The Parkes Southern Pulsar Survey II. Final results and population analysis

Mon. Not. R. Astron. Soc. 295, 743 755 (1998) The Parkes Southern Pulsar Survey II. Final results and population analysis A. G. Lyne, 1 R. N. Manchester, 2 D. R. Lorimer, 1;3 M. Bailes, 2;4 N. D Amico, 5;6 T. M. Tauris, 7 S. Johnston, 8 J. F. Bell 1;9 and L. Nicastro 1;10 1 Nuffield Radio Astronomy Laboratories, University of Manchester, Jodrell Bank, Macclesfield, Cheshire SK11 9DL 2 Australia Telescope National Facility, CSIRO, PO Box 76, Epping, NSW 2121, Australia 3 Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, D 53121 Bonn, Germany 4 Physics Department, University of Melbourne, Parkville, Victoria 3052, Australia 5 Osservatorio Astronomico di Bologna, via Zamboni 33, 40126 Bologna, Italy 6 Istituto di Radioastronomia del CNR, Via Gobetti 101, 40129 Bologna, Italy 7 Institute of Physics and Astronomy, Aarhus University, DK 8000 Aarhus C, Denmark 8 Research Centre for Theoretical Astrophysics, University of Sydney, NSW 2006, Australia 9 Mount Stromlo and Siding Spring Observatories, ANU, Private Bag, Weston Creek, ACT 2611, Australia 10 Istituto di Tecnologie e Studie delle Radiazioni Extraterrestri del CNR, via Gobetti 101, 40129 Bologna, Italy Accepted 1997 September 15. Received 1997 September 15; in original form 1997 May 7 ABSTRACT A survey of the entire southern sky for millisecond and low-luminosity pulsars using the ATNF Parkes radio telescope has now been completed. The survey detected 298 pulsars, of which 101 were previously unknown. The new pulsars include 17 millisecond pulsars. This is the largest sample of both normal and millisecond pulsars detected in any survey. Combining our sample with other recent surveys in the Northern Hemisphere, we present a statistical study of the populations of both normal and millisecond pulsars. We find that the improved statistics allow us to estimate the number and birth-rate of both types of pulsar down to a 400-MHz luminosity limit of 1 mjy kpc 2. The local surface densities of potentially observable normal pulsars and millisecond pulsars are both about 30 kpc ¹2, corresponding to 30 000 potentially observable pulsars of each type in the Galaxy. Once beaming effects are taken into consideration we estimate that the active population of normal pulsars is 160 000. Although there is evidence for flattening of the luminosity function of normal pulsars, this is not evident for millisecond pulsars which probably have a substantial population with luminosities below 1 mjy kpc 2. After correcting for beaming effects, we estimate that a normal pulsar is born with a luminosity greater than 1 mjy kpc 2 between once every 60 and 330 yr in the Galaxy. The birth-rate of millisecond pulsars is at least 3 10 ¹6 yr ¹1 above the same luminosity limit. Modelling the observed transverse speeds of millisecond pulsars using a dynamical simulation, we find their mean birth velocity to be 130 30 km s ¹1, significantly lower than that of the normal pulsars. Key words: methods: statistical surveys pulsars: general Galaxy: stellar content. 1 INTRODUCTION A survey of the entire Southern Hemisphere for millisecond and low-luminosity pulsars has been conducted at a radio frequency of 436 MHz using the Parkes radio telescope of the Australia Telescope National Facility (ATNF). The technical details and the results of the first part of the survey are described by Manchester et al. (1996) (hereafter Paper I). In this paper, we report on the results of the second part of the survey. During the course of the survey, a total of 298 pulsars have been detected, of which 19 are millisecond pulsars (Johnston et al. 1993; Bailes et al. 1994, 1997; Lorimer et al. 1995, 1996; Stappers et al. 1996), including by far the brightest and one of the closest millisecond pulsars known, PSR J0437¹4715 (Johnston et al. 1993). In addition, the discovery of a very low-luminosity pulsar, PSR J0108¹1431, which is one of the closest known neutron stars, has also been announced (Tauris et al. 1994). The survey has detected more normal pulsars and more millisecond pulsars than any other single survey so far, providing a superb data base for statistical studies. We have therefore combined these data with the best surveys of the Northern Hemisphere at similar frequencies in order to study the properties of the local Galactic populations of normal and millisecond pulsars. To assist these and other evolutionary studies, the newly discovered pulsars 1998 RAS

744 A. G. Lyne et al. Figure 1. Distribution in Galactic coordinates of the 43 842 beam pointings which were successfully observed in the survey. The Galactic equator is horizontal and the Galactic Centre is in the middle of the diagram. have been the subject of a programme of timing observations with both the Parkes 64-m and Jodrell Bank 76-m telescopes; the results of these measurements are to be published by D Amico et al. (1998) (hereafter Paper III). The observing system and analysis procedures are fully described in Paper I and we describe them only briefly in Section 2. Results from the second part of the survey are given in Section 3 and the results from the whole survey are discussed in Section 4. Our analysis of the population of normal and millisecond pulsars is presented in Section 5. In Section 6 we assess the velocity distribution of millisecond pulsars and finally, in Section 7, we draw our conclusions. 