PARKES RADIO SEARCHES OF FERMI GAMMA-RAY SOURCES AND MILLISECOND PULSAR DISCOVERIES

2.1.1. Initial Searches of 1FGL Sources

2.1.2. Repeated Searches of Unidentified LAT Sources

Our subsequent searches used the same equipment and methods as Kerr et al. (2012). In brief, total-power measures from the central beam of the 20 cm multibeam receiver were filtered into 512 contiguous 0.5 MHz-wide channels centered on 1390 MHz and sampled 8000 times per second, and then were digitized with 1 bit precision and written to disk for off-line analysis. Individual integration times were about 1 hr, and each LAT source was observed between one and nine times, depending on the then-perceived quality of the source and telescope availability (Table 1).

Table 1.  Radio Searches of FGL Sources at Parkes: Observations

Namea	R.A.b	Decl.b	l	b	Integration Time	${{\rm{DM}}}_{\mathrm{max}}$ c
	(J2000.0)	(J2000.0)	(deg)	(deg)	(minutes)	(pc cm−3)
3FGL J0101.0–6422	${01}^{{\rm{h}}}{00}^{{\rm{m}}}{58}^{{\rm{s}}}$	$-64^\circ 24^{\prime} 06$''	301.2	−52.7	60	270
3FGL J0602.8–4016	${06}^{{\rm{h}}}{03}^{{\rm{m}}}{05}^{{\rm{s}}}$	$-40^\circ 11^{\prime} 04$''	246.8	−25.9	90, 120	270
3FGL J0933.9–6232	${09}^{{\rm{h}}}{34}^{{\rm{m}}}{00}^{{\rm{s}}}$	$-62^\circ 30^{\prime} 17$''	282.2	−7.8	120, 60, 60, 60, 60	435
3FGL J1035.7–6720	${10}^{{\rm{h}}}{36}^{{\rm{m}}}{10}^{{\rm{s}}}$	$-67^\circ 21^{\prime} 05$''	290.4	−7.8	120, 86, 136, 60, 60, 60	435
3FGL J1227.9–4854	${12}^{{\rm{h}}}{27}^{{\rm{m}}}{50}^{{\rm{s}}}$	$-48^\circ 51^{\prime} 57$''	298.9	13.8	120, 65, 60	307
3FGL J1231.6–5113	${12}^{{\rm{h}}}{31}^{{\rm{m}}}{49}^{{\rm{s}}}$	$-51^\circ 18^{\prime} 49$''	299.8	11.4	120	270
3FGL J1514.2–4947	${15}^{{\rm{h}}}{14}^{{\rm{m}}}{06}^{{\rm{s}}}$	$-49^\circ 45^{\prime} 33$''	325.2	6.8	120	270
3FGL J1624.2–4041	${16}^{{\rm{h}}}{24}^{{\rm{m}}}{07}^{{\rm{s}}}$	$-40^\circ 41^{\prime} 20$''	340.6	6.2	120, 120, 120, 60, 60, 60	435
3FGL J1658.4–5323	${16}^{{\rm{h}}}{58}^{{\rm{m}}}{43}^{{\rm{s}}}$	$-53^\circ 17^{\prime} 43$''	335.0	−6.6	83	270
3FGL J1744.1–7619	${17}^{{\rm{h}}}{44}^{{\rm{m}}}{02}^{{\rm{s}}}$	$-76^\circ 20^{\prime} 25$''	317.1	−22.5	41, 90, 80, 72, 71, 78, 60, 60, 60	192
3FGL J1747.6–4037	${17}^{{\rm{h}}}{47}^{{\rm{m}}}{29}^{{\rm{s}}}$	$-40^\circ 36^{\prime} 07$''	350.2	−6.4	86	270
3FGL J1902.0–5107	${19}^{{\rm{h}}}{02}^{{\rm{m}}}{05}^{{\rm{s}}}$	$-51^\circ 09^{\prime} 45$''	345.6	−22.4	70	270
3FGL J2039.6–5618	${20}^{{\rm{h}}}{39}^{{\rm{m}}}{30}^{{\rm{s}}}$	$-56^\circ 20^{\prime} 45$''	341.2	−37.1	75, 112, 60	270
3FGL J2241.6–5237d	${22}^{{\rm{h}}}{41}^{{\rm{m}}}{52}^{{\rm{s}}}$	$-52^\circ 37^{\prime} 38$''	337.4	−54.9	120	270
3FGL J0133.0–4413	${01}^{{\rm{h}}}{33}^{{\rm{m}}}{27}^{{\rm{s}}}$	$-44^\circ 08^{\prime} 29$''	279.2	−71.0	60, 60, 60, 60, 60, 60	77
3FGL J0216.1–7016	${02}^{{\rm{h}}}{14}^{{\rm{m}}}{04}^{{\rm{s}}}$	$-69^\circ 52^{\prime} 06$''	292.9	−45.6	60	115
3FGL J0744.8–4028	${07}^{{\rm{h}}}{44}^{{\rm{m}}}{59}^{{\rm{s}}}$	$-40^\circ 29^{\prime} 49$''	254.6	−8.0	60	717
3FGL J0802.3–5610	${08}^{{\rm{h}}}{02}^{{\rm{m}}}{46}^{{\rm{s}}}$	$-56^\circ 15^{\prime} 32$''	270.0	−13.2	60, 60, 35, 35, 35	627
3FGL J0933.9–6232	${09}^{{\rm{h}}}{34}^{{\rm{m}}}{02}^{{\rm{s}}}$	$-62^\circ 31^{\prime} 34$''	282.2	−7.9	60, 35	627
3FGL J0940.6–7609	${09}^{{\rm{h}}}{40}^{{\rm{m}}}{45}^{{\rm{s}}}$	$-76^\circ 09^{\prime} 34$''	292.2	−17.4	60	269
3FGL J0940.7–6102	${09}^{{\rm{h}}}{40}^{{\rm{m}}}{57}^{{\rm{s}}}$	$-61^\circ 05^{\prime} 10$''	281.9	−6.3	60	653
3FGL J0954.8–3948	${09}^{{\rm{h}}}{55}^{{\rm{m}}}{03}^{{\rm{s}}}$	$-39^\circ 49^{\prime} 25$''	269.9	11.5	60, 60, 60, 60, 60	371
3FGL J0955.6–6148	${09}^{{\rm{h}}}{55}^{{\rm{m}}}{39}^{{\rm{s}}}$	$-61^\circ 48^{\prime} 36$''	283.7	−5.7	60, 60	653
3FGL J1012.0–4235	${10}^{{\rm{h}}}{12}^{{\rm{m}}}{07}^{{\rm{s}}}$	$-42^\circ 35^{\prime} 01$''	274.2	11.2	60	371
3FGL J1025.1–6507	${10}^{{\rm{h}}}{25}^{{\rm{m}}}{00}^{{\rm{s}}}$	$-65^\circ 07^{\prime} 22$''	288.3	−6.5	60	755
3FGL J1035.7–6720	${10}^{{\rm{h}}}{36}^{{\rm{m}}}{09}^{{\rm{s}}}$	$-67^\circ 22^{\prime} 17$''	290.4	−7.8	45, 60, 60	563
3FGL J1036.0–8317	${10}^{{\rm{h}}}{36}^{{\rm{m}}}{20}^{{\rm{s}}}$	$-83^\circ 17^{\prime} 01$''	298.9	−21.5	60, 60, 60	192
3FGL J1057.7–6624	${10}^{{\rm{h}}}{58}^{{\rm{m}}}{43}^{{\rm{s}}}$	$-66^\circ 21^{\prime} 59$''	292.0	−5.9	60, 60, 60, 60	755
3FGL J1136.6–6826	${11}^{{\rm{h}}}{36}^{{\rm{m}}}{47}^{{\rm{s}}}$	$-68^\circ 25^{\prime} 37$''	296.1	−6.5	60	755
3FGL J1227.9–4854	${12}^{{\rm{h}}}{27}^{{\rm{m}}}{42}^{{\rm{s}}}$	$-48^\circ 53^{\prime} 28$''	298.9	13.8	60, 60	371
3FGL J1231.6–5113	${12}^{{\rm{h}}}{31}^{{\rm{m}}}{34}^{{\rm{s}}}$	$-51^\circ 12^{\prime} 31$''	299.8	11.5	60	435
3FGL J1238.3–4543	${12}^{{\rm{h}}}{38}^{{\rm{m}}}{19}^{{\rm{s}}}$	$-45^\circ 43^{\prime} 18$''	300.5	17.1	60	269
3FGL J1306.8–4031	${13}^{{\rm{h}}}{06}^{{\rm{m}}}{52}^{{\rm{s}}}$	$-40^\circ 32^{\prime} 35$''	306.1	22.2	60	192
3FGL J1311.8–3430	${13}^{{\rm{h}}}{11}^{{\rm{m}}}{46}^{{\rm{s}}}$	$-34^\circ 29^{\prime} 19$''	307.7	28.2	60, 58, 60, 58	154
3FGL J1325.2–5411	${13}^{{\rm{h}}}{25}^{{\rm{m}}}{15}^{{\rm{s}}}$	$-54^\circ 11^{\prime} 13$''	307.9	8.4	60	627
3FGL J1326.7–4727	${13}^{{\rm{h}}}{26}^{{\rm{m}}}{40}^{{\rm{s}}}$	$-47^\circ 27^{\prime} 52$''	309.1	15.0	60, 60	371
3FGL J1417.5–4402	${14}^{{\rm{h}}}{17}^{{\rm{m}}}{30}^{{\rm{s}}}$	$-44^\circ 02^{\prime} 40$''	318.9	16.1	60	307
3FGL J1417.7–5026	${14}^{{\rm{h}}}{17}^{{\rm{m}}}{39}^{{\rm{s}}}$	$-50^\circ 25^{\prime} 33$''	316.7	10.1	60	627
3FGL J1518.2–5232	${15}^{{\rm{h}}}{18}^{{\rm{m}}}{27}^{{\rm{s}}}$	$-52^\circ 33^{\prime} 57$''	324.3	4.1	60, 60, 60, 60	1044
3FGL J1536.3–4949	${15}^{{\rm{h}}}{36}^{{\rm{m}}}{29}^{{\rm{s}}}$	$-49^\circ 49^{\prime} 45$''	328.2	4.8	37, 37, 16, 60, 60, 60	883
3FGL J1539.2–3324	${15}^{{\rm{h}}}{39}^{{\rm{m}}}{15}^{{\rm{s}}}$	$-33^\circ 25^{\prime} 42$''	338.7	17.5	60, 15, 52, 60	269
3FGL J1603.7–6011	${16}^{{\rm{h}}}{03}^{{\rm{m}}}{44}^{{\rm{s}}}$	$-60^\circ 11^{\prime} 12$''	324.8	−5.7	60	883
3FGL J1617.4–5846	${16}^{{\rm{h}}}{17}^{{\rm{m}}}{28}^{{\rm{s}}}$	$-58^\circ 46^{\prime} 19$''	327.0	−5.9	60	883
3FGL J1624.2–4041	${16}^{{\rm{h}}}{24}^{{\rm{m}}}{09}^{{\rm{s}}}$	$-40^\circ 40^{\prime} 23$''	340.6	6.2	60, 60	883
3FGL J1702.8–5656	${17}^{{\rm{h}}}{02}^{{\rm{m}}}{33}^{{\rm{s}}}$	$-56^\circ 54^{\prime} 54$''	332.4	−9.2	60, 60, 60, 60	627
3FGL J1717.4–5157	${17}^{{\rm{h}}}{17}^{{\rm{m}}}{35}^{{\rm{s}}}$	$-51^\circ 57^{\prime} 58$''	337.7	−8.1	60	755
3FGL J1736.2–4444	${17}^{{\rm{h}}}{36}^{{\rm{m}}}{13}^{{\rm{s}}}$	$-44^\circ 44^{\prime} 50$''	345.5	−6.7	60	883
3FGL J1744.1–7619	${17}^{{\rm{h}}}{44}^{{\rm{m}}}{11}^{{\rm{s}}}$	$-76^\circ 20^{\prime} 29$''	317.1	−22.5	60	192
3FGL J1753.6–4447	${17}^{{\rm{h}}}{53}^{{\rm{m}}}{38}^{{\rm{s}}}$	$-44^\circ 47^{\prime} 50$''	347.1	−9.4	60	627
3FGL J1803.3–6706	${18}^{{\rm{h}}}{03}^{{\rm{m}}}{25}^{{\rm{s}}}$	$-67^\circ 07^{\prime} 14$''	326.9	−20.4	60	230
3FGL J1808.3–3357	${18}^{{\rm{h}}}{08}^{{\rm{m}}}{22}^{{\rm{s}}}$	$-33^\circ 56^{\prime} 05$''	358.1	−6.7	56, 60, 60, 60	755
3FGL J1831.6–6503	${18}^{{\rm{h}}}{31}^{{\rm{m}}}{45}^{{\rm{s}}}$	$-65^\circ 05^{\prime} 19$''	329.9	−22.4	60	192
3FGL J1903.6–7052	${19}^{{\rm{h}}}{02}^{{\rm{m}}}{43}^{{\rm{s}}}$	$-70^\circ 53^{\prime} 49$''	324.3	−26.4	60	269
3FGL J1946.4–5403	${19}^{{\rm{h}}}{46}^{{\rm{m}}}{24}^{{\rm{s}}}$	$-54^\circ 02^{\prime} 46$''	343.9	−29.6	60, 60, 60, 60, 60, 54	154
3FGL J1959.8–4725	${19}^{{\rm{h}}}{59}^{{\rm{m}}}{57}^{{\rm{s}}}$	$-47^\circ 26^{\prime} 32$''	351.8	−30.9	58, 60, 60, 35, 60	307
3FGL J2039.6–5618	${20}^{{\rm{h}}}{39}^{{\rm{m}}}{51}^{{\rm{s}}}$	$-56^\circ 20^{\prime} 26$''	341.2	−37.2	60, 60, 60, 60	115
3FGL J2043.8–4801	${20}^{{\rm{h}}}{43}^{{\rm{m}}}{49}^{{\rm{s}}}$	$-48^\circ 00^{\prime} 45$''	351.7	−38.3	60, 60	115
3FGL J2112.5–3044	${21}^{{\rm{h}}}{12}^{{\rm{m}}}{36}^{{\rm{s}}}$	$-30^\circ 42^{\prime} 37$''	14.9	−42.4	56, 60, 60, 60, 60, 60	103
3FGL J2131.1–6625	${21}^{{\rm{h}}}{31}^{{\rm{m}}}{06}^{{\rm{s}}}$	$-66^\circ 24^{\prime} 42$''	326.7	−40.3	60, 60	115
3FGL J2133.0–6433	${21}^{{\rm{h}}}{33}^{{\rm{m}}}{30}^{{\rm{s}}}$	$-64^\circ 31^{\prime} 58$''	328.7	−41.3	60, 60, 60, 60, 60, 60, 60	103
3FGL J2200.0–6930	${22}^{{\rm{h}}}{00}^{{\rm{m}}}{00}^{{\rm{s}}}$	$-69^\circ 31^{\prime} 37$''	321.3	−40.9	60	115
3FGL J2220.6–6833	${22}^{{\rm{h}}}{20}^{{\rm{m}}}{24}^{{\rm{s}}}$	$-68^\circ 32^{\prime} 22$''	320.8	−43.0	60, 60, 60	115
3FGL J2333.0–5525	${23}^{{\rm{h}}}{33}^{{\rm{m}}}{04}^{{\rm{s}}}$	$-55^\circ 25^{\prime} 32$''	324.2	−58.3	60, 60	87

