Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Searching for binary and millisecond pulsars : a high-latitude drift-scan survey Kasian, Laura Elizabeth 2005

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_2005-0500.pdf [ 4.14MB ]
JSON: 831-1.0085728.json
JSON-LD: 831-1.0085728-ld.json
RDF/XML (Pretty): 831-1.0085728-rdf.xml
RDF/JSON: 831-1.0085728-rdf.json
Turtle: 831-1.0085728-turtle.txt
N-Triples: 831-1.0085728-rdf-ntriples.txt
Original Record: 831-1.0085728-source.json
Full Text

Full Text

Searching for Binary and Millisecond Pulsars: A High-Latitude Drift-Scan Survey by Laura Elizabeth Kasian B.Sc, The University of Winnipeg, 2003 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science m The Faculty of Graduate Studies (Astronomy) The University Of British Columbia October, 2005 © Laura Elizabeth Kasian 2005 Abst rac t We performed a drift-scan pulsar survey using the Green Bank Telescope (GBT) in West Virginia, USA. The survey is among the most sensitive ever undertaken in the Northern sky, and was collected with the Berkeley-Caltech Pulsar Machine filterbank. It was performed at low frequencies (400, 600, and 800 MHz) and mostly high Galactic latitudes with the aim of detecting new binary and millisecond pulsars. Population models of-pulsars have shown that young pulsars tend to be concentrated in the Galactic plane, while binary and millisecond pulsars appear to have a more isotropic distribution on the sky. Searches at high Galactic latitudes are therefore required in order to discover new binary and millisecond pulsars, and to better understand the distinct population of millisecond (recycled) pulsars. The data from the survey were reduced mainly using the sigproc pulsar signal processing software package. As of the writing of this thesis, ~ 73% of the data have been searched for periodicities. Each pulsar candidate is given a classification based on how closely it resembles a pulsar (Class 1 = probably a pulsar; Class 4 = most likely not a pulsar) . We were successful in re-detecting 4 previously-known pulsars, and we obtained lists of Class-1 and Class-2 candidates. We will confirm these candidates through follow-up observations with the GBT in December 2005. We detected fewer convincing Class 1 candidates than expected, which we attribute to a recently-discovered problem in the sigproc software and persistent radiofrequency interference (RFI) in the data, particularly at 600 MHz. We have corrected the software problem, which has resulted in the discovery our first Class-1 pulsar candidates. We expect that a reanalysis of processed scans and a first analysis of the remaining ~ 27% of data will yield a significantly higher number of promising candidates. iii Contents Abstract ii Contents iii List of Tables v List of Figures vi 1 Introduction 1 1.1 Radio Pulsars - Overview 1 1.1.1 Integrated Profiles 2 1.1.2 Dispersion . . : 4 1.2 Binary and Millisecond Pulsars 5 1.2.1 Millisecond Pulsars versus Normal Pulsars 5 1.2.2 Millisecond Pulsars and the P — P diagram 8 1.2.3 Associations Between Binary and Millisecond Pulsars 10 1.2.4 The Galactic Distribution of Millisecond Pulsars 11 2 Pulsar Surveys and the G B T Drift-Scan Survey 13 2.1 Previous High-Latitude Surveys of the Northern Sky 13 2.2 The Green Bank Drift-Scan Survey 14 2.2.1 Overview 14 2.2.2 Sky Coverage 15 2.2.3 Sensitivity 15 2.3 Expectations 18 3 Data Reduction: Searching 21 3.1 Search Procedure 21 3.1.1 Dedispersion 21 3.1.2 Periodicity Searches 22 3.2 Data Reduction: Searching 25 3.2.1 Drift Time Calculation 25 Contents i v 3 . 2 . 2 R F I E x c i s i o n 2 8 3 . 2 . 3 D M L i s t 3 1 3 . 2 . 4 S e a r c h i n g 3 3 3 . 2 . 5 F o l d i n g . . . 3 3 3 . 2 . 6 R e f i n i n g t h e S e a r c h P a r a m e t e r s 3 5 4 Results: Candidates 3 8 4 . 1 C l a s s i f i c a t i o n S y s t e m 3 8 4 . 1 . 1 C l a s s 1 C a n d i d a t e s . 3 8 4 . 1 . 2 C l a s s 2 C a n d i d a t e s 4 0 4 . 1 . 3 C l a s s 3 C a n d i d a t e s 4 0 4 . 1 . 4 C l a s s 4 C a n d i d a t e s 4 5 4 . 2 R e d e t e c t i o n s 4 5 5 Discussion and Conclusions 4 7 5 . 1 D i s c u s s i o n o f R e s u l t s 4 7 5 . 2 F u t u r e W o r k a n d C o n c l u s i o n s 4 7 Bibliography 4 9 List of Tables 2.1 A comparison of the sensitivities of four previous surveys of the Northern sky with our own 17 2.2 A table of values required in the sensitivity calculation for the drift scan survey. 18 4.1 A list of all observations in the drift-scan survey 39 4.2 A list of two Class 1 candidates from the survey. The detected period, disper-sion measure, and coordinates are shown. The flux densities are estimated from the radiometer equation, and are very rough estimates because we do not know if the pulsars were centered in the telescope beam 40 4.3 A list of the known pulsars that were in the region of sky processed to date in our survey. 42 4.4 Previously known pulsars, re-detected in the drift-scan survey 43 ? v i List of Figures 1.1 An example of an integrated profile (of PSR B0329+54) 3 1.2 Illustration of the dispersion effect (using PSR J1518+4904) 6 1.3 Period histogram of a large sample of the currently-known known pulsars. . 