2 OBSERVING SYSTEM AND ANALYSIS PROCEDURES The 64-m radio telescope at Parkes was used at a central observing frequency of 436 MHz, providing a beamwidth of 0 : 75 at FWHM. Signals from each of two orthogonal polarizations, each of bandwidth 32 MHz, were amplified in cryogenic FET amplifiers which provided a system-noise temperature on cold sky of about 50 K, corresponding to a system equivalent flux density of 90 Jy. The signals were fed to a filterbank which consists of 256 contiguous channels of 125-kHz width for each of the two polarizations. The detected signals were added in polarization pairs and digitized to 1-bit precision every 300 s. Data were stored on Exabyte tapes for subsequent off-line processing. The main analysis was subsequently performed using computing systems at the ATNF and Jodrell Bank. Data were de-dispersed for up to 738 values of dispersion measure (DM) in the range 0 to the smaller of 777 and 42/sin b cm ¹3 pc, where b is the Galactic latitude. The resulting time sequences were Fourier transformed to produce power spectra. After removing known interference features in the spectra, they were searched for high signal-tonoise ratio (S/N) features. These candidate pulsars or suspects were then further verified by forming an inverse transform of the harmonics of the suspect periods and also by doing a limited search of the original time sequence around the nominal period and DM. Surviving suspects were stored for visual inspection and possible confirmation observations. See Paper I for a more detailed description of the observing system. The sky south of the celestial equator was covered with an approximately hexagonal grid of 44 299 observation pointings, each consisting of 157-s or 512-k samples. These pointings are aligned along strips of constant declination (d) separated by 0 : 65, with adjacent pointings separated by approximately 180/cosd s in right ascension. The most southerly declination is ¹89 : 7, with only three observation pointings. About 3 per cent of the pointings produced good suspects and these were re-observed in order to assess the presence or absence of a pulsar. Confirmed pulsars were then scanned in right ascension and declination to give improved positions. Subsequent arrival-time measurements using the Parkes telescope or the 76-m Lovell telescope at Jodrell Bank were used to establish the presence of any binary motion and to refine the position and period evolution. The survey commenced on 1991 July 29. Paper I reported on observations performed before 1993 February 28, during which time a total of 28 465 beam areas were observed. Here we report on the observation of a further 15 377 beam areas, representing the remainder of the survey, which has, in total, covered 99 per cent of the Southern Hemisphere. Fig. 1 represents the Galactic distribution of all 43 842 observed pointings for which good data were obtained. A number of scattered blank areas reported in Paper I represented observations affected by interference, and most of these were successfully re-observed. 2.1 Sensitivity The ultimate sensitivity of the survey is a function of many factors which limit the completeness of the sample of detected pulsars and which must be understood in order to permit statistical studies of the population. The basic survey sensitivity is given by abt sys W 1=2 S min ¼ GðN p Dn t int Þ 1=2 ; ð1þ P ¹ W where a ð 8Þ is the minimum detectable S/N, b ð 1:5Þ is a factor representing digitization and other processing losses, T sys is the sum of receiver temperature and sky temperature, G is the antenna gain

The Parkes Southern Pulsar Survey II 745 parameter which has an approximate value of 0.64 K Jy ¹1, N p is the number of polarizations (two in this case), Dn is the total radio frequency bandwidth (32 MHz), t int is the integration time (157.3 s), P is the pulse period and W is the effective pulse width (Dewey et al. 1985). With the survey parameters given above, equation (1) becomes W 1=2 S min ¼ 0:19 T sys mjy: ð2þ P ¹ W The pulse width W is given by W 2 ¼ W0 2 þ tsamp 2 þ tdm 2 þ tscatt, 2 where W 0 is the intrinsic pulse width, t samp is the sampling interval, t DM is the dispersion smearing across one frequency channel, and t scatt is the broadening of the pulse owing to interstellar multipath scattering. The latter can be very large for distant pulsars near the Galactic plane and is one factor which severely limits the detection of such pulsars in this survey. For a typical pulse duty cycle W=P ¼ 0:08, equation (2) gives S min 3 mjy for a pulsar at the beam centre at high Galactic latitudes. Fig. 6 in Paper I shows the survey sensitivity as a function of pulse period for several DM values for pulsars both on and away from the Galactic plane. 3 RESULTS Analysis of the 43 842 beam positions successfully observed in the course of the survey resulted in the detection of 298 pulsars. Of these, 55 were new discoveries described in Paper I and 46 are further new discoveries which are listed in Table 1. These include nine recently announced millisecond pulsars (Lorimer et al. 1996; Stappers et al. 1996; Bailes et al. 1997). The first column gives the J2000 name, based on the J2000 coordinates which are given to full precision in Paper III. In some cases, the quoted coordinates are sufficiently uncertain that the name may change as improved positions become available. The Galactic longitude and latitude are listed in the second and third columns. The basic pulsar parameters, period and dispersion measure, are listed in the next two columns. The most precise values available for all these parameters are given in Paper III. The sixth column gives the S/N of the detection (not the confirmation) and the seventh column gives the measured flux density at 436 MHz, S 436, based upon the average of the confirmation and subsequent observations to reduce the effects of interstellar scintillation. Observed widths of the mean pulse profile at the 50 and 10 per cent levels are given in the next two columns. The final column contains keys to the previous publications of the millisecond pulsars. Of the nine millisecond pulsars discovered in this part of the survey, five are members of binary systems, the orbital parameters of which are listed in Table 2. All of the systems have nearly circular orbits and all have relatively low-mass companions, probably white dwarfs (Lorimer et al. 1996; Stappers et al. 1996). At this stage, none of the other pulsars in Table 1 is known to have a binary companion. There is no evidence of planet-mass bodies in orbit around any of the millisecond pulsars (Bell et al. 1997; Bailes et al. 1997). Table 3 lists the 51 previously known pulsars detected in this part of the survey. The first two columns are the pulsar J2000 and B1950 names respectively, and the next two columns give the S/N of the detection observation (where a pulsar was detected at more than one grid position, the largest value is taken) and the expected S/N. This latter value was computed using equation (1), taking into account the mean 400-MHz flux density from Taylor, Manchester & Lyne (1993), the effective pulse width, the sky background temperature at the pulsar position (Haslam et al. 1982) and the gain loss resulting from offset of the pulsar from the beam centre. A total of 12 known pulsars with expected S/N above 8.0 were missed by the survey: that is, they were within the half-power beam of an observation, but were not detected by the search procedure. These pulsars are listed in Table 4. Several southern globular cluster pulsars and many high-dm pulsars discovered in the highfrequency surveys along the Galactic plane (Clifton et al. 1992; Johnston et al. 1992a) were expected to be (and were) below our detection limit. In Table 4, the sixth column gives the expected S/N from equation (1), ignoring the effects of interstellar scattering. The four pulsars with DMs > 300 cm ¹3 pc are all likely to suffer significant broadening from interstellar scattering, so reducing the expected S/N and probably explaining their non-detection. Thus the number of missed pulsars that we could have expected to have detected is similar to the number (seven) of detected pulsars with expected values of S/N less than 8.0 (Table 3). 4 STATISTICAL OVERVIEW This survey has clearly been very successful in its main aim of increasing the number of known millisecond pulsars in the Galactic disc. Before the survey began in 1991, only five such pulsars were known, and the survey has added a further 17. Other searches at Arecibo (Foster, Wolszczan & Camilo 1993; Nice, Taylor & Fruchter 1993; Camilo 1995; Ray 1995) and Jodrell Bank (Navarro et al. 1995; Nicastro et al. 1995) have in the meantime added a further 13 such pulsars. The Parkes survey has also discovered a substantial number of low-luminosity normal pulsars, including PSR J0108¹1431 (Tauris et al. 1994). The period distributions of the 101 new discoveries and all the 298 pulsars that were detected are shown in Fig. 2. The pulsars clearly fall into two groups, millisecond pulsars and normal pulsars, with 17 per cent of the new discoveries and 6 per cent of all detections in the millisecond group. Two previously known millisecond pulsars were detected, PSR B1620¹26 in the globular cluster M4 (Lyne et al. 1988) and PSR B1820¹30A in the globular cluster NGC 6624 (Biggs et al. 1994); all other millisecond pulsars lay below our sensitivity limit, not surprisingly, since they were all discovered in deep targeted searches of globular clusters. The median period of the new non-millisecond pulsars, 0.54 s, is somewhat less than that of the detected previously known pulsars, 0.64 s. This reflects this survey s better sensitivity to short-period pulsars compared with many previous large-scale surveys. Fig. 3 shows the DM distributions of the new and all the detected pulsars. The new pulsars are confined to DMs of less than about 270 cm ¹3 pc, whereas the previously known pulsars were detected at DMs extending to more than 500 cm ¹3 pc. This is mainly because of strong selection effects which reduce the sensitivity to such high- DM pulsars, the survey being carried out at the rather low frequency of 436 MHz. The main effects are the high sky brightness temperature in the Galactic plane, where distant, high-dm pulsars are found, and the large pulse smearing resulting from interstellar scattering associated with the large DMs. These effects have been countered in two higher frequency surveys carried out close to the Galactic plane (Clifton et al. 1992; Johnston et al. 1992b), which detected a large number of pulsars with DM values over 200 cm ¹3 pc. These surveys were responsible for most of the previously known high-dm pulsars. The DM distribution of the new millisecond pulsars is much narrower than that of the normal pulsars. This is primarily due to the reduced sensitivity of the survey to high- DM millisecond pulsars (see Paper I, fig. 6).

746 A. G. Lyne et al. Table 1. Observed parameters for 46 pulsars discovered in the second part of the survey. PSR J l b Period DM S/N S 436 W 50 W 10 Ref deg deg ms cm ¹3 pc mjy ms ms 0133¹6957 297.7 ¹46.7 463.474 22.9 22 5 7.8 17.6 0211¹8159 299.6 ¹34.6 1077.333 24.4 21 6 19.3 72.1 0448¹2749 228.3 ¹37.9 450.448 26.2 17 2 10.3 22.0 0520¹2553 228.4 ¹30.5 241.642 33.8 12 8 5.7 15.7 0540¹7125 282.1 ¹31.2 1286.015 29.4 10 5 37.2 70.7 0711¹6830 279.5 ¹23.3 5.491 18.4 10 10 2.3 3.8 1 0857¹4424 265.5 +0.8 326.774 184.4 9 12 14.0 29.7 1006¹6311 285.5 ¹6.0 835.797 196.0 10 4 21.7 26.7 1024¹0719 251.7 +40.5 5.162 6.5 12 5 1.2 1.8 1 1047¹3032 273.5 +25.1 330.328 52.4 12 8 16.5 28.0 1047¹6709 291.3 ¹7.1 198.451 116.2 11 4 9.3 21.4 1123¹4844 288.3 +11.6 244.838 92.9 19 8 11.9 17.1 1123¹6259 293.2 ¹1.8 271.433 223.3 13 11 8.9 16.8 1123¹6651 294.5 ¹5.4 232.976 111.2 13 8 15.6 23.2 1126¹6942 295.6 ¹8.0 579.416 55.3 17 9 13.3 25.4 1130¹6807 295.5 ¹6.4 256.353 148.7 12 8 17.4 59.7 1137¹6700 295.8 ¹5.2 556.216 228.0 17 14 99.0 136.2 1141¹3322 286.6 +27.3 145.734 46.4 9 8 4.8 7.5 1156¹5909* 295.9 +3.0 1037.931 219.2 32 7 14.5 20.7 1159¹7910* 300.4 ¹16.5 525.074 59 10 6 15.7 24.6 1332¹3032 313.4 +31.5 650.434 14.7 13 9 33.8 266.6 1350¹5115 312.2 +10.5 295.699 90.4 19 9 6.2 11.2 1403¹7646 307.1 ¹14.5 653.099 100.6 11 4 16.9 78.3 1603¹5657 326.9 ¹3.3 496.077 264.1 14 8 13.3 29.2 1603¹7202 316.6 ¹14.5 14.842 38.0 14 26 1.7 2.7 2 1604¹7203 316.7 ¹14.5 341.403 54.4 15 10 20.1 38.5 1622¹4332 338.3 +4.3 916.938 230.7 10 16 48.5 93.5 1654¹2713 355.0 +10.3 791.822 92.5 23 9 16.6 32.4 1732¹1930 6.4 +7.6 483.780 73.0 9 11 15.9 22.7 1744¹1134 14.8 +9.2 4.075 3.1 22 13 0.2 0.3 1 1744¹2335* 4.5 +2.9 1683.507 96.7 11 16 33.6 58.9 1759¹2922 1.2 ¹2.9 574.400 79.4 11 13 9.7 20.1 1804¹2717 3.5 ¹2.7 9.343 24.7 15 20 1.6 2.1 2 1809¹3547 356.5 ¹7.8 860.388 194.8 14 21 150.5 198.7 1817¹3837 354.7 ¹10.4 192.243 102.8 12 12 7.8 13.6 1901¹0906 26.0 ¹6.4 890.964 72.7 12 11 13.3 68.6 1904¹1224 23.3 ¹8.5 375.404 118.2 10 6 16.5 25.5 1911¹1114 25.1 ¹9.6 3.626 31.0 13 31 0.3 0.8 2 1932¹3655 2.1 ¹23.5 571.420 59.9 15 9 11.4 27.9 1940¹2403* 15.8 ¹21.0 927.638 63.3 9 1947¹4215* 357.2 ¹27.7 1798.069 35 12 7 62 2038¹3816 3.8 ¹36.7 1577.286 34.0 15 7 45.7 74.1 2051¹0827 39.2 ¹30.4 4.508 20.7 12 15 0.6 1.3 3 2108¹3429 9.7 ¹42.2 1423.102 30.2 12 6 28.4 49.8 2124¹3358 10.9 ¹45.4 4.931 4.6 48 6 2.3 3.5 1 2129¹5721 338.0 ¹43.6 3.726 31.9 12 5 0.6 1.3 2 *The position and name of the pulsar may change. References: (1) Bailes et al. (1997); (2) Lorimer et al. (1996); (3) Stappers et al. (1996).

