to be flat with a one-sided
Gaussian roll-off with width determined by the upper limit measurement error σF i . Again
integrating over a distance PDF f (D), we express the PDF for model Θ given upper limit
measurement i as
R∞
F̂i (Θ,D) dF g(F − Fi ); σF i )
Z Du
Lup,i(Θ) = f (D) dD R∞ (25)
Dl Fi dF g(F − Fi ); σF i )

for F̂i (Θ, D) ≥ Fi . For F̂i (Θ, D) < Fi , the likelihood is constant, so that all models which
predict fluxes less than the measured upper limit are equally likely.

4.3. Diffuse Flux Measurements

To determine the likelihood for model Θ given the measured diffuse flux in the ith
beam, we calculate the population-averaged pulsar luminosity for model parameters α, β,
γ, n, P0 , and B12 from Eq. 7, and then use Eq. 16 to calculate F̂i (Θ), the total predicted
flux from all pulsars in the beam. We compare this to the measured diffuse flux in the beam
Fi , by allowing the pulsars to comprise a maximum fraction ǫd of the total measured diffuse
flux. Choosing a flat function with a one-sided Gaussian roll-off with width determined by
σF i , the error on the measured diffuse flux in beam i, as our likelihood function, the PDF
for model Θ given diffuse measurement i is
R∞
i F̂ (Θ) dF g(F − ǫd Fi ); σF i )
Ldif,i (Θ) = R ∞ (26)
ǫd Fi dF g(F − ǫd Fi ); σF i )

for F̂i (Θ) ≥ ǫd Fi . For F̂i (Θ) < ǫd Fi , the likelihood is constant.
The forms of the likelihood functions for detections, upper limits, and diffuse
background measurements are illustrated in Figure 1.

4.4. Grid Search and Marginalization

To find the parameters which best describe the GRP population, we calculate the total
likelihood (Eq. 20) across a wide range of values for the parameters α, β, γ, n, P0 , and B12
 – 11 –

and find that combination of parameters which maximizes the total likelihood. For each
parameter, we calculate marginalized PDFs as

dΘ L(Θ)
R
exc Θj
fΘj (Θj ) = , (27)
dΘ L(Θ)
R

so that, for a given parameter Θj , the marginalized PDF is the normalized integral over all
other parameters. A confidence interval C on parameter Θj can be calculated from Eq. 27
by finding the range of Θj in which a percentage C of the total probability is contained.

5. Analysis and Results

We make several assumptions about the GRP population at the start of the analysis.
For both the OSSE and the EGRET pulsars, we adopt a spatial distribution (as defined by
Eq. 9) parametrized by rg = 6 kpc, rr = 4 kpc, wr = 1.5 kpc, and η = 0.25 (n.b. η = 0.25
corresponds to a 40% increase of pulsar number density at the radius of the molecular ring
from its value for a Gaussian disk alone). For the OSSE pulsars, all of which have ages less
than 104 years, we adopt a scale height of h = 0.1 kpc. For the EGRET pulsars, the oldest
of which is 106 years old, we adopt a scale height of h = 0.5 kpc, as these older pulsars will
have moved further away from their birth sites. We take R⊙ to be 8.5 kpc throughout. The
corresponding PDF for distance D from the Sun, calculated from Eq. 11, is illustrated in
Figure 2. We take the age of the Galaxy to be 1010 years and take the pulsar birthrate to
be 1 per 100 years. We set ǫγ , the maximum fraction of spin-down luminosity a pulsar may
convert into gamma-ray luminosity, to 0.5, consistent with the highest ǫγ measured for the
known pulsars (ǫγ ∼ 0.3 for PSR B1055-52). Finally, we set ǫd , the maximum fraction of
the diffuse flux attributable to pulsars, equal to 0.5. In Sections 5.1 and 5.2, we discuss the
effects that changing these assumptions have on the analysis.