7 1.4 A P — P diagram for a large sample of the currently-known pulsars. . . . . 11 2.1 The sky coverage for the survey 15 2.2 Sensitivity plot: Signal-to-noise versus pulse period. Plots shown for 400, 600, and 800 MHz, and for several DMs 19 3.1 Illustration of the dedispersion process 23 3.2 An illustration of the path swept out by a source at declination 5 during one Earth rotation. Notice that this apparent path scales as cos 5 with declination. 26 3.3 A schematic explaining how each scan is split up into small files 27 3.4 Examples of the output from quickplot and quickgrid for J1518+4904. . 37 4.1 quickplot / foldsignals output for two Class 1 candidates. See Table 4.2 for more information about the candidates 41 4.2 Detection plots for the re-detected pulsars. 44 1 Chapter 1 Int roduct ion Pulsars are interesting objects, allowing us to study neutron stars with extreme precision. They can act as unique probes into extreme regimes of physics, and can help us understand stellar evolution. As the most stable clocks in nature, pulsars provides some of the most precise astronomical data available. However, before any science can benefit from the study of pulsars, we must first rise to the occasion of locating new pulsars. Searching for new pulsars is not a trivial exercise. Large volumes of data as well as computationally intensive calculations are required, and many hours of telescope time and state-of-the-art hardware are also needed. However, the payoff from searching for pulsars can be very exciting. Exotic binary pulsars such as double neutron-star systems can be highly relativistic systems, thereby providing testing grounds for strong-field gravity. Similarly, the discovery of a pulsar with mass unambiguously greater than two solar masses (c.f. Nice et al. 2005), or a pulsar with a sub-millisecond spin period could constrain the set of theories allowed for the neutron star equation-of-state. 1.1 Radio Pulsars - Overview The study of pulsars has come a long way since their initial discovery by Bell and Hewish in 1967 (Hewish et al. 1968). When the first pulsar was detected, the regularity of its pulses was the most striking feature. It was not immediately obvious that there existed a natural phenomenon that could emit such a regularly pulsed signal, and it was even considered that the signal could be a transmission from extraterrestrial intelligent life. Soon after the initial discovery, however, Pacini (1968) and Gold (1968) independently put forth the idea that the pulsar emission originates from rapidly rotating, highly magnetised neutron stars. In fact, Pacini (1967) had postulated two years prior that a magnetized neutron star could power the Crab nebula. In 1968, the pulsar in the Crab nebula.was discovered (Staelin and Reifenstein), which strengthened the case for the neutron star theory. However, at Chapter 1. Introduction 2 the time of the initial pulsar discovery, neutron stars were known to theorists but were obscure to observers. Neutron stars were first proposed in 1934 by Baade and Zwicky, who had acknowledged that a possible result of a supernova explosion is a small, massive star consisting mainly of neutrons. There were many other ideas proposed to explain the pulsar phenomenon. Theories involving more familiar white dwarfs were favored by some observers. Also, it was not immediately clear that the pulsations were from rotation. Two competing explanations involved stellar oscillations and the orbital motion of an extremely low-mass companion about a massive star (eg, Lyne and Graham-Smith 1998). The discovery of the Crab and Vela pulsars ruled out the possibility that oscillations were responsible for the pulsations. The periods of these pulsars (33ms and 89ms, respectively) fell in a gap between the expected oscillation frequency for neutron stars and that for white dwarfs. The orbital theory was also shown to be. extremely unlikely to produce the stable pulsations seen in the observed pulsars. Rotation or vibration seemed to be the only likely explanation for the pulses. The discovery that the Crab pulsar slows down with time (Richards and Cornelia 1969) finally limited the choice to the rotation scenario, as neutron star rotation is expected to slow down with time, while the vibrations are not. It is currently widely believed that the pulsars are rapidly rotating neutron stars. 