Table 2. Parameters for five binary pulsars newly discovered in this stage of the survey. The Parkes Southern Pulsar Survey II 747 PSR J Binary Period a p sin i Eccentricity Long. of Peri. Epoch of Peri. Ref. d s deg MJD 1603 7202 6.3086 6.8807 <0.00002 0.0 49524.15 1 1804 2717 11.1287 7.2815 0.000035 159.8 49615.11 1 1911 1114 2.7166 1.7629 <0.000013 0.0 49838.95 1 2051 0827 0.0991 0.0451 <0.003 0.0 49642.17 2 2129 5721 6.6255 3.5006 <0.000017 0.0 49528.32 1 References: (1) Lorimer et al. (1996); (2) Stappers et al. (1996). Table 3. Parameters for 51 previously known pulsars detected in this stage of the survey. PSR J PSR B Obs S/N Exp S/N PSR J PSR B Obs S/N Exp S/N 0151¹0635 0148¹06 21.2 138.9 1639¹4604 1635¹45 13.4 3.8 0401¹7608 0403¹76 85.5 66.0 1646¹6831 1641¹68 43.1 25.9 0536¹7543 0538¹75 61.8 28.5 1652¹2404 1649¹23 23.7 17.7 0828¹3417 0826¹34 86.7 40.1 1704¹6016 1659¹60 13.3 25.2 0855¹3331 0853¹33 25.1 8.3 1711¹5350 1707¹53 40.6 9.6 0904¹7459 0904¹74 24.4 30.0 1722¹3207 1718¹32 37.6 27.3 0909¹7212 0909¹71 23.3 18.4 1740¹3015 1737¹30 9.0 3.1 1012¹5857 1011¹58 9.0 15.6 1743¹3150 1740¹31 14.6 2.0 1032¹59 1030¹58 10.9 6.6 1748¹1300 1745¹12 41.2 38.2 1056¹6258 1054¹62 57.6 69.6 1751¹4657 1747¹46 148.6 123.6 1057¹7914 1056¹78 26.7 28.9 1759¹2205 1756¹22 13.1 10.2 1112¹6613 1110¹65 40.8 47.0 1807¹2715 1804¹27 20.5 17.7 1112¹6926 1110¹69 25.9 34.7 1816¹2649 1813¹26 17.0 18.5 1239¹6832 1236¹68 16.6 7.4 1817¹3618 1813¹36 21.0 26.2 1257¹1027 1254¹10 48.1 44.9 1823¹3021A 1820¹30A 8.0 10.0 1259¹6741 1256¹67 10.8 5.8 1824¹1945 1821¹19 28.6 48.9 1312¹5402 1309¹53 30.2 39.9 1833¹6023 1828¹60 15.8 14.8 1320¹5359 1317¹53 43.0 49.6 1834¹0426 1831¹04 51.5 69.9 1326¹5859 1323¹58 18.2 15.4 1849¹0636 1846¹06 20.2 14.1 1430¹6623 1426¹66 195.0 180.9 1854¹1421 1851¹14 25.3 12.5 1456¹6843 1451¹68 167.7 363.2 1946¹2913 1943¹29 30.9 28.1 1553¹5456 1550¹54 11.8 4.5 1949¹2524 1946¹25 67.7 16.6 1555¹3134 1552¹31 29.9 26.5 2053¹7200 2048¹72 49.4 90.6 1600¹5751 1556¹57 22.0 28.3 2127¹6648 2123¹67 30.6 23.0 1623¹0908 1620¹09 15.3 15.0 2330¹2005 2327¹20 25.1 112.7 1623¹2631 1620¹26 12.7 30.8 Figs 4 and 5 give the distributions of the new detections and all the detected pulsars in Galactic coordinates. This sample of millisecond pulsars shows little or no concentration toward the Galactic equator, but the newly discovered normal pulsars tend to be closer to the equator, albeit with a wider scatter than the previously known pulsars. This is also illustrated by Fig. 6 which shows the distributions in Galactic z-distance of the new and all the detected pulsars. Pulsar distances (d) used to calculate the z-distances (z ¼ d sin b) were derived using the model of Taylor & Cordes (1993) and listed in the Taylor et al. (1993) catalogue for the previously known objects. High bins at the ends of the distributions in Fig. 6 are a result of the lower limits assigned by the distance model. The z-distribution of the new pulsars is significantly wider than that for the previously known pulsars; the median values of jzj are 0.58 and 0.35 kpc respectively, reflecting the good sensitivity of the survey at high Galactic latitudes. The median values of jzj for the detected millisecond and non-millisecond pulsars are 0.35 and 0.44 kpc respectively; there is no statistically significant difference between the z-distributions of the two classes of pulsars. Possible associations of the new pulsars with Galactic supernova remnants (SNRs) and globular clusters (GCs) were checked by comparison with the catalogues of Green (1996) and Webbink (1985). Three pulsars detected in the survey lie within a degree of SNRs: J0905¹5127 and J0907¹5157 lie 35 and 6 arcmin respectively from G272¹3.2 and, as noted in Paper I, J1835¹1106 lies 41 arcmin from G21.5¹0.9, although there is nothing else to commend these associations. While two previously known pulsars detected in the survey, PSRs B1620¹26 and B1820¹30A, are associated with globular clusters, there is no clear association of any new pulsar discovered by this survey with a globular cluster.