5.1. OSSE Results

Using the 3 OSSE detections, 26 upper limits, and 3 diffuse measurements described
in Section 2.1, we calculate the total likelihood across a wide range of values for α, β, γ,
n, P0 , and B12 . The likelihood maximizes for one combination of parameters α, β, and γ,
and is a maximum across a range of values for n, P0 , and B12 . Figure 3 shows contours of
equal log likelihood as a function of α, β, and log γ, taken in pairs. Figure 4 shows the
marginalized PDFs for all 6 parameters. Best parameter values for α, β, and log γ, with
95% confidence intervals calculated from the marginalized PDFs, are 8.3+1.6 +1.9
−1.2 , 7.6−1.1 , and
 – 12 –

19.4+2.0
−3.6 , respectively, leading to a best-fit luminosity law of

L = 1019.4 P −8.3 B12 7.6 ergs/s. (28)

The marginalized PDFs allow us to place a 95% confidence upper limit on n of 4.75, a
95% confidence lower limit on P0 of 6.5 ms, and a 95% confidence upper limit on B of 1014
Gauss. Changing ǫd , the fraction of the total diffuse flux that is allowed to be attributable
to pulsars, from our assumed value of 0.5 alters these limits slightly. Allowing pulsars to
comprise up to 100% of the total diffuse flux results in a lower limit of 4.5 ms on P0 , while
allowing pulsars to comprise no more than 10% of the total diffuse flux leads to an upper
limit of 3.75 on n. In Figure 5, we show how the fraction of OSSE diffuse flux due to pulsars
varies with n, P0 , and B12 . For best-fit values of α, β, and γ, the population-averaged pulsar
luminosity (Eq. 7), and hence the diffuse flux from all pulsars, increases with increasing n,
decreasing P0 , and increasing B12 . Figure 5 shows that unidentified pulsars may be quite
important in contributing to the OSSE diffuse background and should be considered in
future models of diffuse Galactic emission.
Given our best-fit luminosity law and assuming values for n, P0 , and B12 of 2.5, 10 ms,
and 2.5 × 1012 Gauss, respectively, we use Eq. 15 to calculate the distribution of predicted
OSSE flux for our model Galactic GRP population. According to this histogram, shown
in Figure 6, and given OSSE’s point source sensitivity of 6.3 × 1031 ergs/s (in the 50−200
keV energy range, and for an integration time of 5 × 105 seconds (Gehrels, Chipman, &
Kniffen 1993)), OSSE could have detected 20 GRPs as steady sources in a full-sky survey.
In Figure 6, we also plot the model-predicted OSSE fluxes for all known spin-driven pulsars
with measured period and period derivative (includes 546 pulsars (Taylor, Manchester, &
Lyne 1993)). The pulsar with the fourth highest predicted flux in the OSSE energy range is
PSR B0540−69, the Crab-like pulsar in the Large Magellanic Cloud at a distance of 50 kpc.
This illustrates the enormous capability a more sensitive low-energy gamma-ray instrument
would have for detecting young, distant pulsars that are unlikely to be discovered through
radio searches5 .
In Figure 7, we list the 50 spin-driven pulsars with highest predicted OSSE fluxes in
order of decreasing predicted flux, along with measured fluxes or upper limits. Error bars
are derived from the uncertainties on both the luminosity law parameters α, β, and γ and
on the pulsar’s distance. This plot allows us to predict which pulsars are good candidates
to search for gamma-ray emission with more sensitive low-energy gamma-ray instruments.
Because our analysis assumes that pulsars’ beams point directly at us, a pulsar’s actual

5
We note that 10 hours of integration are necessary to detect B0540−69 in the radio (Manchester et al.
1993).
 – 13 –

detectable gamma-ray flux may be lower than that listed in Figure 7, depending on the
orientation of the beam and our line of sight. However, if the successfully detected pulsars
have beams that are somewhat skewed from our line of sight, then the actual flux of other
pulsars may be larger than the prediction.

