FIVE NEW MILLISECOND PULSARS FROM A RADIO SURVEY OF 14 UNIDENTIFIED FERMI-LAT GAMMA-RAY SOURCES

After the discovery observations on 2009 November 25 and confirmation on December 8, we began regular timing observations of PSR J0101–6422 at Parkes. We observed the pulsar on 35 days through 2011 November 9, detecting it on 28 occasions; on the remaining 7 days it was too faint to detect, due to interstellar scintillation (DM = 12 pc cm⁻³). Each observation was typically 1 hr, with the same receiver and data acquisition system used in the search observations. Using TEMPO2¹⁷ with the 28 times of arrival, we obtained a phase-connected timing solution whose parameters are given in Table 2.

Table 2. Measured and Derived Parameters for PSR J0101–6422

Parameter	Value
Right ascension, R.A. (J2000.0)	$01^{\rm h}01^{\rm m}11\mbox{$.\!\!^{\mathrm s}$}1163(2)$
Declination, decl. (J2000.0)	−64°22'30171(2)
Proper motion in R.A. × cos  decl. (mas yr−1)	10(1)
Proper motion in decl. (mas yr−1)	−12(2)
Galactic longitude, l (deg)	301.19
Galactic latitude, b (deg)	−52.72
Spin period, P (ms)	2.5731519721683(2)
Period derivative (apparent), $\dot{P}$	5.16(3) × 10−21
Epoch of period (MJD)	55520
Orbital period, Pb (days)	1.787596706(2)
Epoch of periastron (MJD)	55162.4011764(3)
Semimajor axis (lt s)	1.701046(2)
Eccentricity	<1.5 × 10−5
Dispersion measure, DM (pc cm−3)	11.926(1)
Spin-down luminositya , $\dot{E}$ (erg s−1)	1.0 × 1034 −16%
Characteristic agea , τc (yr)	9 × 109 +20%
Surface dipole magnetic field strength (G)a	1.1 × 107 −9%
Mean flux densityb at 1.4 GHz, S1.4 (mJy)	0.28 ± 0.06
Radio–γ-ray profile offset, δ (P)	0.35 ± 0.01 ± 0.05
γ-ray profile peak-to-peak separation, Δ (P)	0.26 ± 0.03 ± 0.02
γ-ray (>0.2 GeV) photon index, Γ	0.9 ± 0.3 ± 0.2
γ-ray cutoff energy, Ec (GeV)	1.6 ± 0.4 ± 0.3
Photon flux (>0.1 GeV) (10−9 cm−2 s−1)	9.5 ± 1.8+1.5− 2.0
Energy flux (>0.1 GeV), Fγ (10−11 erg  cm−2 s−1)	1.1 ± 0.1 ± 0.1

Notes. Timing solution parameters given in TDB relative to the DE405 planetary ephemeris. Numbers in parentheses represent the measured 1σ TEMPO2 timing uncertainties on the last digits quoted. For γ-ray parameters, the first uncertainty is statistical and the second systematic. ^aCorrected for Shklovskii effect at 0.55 kpc; the percentage change from the nominal value is given. ^bAverage flux for 28 timing detections.

Download table as:  ASCIITypeset image

The Shklovskii effect (Shklovskii 1970) increases the measured value of $\dot{P}=5.2\times 10^{-21}$ beyond that intrinsic to the pulsar by $\Delta \dot{P}=P\,v_{\perp }^2/d\,c$, with v_⊥ the source velocity transverse to the line of sight. For nearby objects, it may dominate the true spin-down rate and lead to overestimates of the spin-down luminosity (Camilo et al. 1994). The observed proper motion, at the 0.55 kpc DM distance (Cordes & Lazio 2002), implies a true $\dot{P}_i=(4.3\pm 0.1)\times 10^{-21}$. At this distance, the indicated v_⊥ = 41 km s⁻¹ is typical for an MSP (see Nice & Taylor 1995).

We made polarimetric and flux-calibrated observations of PSR J0101–6422, using the PDFB3 digital filter bank. The data were analyzed with PSRCHIVE (Hotan et al. 2004). The best full-Stokes profile that we obtained is shown in Figure 1 (the flux density of this profile is not necessarily representative of the average pulsar intensity, which varies greatly owing to interstellar scintillation). The main pulse is linearly polarized at the ∼15% level, but the low signal-to-noise ratio prevents a useful determination of rotation measure (the best-fit value is ∼+10 rad m⁻², but within the uncertainties is consistent with zero). These data show that the subsidiary pulse component is composed of two outer peaks joined by a low bridge of radio emission.

Zoom In Zoom Out Reset image size

To characterize the γ-ray profile, we selected photons collected between 2008 August 4 and 2011 August 1 with reconstructed energies 0.3–3 GeV lying within 11 of the timing position. This selection balances completeness (including as many pulsed photons as possible) and pulsed signal-to-noise ratio as measured by the H-test (de Jager et al. 1989). We apply the same data processing as described in Section 4.4, omitting the horizon cut to increase the live time by about 20%, at the expense of a slightly increased background.