748 A. G. Lyne et al. Table 4. Parameters of 12 known pulsars not detected. PSR J PSR B Period DM S 400 Exp msec cm ¹3 pc mjy S/N 1119¹7936 1118¹79 2280.6 27 7 9.7 1306¹6617 1303¹66 473.02 437 21* 14.9 1326¹6700 1322¹66 543.00 211 28 10.1 1327¹6301 1323¹627 196.47 295 22* 8.9 1357¹62 1353¹62 455.76 417 45* 23.0 1402¹51 1359¹51 1380.18 39 10 26.5 1507¹6640 1503¹66 355.65 130 13 13.8 1610¹1322 1607¹13 1018.39 48 10 18.1 1701¹3726 1658¹37 2454.61 303 29* 19.8 1745¹3040 1742¹30 367.42 88 66 11.6 1848¹1952 1845¹19 4308.18 18 17 16.4 1900¹7951 1851¹79 1279.19 136 6 12.5 (*) S 400 estimated by extrapolation from higher frequencies. Figure 3. Dispersion measure distributions of the new millisecond pulsars (dash dotted line), all new pulsars (solid line), and all the pulsars detected by the survey (dotted line). studying the low-luminosity members of the population which, of course, comprise the large majority. The subsequent timing observations (see Paper III and references therein) have provided period derivatives for most of the detected pulsars, allowing studies of the local birth-rates of both populations. Since there are extreme selection effects which prevent the detection of distant pulsars at this low frequency and which are difficult to model accurately, we use the new data to establish only the local space density and local birth-rate of pulsars, and rely on the higher frequency surveys to extrapolate to the whole of the Galactic population. We begin by discussing each of the two populations separately. As we will see, the luminosity functions of both populations are now well determined down to a limit of about 1 mjy kpc 2. This allows us, for the first time, to make a reliable quantitative estimate of the relative abundances of the two populations. Figure 2. Period distributions of the new discoveries (solid line) and of all the pulsars detected by the survey (dotted line). The radio luminosities of the new detections and all the detected pulsars are shown in Fig. 7. Here the luminosity is calculated as L 436 ¼ S 436 d 2 mjy kpc 2. The survey has increased the number of low-luminosity pulsars so that there are now four normal pulsars in the Southern Hemisphere with luminosity less than 3 mjy kpc 2, compared with only one known previously. Even though such weak pulsars are very numerous in the Galaxy, only a small number are known because of the small volume of space close to the Sun in which they can be detected. 5 THE NORMAL AND MILLISECOND PULSAR POPULATIONS In this section, we attempt to establish the local space densities of normal and millisecond pulsars. The high sensitivity and large sky coverage of the survey make it by far the best survey to date for 5.1 The population of normal pulsars One aim of this survey was to revisit the statistics of the lowluminosity pulsar population which was investigated by Lorimer et al. (1993, hereafter LBDH) just before this survey. LBDH demonstrated that previous surveys could not make reliable statements about the population and birth-rate of pulsars with luminosities lower than about 10 mjy kpc 2, mainly because of the small numbers resulting from the poorer sensitivity of previous surveys to faint pulsars. The improved sensitivity of this survey means that we should now be able to put better constraints on the population of pulsars with luminosities below 10 mjy kpc 2. Any estimate for the number of pulsars in the Galaxy must take account of the fraction of pulsars missed by the surveys as a result of the known selection effects. This is most readily achieved by computing a weight, or scale factor, for each pulsar by placing it at a large number of randomly selected locations in a model Galaxy and recording the number of detections, i.e. those positions for which the predicted S/N exceeds the survey limit (Section 2.1). The scale factor is then defined as the ratio of the number of locations

The Parkes Southern Pulsar Survey II 749 Figure 4. Distribution in Galactic coordinates of the new discoveries. The celestial equator is indicated by the dashed line, solitary millisecond pulsars by an asterisk, binary millisecond pulsars by a dot within an open circle and normal pulsars by a dot. Figure 5. Distribution in Galactic coordinates of all the pulsars detected by the survey. Symbols are as in Fig. 4. tried to the number of detections. In this way, for each pulsar, the scale factor gives us an estimate of the number of similar pulsars in the Galaxy. LBDH verified the validity of this approach using synthetic populations of pulsars with well-defined intrinsic properties. One can also correct for the effects of beaming of the pulsar radiation; because of beaming, only a fraction of the active pulsars are potentially observable, that is, have their beams directed toward us. To make this correction, the scale factor is divided by the beaming fraction estimated from an appropriate model, such as that described by Biggs (1990). As in LBDH, we estimate the numbers of pulsars with and without such a beaming correction in order to distinguish between the properties of the population of potentially observable pulsars, which is independent of any assumptions made about the beaming, and the total active pulsar population, which depends on the validity of the adopted beaming model. As mentioned above, because of severe selection effects in the inner Galaxy, we choose to focus here on only the local population. We do this by considering only those pulsars within a cylinder of radius 1.5 kpc centred on the Sun and with axis normal to the Galactic plane. A total of 84 normal pulsars satisfy this criterion. To simulate this sub-set of the population, we generate 10 5 positions within this cylinder distributed with a uniform surface density along the Galactic plane. To model the dispersion of pulsars with respect to the Galactic plane, each position is assigned a height z above the Galactic plane which is chosen from a Gaussian distribution in z with a scaleheight j z. The rest of the procedure is similar to that described by LBDH with the exception of the distance model. Here we use the Taylor & Cordes (1993) model to derive pulsar distances, DMs and scattering time-scales for the model pulsars at each location. In addition to the present survey, we consider the Jodrell Bank A survey (Davies, Lyne & Seiradakis 1972, 1973), the U. Mass Arecibo survey (Hulse & Taylor 1974), the second Molonglo survey (Manchester et al. 1978), the U. Mass NRAO survey (Damashek, Taylor & Hulse 1978), the Princeton NRAO survey phase I (Dewey et al. 1985) and phase II (Stokes, Taylor & Dewey 1985), the Princeton Arecibo survey (Nice, Fruchter & Taylor 1995), and several high-latitude searches at Arecibo (Thorsett et al. 1993; Foster et al. 1995; Ray et al. 1995; Camilo, Nice & Taylor 1996a; Camilo et al. 1996b), as well as the Jodrell Bank millisecond pulsar survey (Nicastro et al. 1995). An important check on the validity of the assumed dispersion of the model population with respect to the Galactic plane is obtained by comparing the z-distribution of the detected model positions during the scale factor analysis with that of the real observed pulsars