5.2. EGRET Results

Using the 7 EGRET pulsar detections, 353 upper limits, and 3 diffuse background
measurements detailed in Section 2.2, we calculate the likelihood for a wide range of
parameter values and again find a well-defined likelihood maximum in α, β, and γ. Figure
8 shows contours of log likelihood vs. pairs of these parameters. The best-fit luminosity law
is

L = 1032.0 P −1.8 B12 1.5 ergs/s, (29)

quite different from the OSSE best-fit law. Figure 9 shows the marginalized PDFs for all 6
model parameters. Best parameter values for α, β, and log γ, with 95% confidence intervals
calculated from the marginalized PDFs, are 1.82+0.13 +0.13 +0.07
−0.11 , 1.54−0.15 , and 32.04−0.15 , respectively.
We can place an upper limit at 95% confidence on n of 3.6, a 95% confidence lower limit on
P0 of 7.1 ms and a 95% confidence lower limit on B12 of 1.9 × 1010 Gauss.
In Figure 10, we show the fraction of EGRET diffuse flux due to pulsars for our
best-fit model and for various values of n, P0 , and B12 . Because the dependence of the
population-averaged luminosity (Eq. 7) on period and magnetic field is much different than
for the OSSE pulsars, the total diffuse EGRET flux varies differently with these parameters
and actually increases with increasing P0 and decreases with increasing B12 for n = 2.5.
Figure 10 show that the contribution from unidentified pulsars to the diffuse flux is, at
most, a few percent, and therefore is not important for models of the diffuse background.
There are, however, several factors which influence this conclusion. For instance, as noted
in Section 2.2, if the spectral index of the diffuse emission is −1.5 and not −2, integrated
diffuse fluxes would be 3 times greater. In this case, pulsars might comprise as much as
10% of the diffuse Galactic emission. Additionally, if the GRP scale height is lower than the
0.5 kpc that we have assumed, a greater fraction of the diffuse flux could be due to pulsars.
However, even with a scale height of 0.1 kpc, as used for the OSSE population, this fraction
is always lower than 10% for reasonable values of n, P0 , and B12 .
In Figure 11, we show the predicted flux distribution for a model population of GRPS
given our best-fit luminosity law (Eq. 15) and for values of n, P0 , and B12 of 2.5, 15 ms,
and 1012 Gauss, respectively. Predicted fluxes for all known radio pulsars with measured P
 – 14 –

and Ṗ are also shown. The point source sensitivity of EGRET (SE = 1 × 10−7 ph cm−2
s−1 = 1031.54 ergs kpc−2 s−1 ) and the expected sensitivity of GLAST (SG = 4 × 10−9 ph
cm−2 s−1 = 1031.54 ergs kpc−2 s−1 ) are marked by dashed lines. We take SG to be twice
that calculated by Gehrels & Michelson (1999) for a 2-year all sky survey, recognizing that
because pulsars are primarily a low-latitude population, surveys for them will suffer from the
increased diffuse background in the Galactic plane. We estimate that EGRET should have
detected roughly 20 pulsars. This number is tantalizing as, of the 73 unidentified EGRET
sources within 10 degrees of the Galactic plane, 17 have non-variable fluxes, as we expect
for pulsars (McLaughlin et al. 1996; McLaughlin & Cordes 1999). We therefore suggest, as
have others, that these unidentified EGRET sources are GRPs. Unfortunately, due to the
long integration times needed to detect only a small number of gamma-ray photons from
these sources, searches for pulsars of unknown period in the EGRET data are prohibitive.
Searching for radio pulsar counterparts will help to determine the nature of some of these
sources, but because radio pulsar beams are evidently narrower than gamma-ray beams,
many of these pulsars will be radio-quiet. Therefore, detecting these EGRET unidentified
sources as pulsars will likely have to wait for a more sensitive high-energy instrument which
will allow blind periodicity searches.
According to Figure 11, GLAST should detect roughly 750 GRPs as steady sources.
We estimate that 140 of these pulsars (including all unidentified EGRET sources, if they
are indeed GRPs) could be detected through a blind periodicity search of GLAST data6 .
We note that, using the same detection criterion, only 3 EGRET pulsars are detectable
through a blind periodicity search of EGRET data. Figure 11 also shows that about
120 of the approximately 1000 currently known radio pulsars should be seen by GLAST.
With this large sample of radio and gamma-ray emitting pulsars, which will undoubtedly
increase as the Parkes survey and other surveys near completion, one may compare radio
and gamma-ray pulse profiles, enabling us to determine where in the magnetosphere the
gamma-ray emission is coming from and providing important insights on the merits of
outer gap and polar cap models. For the pulsars which are not detectable in the radio and
are too faint to perform a blind search with GLAST data, x-ray observations may yield
counterparts. As the angular resolution of GLAST will be better than that of EGRET,
these searches should be much more feasible because fewer trial sky positions need to be
considered in referencing photon arrival times to the solar system barycenter.
Our prediction of 750 GLAST-detectable pulsars is dependent upon several