The light curve corresponding to this extraction appears in Figure 2. If γ-rays are produced above the null charge surface (the locus of points where the magnetic field is orthogonal to the pulsar spin axis), they appear to an observer in the opposite hemisphere, whereas radio emission from low altitudes is beamed into the same hemisphere (see Romani & Yadigaroglu 1995). In this scenario, the line of sight to PSR J0101–6422 would intercept a cone of radio emission in one hemisphere to produce the weaker radio peak at ϕ ∼ 0.5 and a cone of radio and γ emission in the other to produce the two γ peaks and bright radio peak at ϕ ∼ 1. In keeping with this picture and the convention established in the literature, we identify the γ peak at ϕ ∼ 0.8 as "P1" and that at ϕ ∼ 0.1 as "P2."

Zoom In Zoom Out Reset image size

The relative phasing of the peaks is of interest (e.g., Watters et al. 2009). The radio-to-γ offset, δ, is measured from the center of the leading radio peak (taken as ϕ ≡ 0.5, see Figure 2). We estimated the phase of the γ-ray peaks using unbinned maximum likelihood with a two-sided Lorentzian model for the peaks. These estimates yield δ = 0.33 ± 0.01_stat ± 0.03_syst and a γ-ray peak-to-peak separation of Δ = 0.26 ± 0.03_stat ± 0.02_syst. The DM-induced uncertainty in δ is <0.001P. The systematic uncertainty for both δ and Δ is estimated from the scatter obtained by fitting alternative functional forms for the peaks.

To measure the spectrum of PSR J0101–6422, we analyzed low background DIFFUSE class "Pass 6" (Atwood et al. 2009) LAT data collected between 2008 August 4 and 2011 April 17. We filtered periods where the observatory's rocking angle exceeded 52° or Earth's limb impinged upon the field of view (horizon cut, requires zenith angle <100°). We modeled the sensitivity to these events with the flight-corrected P6_V11_DIFFUSE instrument response function. To compute background contributions, we used a preliminary version of the 2FGL catalog and its accompanying diffuse emission models.¹⁸ We selected events within 10° and re-fit the spectra of point sources within 8° and the normalizations of the diffuse models. Fits employed pointlike (Kerr 2011), a binned maximum likelihood algorithm.

The phase-averaged spectrum of PSR J0101–6422 is well described by an exponentially cutoff power law, dN/dE = N₀ (E/E₀)^−Γ exp (− E/E_c), whose parameters appear in Table 2. We also considered separately the spectra of the three major components, viz., P2 (0 < ϕ < 0.33), the off-pulse (0.33 < ϕ < 0.63), and P1 (0.63 < ϕ < 1). The parameter uncertainties are relatively large, and accordingly we found no significant difference in the spectral shapes of the two peaks. Further, we detected no significant emission in the off-pulse phase window and derived a 95% confidence upper limit on the energy flux of a point source with a power-law spectrum with Γ = 2.2, obtaining 6 × 10⁻¹² erg cm⁻² s⁻¹ (>100 MeV; scaled up by 3.3 to the full phase window).

The observed flux is related to the γ-ray luminosity by F_γ = L_γ/4π f_Ω d², where f_Ω is an unknown beaming factor (typically of order unity; e.g., Watters et al. 2009) implying a phase-averaged, isotropic (f_Ω = 1) γ-ray luminosity L_γ = 4.0 × 10³² erg s⁻¹ $= 0.04\,\dot{E}$ at d = 0.55 kpc. Best-fit values of f_Ω from the two-pole caustic (TPC) and outer gap (OG) models (Section 5.1) are 0.7 and 0.9. This γ-ray efficiency is in accord with other LAT-detected MSPs (Abdo et al. 2009a).

A 3.3 ks Swift observation with the X-ray Telescope in PC mode was obtained on 2009 November 21 to search for an X-ray counterpart to the then-unidentified 1FGL source. No significant counterpart is detected, with a 3σ upper limit at the pulsar position of 2.0 × 10⁻³ counts s⁻¹ (0.5–8 keV). Assuming a power-law spectrum with photon index Γ = 1.5 for a column with N_H = 3.7 × 10²⁰ cm⁻² (10 HI atoms per free electron along the line of sight) yields an unabsorbed flux limit of 1.1 × 10⁻¹³ erg  cm⁻² s⁻¹, or efficiency $L_X/\dot{E}<5.0\times 10^{-4}$ at 0.55 kpc. This limit is comparable to the detected X-ray flux from other MSPs (e.g., Marelli et al. 2011).

We searched four DSS plates and found no optical counterpart consistent with the pulsar's position, suggesting m > 21.