750 A. G. Lyne et al. Figure 6. Distribution in Galactic z-distance of the new millisecond pulsars (dash dotted line), all the new pulsars (solid line) and all the pulsars detected by the survey (dotted line). Figure 8. The observed and underlying distributions of pulsar luminosity at 400 MHz, based upon a sample of 84 normal (non-millisecond) pulsars lying within a cylindrical radius of 1.5 kpc from the Sun. The observed distribution is given by the dashed line. The distribution of potentially observable pulsars, obtained by correcting the observed sample for selection effects, is given by the dash dotted line, whilst the solid line shows the distribution of active pulsars, obtained by applying the beaming model of Biggs (1990) to the potentially observable distribution. The error bars on the latter distribution represent the statistical uncertainties in the scale factor calculation. The straight line shows a fit to the data above 10 mjy kpc 2 assuming a slope of ¹1. Figure 7. Distribution of observed 436-MHz luminosities for the new millisecond pulsars (dash dotted line), all the new pulsars (solid line) and all the pulsars detected by the survey (dotted line). Fig. 8 shows the observed luminosity function and the derived luminosity function before and after beaming corrections have been applied. As before (LBDH), the luminosity function was formed by summing the scale factors for the pulsars contained in each luminosity bin. The difference between the observed and derived distributions below 10 mjy kpc 2 highlights the severe undersampling of low-luminosity pulsars in the observed population. Accordingly, the distribution below 1 mjy kpc 2 becomes highly uncertain because of small numbers, and is not displayed. Based upon Fig. 8, we estimate the number of potentially observable pulsars with luminosities greater than 1 mjy kpc 2 within a cylindrical radius of 1.5 kpc to be 210 40, corresponding to a local surface density of 30 6 kpc ¹2. The errors here represent the statistical uncertainty in the scale factors (Vivekanand & Narayan 1981; LBDH). Applying the beaming model of Biggs (1990), the corresponding numbers are 1100 220 pulsars with a mean surface density of 156 31 kpc ¹2. We now estimate the number of pulsars in the whole Galaxy by assuming that the distribution has azimuthal symmetry, and using a scaling factor of 1000 250 kpc 2 for the ratio of the Population I content of the whole Galaxy to the local surface density at the position of the Sun (Ratnatunga & van den Bergh 1989). Thus we estimate the Galactic population of potentially observable pulsars with luminosities greater than 1 mjy kpc 2 to be ð30 10Þ 10 3, corresponding to ð160 50Þ 10 3 active pulsars. (cf. Lorimer 1995). The best-fitting scaleheight was found to be between 400 and 500 pc, consistent with the observed value (Taylor et al. 1993). Models with lower scaleheights produce an excess of pulsars close to the Galactic plane, whereas models with larger scaleheights produce too many high-z pulsars. In the remainder of this section, therefore, we adopt 450 pc as the underlying scaleheight of the normal pulsar population. 5.2 The birth-rate of normal pulsars A relatively model-independent technique to determine the birthrate of pulsars is the pulsar current analysis (Phinney & Blandford 1981; Vivekanand & Narayan 1981). The idea here is to compute the flow ( current ) of pulsars from short to long periods as a function of period. In this scheme, provided that the population is in a steady state, the birth-rate required to sustain the population is just

The Parkes Southern Pulsar Survey II 751 Figure 9. Pulsar current for potentially observable pulsars based upon the 82 pulsars with measured period derivatives within 1.5 kpc of the Sun. the current at short periods. In this section, we use this technique to calculate the local birth-rate of pulsars using a pulsar current analysis on the sub-set of 82 pulsars from the above sample that have measured values of Ṗ. Our analysis follows a similar style to that of LBDH, i.e. we estimate the contribution made to the gross pulsar current by each pulsar to be its scale factor multiplied by its Ṗ. The distribution of pulsar current is shown in Fig. 9 before and after the application of the Biggs (1990) beaming factor to the scale factors. The corresponding birth-rates are ð2:8 1:7Þ 10 ¹4 PSR century ¹1 kpc ¹2 for the potentially observable pulsars and ð10 6Þ 10 ¹4 PSR century ¹1 kpc ¹2 after applying the beaming model. Because of our sample definition, these birth-rates apply just to the population of pulsars with luminosities greater than 1 mjy kpc 2. The birth-rate in the Galaxy as a whole is obtained by multiplying these values by the same factor of 1000 discussed in the previous section, giving values of 0:3 0:2 and 1:0 0:7 per century respectively. The latter figure corresponds to a minimum birth-rate of one every 60 to 330 yr in the Galaxy. So what does this survey tell us about the population of faint pulsars? LBDH concluded that relatively little could be said about the population of pulsars below 10 mjy kpc 2 because of their overwhelming effect on the scale factors derived from previous surveys. It was proposed by LBDH that there may be no need for a significant number to be born with luminosities below 30 mjy kpc 2 in order to explain the observed population of pulsars. The improvement to this analysis using the Parkes and Arecibo surveys over that of LBDH has allowed us to reduce our luminosity limit to 1 mjy kpc 2. Despite this order-of-magnitude reduction, the number and birth-rate estimates are only a factor of 2 3 larger than those found by LBDH. This reflects the significant flattening of the luminosity function below 20 mjy kpc 2 seen in Fig. 8. 5.3 The population of millisecond pulsars Although this survey has found many millisecond pulsars, its sensitivity to these objects is generally somewhat poorer than to normal pulsars because of the pulse broadening arising from instrumental effects and interstellar dispersion. The effects, which Table 5. Parameters for 21 nearby millisecond pulsars. PSR Period L 400 W/P Scale V t jzj Reference (ms) mjy kpc 2 % Factor km s ¹1 kpc for V t J0034¹0534 1.877 16 41 6 <80 0.909 3 J0437¹4715 5.757 19 20 3 120 0.095 2 J0711¹6830 5.491 11 41 7 140 0.411 1 J1012þ5307 5.256 5 14 9 60 0.401 4 J1022þ1001 16.45 5 27 10 <120 0.466 J1024¹0719 5.162 0.6 24 180 <120 0.230 1 B1257þ12 6.219 8 11 5 279 0.640 5 J1455¹3330 7.987 5 19 9 62 0.283 3 J1713þ0747 4.570 15 17 4 30 0.469 6 J1730¹2304 8.123 4 13 10 74 0.053 3 J1744¹1134 4.075 0.4 5 210 <40 0.027 1 J1804¹2717 9.343 14 7 2 <220 0.056 B1855þ09 5.362 12 14 3 26 0.048 7 J2019þ2425 3.934 2 24 40 100 0.105 8 J2033þ1734 5.949 28 6 <150 0.314 J2051¹0827 4.509 9 27 7 <180 0.647 J2124¹3358 4.931 0.4 42 520 67 0.176 1 J2145¹0750 16.05 18 4 2 29 0.335 3 J2229þ2643 2.978 13 19 5 <103 0.634 9 J2317þ1439 3.445 68 25 2 68 1.275 9 J2322þ2057 4.808 1 16 70 89 0.474 8 References: (1) Bailes et al. (1997); (2) Bell et al. (1997); (3) Camilo et al. (1997); (4) Camilo et al. (in preparation); (5) Wolszczan (1994); (6) Camilo et al. (1994b); (7) Kaspi, Taylor & Ryba (1994); (8) Nice & Taylor (1995); (9) Camilo et al. (1996a).