Notes. Boldfaced entries denote observations with detection of MSPs. Discoveries in single-observation searches of the first 14 entries (listed above the horizontal line) were reported in Kerr et al. (2012). Italicized entries denote sources re-observed in 2012 at the improved locations listed below the horizontal line.

^aThe names given are of the 3FGL sources closest to our pointing locations (the offset between the pointing positions given here and the 3FGL positions are provided in Table 2). ^bParkes telescope pointing position. ^cMaximum trial dispersion measure used in our analysis of the respective data set(s), corresponding approximately to twice the maximum DM predicted by the Cordes & Lazio (2002) model for the corresponding line of sight. The first observation of each source above the horizontal line was analyzed with ${{\rm{DM}}}_{\mathrm{max}}=270$ pc cm⁻³ (Kerr et al. 2012). ^dMSP discovered independently by Keith et al. (2011).

Download table as:  ASCIITypeset images: 1 2

All the data were analyzed using PRESTO (Ransom 2001), which implements standard pulsar search techniques including radio-frequency interference excision and optimization for signals with changing apparent spin periods caused by orbital motion. Finite sampling time and smearing from dispersive propagation delays within finite-width filterbank channels both unavoidably degrade sensitivity to pulsed signals. Additional smearing can result from the use of an incorrect dispersion measure (DM) to remove the delays between channels. We dedispersed each observation using a set of trial DMs such that this effect was negligible. This was done up to twice the maximum DM predicted by the Cordes & Lazio (2002) model for the corresponding line of sight (see Table 1). The maximum acceleration searched for corresponded to signals drifting by $\pm 200/{n}_{{\rm{h}}}$ bins in the Fourier domain, where n_h is the largest harmonic at which a signal is detected (up to 16 harmonics were summed, in powers of two). This was parameterized by ${\mathtt{zmax}}=200$ within PRESTO (Ransom et al. 2002).

In the 2009 searches (Section 2.1.1), we detected six MSPs in single observations of 14 LAT sources (Table 1, above the horizontal dividing line). In 2010–2011, we re-observed seven of the remaining eight 1FGL sources (Table 1, above the dividing line), but no new pulsars were detected.¹⁴

The 1FGL catalog was based on 11 months of LAT data. The 2FGL catalog (Nolan et al. 2012) was based on two years of data. In 2012, we selected new search targets based on a three-year source list developed by the LAT collaboration but never published (the subsequent 3FGL catalog is based on four years of data and a different pipeline; Acero et al. 2015). As for the 1FGL searches (Kerr et al. 2012), we restricted ourselves to non-variable southern sources with no plausible known blazar counterparts and with a LAT positional uncertainty (95% confidence level error radius) ≤7' to fit within the 1.4 GHz primary beam of the Parkes telescope. The remaining sources were classified by visual inspection of the gamma-ray spectral energy distribution to pick out candidates with a spectral shape resembling those of known pulsars, which typically have exponentially cutoff power-law spectra (Abdo et al. 2013). Spectral modeling and source localization are more difficult for LAT sources near the Galactic plane, and we only considered those with $| b| \gt 4^\circ$.