At first blush, PSR J0101–6422 is an unremarkable member of the growing population of γ-ray MSPs. The orbital characteristics suggest an epoch of Roche lobe overflow from an evolved 1–2 M_☉ companion which later became the ∼0.2 M_☉ He white dwarf currently inferred in the system (e.g., Tauris & Savonije 1999). However, as we discuss below, PSR J0101–6422's light curve is challenging to explain with simple geometric models that tie the radio and γ emission to particular regions of the magnetosphere.

Remarkably, a simple picture of γ-ray emission arising in the outer magnetosphere appears to describe the majority of both young (though see Romani et al. 2011) and recycled pulsars (e.g., Venter et al. 2009). On the other hand, while radio emission in young pulsars and some MSPs appears to come from lower altitudes, several MSPs have phase-aligned γ and radio peaks (e.g., Abdo et al. 2010a), implying a joint emission site at high altitude. These observations raise the prospect of MSPs whose radio emission combines "traditional" polar cap emission with a higher-altitude component.

To determine which scenario best describes the emission pattern of PSR J0101–6422, we performed joint fits to the radio and γ-ray data following the method of Johnson (2011). We modeled the radio as a simple "cone" component (Story et al. 2007) and the γ-rays using both the TPC model of Dyks & Rudak (2003; note that we adopt a larger maximum cylindrical radius of 0.95 of the light cylinder) and the OG model (Cheng et al. 1986; Romani & Yadigaroglu 1995). The angle between the pulsar spin axis and the line of sight, ζ, is unknown a priori, but if the radio emission arises from low altitudes, then the presence of two peaks separated by roughly half of a rotation indicates the inclination of the magnetic axis from the spin axis (α) cannot be small. The best-fit light curves for the two γ-ray models appear in Figure 3. The low α = 26° preferred for the TPC model cannot produce the radio interpulse and produces too much off-pulse γ-ray emission. The OG model, with a best fit at α = 90° (orthogonal rotator), produces two radio peaks but fails to produce two clear γ peaks and the radio interpulse morphology.

Zoom In Zoom Out Reset image size

Since neither model faithfully reproduces the observed light curves, we propose that one or both of the radio peaks may originate at higher altitude. Emission over a range of altitudes would also explain the low observed linear polarization of PSR J0101–6422 (e.g., Dyks et al. 2004) and admit smaller α values. Alternatively, the failure of geometric models may indicate that the physical details of the MSP magnetosphere—e.g., currents, multipolar fields, and plasma loading of the field lines—and the radio- and γ-emission mechanisms play an important role.

With 6/14 pointings resulting in unbiased MSP detections, this survey was fruitful, though this high efficiency may be luck. Indeed, in our timing campaign, we only detected PSR J0101–6422 on 28 of 35 attempts, and several observations required prior knowledge of the ephemeris for detection. Thus, because of scintillation, we estimate that PSR J0101–6422 is only detectable in 2/3 observations, depending somewhat on integration time. Many field MSPs are in eclipsing binaries, further diminishing the probability of detection in a single observation.

We believe the strongest factor in our search's success was the source selection criteria. For example, Ransom et al. (2011) selected nonvariable LAT sources (but including no spectral shape information) and found MSPs in 3/25 pointings. But when considering only targets with |b| > 5°, their efficiency jumps to 3/8. Keith et al. (2011) employed some spectral information and detected MSPs unbiasedly in 1/11 pointings (1/4 for |b| > 5°). A short list of seven spectrally exceptional candidates drawn up for a Nançay radio telescope survey included five MSPs, three of which were discovered from Nançay observations (Cognard et al. 2011; Guillemot et al. 2012). Hessels et al. (2011) used nearly identical selection criteria as this work in a Green Bank Telescope survey at 350 MHz and found MSPs in 13/49 pointings, all but one at |b| > 5°.

Targeting pulsar-like sources is clearly efficient. However, new LAT sources will lie at the sensitivity threshold and the limited statistics will not admit classification of spectral shape and variability. Nonetheless, we argue that future radio searches should omit detailed ranking schemes and target any LAT source off the Galactic plane that has not been associated with a known blazar.

We believe this substantially increased target list (and telescope time) is justified. In the 2FGL source catalog (Abdo et al. 2012), there are ∼350 unassociated sources more than 5° off the Galactic plane. MSPs account for roughly 5% of associated LAT sources, so 15–20 MSPs may remain to be found in these unassociated 2FGL sources. With only 350 positions to monitor, multiple deep observations can ameliorate the confounding effects of scintillation and eclipses while requiring a fraction of the time of a comparably complete all-sky survey. By comparison, the otherwise prodigious Parkes multibeam survey found "only" 17 MSPs in over 40,000 pointings (e.g., Faulkner et al. 2004). In addition to the intrinsic interest in discovering new MSPs, by thoroughly searching all such LAT sources, we will gain confidence that the MSPs we have found can truly provide a volume-limited sample of the energetic, γ-loud MSPs, an invaluable step in the study of the MSP population of the Milky Way.