1.1.1 Integrated Profiles We study pulsars is though a time series, a series of voltages sampled at a certain sampling rate, tsamp. The time series of a typical pulsar has pulsations that are regularly spaced at the spin period of the neutron star (with the exception of nulling pulsars (Backer, 1970), and some pulsars exhibiting subpulses (Drake and Craft 1968)).. Although the individual pulses of some pulsars can be seen by eye in the time series, most pulsars do not have individual pulses that are visible above the noise background of the receiver. In order to gain any useful information about the neutron star and its environment, it is a common practice (once the period of the pulsar is determined) to fold the time series at the spin period. Folding the data consists of adding the pulsar power modulo the topocentric period (i.e., the period of the pulsar as seen by the receiver). The signal increases approximately linearly with the number of pulses np; i.e., S oc np, while the the noise increases as ./V oc ^ /n^. The signal-to-noise ratio (SNR) therefore increases as SNR oc y/n^. Since the number of pulses increases linearly with time, SNR oc y/time. The integrated profile of a pulsar is the Chapter 1. Introduction 3 result of folding its signal at the topocentric period. Figure 1.1: An example of an integrated profile, created using data from the EPN database ( This is a 600-MHz signal of PSR B0329+54, from Gould and Lyne (1998). The plot on the left shows the entire pulse phase. The plot on the right zooms in on the pulse. For a given time series, the shape of each individual pulse varies from one pulse to the next. However, once an integrated profile is created by summing several thousand pulses, the shape of the profile is generally stable with no detectable long-term variations (Helfand et al. 1975). Although this is true for most pulsars, there are exceptions, including some double neutron star (DNS) systems in which profile changes are seen over long-term measurements due to geodetic precession (e.g., Weisberg et al. 1989), and some isolated pulsars apparently showing free precession (Stairs et al. 2000). The duty cycle of a normal pulsar is typically between 2-5% (Hankins and Rickett 1975), but millisecond pulsars have significantly higher duty cycles, which can be up to ~ 30% (Kramer et al. 1998). The shape of the pulse profile of MSPs are also, on average, slightly more complex. Using a sample of 180 pulsars, Kramer et al. (1998) determined that the median number of components required to represent the profiles of normal pulsars is 3, while the MSPs can be described by 4. The shape of the integrated profile is thought to contain information about the pulse emission mechanism at the polar regions of the neutron star. There are several simple models of the radio emission beam. Although these models can help explain and predict the emission properties of pulsars, they are empirical models - a firm theoretical understanding of the pulsar magnetosphere and emission mechanism is still lacking. The geometry of Chapter 1. Introduction 4 the emission beam is often modeled as a hollow cone centered about the magnetic axis of the neutron star, and was suggested early on by Radhakrishnan and Cooke (1969) and Komesaroff (1970). This model can somewhat account for the multiple components within the integrated profiles: we see more emission when our line-of-sight passes through the outer edge of the emission cone, and we see a drop in emission when we are not looking directly along an edge. If, as the neutron star rotates, our line-of-sight happens to slice through the cone, then we will see two edges and therefore two components. We see only one component if our line-of-sight happens to be right at the edge of the emission cone. In order to account for more complex emission beams, Backer (1976) extended this idea to include a core component in the centre of a hollow emission cone. Others further included nested cones (eg. Rankin 1993) in order to allow for yet more complex profiles. As an alternative to the cone-core model, Lyne and Manchester (1988) suggested a 'patchy beam' model for the emission beam. In this model, random regions of open field lines emit radiation, making for a patchy emission region that can account for more complex profiles better than the nested cone model. However, the patchy beam model does not explain the evolution of the profile with frequency as well as the nested cone model. Although these models can help understand some aspects of the pulsar emission process, there are still many open questions that need to be answered before we can understand it more completely. In particular, the coherent radiation process responsible for the observed radio emission is still not understood. 1.1.2 Dispersion 1 Dispersion by the interstellar medium plays a fundamental role in the study of pulsars. Generally, when an integrated profile is created for a given pulsar, we sum not only over all sub-integrations in the time series, but also over a range of frequencies. Although the emission mechanism for radio pulsars is still a heavily debated topic, it is commonly assumed that each pulse is emitted from the emission cone of the neutron star nearly simultaneously at all radio frequencies within a given, relatively narrow bandwidth. However, the radiation from each pulse is dispersed by the interstellar medium that lies between the Earth and the pulsar. Dispersion introduces a relative delay in the pulsar's signal between high- and low-observing frequencies. Over a range of frequencies from v\ and