The new target set consisted of 49 sources, each observed between one and seven times, for 122 integrations in the aggregate. Seven of the 49 targets had also been observed in their prior 1FGL incarnation (italicized in Table 1). Among the remaining 42, we discovered five MSPs (Table 1).

Overall, we searched 56 individual LAT sources in our Parkes survey using the analog filterbank system. We detected 11 MSPs, of which 10 were discoveries (Table 1). In fact, radio and/or gamma-ray pulsars are now known in 17 of those 56 sources (Table 2). We discuss the sources, in particular those which might still be good pulsar candidates, in Section 3.

Table 2.  Radio Searches of FGL Sources at Parkes: Source Information

3FGL Namea	${\rm{\Delta }}\theta$ b	r95c	Classd	Sige	Curvef	Varg	Spectrum	${N}_{\mathrm{obs}}$ i
	(deg)	(deg)		(σ)	(σ)		Notesh
J0101.0–6422j	0.03	0.04	PSR	26.8	6.8	55	1 CpR	1
$\text{J0133.0}\kern-2pt-\kern-2pt\text{4413}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.12	0.09	bll	7.7	0.5	48	4 lhr	6
$\text{J0216.1}\kern-2pt-\kern-2pt\text{7016}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.44	0.12	bcu	6.0	0.5	49	4 ld	1
$\text{J0602.8}\kern-2pt-\kern-2pt\text{4016}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.10	0.04	bcu	15.3	1.8	49	4 lihr	2
$\text{J0744.8}\kern-2pt-\kern-2pt\text{4028}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.04	0.08	bcu	7.4	1.9	39	3 ld	1
$\text{J0802.3}\kern-2pt-\kern-2pt\text{5610}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.11	0.10	⋯	12.3	3.6	44	4 ld	5
J0933.9–6232	0.01	0.04	⋯	22.0	8.8	59	1 CPr	2
J0940.6–7609	0.01	0.10	⋯	10.0	1.1	48	3 ld	1
$\text{J0940.7}\kern-2pt-\kern-2pt\text{6102}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.05	0.21	bcu	6.4	2.7	51	3 pr	1
J0954.8–3948	0.04	0.09	⋯	18.8	3.9	51	3 D*	5
J0955.6–6148k	0.00	0.11	psr	7.0	2.2	39	2 cr	2
J1012.0–4235k	0.02	0.07	psr	8.8	3.6	38	3 ?pr	1
J1025.1–6507	0.02	0.09	⋯	7.3	2.7	52	3 ld	1
J1035.7–6720l	0.06	0.04	⋯	30.4	8.3	47	1 CpR	3
J1036.0–8317k	0.01	0.12	psr	6.6	2.3	42	3 ?p	3
$\text{J1057.7}\kern-2pt-\kern-2pt\text{6624}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.10	0.06	⋯	9.1	1.2	29	4 LD	4
$\text{J1136.6}\kern-2pt-\kern-2pt\text{6826}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.01	0.11	bcu	7.7	0.7	43	4 LD	1
J1227.9–4854m	0.04	0.04	psr	38.6	4.6	74	3 Dh*	2
J1231.6–5113	0.01	0.11	⋯	13.4	5.6	46	1 CPr	1
$\text{J1238.3}\kern-2pt-\kern-2pt\text{4543}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.01	0.08	bcu	8.0	0.3	61	4 ?lh	1
$\text{J1306.8}\kern-2pt-\kern-2pt\text{4031}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.01	0.07	⋯	14.9	0.3	43	4 lD	1
J1311.8–3430n	0.02	0.02	PSR	62.3	9.0	52	1 CR	4
J1325.2–5411	0.00	0.10	⋯	7.0	2.5	31	3 ?ld	1
J1326.7–4727	0.02	0.05	glc	11.3	6.1	66	1 Cpr	2
J1417.5–4402	0.00	0.06	⋯	12.5	1.4	54	2 cd	1
$\text{J1417.7}\kern-2pt-\kern-2pt\text{5026}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.02	0.11	⋯	6.1	0.4	40	4 Ld	1
J1514.2–4947j	0.03	0.02	PSR	39.4	9.3	35	1 CpR	1
J1518.2–5232	0.04	0.07	⋯	10.3	3.6	62	2 cr	4
J1536.3–4949o	0.02	0.02	psr	60.0	8.5	51	3 pRh	6
J1539.2–3324	0.02	0.04	⋯	19.2	9.6	57	2 cPr	4
$\text{J1603.7}\kern-2pt-\kern-2pt\text{6011}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.01	0.09	⋯	5.0	3.6	50	4 ?h	1
$\text{J1617.4}\kern-2pt-\kern-2pt\text{5846}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.00	0.08	fsrq	15.8	1.6	128	5 LDV	1
J1624.2–4041l	0.02	0.04	⋯	19.2	7.3	50	1 cpR	2
J1658.4–5323j	0.10	0.06	PSR	16.9	6.5	38	1 Cr	1
J1702.8–5656	0.05	0.04	⋯	28.8	6.1	58	3 rdh*	4
$\text{J1717.4}\kern-2pt-\kern-2pt\text{5157}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.02	0.09	fsrq	12.2	2.5	383	5 LDV	1
J1736.2–4444	0.00	0.05	glc	16.9	3.3	46	2 cd	1
J1744.1–7619l	0.01	0.03	⋯	32.8	9.9	51	1 CPR	1
J1747.6–4037j	0.05	0.07	PSR	10.0	2.8	46	2 ?c	1
J1753.6–4447	0.01	0.08	⋯	11.1	4.0	41	1 cpr	1
J1803.3–6706	0.01	0.08	⋯	10.4	1.8	43	3 ldh	1
J1808.3–3357	0.03	0.09	⋯	8.7	4.4	50	1 cpr	4
J1831.6–6503	0.02	0.10	⋯	8.5	4.5	35	2 cPr	1
J1902.0–5107j	0.04	0.04	PSR	28.9	6.2	50	1 CR	1
J1903.6–7052k	0.08	0.05	PSR	16.3	1.9	52	2 cd	1
J1946.4–5403k	0.01	0.06	⋯	20.2	6.8	39	1 pr	6
$\text{J1959.8}\kern-2pt-\kern-2pt\text{4725}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.02	0.03	bcu	19.2	3.9	49	4 pihr	5
J2039.6–5618	0.04	0.04	⋯	25.3	5.1	34	1 CpR	4
J2043.8–4801	0.01	0.08	⋯	9.5	3.0	35	2 cr	2
J2112.5–3044	0.02	0.04	⋯	30.1	7.6	51	1 CpR	6
J2131.1–6625	0.01	0.11	⋯	10.1	3.1	52	2 rl	2
J2133.0–6433	0.05	0.10	⋯	10.1	4.7	49	2 Pr	7
J2200.0–6930	0.02	0.13	⋯	9.5	0.6	41	3 ld	1
$\text{J2220.6}\kern-2pt-\kern-2pt\text{6833}\kern-0.9in\raise3pt{\Rule{0.9in}{0.1mm}{0.1mm}}$	0.02	0.11	⋯	5.7	1.1	50	4 lh	3
J2241.6–5237p	0.03	0.03	PSR	51.7	12.6	60	1 CpR	1
J2333.0–5525	0.00	0.08	⋯	10.7	2.5	42	2 cr	2

Notes.

^aBoldfaced names denote 17 3FGL sources with now known associated radio and/or gamma-ray pulsars; struck-through names indicate 16 sources that we believe are no longer viable pulsar candidates (see the next-to-last column). ^bOffset between 3FGL position and pointing position in Table 1 (where there are two of the latter, only the second is listed here); the Parkes beam HWHM is 012. ^cSize of 3FGL source error box (95% confidence level semimajor axis; all 3FGL properties listed here are taken from the 3FGL catalog; Acero et al. 2015). ^dClassification from 3FGL pipeline. "PSR" is a pulsar with LAT pulsations; "psr" is a positionally coincident pulsar so far without LAT pulsations; "bll" is a BL Lac object; "bcu" is an unclassified blazar; "glc" is a globular cluster; "fsrq" is a flat-spectrum radio quasar. ^e3FGL source significance. ^fSignificance of curvature of 3FGL source spectrum when fit to a log-parabolic model. ^gVariability index of source ($\gt 73$ indicates variability at the $\gt 99$% C.L.). ^hSee Section 3.5 for a description of these classifications and characteristics. ⁱNumber of observations of each target at the position closest to the 3FGL source (from Table 1), ${\rm{\Delta }}\theta$ offset from it. ^jRadio MSP discoveries from our Parkes survey, first reported in Kerr et al. (2012). ^kFirst reported in this work. ^lPulsar discovered via gamma-ray pulsations (H. J. Pletsch 2015, private communication). ^mIntermittently radio-emitting MSP (Roy et al. 2015). ⁿMSP in 93 minute orbit discovered via gamma-ray pulsations (Pletsch et al. 2012a), subsequently detected at the GBT (Ray et al. 2013). ^oRadio MSP discovered at the GMRT (Ray et al. 2012). ^pRadio and gamma-ray MSP, discovered at Parkes (Keith et al. 2011).