6
We assume a one month integration, a duty cycle of 1/2, and require a minimum Ns2 /Nt of 50, where
Ns is the number of pulsed source counts and Nt is the number of total counts, for detection, as in Mattox
et al. (1996).
 – 15 –

assumptions. We are taking ǫγ , the maximum fraction of spin-down luminosity a pulsar

may convert to gamma-ray luminosity, to be 0.5. If the actual value of ǫγ is lower than
this, fewer pulsars will be detectable by GLAST. For example, if ǫγ is actually 0.1, we
expect 610 GLAST-detectable pulsars. In Figure 11, we plot the predicted flux distribution
for a population with ǫγ of 0.01. While this value is so low as to be inconsistent with the
measured gamma-ray luminosities of the known EGRET pulsars, the figure illustrates the
effect a lower ǫγ has on the expected flux distribution. We note that, because the scaling of
the best-fit OSSE luminosity law with P and B12 is far steeper than the scaling of Ė with
these values, the predicted OSSE flux distribution is insensitive to ǫγ .
Our GLAST prediction is also dependent upon assuming a birthrate of 1/(100 yr). It
is possible, however, that the actual pulsar birthrate is substantially higher or lower than
this value. Adopting a much higher or lower birthrate while keeping all other parameters
fixed would cause the histogram in Figure 11 to be inconsistent with the EGRET pulsar
detections. However, because the flux histogram of Figure 11 depends not only on the
number of pulsars but also on the average luminosity of a pulsar, changing the braking
index, magnetic field, and initial spin period can compensate for a reduced birthrate. For
reasonable values of these parameters (i.e. 2 < n < 3; 1 ms < P0 < 30 ms; 1 < B12 < 5),
birthrates as low as 1/(500 yr) and as high as 1/(25 yr) can be made consistent with the
observed GRP flux distribution.
Our histogram of predicted fluxes also depends on the spatial distribution model we
choose for the pulsar population. Our model does not account for the location of the spiral
arms or for local density variations, which is evidenced by the discrepancy between our
model’s predictions and the detections of the bright Vela, Geminga, and Crab pulsars.
Adopting a model with a larger radial scale and/or including spiral arms would cause the
number of GLAST-detectable pulsars and the expected density of pulsars in our region of
the Galaxy to increase.
Furthermore, and perhaps most significantly, because GLAST has not yet been built,
its actual sensitivity is not yet known. Additionally, this sensitivity will be latitude
dependent and will also depend upon models for the diffuse Galactic background. While
we defer a more thorough analysis which includes these effects to a later paper, we do find
that the number of detectable pulsars varies with GLAST sensitivity as SG−1 . Therefore, if
the actual sensitivity is half our assumed value, we expect roughly 1500 GLAST-detectable
pulsars. For a sensitivity twice that we have assumed, we expect 500 GLAST-detectable
pulsars.
Figure 12 lists the known pulsars with the top 50 predicted EGRET fluxes, along
with measured fluxes and upper limits, indicating which pulsars are good candidates for
 – 16 –