v2 we observe a relative delay of: At = - l— T x [y72 - v~2) • DM ' (1.1) 2.41 x 10"4 s-1 v 1 2 ' K 1 Chapter 1. Introduction 5 DM = ['Nedz Jo where DM is the dispersion measure of the pulsar, equal to the integral of the electron density Ne along the propagation path (eg, Hankins and Rickett 1975). See Figure 1.2 for an example of a dispersed signal. Although the effect, of dispersion is second-order in frequency, for the purpose of searching it often suffices to approximate this effect as linear over a small range of frequencies. For a more formal discussion of dispersion, including 'the propagation of the pulsar signal considering the transfer function of the interstellar medium, see Hankins and Rickett (1975). Not only can dispersion be seen between individual frequency channels (as in Figure 1.2), but it also has implications for the pulse shape itself. Within a given channel of bandwidth Av, the pulse will be a sum of all frequencies within the bandwidth. In fact, the pulse will be smeared over a time interval , which can be approximated as (eg, Lyne and Graham-Smith 1998): o DM T D M « 8.3 x 10 3 -3 sec (1.2) uMHz In order to minimize the effect of dispersion smearing within a given bandwidth, the simplest solution is to split the bandwidth into many narrow channels. When this is done, it is possible to improve the signal-to-noise ratio of the pulsar by correcting for the relative delay of the pulsar's signal at different frequencies. This process is called dedispersion and will be discussed in more detail in §1.1.2. If the DM of the pulsar is known, then we add an appropriate time delay to the time series at each frequency channel (as in equation 1.1), and then sum all frequency channels. Dedispersion can dramatically increase the signal-to-noise ratio. The effect is illustrated in Figure 1.2: if the time series at each frequency were added up without adding a time delay, all the individual pulses would not add coherently, and would thus produce no net signal. However, when the data is dedispersed before summing over all frequencies channels, we obtain the pulse shown at the bottom of the figure.. 1.2 Binary and Millisecond Pulsars 1.2.1 Millisecond Pulsars versus Normal Pulsars There are about 1600 radio pulsars known to date. The spin periods range from 8.5 seconds (Young et al. 1999) for the slowest known pulsar to 1.558 milliseconds (Backer et al. 1982) Chapter 1. Introduction 6 e Pulse Phase Figure 1.2: 130 s observation of J1518+4904 using the GBT. The x-axis is the pulse phase, and the y-axis is the sub-band number. We have plotted the (dedispersed) folded time series for 48 sub-bands, with the centre observing frequency being 590 MHz. The integrated profile is shown at the bottom of the plot. The period and dispersion measure of this pulsar, respectively, are ~ 40.9447 ms and 11.7 cm~3pc. for the fastest. If we plot a histogram of the spin periods of known pulsars (see Figure 1.3), we notice that the distribution is bimodal. The smaller, more rapidly-spinning population is what is now known as the class of millisecond pulsars (MSPs). The slower pulsars are referred to as normal pulsars. The first millisecond pulsar to be discovered, B1937+21 (Backer et al. 1982), is currently still the fastest pulsar known. It has become clear that the population of millisecond pulsars is phenomenologicalfy different from that of normal pulsars. Within a year of the discovery of the first MSP, both Alpar et al. (1982) and Radhakrishnan and Srinivasan (1982) had suggested accreting binary systems (in particular, the low-mass X-ray binaries, or LMXBs) as a progenitor for Chapter 1. Introduction 7 0.01 0.01 0.1 1 10 Figure 1.3: A histogram of the periods of a large sample of the known pulsars. Pulsars detected in globular clusters are omitted. Notice the bimodality of the distribution that distinguishes the two populations, normal and millisecond classes. Data from the ATNF pulsar catalogue (Hobbs et al. 2004, MSPs,. LMXBs are close binary systems in which a neutron star accretes matter from a low-mass companion, and X-rays are emitted due to gravitational interactions of infalling material onto the neutron star's surface. During this stage, the companion has evolved into a giant, filling its Roche lobe and transferring matter and angular momentum onto the neutron star. This causes the neutron star to be spun up or 'recycled' to a millisecond rotation period by the end of the accretion process. Through some (currently) unknown mechanism, the magnetic field of the neutron star, and hence the spin-down rate of the pulsar, become low. This gives the millisecond pulsar an extremely stable rotational period. The argument for this accretion scenario has recently been strengthened by the detection of millisecond oscillations in LMXB systems. In 1996, observations of two types of quasi-periodic oscillations were detected in LMXBs, whose oscillation frequencies hinted towards a connection to MSPs. In 1996, kilohertz quasi-periodic oscillations (kHz QPOs) were seen in Sco X - l