Download table as:  ASCIITypeset image

Three of the five new MSPs were detected unbiasedly in all their search observations (i.e., without prior knowledge of their DM or approximate spin period). However, PSR J1946–5403 was detected in only three of six observations, and PSR J1036–8317 was detected in only the third of three search observations. We now consider the reasons behind these failures to consistently detect some new pulsars.

The nominal sensitivity of our survey to a P = 2 ms pulsar with a duty cycle of 25% and ${\rm{DM}}\lesssim \;40$ pc cm⁻³, for a typical integration time of 1 hr (Table 1) and for a sky temperature corresponding to the average at the locations of the targets searched, was 0.2 mJy—provided that the dilution of power in the Fourier domain caused by orbital motion in a putative binary was ideally corrected by our acceleration search analysis. However, two of the newly discovered MSPs have orbital periods P_b such that the discovery integrations were 0.1–0.3 P_b (Section 2.3), for which this idealized correction breaks down badly; how badly, for a given system, depends on the observed orbital phases (see Figure 1 and Johnston & Kulkarni 1991; Bagchi et al. 2013).

Zoom In Zoom Out Reset image size

As well, a considerable fraction of LAT-selected MSPs are in eclipsing binary systems (e.g., Hessels et al. 2011). Multiple observations of a promising source can thus help combat these factors affecting detectability. In addition, the flux density received from many MSPs is severely affected by interstellar scintillation (see, e.g., Figure 1 of Levin et al. 2013). As we now show, this has played a very significant role in our survey, and the early realization of the magnitude of some of these selection effects is what led us to do multiple observations of many unidentified LAT sources starting in 2010 (Section 2.1.2).

2.2.1. Interstellar Scintillation and Detection Statistics

In order to acquire the detections required to determine the timing solution for PSR J1514–4946 (Section 2.3), we used 100 hr of Parkes telescope time in 79 observations on 54 days spread over 2 years. In only 27 of those observations did we detect the pulsar in a relatively unbiased manner—by dedispersing the raw data using the known DM, performing an acceleration search, and looking for a signal with period 3.589 ms. In several of those observations the pulsar was not detectable without prior knowledge of the DM and approximate period. A pulsar like PSR J1514–4946 is therefore discoverable in a search like ours at Parkes less than one-third of the time! This is due to its small intrinsic flux density combined with very large modulation owing to propagation through the dynamic and inhomogeneous interstellar medium (see Figure 2).

Zoom In Zoom Out Reset image size

This large modulation of observed flux density is primarily caused by strong diffractive interstellar scintillation. This causes the detected pulsar signal strength in the time–frequency plane to form patches (or "scintles") of characteristic size ${\rm{\Delta }}\nu$ and ${\rm{\Delta }}t$. For a given observing frequency, ${\rm{\Delta }}\nu$ increases strongly with decreasing pulsar distance d, while ${\rm{\Delta }}t$ increases particularly with decreasing pulsar velocity in the plane of the sky ${V}_{\perp }$ (see, e.g., Johnston et al. 1998; Nicastro et al. 2001; Lorimer & Kramer 2005). When ${\rm{\Delta }}\nu$ or ${\rm{\Delta }}t$ are large compared to the observing bandwidth and integration time, deep fluctuations in observed flux density result. While we have not measured ${\rm{\Delta }}\nu$ or ${\rm{\Delta }}t$ for PSR J1514–4946, the observations (e.g., Figure 2) are consistent with large values for both (e.g., $\gtrsim 100$ MHz and $\gtrsim 1$ hr, respectively), accounting for the very large observed modulations.

The measured flux density of PSR J1658–5324 is affected by scintillation even more, ranging over a factor $\gt 50$. It was detected in a relatively unbiased manner only 70% of the time, and less than one-half of the time in the absence of prior DM and period information. PSR J0101–6422, another MSP discovered in the first phase of our survey, for which we have good statistics, is detectable 80% of the time (Kerr et al. 2012). These detection statistics are based on typical 1 hr individual Parkes timing observations (Section 2.3).

For PSR J1903–7051, the observed flux densities range over a factor $\gt 10$. PSR J1902–5105 varies less: its recorded flux density has a standard deviation of 40% of the mean. These MSPs are reasonably bright, with relatively narrow pulse profiles, and were always detected in typical 20 minute Parkes timing observations. For PSR J1747–4036, with a large DM = 153 pc cm⁻³, the effects of scintillation modulate the flux density that we record by only 25% about the mean, and it was detected 100% of the time at Parkes and the Robert C. Byrd Green Bank Telescope (GBT; Section 2.3).

Away from the Galactic plane, putative pulsar counterparts to unidentified gamma-ray sources are largely expected to be relatively nearby MSPs, which as a class have relatively small space velocities (e.g., Gonzalez et al. 2011). Depending on observing parameters (frequency, bandwidth, and integration time), these characteristics can make them particularly susceptible to deep flux density fluctuations, in turn with important implications for radio searches. Scintillation modulations in the strong regime have exponential statistics (e.g., Rickett 1990), with median flux density measurements less than the mean. Two of the six MSPs for which we have already obtained timing solutions (Section 2.3) display flux densities that vary by a factor of $\gt 40$, with small median values and, given scintillation statistics, not detectable in an unbiased manner most of the time for the parameters of our Parkes survey.

The foregoing strongly suggests that an unidentified LAT source that is a good pulsar candidate should be searched repeatedly before much of a statement can be made about the likelihood of it being a radio MSP beaming toward the Earth. Eight of the 10 MSPs that we discovered were detected on their first observations (Table 1), but as the preceding account illustrates, in some cases (particularly for PSRs J1514–4946 and J1658–5324), this was fortuitous. Based on our empirical evidence, we judge that every promising LAT source should be observed a minimum of three or four times in a survey such as the one we did at Parkes before it can reasonably be considered searched in an average sense.

As indicated in Table 2, a number of our survey targets remain promising pulsar candidates and at least some should be re-searched in radio. We discuss this further in Sections 3.5 and 3.6.

We began timing observations of all 10 MSPs immediately following their discovery. In Table 3, we present initial parameters for the four most recent discoveries. We have determined phase-connected rotational ephemerides for the six remaining MSPs. That for PSR J0101–6422 was reported in Kerr et al. (2012); the other five are given in Tables 4 and 5, which also list available flux density measurements.

Table 3.  Preliminary Parameters for Four Millisecond Pulsars

	PSR J0955–6150	PSR J1012–4235	PSR J1036–8317	PSR J1946–5403
Right ascensiona, R.A. (J2000.0)	${09}^{{\rm{h}}}{55}^{{\rm{m}}}{39}^{{\rm{s}}}$	${10}^{{\rm{h}}}{12}^{{\rm{m}}}{07}^{{\rm{s}}}$	${10}^{{\rm{h}}}{36}^{{\rm{m}}}{20}^{{\rm{s}}}$	${19}^{{\rm{h}}}{46}^{{\rm{m}}}{24}^{{\rm{s}}}$
Declinationa, decl. (J2000.0)	$-61^\circ 48^{\prime} 36$''	$-42^\circ 35^{\prime} 01$''	$-83^\circ 17^{\prime} 01$''	$-54^\circ 02^{\prime} 46$''
Spin period, P (ms)	1.999	3.101	3.408	2.710
Dispersion measure, DM (pc cm−3)	160.7	71.6	27.0	23.7
Orbital period, Pb (days)	24.578	37.972	0.335	0.130
Projected semimajor axis, x (l-s)	13.283	21.263	0.506	0.0435
Eccentricity, e	0.11	$\lt 0.001$	$\lt 0.001$	$\lt 0.001$
Companion massb, m2 (${M}_{\odot }$)	$\gt 0.21$	$\gt 0.26$	$\gt 0.14$	$\gt 0.021$
Galactic longitude, l (deg)	283.7	274.2	298.9	343.9
Galactic latitude, b (deg)	−5.7	11.2	−21.5	−29.6
DM-derived distancec, d (kpc)	3.8	2.5	1.0	0.9

Notes. The listed P and DM values are from the discovery observations. Orbital parameters are from fits to sets of Doppler-shifted barycentered periods.

^aThese discovery pointing positions have $\pm 7^{\prime}$ uncertainties. On the assumption that the MSPs are associated with the target gamma-ray sources, in some cases the position is better constrained (see Table 2). ^bDerived from the pulsar mass function f₁ assuming ${m}_{1}=1.35$ ${M}_{\odot }$ (Thorsett & Chakrabarty 1999) and $i\lt 90^\circ$. ${f}_{1}={x}^{3}{(2\pi /{P}_{b})}^{2}{T}_{\odot }^{-1}={({m}_{2}\mathrm{sin}i)}^{3}/{({m}_{1}+{m}_{2})}^{2}$, where ${T}_{\odot }\equiv {{GM}}_{\odot }/{c}^{3}=4.925\;\mu$s, m₁ and m₂ are the pulsar and companion masses, respectively, and i is the orbital inclination angle. ^cUsing the Cordes & Lazio (2002) Galactic free electron density model. The individual estimates have substantial uncertainties.