gamma-ray detection with future instruments. Error bars are derived from the uncertainties
on α, β, γ, and distance. Again, as discussed in Section 5.1, a pulsar’s actual detectable
gamma-ray flux may be higher or lower than that listed in Figure 12, depending upon the
correctness of our assumptions about beaming. We note that the measured and predicted
fluxes for PSR B1055-52 differ significantly. Some of this may be due to the uncertain
alignment between the rotation axis and the line of sight to this pulsar (Lyne & Manchester
1988; Rankin 1993). There is also evidence that this pulsar is much closer than we have
assumed. While we derive a distance of 1.5 kpc from its dispersion measure and the Taylor
and Cordes (1993) model, Ögelman and Finley (1993), in their analysis of the ROSAT
x-ray data, found that a distance of 500 pc would produce a more realistic estimate of the
neutron star radius (15 km compared to 30 km for the 1.5 kpc estimate). Furthermore,
Combi, Romero, and Azcárate (1997) derive a distance estimate of 700 pc from a study
of the extended nonthermal radio source around the pulsar. Therefore, in Figure 12, we
also mark the predicted flux for PSR B1055-52 for a revised distance of 500 pc. It is
interesting that adopting a distance of 500 pc removes the constraint that this pulsar must
be converting 30% of its spin-down energy into observable gamma-ray luminosity. With
the revised distance estimate, a conversion percentage of only 3% is necessary. Similarly,
we may also be overestimating the distance to Vela. While we are using the commonly
quoted distance of 0.5 kpc for Vela (Taylor, Manchester, & Lyne 1993), recent spectroscopic
evidence suggests that this pulsar may actually be only 0.25 kpc away (Cha et al. 1999).
Adopting this revised distance would alter our best-fit model modestly, but would have a
large effect on the implied efficiency of gamma-ray production for the Vela pulsar, lowering
it from 1.5% to 0.4%. Incidentally, assuming a 0.5 kpc distance for B1055-52, the Geminga
pulsar is the most efficient converter of spin-down energy to gamma-rays, with an implied
efficiency of 22%. The Crab pulsar, with an implied efficiency of 0.01%, is the least efficient
of the EGRET pulsars.

6. Discussion

We have used OSSE and EGRET pulsar detections, measured upper limits, and
diffuse background measurements to constrain the luminosity law and some properties of
the gamma-ray pulsar population. The luminosity laws are quite different at OSSE and
EGRET energies, suggesting that different mechanisms are responsible for pulsar low- and
high-energy gamma-ray emission. The OSSE law (L ∝ P −8.3 B12 7.6 ) is roughly consistent
with optical pulsar detections (Pacini & Salvati 1983) and is steeper than that found for
ROSAT/ASCA pulsar detections (Ögelman et al. 1993; Becker & Trumper 1996; Saito
1997). It is quite different from models which predict that a pulsar’s gamma-ray luminosity
 – 17 –

should be some fraction of the total spin-down luminosity (i.e. L ∝ P −4B12 2 ). The EGRET
law (L ∝ P −1.8 B12 1.5 ) is more similar to this spin-down law and even more consistent with
a luminosity law which scales with the open-field line voltage drop (i.e. L ∝ P −2B12 ). It is
consistent with the basic polar cap (Daugherty & Harding 1996) and outer gap (Zhang &
Cheng 1997) models that have been published. More gamma-ray pulsar detections and a
comparison between radio and gamma-ray pulsar beam shapes and properties are necessary
to discriminate between models and the many versions of them.
We note that a pulsar’s age is proportional to P 2 B12 −2 , so that the gamma-ray
efficiency, or the fraction of total spin-down energy converted to gamma-ray luminosity,
decreases with increasing age for OSSE pulsars and increases with increasing age for
EGRET pulsars. Because both the OSSE and EGRET best-fit luminosity laws differ from
the standard Ė model, they produce a new ranking of pulsars as candidates for gamma-ray
detection and result in new predictions for the pulsar population we expect to see with
future more sensitive gamma-ray instruments.
We find that a more sensitive low-energy gamma-ray instrument (50−200 keV energy
range) would detect young, distant pulsars which are most likely not detectable through
their radio emission. This would be enormously helpful for studies of young pulsars, pulsar
populations, and supernova physics. A more sensitive high-energy gamma-ray instrument
(0.1−100 GeV energy range) such as GLAST would detect roughly 750 GRPs as steady
sources, most of which will have radio beams misaligned from our line of sight. Of these
750 GRPs, roughly 120 will be known radio pulsars. About 140 of the 750 GRPs will be
detectable as pulsars through blind periodicity searches, giving us a much more complete
picture of the GRP population. GLAST will be able to do blind periodicity searches on
all of the unidentified EGRET sources, of which 20 are probably GRPs. We note that all
predictions are dependent upon an assumed ǫγ and birthrate and will vary for different
values of these parameters. While there is currently a large range of acceptable values for
these parameters, comparing model predictions and actual GLAST detections will enable
us to constrain these parameters much more tightly. Our prediction will also obviously
depend upon the final sensitivity of GLAST, with the number of detectable pulsars scaling
roughly as SG−1 .
We estimate the contribution of pulsars to the measured diffuse background in the
OSSE and EGRET energy ranges, and find that while pulsars may be an important
component of the diffuse background at OSSE energies, they are most likely not important
at EGRET energies. We are able to use measurements of the diffuse background to
constrain the braking index, initial spin period, and magnetic field of the pulsar population.
The OSSE background allows us to place a 95% confidence upper limit of 4.75 on the
 – 18 –