by Van Der Klis et al. (1996), and in 4U 1728-34 by Strohmayer et al. (1996). These oscillations were interpreted as a beat-frequency between the neutron star rotation period and the orbital frequency of accretion-disk material around the neutron star. During the same year, oscillations were detected in 4U 1738-34 (Strohmayer et al. 1996) during a Type-I X-ray burst (a type of burst that is attributed to a thermonuclear flash on the surface of an Chapter 1. Introduction 8 accreting neutron star). The burst was modulated by a frequency that initially increased but asymptotically approached a single value in the tail of the burst. The oscillation was thought to be related to the neutron star spin period. In 1998, a detection was finally made of coherent oscillations within the persistent emission of an LMXB. The transient X-ray source SAX J1808.4-3658 exhibited 2.49 millisecond oscillations which showed a clear periodicity, highly stable to within 0.01%, at the spin frequency of the accreting neutron star (Wijnands and van der Klis 1998). Further, in 2003 burst oscillations were also seen in J1808.4-3658 (Chakrabarty et al. 2003) at the spin period of the neutron star, as determined by the coherent oscillations in 1998. Kilohertz QPOs were also detected in the millisecond X-ray pulsars J1808.4-3658 (Wijnands et al., 2003) and XTE J1807-294 (Linares et al., 2005). These discoveries effectively closed the observational gap between the LMXBs and MSPs, by providing evidence that neutron stars within LMXBs are able to spin at very fast periods. The term 'millisecond pulsar' has come to include those with periods from 1.5 to 30 ms, more or less. It refers to pulsars that have been spun up or recycled through the process described here, which typically have spin periods in the aforementioned range. The heavily recycled MSPs (i.e. those that have been spun up to very low spin periods) are thought to be those with LMXB progenitors; however, more massive systems (eg, double neutron star systems) which are not as recycled are still included as MSPs. 1.2.2 Millisecond Pulsars and the P — P diagram It is useful to look at a 'P — P' diagram such as in Figure 1.4 in order to understand, at least observationally, the difference between the two populations of pulsars. It is clear that there are two distinct islands of pulsars within the period-period derivative space: the normal pulsars are concentrated in the middle-to-upper right area of the plot, while millisecond pulsars appear towards the lower left-hand corner. We can gain a rough idea of the magnetic field B and the characteristic age r of a pulsar from its position on the P — P diagram. Since the pulsar has a strong dipole magnetic field, it is convenient to make the assumption that it is a perfect magnetic dipole. If we assume the spin-down luminosity of the pulsar is strongly dominated by dipole radiation, we can Chapter 1. Introduction 9 obtain an expression for the surface magnetic field B of the pulsar: (eg, Lorimer and Kramer 2004). For a neutron star that has radius R — 10 km, moment of inertia I = 1045 g cm , and orthogonal spin and magnetic axes (i.e., a = n), we obtain the following: Bs = 3.2 x 101 9 G x \[pP (1.4) If we assume that the spin frequency v and frequency derivative z> follow a general power law with index n, v — —Kvn, then in terms of the period P and period derivative P we can write, P = KP2~N. Integrating this expression over time, we obtain the following characteristic age of a pulsar born with period PQ: T = (n-l)P (1.5) (eg, Lorimer and Kramer 2004)'. If we include the assumption that a dipole magnetic field is the dominant source of rotational energy loss in the pulsar, then we have the braking index n = 3. If we also assume also that PQ « P, then Equation 1.5 simplifies to: P TC = A- (1.6) 2P K J The characteristic age is often not a good estimation of the true age of a pulsar. For example, Camilo et al. (1994) found that the characteristic ages of some millisecond pulsars exceeded the age of the galactic disk, which does not make sense. However, r c is still a useful quantity to measure, since there is often no better estimate available. It is possible to measure the braking index n in systems where measurements of the second period derivative are available, and provided we are confident it is not timing noise. This can then be compared with the value of 3 which is assumed in the case of a magnetic dipole. For rotation frequency Q, the braking index can be expressed as n = QQ/Q.2. We now have a rough estimate for the surface magnetic field Bs of the pulsar, and its characteristic age r c . Both are related to the period and period derivative through a power law. When we plot curves of constant characteristic age and of constant surface magnetic Chapter 1. Introduction 10 field, they appear on the log-log P — P diagram as lines. In Figure 1.4, lines are plotted for several values of constant r c and B$-Young pulsars begin their lives in the top left region of the plot. As the pulsar radiates away