Download table as:  ASCIITypeset image

Table 4.  Parameters for Four Millisecond Pulsars with Coherent Timing Solutions

	PSR J1514–4946	PSR J1658–5324	PSR J1747–4036	PSR J1902–5105
Timing Parameters
Right ascension, R.A. (J2000.0)	${15}^{{\rm{h}}}{14}^{{\rm{m}}}19\buildrel{\rm{s}}\over{.} 1141(1)$	${16}^{{\rm{h}}}{58}^{{\rm{m}}}39\buildrel{\rm{s}}\over{.} 34359(9)$	${17}^{{\rm{h}}}{47}^{{\rm{m}}}48\buildrel{\rm{s}}\over{.} 71692(3)$	${19}^{{\rm{h}}}{02}^{{\rm{m}}}02\buildrel{\rm{s}}\over{.} 84821(9)$
Declination, decl. (J2000.0)	$-49^\circ 46^{\prime} 15\buildrel{\prime\prime}\over{.} 516(5)$	$-53^\circ 24^{\prime} 07\buildrel{\prime\prime}\over{.} 003(1)$	$-40^\circ 36^{\prime} 54\buildrel{\prime\prime}\over{.} 773(1)$	$-51^\circ 05^{\prime} 56\buildrel{\prime\prime}\over{.} 9695(8)$
Proper motion in R.A., $\dot{\alpha }\mathrm{cos}(\delta )$ (mas yr−1)	−0.3(32)	0.2(8)	−0.8(6)	−4.8(13)
Proper motion in decl., $\dot{\delta }$ (mas yr−1)	−30.0(66)	4.9(23)	−4.9(16)	−4.4(16)
Spin frequency, f (Hz)	278.60300920296(5)	409.95436264371(4)	607.67753906573(2)	573.92104496683(5)
Frequency derivative, $\dot{f}$ (Hz s−1)	$-1.4473(8)\times {10}^{-15}$	$-1.8746(6)\times {10}^{-15}$	$-4.8510(5)\times {10}^{-15}$	$-3.0301(4)\times {10}^{-15}$
Epoch (MJD)	55520.0	55520.0	55520.0	55520.0
Dispersion measure, DM (pc cm−3)	31.05(2)	30.81(3)	152.98(1)	36.25(1)
Orbital period, Pb (days)	1.922653523(5)	⋯	⋯	2.0118037388(9)
Projected semimajor axis, x (l-s)	1.933268(2)	⋯	⋯	1.9019570(7)
Time of ascending node, ${T}_{\mathrm{asc}}$ (MJD)	55585.8605555(3)	⋯	⋯	55162.2815604(1)
$e\mathrm{sin}\omega$ a, EPS1	$6.453587(2)\times {10}^{-6}$	⋯	⋯	$5.5239429(7)\times {10}^{-6}$
$e\mathrm{cos}\omega$ a, EPS2	$8.789469(3)\times {10}^{-6}$	⋯	⋯	$-1.9671264(9)\times {10}^{-6}$
Span of timing data (MJD)	55160–57013	55166–57057	55161–56993	55161–57014
rms timing residual (μs)	8.3	2.9	2.4	3.8
Flux Densitiesb and Rotation Measures
0.8 GHz flux density, ${S}_{0.8}$ (mJy)	⋯	⋯	4.8	⋯
1.4 GHz flux density, ${S}_{1.4}$ (mJy)	0.17 ± 0.11 (N = 71)	0.7 ± 0.5 (N = 37)	⋯	1.2 ± 0.5 (N = 69)
1.5 GHz flux density, ${S}_{1.5}$ (mJy)	⋯	⋯	0.9	⋯
2 GHz flux density, S2 (mJy)	⋯	⋯	0.5 ± 0.1 (N = 28)	⋯
Rotation measure, RM (rad m−2)	35 ± 15	4 ± 7	$-39\pm 2$	⋯
Derived Parametersc
Spin period, P (ms)	3.589	2.439	1.645	1.742
Characteristic age, ${\tau }_{{\rm{c}}}$ (109 years)	4.9	3.5	2.0	3.1
Spin-down luminosity, $\dot{E}$ (1034 erg s−1)	1.0	3.0	11.3	6.7
Surface dipole magnetic field strength, B (108 G)	2.1	1.7	1.5	1.3
Eccentricitya, e	$(1.1\pm 0.2)\times {10}^{-5}$	⋯	⋯	$(5.9\pm 0.7)\times {10}^{-6}$
Mass function, f1 (${M}_{\odot }$)	0.00210	⋯	⋯	0.00183
Companion mass, m2 (${M}_{\odot }$)	$\gt 0.17$	⋯	⋯	$\gt 0.16$
Spectral indexd, α	⋯	⋯	$\approx -2.5$	⋯
Galactic longitude, l (deg)	325.25	334.87	350.21	345.65
Galactic latitude, b (deg)	6.81	−6.63	−6.41	−22.38
DM-derived distance, d (kpc)	0.9	0.9	3.4	1.2
Composite proper motion, μ (mas yr−1)	30.0 ± 6.6	4.9 ± 2.2	5.0 ± 1.6	6.5 ± 1.4
Transverse velocity, ${V}_{\perp }$ (km s−1)	$\approx 130$	$\approx 20$	$\approx 80$	$\approx 40$

Notes. All uncertainties are reported at the $1\;\sigma$ level. Numbers in parentheses represent the TEMPO2 timing uncertainties on the last digits quoted.

^aThe orbital eccentricities were derived from the parameters $(e\mathrm{sin}\omega ,e\mathrm{cos}\omega )$, fitted using the TEMPO2 ELL1 binary model (Lange et al. 2001). ^b ${S}_{1.4}$ and S₂ values are averages and standard deviations for N detections. Individual flux densities were determined by computing the area under each pulse profile compared to its off-pulse rms, scaled using the measured system equivalent flux density at the location of the pulsar. Values for PSR J1747–4036 are from GBT observations; ${S}_{0.8}$ and ${S}_{1.5}$ are from single flux-calibrated observations (Figure 3(c)) and have uncertainties $\lesssim 10\%$. For one similar observation at 2 GHz, S₂ = 0.45 mJy. ^cThe following have been used: ${\tau }_{{\rm{c}}}=P/(2\dot{P})$, $\dot{E}=4{\pi }^{2}\times {10}^{45}\dot{P}/{P}^{3}$ erg s⁻¹, and $B=3.2\times {10}^{19}{(P\dot{P})}^{1/2}$ G, with P in s, where $P=1/f$. The listed values of these parameters include corrections for acceleration effects (mainly due to proper motion; see Camilo et al. 1994). ^d ${S}_{\nu }\propto {\nu }^{\alpha }$, where ${S}_{\nu }$ is the flux density at frequency ν.

Download table as:  ASCIITypeset image

Table 5.  Radio and Gamma-Ray Parameters of PSR J1903–7051

Parameter	Value
Timing Parameters
Right ascension, R.A. (J2000.0)	${19}^{{\rm{h}}}{03}^{{\rm{m}}}38\buildrel{\rm{s}}\over{.} 7935(3)$
Declination, decl. (J2000.0)	$-70^\circ 51^{\prime} 43\buildrel{\prime\prime}\over{.} 461(2)$
Proper motion in R.A., $\dot{\alpha }\mathrm{cos}(\delta )$ (mas yr−1)	$-8.8(16)$
Proper motion in decl., $\dot{\delta }$ (mas yr−1)	$-16(2)$
Spin frequency, f (Hz)	277.94006243351(8)
Frequency derivative, $\dot{f}$ (Hz s−1)	$-8.06(4)\times {10}^{-16}$
Epoch (MJD)	56526.0
Dispersion measure, DM (pc cm−3)	19.66(1)
Orbital period, Pb (days)	11.05079833(2)
Projected semimajor axis, x (l-s)	9.938869(2)
Time of ascending node, ${T}_{\mathrm{asc}}$ (MJD)	56027.2292914(7)
$e\mathrm{sin}\omega$, EPS1	$2.0261968(5)\times {10}^{-6}$
$e\mathrm{cos}\omega$, EPS2	$1.1855635(5)\times {10}^{-7}$
Span of timing dataa (MJD)	54760–57013
rms timing residuala (μs)	12.8
Flux Densitiesb and Rotation Measure
1.4 GHz flux density, ${S}_{1.4}$ (mJy)	$\approx 0.6$ (N = 9)
3.1 GHz flux density, ${S}_{3.1}$ (mJy)	∼0.14 (N = 3)
Rotation measure, RM (rad m−2)	11 ± 24
Derived Parametersc
Spin period, P (ms)	3.597
Characteristic age, ${\tau }_{{\rm{c}}}$ (109 years)	7.1
Spin-down luminosity, $\dot{E}$ (1033 erg s−1)	6.8
Surface dipole magnetic field strength, B (108 G)	1.7
Eccentricity, e	$(2.0\pm 0.5)\times {10}^{-6}$
Mass function, f1 (${M}_{\odot }$)	0.00863
Companion mass, m2 (${M}_{\odot }$)	$\gt 0.28$
Spectral index, α	$\sim -1.8$
Galactic longitude, l (deg)	324.39
Galactic latitude, b (deg)	−26.51
DM-derived distance, d (kpc)	0.8
Composite proper motion, μ (mas yr−1)	18.3 ± 2.0
Transverse velocity, ${V}_{\perp }$ (km s−1)	$\approx 70$
Gamma-ray Parameters
Gamma-ray–radio profile lagd, δ (P)	0.55 ± 0.05
Gamma-ray ($\gt 0.1$ GeV) photon index, Γ	1.90 ± 0.16
Gamma-ray cutoff energy, Ec (GeV)	7.8 ± 4.0
Photon flux ($\gt 0.1$ GeV) (${10}^{-8}\;{\mathrm{cm}}^{-2}\;{{\rm{s}}}^{-1}$)	1.2 ± 0.3
Energy flux ($\gt 0.1$ GeV) (${10}^{-11}\;\mathrm{erg}\;{\mathrm{cm}}^{-2}\;{{\rm{s}}}^{-1}$)	0.9 ± 0.2