pulsar braking index, a lower limit of 6.5 ms on the pulsar initial spin period, and an upper
limit of 1014 Gauss on the pulsar magnetic field, while the EGRET background implies a
95% confidence upper limit of 3.6 on n, a lower limit of 7.1 ms on P0 , and a lower limit of
1.9 × 1010 Gauss on B. We note that there are large errors on these parameters, and that
the real distributions of these quantities are likely to be more complicated than we have
assumed. Again, GLAST should allow us to constrain these parameters far more tightly
than currently possible.
In order to do a more sophisticated analysis of the GRP population, more GRP
detections are necessary. Large strides in our understanding of GRPs will likely have to
wait until the launch of GLAST. In the meantime, however, there are useful things that can
be done to further our understanding. While searches for periodicities in EGRET data are
very difficult, they may not be impossible if one takes an iterative approach (i.e. searching
small, sequential intervals of data sets where fewer trial periods and positions are necessary)
and uses Bayesian methods to detect pulses of unknown shape (e.g. McLaughlin & Cordes
1999). Searches for radio counterparts of the unidentified EGRET sources likely to be
GRPs are also important, as some GRPs will emit in the radio and even non-detections
provide useful information. Furthermore, whenever new x-ray or radio pulsars are detected,
especially young or nearby pulsars, one should return to the OSSE and EGRET data and
fold them. An accurate ephemeris may not be available but a known period allows a smaller
range of parameters to search and a reduction of the search threshold. Some of the Parkes
multibeam pulsars (Camilo et al. 1999), for instance, are excellent gamma-ray candidates.

We thank M. Ulmer for providing OSSE upper limits and for useful discussions. We
thank Z. Arzoumanian for providing updated EGRET upper limits, for useful discussions,
and for a careful reading of the manuscript. We also thank D. Helfand and D. Thompson
for thought provoking discussions, and the anonymous referee for useful comments. M.
A. McLaughlin acknowledges support from the NASA/NY Space Grant Program. The
research was supported by NSF Grants AST-9528394 and AST-9819931 and NASA Grant
NAG5-3515 to Cornell University.
 – 19 –

REFERENCES

Arzoumanian, Z., 1997, personal communication

Becker, W. & Trumper, J., 1996, A&AS, 120, 69

Camilo, F. et al. 1999, “The Parkes Multibeam Pulsar Survey”, IAU Coll. 177, PASP

Cha, A. N., Sembach, K. R., & Danks, A. C., ApJ, 515, L25-L28

Combi, J. A., Romero, G. E., & Azcárate, I. N., 1997, ApSS, 250, 1

Daugherty, J. K., & Harding, A. K., 1996, ApJ, 458, 278

Fierro, J. M., et al., 1995, ApJ, 447, 807

Gehrels, N., Chipman, E., & Kniffen, D. A., 1993, “The Compton Observatory in
Perspective”, AIP Conference Proceedings 280, New York, NY