rotational energy and loses angular momentum, its period increases to somewhere near the largest island of pulsars in the plot, with typical periods on the order of one second, and typical spin-down rate of 1 0 - 1 5 s s _ 1 (eg. Lorimer 2001). After many (~ 108) years, the pulsar has slowed down to a period that no longer suffices to produce the electron-positron pairs that provide the polar emission of radiation. We can no longer see the neutron star as a pulsar, and this region of the P — P space is often called the pulsar graveyard (Chen and Ruderman 1993). The millisecond pulsar population are those that have been spun up to very stable, fast rotational periods. It is clear from the position of the millisecond pulsars in Figure 1.4 that they are a substantially older population with lower magnetic fields than normal pulsars. Typical millisecond pulsars have magnetic fields ~ 108 Gauss and characteristic ages of ~ 109 years. By comparison, normal pulsars have typical magnetic fields of ~ 1012 Gauss, and characteristic ages of ~ 101 0 years (eg. Lorimer 2001). During the accretion process described above, the pulsar is thought to move in the diagram from below the death line to the lower left-hand corner, where it will remain for a long time due to its low period derivative. 1.2.3 Associations Between Binary and Millisecond Pulsars Millisecond pulsars are much more likely than normal pulsars to have a binary companion. This is advantageous for learning about the supernova and mass-transfer histories of these systems: the vast majority of MSPs are in binary systems (around 80% of the observed MSPs have binary companions, as compared to less than one percent for normal pulsars (Lorimer 2001)), which means we can learn more about MSPs than about isolated pulsars. Through timing, the orbital parameters can be obtained, which provide clues as to how these systems formed. For example, in long-orbit low-mass MSP binaries, the theory of a red giant overflowing its Roche lobe and transferring matter to its companion through accretion, in combination with nonuniform plasma density within the envelope of the red giant, has been able to predict the orbital eccentricities of the systems fairly well (Phinney 1992; c.f. Stairs et al. 2005). Chapter 1. Introduction 11 Figure 1.4: A P — P diagram, which plots the period derivative (or spin-down rate) of each pulsar versus its spin period. Data from the ATNF pulsar catalogue (Hobbs et al. 2004, Pulsars in globular clusters are omitted. Millisecond pulsars in binary systems can be extremely useful to study. They have extremely stable rotation periods, which makes them a valuable resource in testing general relativity (eg. Stairs 2003). However, only around 90 millisecond pulsars are currently known, many of them having been discovered in globular, clusters (in which it is much more likely for the MSPs to interact with other systems than the rest of the MSP population). One of the goals of pulsar astronomy has therefore been to increase the known sample of millisecond and binary pulsars for further study. 1.2.4 The Galactic Distribution of Millisecond Pulsars In order to further study millisecond pulsars, we must increase our sample; however, to do this we first need to know where to find them. Studies of the known population of MSPs can be used to guide our our surveys towards detecting new ones. The first MSP discovered at a high galactic latitude (Wolszczan 1990) provided evidence that the distribution of observable MSPs is roughly isotropic. This is a contrast from the population of young pulsars, which Chapter 1. Introduction 12 are concentrated in the Galactic plane where they are born. Also, the inferred flux density 5„ for each pulse is related to the observing frequency u by a power law S„ oc ua. with typical spectral index a = —2 (eg, Lorimer and Kramer 2004) or a = —1.6 (Lorimer et al., 1995) so that pulsars are brighter at low frequencies. It was soon suggested (Johnston and Bailes 1991) that all-sky surveys at low frequencies would be an efficient way to discover new MSPs. Since then, many surveys at high- and intermediate- galactic latitudes have been performed, and have discovered ~ 50 MSPs (eg. Lorimer et al. 2005). Population statistics of currently-known samples of pulsars have suggested that the num-ber of active normal pulsars in the Galaxy is approximately equal to the number of active MSPs in the Galaxy (Lyne et al. 1998). However, MSPs are often heavily selected against in surveys. The average luminosity of MSPs appears to be lower than for normal pulsars. Also, DM smearing due to inadequate time sampling and/or coarse channelization has made them difficult to detect. In addition, acceleration within the binary systems we hope to dis-cover creates another selection effect. Also, many surveys have concentrated on the Galactic plane, which is not the most efficient way to search for MSPs, as mentioned above. It is clear that a high-latitude, extremely sensitive survey conducted at low frequencies would have the potential to discover new millisecond pulsars quite efficiently. We have conducted such a survey with the Green Bank Telescope, which will be discussed in more detail in Chapter 2. 