Notes. All uncertainties are reported at the $1\;\sigma$ level. Numbers in parentheses represent the TEMPO2 timing uncertainties on the last digits quoted. Uncertainties for gamma-ray spectral parameters are statistical only (for a discussion of systematic errors, see Acero et al. 2015).

^aWe have used both radio and gamma-ray TOAs to derive this timing solution. ^bThe listed values are median flux densities for N calibrated detections. Individual ${S}_{1.4}$ values ranged over 0.13–1.5 mJy. In four 3.1 GHz observations, we did not detect the pulsar. ^cThe listed values of ${\tau }_{{\rm{c}}}$, $\dot{E}$, and B include corrections for acceleration effects. ^dThis is measured from the profiles in Figure 4, with the gamma-ray profile centroid at phase ϕ = 0.27 and the radio profile reference phase mid-way between its full observed span ($0.55\lt \phi \lt 0.9$).

Download table as:  ASCIITypeset image

PSR J1747–4036 was observed mainly at the GBT, using the GUPPI spectrometer¹⁵ to sample a bandwidth of 800 MHz centered at 2 GHz, with typical integration times of 5 minutes.

The remaining MSPs were observed exclusively at Parkes, where we first used the analog filterbank/PMDAQ data acquisition system as employed in the discovery observations, and later a digital filterbank (PDFB3/4), in all cases centered at 1.4 GHz. Each observation typically lasted for 1 hr, except for the brighter PSRs J1902–5105 and J1903–7051 (about 20 minutes). Apart from a dense set of observations to obtain orbital parameters for the binary pulsars and to unambiguously establish pulse numbering, we aimed to detect each pulsar approximately monthly over a span of 1–2 years. Particularly for PSR J1514–4946, this required a very large number of observations because the pulsar is very faint on average and its received radio flux density varies enormously due to interstellar scintillation (Section 2.2.1). We improved the precision of the timing solutions by substantially extending the measurement baselines with a few additional observations in 2014 and early 2015.

We used the PRESTO and PSRCHIVE (Hotan et al. 2004) software to analyze the raw data and obtain pulse times of arrival (TOAs). For the MSPs with sparse radio detections, we obtained initial estimates of the orbital parameters using the method described in Freire et al. (2001). We then used TEMPO¹⁶ and TEMPO2 (Hobbs et al. 2006) to obtain phase-connected timing solutions. Starting with solutions spanning $\approx 1$ year for each pulsar, we detected gamma-ray pulsations for six MSPs. For PSR J1903–7051, we then obtained a few LAT TOAs. While the radio TOAs have much higher precision, they span only 3 years. Adding gamma-ray TOAs increases the solution span and improves some results, in particular the proper motion measurement (see Section 2.5.2).

In principle, study of the polarized radio emission from pulsars can constrain the magnetic field geometry and, particularly when considered together with gamma-ray profile characteristics, can elucidate emission locations and mechanisms (cf. Section 3.2).

In order to measure the polarization characteristics of each pulsar, we used the digital filterbank PDFB3 at Parkes, and GUPPI at GBT in coherent dedispersion mode, to do a few observations recording calibrated, folded full-Stokes pulse profiles. These data were analyzed in standard fashion with PSRCHIVE. The resulting profiles are shown in Figures 3 and 4, and rotation measures (RMs) are listed in Tables 4 and 5.

Zoom In Zoom Out Reset image size

The main profile component of PSR J1514–4946 is about 80% linearly polarized, with a flat position angle (P.A.) of linear polarization (Figure 3(a)).

PSR J1658–5324 has a complex profile, with a high (up to 100%) linear polarization fraction in most pulse components (Figure 3(b)). Such a high level of linear polarization is relatively uncommon among MSPs (see Yan et al. 2011; Dai et al. 2015). Despite the P.A. being measured over a large span of the rotation phase, we were not able to obtain a satisfactory rotating vector model fit (Radhakrishnan & Cooke 1969).

The profile of PSR J1747–4036 is curious. This MSP has the second largest value of ${\rm{DM}}/P$ among those known in the Galactic disk, and so was particularly well suited for observing with a coherent dedispersion system. We show two such observations in Figure 3(c). At 1.5 GHz, the profile has two featureless components, with a very flat P.A. throughout. Lack of variation in P.A. across the entire profile is unusual (there is no such example in a well-studied sample of 24 MSPs; Dai et al. 2015). The profile observed at 2 GHz (which overlaps in frequency with that at 1.5 GHz) looks similar. At 0.8 GHz, however (bottom panels of Figure 3(c)), the second component is brighter than the first, signifying that it has a steeper spectrum, and a third component appears (more easily discernible in linear polarization), with P.A.s offset by about $90^\circ$ from the rest of the profile. At this frequency, emission is detectable across 85% of the pulse period (but the observed profile may be somewhat affected by multi-path propagation; Cordes & Lazio 2002 predict pulse broadening of $0.02\;P$).

The PSR J1902–5105 profile is relatively unusual in showing no discernible linear polarization (there is possibly a very small amount of circular polarization in both principal components; Figure 3(d)).

The radio profile of PSR J1903–7051 is shown in the bottom panel of Figure 4. It displays a weak, highly polarized component (pulse phase $\phi \approx 0.58$) leading a much stronger double-peaked component ($0.65\lesssim \phi \lesssim 0.85$). The first peak is highly linearly polarized and also shows a degree of circular polarization, while the trailing (brightest) peak is completely unpolarized.

Zoom In Zoom Out Reset image size

The gamma-ray properties of the first five MSPs we discovered (PSR J0101–6422 and those in Table 4) have been reported elsewhere (Kerr et al. 2012; Abdo et al. 2013), and we focus here on the newly detected gamma-ray pulsations of PSR J1903–7051. To characterize its properties, we analyzed "reprocessed Pass 7" SOURCE class Fermi-LAT events collected between 2008 August 4 (the start of nominal operations) and 2015 February 1 and with energies between 0.1 and 30 GeV. The data were filtered to exclude events whose reconstructed direction exceeds a zenith angle of 100° and those taken when the observatory was rocked more than 52° from the zenith, or when the region of interest (ROI) around the pulsar approached the Earth's limb. These cuts minimize the bright gamma-ray background contribution from the limb of the Earth.

2.5.1. Spectral Analysis and Photon Weights

2.5.2. Pulsar Timing

2.5.3. Light Curve

In an attempt to assist in determining accurate positions for the pulsars before timing solutions existed, we undertook X-ray observations of the fields of PSRs J1514–4946 and J1658–5324 with the ACIS-S camera on board the Chandra X-ray Observatory (CXO). For both observations, the (then) best pulsar positions were centered on the back-illuminated S3 chip, and we analyzed the event data with the standard CIAO version 4.7 software (Fruscione et al. 2006). After excluding events outside the energy range 0.3–7.0 keV, we searched the fields for any point-like X-ray source possibly associated with the pulsars using the celldetect source search tool. In each case, we detected a faint source $0\buildrel{\prime\prime}\over{.} 3\pm 0\buildrel{\prime\prime}\over{.} 6$ away from the pulsar timing position. We name these, respectively, CXOU J151419–494615 and CXOU J165839–532406, which we identify as the pulsar counterparts on the basis of positional coincidence. Observation and source parameters are listed in Table 6.