752 A. G. Lyne et al. Figs 8 and 10 that millisecond pulsars are, on average, less luminous than normal pulsars. The median luminosities of the observed samples are, respectively, 9 and 15 mjy kpc 2 not so different. However, low-luminosity millisecond pulsars are much more heavily selected against by current surveys. The luminosity function for normal pulsars flattens off at luminosities less than about 20 mjy kpc 2, whereas that for millisecond pulsars remains consistent with a slope of ¹1 to the lowest luminosities sampled, 0:3 mjy kpc 2. Therefore, at low luminosities, the number of potentially observable millisecond pulsars is much greater than the number of potentially observable normal pulsars. On the other hand, the beams of millisecond pulsars are, on average, wider than those of normal pulsars. Whilst it is hard to quantify the behaviour of these functions down to low luminosities, it seems that the populations of active millisecond and normal pulsars are comparable. Figure 10. The observed (dashed line) and underlying (solid line) distributions of the luminosity of millisecond pulsars at 400 MHz, based upon a sample of 21 objects. No beaming correction has been applied. The straight line shows a fit to the data assuming a slope of ¹1. are a strong function of pulsar period and dispersion measure, are described in Paper I. The form of the underlying period distribution of millisecond pulsars has been discussed by Lorimer et al. (1996). Here we discuss what can be learned about their local space density and birth-rate, and extrapolate the results to the whole of the Galaxy. We finally use a dynamical simulation to assess the likely velocity distribution at birth. To estimate the local surface density of millisecond pulsars, we used the same analysis as described for the normal pulsars. The sample used was defined as being all known objects within the 1.5- kpc cylindrical radius of the Sun, and contains the 21 pulsars listed in Table 5. Summing the scale factors in Table 5, we estimate there to be 1110 600 millisecond pulsars in the local solar neighbourhood or a local surface density of 157 85 kpc ¹2 for L 436 > 0:3 mjy kpc 2. This estimate is significantly influenced by the three very lowluminosity pulsars: J1024¹0719, J1744¹1134 and J2124¹3358 (Bailes et al. 1997). For pulsars with L 436 > 1:0 mjy kpc 2,we estimate 200 85 pulsars, or a surface density of 28 12 kpc ¹2. The observed and derived luminosity distributions for the full sample of 21 objects are shown in Fig. 10. Since we have no cause to suspect that millisecond pulsars have a substantially different intrinsic galactocentric radial distribution from that of normal pulsars, we use the same scale factor to derive the total population from the local surface density. This gives Galactic populations of potentially observable millisecond pulsars of 30 10 3 and 200 10 3 for luminosities exceeding 1 and 0.3 mjy kpc 2 respectively. Because of beaming effects, the total active population will be larger. 5.4 Relative abundances of normal and millisecond pulsars Our calculations for the number of normal and millisecond pulsars in the local solar neighbourhood allow, for the first time, a quantitative estimate of the relative abundances of the two populations. For the number estimates of potentially observable pulsars of both types, we find the ratio of normal-to-millisecond pulsars with luminosities greater than 1 mjy kpc 2 to be 1:1 0:5. Thus, for this common luminosity limit, the number of millisecond pulsars with beams that intersect our line of sight is remarkably similar to the number of normal pulsars. However, it is clear from a comparison of 5.5 The birth-rate of millisecond pulsars It is generally believed that the ages of the millisecond pulsars are much greater than those of normal pulsars and probably similar to the age of the Galaxy. The evidence for this comes from their large characteristic ages (Camilo, Thorsett & Kulkarni 1994a) and, for those millisecond pulsars in binary systems, optical observations of cool white dwarf companions (Bell, Bailes & Bessell 1993; Bell et al. 1995; Lundgren et al. 1996). We attempted to derive the birthrate using the pulsar current approach described above. The small size of the sample meant that we were not able to derive a statistically significant birth-rate for the population using this method. We can, however, set a firm lower limit to the birth-rate of millisecond pulsars, noting that their ages must not exceed the age of the Galactic disc, which we take to be 10 10 yr. Based on the population estimates calculated in Section 5.3, and for a beaming fraction of unity, we find the birth-rate of millisecond pulsars to be at least 3 10 ¹6 and 20 10 ¹6 yr ¹1 above luminosities of 1 and 0.3 mjy kpc 2 respectively. This is consistent with earlier work which was based on a slightly smaller sample (Lorimer et al. 1995). 6 THE SPACE VELOCITIES OF MILLISECOND PULSARS The velocities given to pulsars at birth result in a general movement of the population away from the Galactic plane. The velocities of normal pulsars have already been extensively discussed (Dewey & Cordes 1987; Bailes 1989; Lyne & Lorimer 1994; van den Heuvel & van Paradijs 1997; Lorimer, Bailes & Harrison 1997), showing that they have birth space velocities of 400¹500 km s ¹1, most probably from kicks resulting from asymmetry in the supernova events of formation. We focus our attention here on the velocities of millisecond pulsars. The sample of millisecond pulsars used throughout this section and the surveys used to discover them are the same as those used in the previous section (see also Lorimer et al. 1996). Timing measurements have already given the proper motions for 13 of the millisecond pulsars in our sample. The transverse speed of a pulsar may be calculated from its proper motion m (mas yr ¹1 ) and DM-derived distance d (kpc) using the following expression: V t ¼ 4:74m d km s ¹1 : ð3þ The transverse speeds derived from these data are given in Table 5. For those seven millisecond pulsars that presently do not have a measured proper motion, we also include in Table 5 upper limits to V t inferred from their measured period derivatives (Shklovsky