13 Chapter 2 Pulsar Surveys and the G B T Drift-Scan Survey 2.1 Previous High-Latitude Surveys of the Northern Sky There have been several groups who have undertaken high-latitude surveys of the Northern sky in the search for new MSPs in the past two decades. Four of these surveys are briefly outlined here. • A survey at 411 MHz was performed using the 76-metre Lovell telescope at Jodrell Bank (Nicastro et al. 1995). This survey detected both circular polarizations, and used 8 MHz of total bandwidth which was split into 64 channels of 125 kHz each. The data was one-bit sampled at a rate of 0.3 ms. The survey yielded only one pulsar, PSR J1012+5307, which is a fast MSP in a 14.5 hour binary orbit around a white dwarf (Callanan et al. 1998; van Kerkwijk et al. 1996). The sensitivity to millisecond pulsars was 5 mJy. • A 81.5-MHz survey was undertaken using the same Cambridge 3.6 Hectare dipole array that discovered the first pulsars (Shrauner et al. 1998). The array was operated as a total-power phased array, and had a total bandwidth of 1MHz and one-bit sampling at a rate of once per 0.768 ms. Declinations between —20° and +86° were surveyed. The survey was successful at re-detecting 20 previously-known pulsars, but did not discover any new ones. The survey was sensitive to MSPs down to 200 mJy. • A search was performed at 390 MHz using the 92-m telescope at Green Bank (Dewey et al. 1985). This survey covered ~ 1.8 sr of the Northern sky, and sought to detect low-luminosity pulsars. The survey resulted in the detection of 83 pulsars, including 34 new ones. The bandpass of 16 MHz was split into 8 channels, at a sampling rate of Chapter 2. Pulsar Surveys and the GBT Drift-Scan Survey 14 60 Hz. This survey was not sensitive to MSPs with periods below 30 ms. A companion survey (Stokes et al. 1986) was sensitive to MSPs but covered primarily the Galactic plane. • A survey at 370 MHz was performed using the Green Bank 140-foot telescope (Sayer et al. 1997). The survey had a bandwidth of 40MHz and sampled once every 0.256 ms. Its sensitivity to MSPs was 8 mJy. Its goal was to detect the strongest millisecond pulsars in the Northern sky, and it succeeded in detecting 8 new pulsars including two recycled pulsars, PSR J0122+1001 (detected first by Camilo et al. (1996a)) and PSR J1518+4904 (Nice et al. 1996). 76 previously-known pulsars were also detected. 2.2 The Green Bank Drift-Scan Survey 2.2.1 Overview Our drift-scan pulsar survey was done using the 100-metre Green Bank Telescope (GBT) during 2002-2003. Observations were made at times when the dish needed to be station-ary for maintenance that was required for investigation and repair of problems with the azimuth track. We were able to take advantage of unplanned downtime that resulted from unpredictable scheduling due to the maintenance. The data was taken at frequencies of 400, 600 and 800 MHz, depending on which prime-focus receiver happened to be in place at the time of the observation. Data was taken using the Berkeley-Caltech Pulsar Machine (BCPM) filterbank (see Backer et al. (1997) for a description, and Camilo et al. (2002) for the discovery of the pulsar in the supernova remnant 3C58, the first pulsar discovery made using the BCPM). This provided 4-bit sampling every 72^s over a bandwidth of 48 MHz split into 96 channels. Due to the drift-scan nature of our survey, we were not able to choose the integration time at each pointing. The integration time for any particular source is given by the drift time of the telescope for the declination of the source. This is the amount of time it takes a point source to move through the beam of the telescope on the sky, and can be calculated for any particular observing frequency and declination. See §3.2.1 for a calculation of the drift time for this survey. Chapter 2. Pulsar Surveys and the GBT Drift-Scan Survey 15 2.2.2 Sky Coverage Figure 2.1 shows the sky coverage of the survey. The total observing time was about 200 hours, and nearly 600 square degrees were covered. The telescope was pointed North as much as possible, so as to cover the largest range of right ascensions. The colours in the figure represent the three observing frequencies. The blue (darkest) strips were observed at 400 MHz; yellow (lightest) strips at 600 MHz; and green (medium) strips at 800 MHz. The survey consisted of 31 scans taken at a variety of declinations, for different lengths of time ranging from 2 to 14.3 hours. 40 l-3 0 I I , , i , I I 0 5 10 15 20 25 Right Ascension (hours) Figure 2.1: The sky coverage for the survey. The colours (grayscale) represent the three observing frequencies: blue (darkest) for 400 MHz; red (lightest) for 600 MHz; and green (medium) for 800 MHz. Declination is plotted versus Right Ascension. 