Gehrels, N., & Michelson, P., 1999, APh, 11, 277

Gouiffes, C., Finley, J., & Ögelman, H., 1992, ApJ, 394, 581

Groth, E.J., 1975, ApJS, 29, 431

Hunter, S. D., et al., 1997, ApJ, 481, 205

Kaspi, V. M., et al., 1994, ApJ, 422, L83

Kaspi, V. M., et al. 1999, submitted

Kinzer, R. L., Purcell, W. R., & Kurfess, J. D., 1999, ApJ, 515, 215

Kuiper L., Hermsen W., Verbunt F., et al., 1999, Proc. 3rd INTEGRAL Workshop: “The
Extreme Universe”, Taormina, Italy

Lyne, A. G., Pritchard, R. S., & Graham-Smith, F., 1993, MNRAS, 265, 1003

Lyne, A. G., Pritchard, R. S., Graham-Smith, F., & Camilo, F., 1996, Nature, 381, 497

Lyne, A. G., & Manchester, R. N., 1988, MNRAS, 234, 477

Malofeev, V. M. & Malov, O. I., 1997, Nature , 389, 697

Manchester, R. N., Newton, L. M., & Durdin, J. M., 1985, Nature, 313, 374

Manchester, R. N. & Peterson, B. A., 1989, ApJ, 342, 23

 – 20 –

Manchester, R. N., et al., 1993, ApJ, 403, L29

Mattox, J. R., et al. 1996, A&AS, 120, 95

Matz, S. M., et al. 1994, ApJ, 434, 288

McLaughlin, M. A., Mattox, J. R., Cordes, J. M., & Thompson, D. J., 1996, ApJ, 473, 763

McLaughlin, M. A., & Cordes, J. R., 1999, in preparation

McLaughin, M. A. Cordes, J. M., Hankins, T. H., & Moffett, D.A., 1999, ApJ, 512, 929

Nel, H. I., et al., 1996, ApJ, 465, 898

Nolan, P. L., et al., 1996, A&AS, 120, 61

Ögelman, H. B., et al., 1993, AdSpR, 131, 351

Ögelman, H. B. & Finley, J. P., 1993, ApJ, 413, L31

Pacini, F., & Salvati, M., 1983, ApJ, 274, 369

Rankin, J. M., 1993, ApJS, 85, 145

Saito, Y., 1997, PhD. Thesis

Schroeder, P. C., et al., 1995, ApJ, 450, 784

Skibo, J. G., et a., 1997, ApJ, 483, L95

Strickman, M. S., 1996, ApJ, 460, 735

Taylor, J. H., & Cordes, J. M., 1993, ApJ, 411, 674

Taylor, J. H., Manchester, R. N., & Lyne, A. G., 1993, ApJS, 88, 529

Hartman, R. C., et al., 1999, ApJS, 123, 79

Ulmer, M. P., Purcell W. R., Wheaton, W. A., & Mahoney, W. A., 1991, ApJ, 369, 485

Ulmer, M. P., et al. 1994, ApJ, 432, 228

Ulmer, M. P., et al., 1999, in preparation

Wilson, R. B., et al. 1993, in Isolated Pulsars, ed. K. van Riper, R. Epstein, & C. Ho
(Cambridge: Cambridge University Press), 257
 – 21 –

Zhang, L. & Cheng, K. S., 1997, ApJ, 487, 370

This preprint was prepared with the AAS LATEX macros v4.0.
 – 22 –

Fig. 1.— Likelihood functions for detections, upper limits, and diffuse measurements are
plotted against predicted model flux, F̂i . Dotted lines mark where predicted flux F̂i equals
the measured detection, upper limit, or diffuse measurement.