The Parkes Southern Pulsar Survey II 753 Figure 11. The observed (stepped solid line) and model-observed cumulative distributions for millisecond pulsars in V t for each of five models with different initial velocity distributions (see text). 1970; Camilo et al. 1994a). The mean V t for the 13 millisecond pulsars that have measured values is 88 19 km s ¹1. We note that these upper limits are mostly greater than this mean and are consistent with the true velocities of these pulsars being similar to those presently measured. In the discussion that follows, we consider only those transverse speeds that have been measured. Since the ages of all the millisecond pulsars are almost certainly much greater than their oscillation time across the plane of the Galaxy ( 60 Myr: Mihalas & Binney 1981), the population will now have a scaleheight above the plane which depends upon the magnitude of their birth velocities. From our vantage point close to the plane of the Galaxy, surveys will therefore detect an excess of pulsars at small z-distance which are likely to have preferentially small z-velocities. On the other hand, detected pulsars are close to the plane where their kinetic energy is greatest. The net effect on the derived total space and transverse velocities of the observed pulsars is, in fact, small. To determine what initial velocity distribution is consistent with the observed transverse speeds, we performed a number of simulations in which we followed the orbits of pulsars in the Carlberg & Innanen (1987) model of the Galactic potential for up to 10 10 yr using a numerical integration technique to solve the equation of motion. After applying our model of the survey selection effects, we obtained model observed samples with distributions in transverse velocity V t and height above the Galactic plane z. As in previous work (Lorimer et al. 1996) we took care to ensure that the modelobserved samples contained a large number of pulsars (typically 1500) in order to minimize the effect of statistical noise. To compare these with the observed sample in Table 5, we used the Kolmogorov Smirnov (KS) test (Press et al. 1986) which returns a probability that the two distributions are the same for both V t and z. In the simulations, we considered five different Maxwellian velocity distributions with 1D dispersions of j v ¼ 0, 50, 75, 100 and 150 km s ¹1, which we shall refer to as models (a) to (e) respectively. For each model, the resulting model observed, transverse velocity distribution is compared directly with the observed sample in Fig. 11. It is evident from model (a), which has no intrinsic velocity dispersion, that velocities of millisecond pulsars clearly require more than just Galactic rotation to explain their observed distribution, with the model distribution producing too few high-velocity pulsars. Conversely, model (e), with a 1D velocity dispersion of 150 km s ¹1, produces too many high-velocity pulsars. Model (c) agrees best with the observations, with a KS probability of 85 per cent. These models suggest that the data are most consistent with a 1D velocity dispersion of 80 20 km s ¹1. Such a model sample has a mean observed value of V t of 85 km s ¹1, consistent with that of the observed sample. It is interesting to note that the model underlying, as opposed to the model observed, sample has a mean transverse speed of 100 km s ¹1, showing the strength of the selection effect against detecting millisecond pulsars with larger space velocities. As a consistency check, we can compare the observed z-distribution with the model sample. For model and observed samples these are calculated using the distances obtained from the Taylor & Cordes (1993) electron density model. For the observed sample, we find the mean jzj to be 380 70 pc. The adopted model has a mean value of jzj of 393 pc, in satisfactory agreement. To summarize, we find that a population with a Maxwellian initial 1D velocity dispersion of 80 20 km s ¹1 agrees best with the existing data, corresponding to a mean space velocity of p130 30 km s ¹1 (the mean of a 3D Maxwellian distribution is 8= times the 1D dispersion). This velocity distribution must be that of those binary systems that survived the supernova explosion that formed the neutron star and went on to produce a millisecond pulsar by accreting matter from the companion star. Very few binaries are likely to survive the supernova explosions that result in larger kick velocities. This distribution has been discussed by Tauris & Bailes (1996) and by Ramachandran & Bhattacharya (1997). Tauris & Bailes (1996) developed a detailed binary evolution code which predicts the velocity distribution of the surviving binary systems as a function of the initial kick velocity imparted to the neutron star at birth. We have extended the analysis of Tauris & Bailes (1996) using the Lyne & Lorimer (1994) distribution of kick velocities. This predicts a 3D velocity distribution for the surviving binary systems with a mean value of 140 km s ¹1 agreeing well with the results of our study. This demonstrates that the observed velocities of millisecond pulsars are consistent with birth kick velocities lying at the low end of the global neutron star distribution. Ramachandran & Bhattacharya (1997) reached similar conclusions by comparing the kinematics of millisecond pulsars and low-mass X-ray binaries. Although the velocity dispersion of millisecond pulsars is lower than that of the normal pulsars, their underlying z-distribution is very broad. From our dynamical simulation of the orbits of millisecond pulsars in the local solar neighbourhood we find, for instance, that, for a Maxwellian birth velocity distribution with a mean 3D velocity of 130 km s ¹1, the median and mean values of jzj are 570 pc and 1.1 kpc respectively. Many of these pulsars will therefore lie above the layer of ionized material, which has an effective half-width of 1:6 kpc (Taylor & Cordes 1993). The observed z-distribution of the sample of millisecond pulsars in Table 5 has median and mean jzj values of 314 and 383 pc respectively, reflecting the bias against detecting pulsars at large heights above the Galactic plane. 7 CONCLUSIONS We have designed and implemented a system using the Parkes radio telescope and workstation networks to search for millisecond and low-luminosity pulsars over the entire southern sky. A total of 17 millisecond pulsars and 84 other pulsars were discovered, and a