Table 6.  X-Ray Observations of Five Millisecond Pulsars

	PSR J1514–4946	PSR J1658–5324	PSR J1747–4036	PSR J1902–5105	PSR J1903–7051
Telescope/Instrument	CXO/ACIS-S	CXO/ACIS-S	Swift/XRT PC	Swift/XRT PC	Swift/XRT PC
Exposure (ks)	9.9	9.9	3.3	4.2	3.2
Background-subtracted counts	9	23	$\lt 6.6$ a	$\lt 12.7$ a	$\lt 6.6$ a
${N}_{{\rm{H}}}$ b (1021 cm−2)	1	1	5	1	0.6
X-ray fluxc, ${f}_{0.1-2.4\;\mathrm{keV}}$ (10−14 erg cm−2 s−1)	1.1	2.9	$\lt 23$	$\lt 12$ d	$\lt 7$
X-ray luminositye, ${L}_{0.1-2.4\;\mathrm{keV}}$ (${10}^{-4}\dot{E}$)	1.1	0.9	$\lt 28$	$\lt 3$	$\lt 8$

Notes. A blackbody model with kT = 0.2 keV is assumed in all cases, absorbed by the indicated column density ${N}_{{\rm{H}}}$. Considering instead a power-law spectrum with photon index Γ = 2 and calculating flux/luminosities for the 2–10 keV range does not fundamentally alter our conclusions (Section 2.6).

^aWe used $47^{\prime\prime}$ radius extraction regions around each pulsar (90% PSF radius for XRT) and $141^{\prime\prime}$ radii to estimate backgrounds. For two sources, we obtained zero background-subtracted counts, while for PSR J1902–5105 we obtained three counts. In each instance, we convert to a $3\sigma$ upper limit (see Gehrels 1986). ^bAbsorbing columns are estimated from the DMs according to ${N}_{{\rm{H}}}\;({10}^{21}\;{{\rm{cm}}}^{-2})=0.03{\rm{DM}}$ (see He et al. 2013). ^cWe used http://heasarc.gsfc.nasa.gov/Tools/w3pimms.html to obtain these unabsorbed flux estimates (or $3\sigma$ limits). ^dTakahashi et al. (2012) report a somewhat lower flux limit based on a 38 ks Suzaku XIS observation. ^eIsotropic 0.1–2.4 keV luminosities are given in terms of the acceleration-corrected values of $\dot{E}$ and for the DM-derived distances listed in Tables 4 and 5.

Download table as:  ASCIITypeset image

For three other MSPs, Swift X-Ray Telescope (XRT) observations have been undertaken as part of a Swift campaign of observations of Fermi-LAT unassociated sources. No sources were detected at the pulsar positions. More information, including flux and luminosity upper limits, is given in Table 6.

All these X-ray detections and upper limits are consistent with the known distribution of MSP X-ray luminosities (e.g., Possenti et al. 2002 and references therein).

Two of the 10 new pulsars are isolated (PSRs J1658–5324 and J1747–4036; Table 4). Another five are in circular orbits ($e\lt {10}^{-3}$) with 0.2–0.3 ${M}_{\odot }$ companions, likely helium-core white dwarfs (PSRs J0101–6422, J1012–4235, J1514–4946, J1902–5105, and J1903–7051; Kerr et al. 2012 and Tables 3–5). Three of these have orbital periods of about 2 days, while the other two have periods of 11 and 38 days. These characteristics are typical of MSPs known in the Galactic disk. The timing precision of PSR J1747–4036 is good enough that it has been added to the NANOGrav gravitational wave pulsar timing array (McLaughlin 2013).

The remaining three systems (see Table 3) are less common. PSR J0955–6150 is in a P_b = 24 day orbit with a ${m}_{2}\approx 0.25$ ${M}_{\odot }$ companion, but with a very significant eccentricity (e = 0.11). It joins four Galactic MSP systems with broadly similar parameters: $0.2\lt {m}_{2}\lt 0.3$ ${M}_{\odot }$, $0.03\lt e\lt 0.13$, $22\lt {P}_{b}\lt 32$ days (see Antoniadis 2014; Knispel et al. 2015 and references therein). Such eccentric systems are not predicted through the standard MSP formation channels (see Phinney & Kulkarni 1994), leading to alternative scenarios (Antoniadis 2014; Freire & Tauris 2014).

PSR J1036–8317 is in an 8 hr orbit with a $\approx 0.16$ ${M}_{\odot }$ companion, similar to so-called "redback" systems (where outflows from $\gtrsim 0.15$ ${M}_{\odot }$ non-degenerate companions can cause irregular radio eclipses of the pulsar; e.g., D'Amico et al. 2001). Despite many observations at essentially all orbital phases, no radio eclipses have been detected. This is therefore unlikely to be a redback. It is more likely that the companion is a white dwarf. If its distance is $\approx 1.0$ kpc, as inferred from the DM, this could be a good target for optical studies; it may be a system similar to PSR J1012+5307, an MSP in a 14 hr orbit with a spectroscopically identified 0.16 ${M}_{\odot }$ white dwarf (van Kerkwijk et al. 1996), or to PSR J1738+0333 (Antoniadis et al. 2012).

PSR J1946–5403 is in a 3 hr orbit with a $\gt 0.021$ ${M}_{\odot }$ companion. These parameters suggest a "black widow" interacting system (sub-day binaries with degenerate $\lesssim 0.05\;{M}_{\odot }$ companions; e.g., Fruchter et al. 1988). Most such systems show radio eclipses near superior conjunction, but so far we have not detected any (when folded using the known orbital parameters, the pulsar is detected in all six data sets listed in Table 1, including two observations of superior conjunction). This could be due simply to geometry: if the actual companion mass is slightly larger than the minimum value inferred from the mass function, e.g., if ${m}_{2}\gtrsim 0.025$ ${M}_{\odot }$, the orbital inclination angle $i\lesssim 60^\circ$, and we could be viewing the system relatively face-on. In any case, with a DM-derived distance of 0.9 kpc, this may also be an interesting optical target.

It seems curious that among the 10 MSPs discovered in this Parkes survey only one is in an interacting binary system (either black widow or redback), when about 50% of the 67 MSPs so far discovered in searches of unidentified LAT sources worldwide are in such systems (see Roberts 2013). For instance, in a recent survey at Arecibo, five of six MSPs discovered are either black widows or redbacks (H. T. Cromartie et al. 2015, in preparation). However, the Arecibo searches have integration times of T = 15 minutes, which are far more suitable for the discovery of few-hour binaries than the 1 hr Parkes integrations: the maximum accelerations probed by our searches scale as ${T}^{-2}$. At the GBT, where half of these 67 MSPs were discovered, typical integration times are 30–45 minutes (e.g., Ransom et al. 2011; P. Bangale et al. 2015, in preparation). In any case, the pulsars that we did not detect unbiasedly in every one of our Parkes survey observations are the two sub-day binaries (Table 1), at least in part due to large and rapidly changing orbital acceleration (cf. Figure 1).

Radio timing observations since 2009 have yielded rotational ephemerides for 43 of the 67 MSPs discovered in LAT-guided searches. Of these, 39 are now confirmed as gamma-ray MSPs (the other four are unrelated to the LAT sources; e.g., Keith et al. 2011), including the six discovered in our Parkes survey for which we already have timing solutions. In this paper, we have for the first time presented the ephemerides and polarimetry for five of these MSPs and gamma-ray properties for PSR J1903–7051. In some cases, these results add to our understanding of the pulsars summarized in the second LAT catalog of gamma-ray pulsars (2PC; Abdo et al. 2013).

For PSR J1903–7051, the radio peak leads the gamma-ray profile by $\delta \approx 0.5\;P$. Other pulsars with such values of δ have gamma-ray profiles composed of only one peak (2PC), which is also the case for PSR J1903–7051 (Figure 4). This is one of the few MSPs for which spectral curvature (deviation from a power-law spectrum) is not apparent even with 4 years of LAT data (Table 2). Our spectral fits to 6.5 years of data yield the largest cutoff energy of any pulsar, even if with large uncertainty: ${E}_{{\rm{c}}}=(7.8\pm 4.0)$ GeV (Table 5; see also Figure 5(f)). The MSPs with the next largest values of E_c are also from our Parkes sample: PSRs J1747–4036 and J1514–4946 (2PC).

Zoom In Zoom Out Reset image size

With our proper motion measurement for PSR J1514–4946 (Table 4), its intrinsic $\dot{E}$ is 1.6 times smaller than previously thought. Its computed gamma-ray efficiency, already high in 2PC (30%), now increases to 50%. As usual, this assumes a geometry-dependent beaming correction of ${f}_{{\rm{\Omega }}}=1$ (see 2PC for definition), i.e., assuming isotropic emission. Perhaps for PSR J1514–4946, ${f}_{{\rm{\Omega }}}\ll 1$. Or maybe its DM-derived distance of 0.9 kpc is an overestimate (its nominal transverse velocity is the largest of those listed in Tables 4 and 5, although it is not unusually large for an MSP; e.g., Gonzalez et al. 2011).

PSR J1747–4036 is notable in several respects. Spectral curvature is not apparent in the 3FGL catalog (see Table 2). Its $\dot{E}$ is large (fourth highest among the 2PC MSP sample). Its nominal distance (Table 4) is also among the largest in the 2PC MSP sample, and it is a relatively faint gamma-ray source ($10\;\sigma$ in 3FGL; Table 2). If it had a typical $\dot{E}$ at that distance, presumably it would not be a detectable LAT pulsar. This large $\dot{E}$ ultimately is due to its very short spin period (third shortest among disk MSPs). While PSR J1747–4036 may be somewhat extreme in this regard, it does appear that LAT-detected MSPs are a smaller-P, higher-$\dot{E}$ population than radio-selected MSPs (Ray et al. 2012). In our Parkes sample, PSRs J1902–5105 and J0955–6150 (the latter apparently quite distant) are also short-period MSPs, with $P\lt 2$ ms.