0.1

0.08

0.06

0.04

0.02

0
0 10 20 30
D (kpc)

Fig. 2.— The PDF for distance D from the Sun for a model population of GRPs, calculated
for a model parametrized by rg = 6 kpc, h = 0.1 kpc, rr = 4 kpc, wr = 1.5 kpc, and η = 0.25,
and for R⊙ = 8.5 kpc.
 – 23 –

Fig. 3.— Contours of equal log likelihood vs. pairs of parameters for the OSSE data. Crosses
mark the maximum of the log likelihood.
 – 24 –

n log B (Gauss)

Fig. 4.— Marginalized PDFs for all parameters for the OSSE data.
 – 25 –

0.8

0.6

0.4

0.2

0
2.2 2.4 2.6 2.8 3 10 15 20 12 12.5
n log B (Gauss)

Fig. 5.— The fraction of OSSE diffuse flux attributable to unidentified pulsars as a function
of braking index n, initial spin period P0 , and surface magnetic field strength B12 . One
parameter is varied and other parameters are kept constant at nominal values of n = 2.5,
P0 = 15 ms, and B12 = 1.0. The dashed line shows where the fraction of diffuse flux
attributable to pulsars reaches unity and defines a range of parameter values which are
disallowed.
 – 26 –

10

d c b a
1

10 15 20 25 30 35

Fig. 6.— Histogram of predicted OSSE pulsar flux for α = 8.3, β = 7.6, log γ = 19.4, n =
2.5, P0 = 10 ms, and B12 = 2.5. The thick solid line shows the predicted flux distribution for
a model population of GRPs. The dashed line denotes the point source sensitivity of OSSE.
The thin solid line shows the predicted fluxes for known pulsars with measured P and Ṗ .
Bins with highest predicted fluxes are labeled and correspond to the following pulsars: a)
Crab; b) B1509-59; c) Vela; d) B0540-69; e) B1610-50, B1706-44.
 – 27 –

Fig. 7.— Known radio pulsars listed in order of decreasing predicted OSSE flux. Crosses
denote predicted gamma-ray fluxes, with error bars. Solid dots mark OSSE-detected fluxes,
while open circles indicate upper limits. If the distance to B1055−52 is indeed 500 pc,
instead of the assumed 1.5 kpc, this pulsar’s predicted flux would be 9 times higher, making
it the 35th (as opposed to 47th) entry in this table.
 – 28 –

Fig. 8.— Contours of equal log likelihood vs. pairs of parameters for EGRET data. Crosses
mark the maximum of the log likelihood.
 – 29 –

n log B (Gauss)

Fig. 9.— Marginalized probability density functions for all parameters for EGRET data.
 – 30 –

1 0.15
0.004
0.8
0.003 0.1
0.6

0.4 0.002 0.05

0.2
0.001
0
0
2 2.5 3 3.5 10 20 30 40 10 11 12 13
n log B (Gauss)

Fig. 10.— The fraction of EGRET diffuse flux attributable to unidentified pulsars as a
function of braking index n, initial spin period P0 , and surface magnetic field B12 . As one
parameter is varied, other parameters are kept constant at values of n = 2.5, P0 = 15 ms,
and B12 = 1.0. The dashed line shows where the fraction of diffuse flux attributable to
pulsars reaches unity and defines a range of disallowed values for n.
 – 31 –

0.5

10

e d a
1

30 35

Fig. 11.— Histogram of predicted EGRET pulsar flux for α = 1.8, β = 1.5, log γ = 32.0, n
= 2.5, P0 = 15 ms, and B12 = 1.0. The thickest solid line shows the predicted pulsar flux
distribution of a model population of GRPs, assuming ǫγ = 0.5. The thinner line illustrates
the predicted distribution for ǫγ = 0.01. Dashed lines show the sensitivities of EGRET
and GLAST to steady sources and for a blind periodicity search. The thin solid line shows
the predicted fluxes for known pulsars. Bins with highest predicted fluxes are labeled and
correspond to the following pulsars: a) Vela; b) Geminga, Crab; c) J0631+105, B0656+14,
B1509-58, B1706-44; d) B0950+08; e) B1046-58.
 – 32 –

Fig. 12.— Known radio pulsars listed in order of decreasing predicted EGRET flux. Crosses
denote predicted gamma-ray fluxes, with error bars. Solid dots mark detected fluxes, while
open circles indicate upper limits. The triangle marks the predicted flux for PSR B1055-52
for a distance of 500 pc.