PSR J1902–5105 is one of only six MSPs known with phase-aligned gamma-ray and radio light curves (2PC). This points to emission that is either extended and caustic in nature or originates near the neutron star surface (Venter et al. 2012). In order to constrain their emission and viewing geometry, Johnson et al. (2014) have jointly modeled the gamma-ray and radio profiles of 40 LAT-detected MSPs. They fit gamma-ray light curves using standard outer magnetosphere gap (OG) and two-pole caustic (TPC) geometric models assuming a vacuum-retarded dipole magnetic field. For PSR J1902–5105, Johnson et al. (2014) consider altitude-limited (al) versions of the standard models as well as a low-altitude slot gap model (laSG). They find that all three models match the gamma-ray and radio profiles fairly well. However, the complete lack of polarization that we observe (Figure 3(d)) strongly favors the alTPC and alOG models, where high-altitude caustics are predicted to have a strong depolarizing effect (Dyks et al. 2004), over the laSG model, where one expects non-zero polarization from near-surface emission.

For PSR J1514–4946, the high level of observed linear polarization and modest P.A. swing (Figure 3(a) and Section 2.4) suggest we are viewing the edge of a cone beam at a relatively low altitude. For PSR J1658–5324, the very high level of linear polarization suggests that at least some components originate at relatively low altitude (the small peak with low polarization, which might also originate from the opposite pole, could be a high-altitude caustic; Figure 3(b)). PSR J1747–4036 has a very unusual polarization pattern (Figure 3(c)), which does not fit either standard radio core/cone or high-altitude caustic emission models. For these MSPs, Johnson et al. (2014) find models that plausibly match the gamma-ray light curves, but the radio profiles are not well reproduced.

Our survey of unidentified LAT sources had a success rate of 20% (11 MSPs detected unbiasedly in 56 sources searched; Section 2.1), despite significant selection effects (Section 2.2). This is very encouraging—a substantial number of unidentified LAT sources contain previously unknown MSPs that can be detected at Parkes, and our source selection criteria have allowed us to target them with satisfactory efficiency. However, comparing the initial segment of the survey (Section 2.1.1) with the latter portion (Section 2.1.2) reveals a drop in efficiency: we had to search 3.5 times as many sources in 2012, for twice as long in the aggregate, to discover as many pulsars as earlier on.

Since our observations, another six pulsars have been discovered in these very same sources: four via direct pulsation searches of the gamma-ray photons, one of which was subsequently detected as a very faint radio source with the GBT (the others might be significantly affected by scintillation, or perhaps they may yet prove to be gamma-ray-only pulsars); one at the Giant Metrewave Radio Telescope (GMRT) at a lower frequency than at Parkes (perhaps the Parkes non-detections reflect a steep radio spectrum); and one that is a transient radio source—not active when we did our Parkes searches, but now detectable there (see notes l–o in Table 2). Considering these additional detections, a full 10 of the original 14 targets of our survey are now known to contain pulsars! This reflects superb source selection criteria for those targets. Why has the success rate decreased since then? The original (1FGL-based) sources were brighter in gamma-rays, allowing for relatively unambiguous determination of spectral characteristics, on which we based our target selection. There might be secondary effects related to this: e.g., the brighter 1FGL-based targets might be nearer to the Earth on average, and any associated radio pulsars could then also be brighter on average, although we do not see such an effect among our small sample of detected MSPs.

Now that the third catalog of LAT sources is in hand, based on substantially more data and a much better understanding of the instrument and background than was available earlier, we can usefully revisit our 56 sources and the MSPs found amid them. In Table 2, we have summarized some properties of these sources as obtained from the 3FGL catalog. Given that gamma-ray pulsars typically have exponentially cutoff power-law spectra (Abdo et al. 2013), in contrast to power-law spectra for active galactic nuclei, it is not surprising that the spectral curvature gleaned from 3FGL correlates well with gamma-ray pulsars: of the 12 pulsars with known gamma-ray pulsations listed in Table 2 (the eight classified as "PSR," as well as J1035.7–6720, J1227.9–4854, J1624.2–4041, and J1744.1–7619), 10 display significant curvature according to 3FGL (${\rm{Curve}}\gt 4\;\sigma$; Table 2). However, two other established gamma-ray MSPs (PSRs J1747–4036 and J1903–7051, reported here) show no curvature according to 3FGL (${\rm{Curve}}\lt 3\;\sigma$). Now that 170 gamma-ray pulsars are known, it is to be expected that some may depart from the norm. In addition, "Curve" in 3FGL tests against a log-parabolic spectral model, which may not be a good proxy for testing against exponentially cutoff power laws, especially when the source is not very bright or the background is problematic (as it happens, these two MSPs are the faintest in gamma-rays of the 12 under consideration; see "Sig" in Table 2). Thus, in considering whether a particular 3FGL source is a good pulsar candidate, we should do more than simply look for significant cataloged curvature.

Flux information is available from the 3FGL pipeline processing in five energy bins (two within 0.1–1 GeV, two within 1–10 GeV, and one for 10–100 GeV). After familiarizing ourselves with how known gamma-ray pulsars (and control sources) appear in five-bin spectra by visual inspection of many 3FGL plots, we have classified the 56 target sources of our Parkes survey according to a heuristic ranking scheme that we describe next.

We qualitatively assess the likelihood of a source being a pulsar. According to this scheme, a classification of "1" denotes near certainty of being a pulsar; "2" is less conclusive but quite plausible; "3" is a poorer pulsar candidate; "4" is very likely not a pulsar; "5" is not a pulsar.

These classifications are based on combinations of the following spectral characteristics (both are listed for each source in Table 2): strong energy cutoff (c); parabolic (peak at center energies, p); flat or rising spectrum at $\lt 1$ GeV (r)—c, p, r are "positive" features leading to higher likelihood of a source being a pulsar (for examples, see Figures 5(c) and (e)). Power law (l); variability (v); monotonically decreasing (d) or increasing (i); excess high-energy emission (flat or rising spectrum; h)—l, v, d, i, h are "negative" features, not ordinarily associated with pulsars (see Figures 5(a) and (d)). Capitalization means that the features are more certain, except for p/P which indicates the sharpness of the parabola. A "?" indicates poor spectral quality, increasing uncertainty in classification. An asterisk represents an odd spectrum with "banana" shape to high energy (see Figure 5(b)). In Figure 5, we show 3FGL spectral plots for six of our Parkes targets that illustrate the features described above.

Sixteen of our 56 sources are classified as 5 or 4 (or 3 with a blazar association), and we regard them as no longer viable pulsar candidates.

Of the 12 known gamma-ray pulsars in Table 2, nine are classified as 1 and two others (already noted in Section 3.4 as not curved in 3FGL) as 2. The one confirmed gamma-ray MSP classified as a 3, PSR J1227–4853, is a spectral outlier (Figure 5(b)). In a rare state-changing binary system, it is borderline variable within 3FGL and recently displayed significant variability (see Johnson et al. 2015). This and its sister system J1023+0038 (Stappers et al. 2014), as well as the young pulsar J2021+4026 (Allafort et al. 2013), are counterexamples to the usual assumption that pulsars are steady gamma-ray emitters. Of the other five MSPs known in Table 2 (none of which has yet a reported long-term timing solution or detected gamma-ray pulsations), one is classified as 1, one as 2, and three as 3. Based on prior statistics, we expect that most, perhaps all, of these five MSPs will eventually be established as gamma-ray pulsars. That most have a relatively poor classification in our scheme likely reflects their faintness (three, all discovered in our survey, have a 3FGL significance of 6.6–8.8 $\;\sigma$) or location close to the Galactic plane (PSR J1536–4948 is at $b=4\buildrel{\circ}\over{.} 8$).

Following the discussion in Section 3.5, while we certainly recommend additional searches first of the "1" sources that remain without coincident pulsars (of which there are seven in Table 2), followed by the "2" sources (nine in Table 2), the "3" sources (seven in Table 2) are also reasonable candidates for further searches.¹⁹ Among these 23 sources, two are coincident with globular clusters that currently have no known pulsars, and a population of MSPs in those clusters is a plausible origin for the gamma-ray emission. We expect further MSP discoveries among these 23 targets: while the largest offset between our optimized search and the 3FGL positions is 3', which is not very large compared to the $7\buildrel{\,\prime}\over{.} 2$ HWHM Parkes beam, many promising targets have not been searched enough to counter the selection effects presented in Section 2.2—e.g., four of the seven "1" sources have been searched at Parkes only once or twice at locations not far from their 3FGL positions (see Table 2). In addition, in the near future we intend to re-analyze our existing data sets with "jerk searches," i.e., accounting for changing accelerations that are relevant for short orbital periods (cf. Figure 1).

After all such searches are done, it remains a possibility that through a combination of faintness and extreme orbital and spin parameters, some of those promising LAT sources could still harbor undetected radio MSPs beamed toward the Earth. It is also possible that some of these sources may be MSPs that are detectable only via their gamma-ray emission, although this fraction is known to be small (e.g., Romani 2012).