2.2.3 Sensitivity Our survey is the most sensitive of its kind in this region of the Northern sky. The previous surveys discussed above have a range of sensitivities and sampling times, giving each one a different sensitivity to MSPs. In Table 2.1, the minimum detectable flux and sampling time of our survey is compared with those of the four surveys discussed above. It is clear from Chapter 2. Pulsar Surveys and the GBT Drift-Scan Survey 16 this comparison, especially considering that we sampled at 4 bits as opposed to 1 with the other instruments, that our survey is far more sensitive to MSPs than the others. The sensitivities in Table 2.1 for our survey were calculated using an expression derived from the radiometer equation (Dicke 1946). The radiometer equation expresses the root-mean-square fluctuations in the system temperature ATsys as: Psys + Tsky(l, b) ^/npt0bsAu Ann _ -^sys T J-sicyy, <Jj . , L±l-sys — rZT~i~.—7TT. l z - 1 J where Tsys is the system temperature, Tsky(l,b) is the sky temperature for given Galactic coordinates 1 and b, np is the number of orthogonal polarizations sampled, t0bs is the integration time, and Au is the total bandwidth of the receiver. Using this definition of the RMS noise in the receiver, we obtain (see, for example, Lorimer and Kramer (2004)) an expression for the minimum detectable flux density Smin (Dewey et al. 1985): „ _ JS/Nmin)[Tsys + Tsky(l,b)} We b m m ~ P GJnptobsAv iP-We in terms of the threshold signal-to-noise ratio S/Nmin (which we take to be 8), antenna gain G, effective pulse width We, period P, and quantization correction factor (3. We take (3 — 1, following the choice of Camilo et al. (1996b) for their survey which used 3-bit sampling. We include the following contributions in the effective pulse width We: the intrinsic pulse width of the pulsar W; the dispersion broadening across a single frequency channel TDM (given by Equation 1.2); the sampling time of the survey TSAMP; and the broadening of a pulse due to interstellar scattering rscatt. We obtain: We = ^W2 + rlM + T l m p + T l a t t (2.3) (Dewey et al. 1985; Lyne and Graham-Smith 1998). Both Tsamp and Tscatt are assumed to be small in comparison to W and T^M- This is clear for the sampling time (which is at. least an order of magnitude smaller than the fastest pulsar), and the scattering broadening can be expressed as a function of DM and frequency as: /DM\35 ( 400 4 w = U b o J {^MVJ SECONDS (2-4) Chapter 2. Pulsar Surveys and the GBT Drift-Scan Survey 17 This relation was discovered empirically by Slee et al. (1980; eg, Lyne and Graham-Smith 1998). It can be seen from this expression that for our survey, which is is concerned with DMs smaller than 1000 and frequencies greater than or equal to 400 MHz, we obtain a very small value for TSCATT: This reduces Equation 2.3 to: We = JVp + T\ 2 _ DM \ W2 + .3 x 103 D Z M Av sec vMEz (2.5) Note that the sky temperature Tsky(l,b) is low off the Plane, but it is high on Plane (Haslam et al. 1982). The sensitivities in Table 2.1 and in Figure 2.2 (discussed below) represent the off-Plane best-case scenario, and must be modified when pointing near the Plane. For our observations, two polarizations and the total bandwidth of 48 Hz were used. The system temperature and gain for each receiver were obtained from the GBT website, and are shown in Table 2.2. The integration times are also listed in the table, as calculated in §3.2.1 for a declination 5. We chose 5 — 45° in Equation 2.1 to obtain the sensitivities in Table 2.1. Note that we calculated the sensitivity of our (100-m) GBT survey only; the values for the other surveys were found in the literature. Survey Frequency Sampling Time Sensitivity to MSPs (MHz) (ps) (mJy) Jodrell Bank 411. 300 5 Cambridge 81.5 768 200 Green Bank (92-m) 390 5555 2** Green Bank (43-m) 370 256 8 Green Bank (100-m) 400 72 1.5 600 72 1.1 800 72 0.7 Table 2.1: A comparison of the sensitivities of four previous surveys of the Northern sky with our own. Listed are the frequency, sampling time, and minimum detectable flux for each of the surveys. A high sampling rate and a high sensitivity are both required for detecting faint MSPs. ** Note: The minimum flux density for the 92-m Green Bank survey is listed here for normal pulsars; the survey was not sensitive to MSPs. In Figure 2.2 we plot the minimum detectable flux density at each of our observing Chapter 2. Pulsar Surveys and the GBT Drift-Scan Survey 18 Receiver Frequency ± sys Gain Integration Time (MHz) (K) (K/Jy) (sec) 450 57 2.0 126/cos 5 600 48 2.0 84/ cos S 800 25 2.0 63/cos 5 Table 2.2: A table of values required in the sensitivity calculation for the drift scan survey. The values for Tsys and the gain were obtained from the GBT website http://www. nrao. edu//GB T/proposals/ TPROPOSERSG UIDE.pdf, and the integration time is calculated from Equation 3.2.1. frequencies as a function of pulse period for five different dispersion measures. We assumed an integration time equal to the drift time for each frequency and a declination of 45°. We also assumed a duty cycle of 30% for periods less than 10 ms, and scaled the duty cycle as P~1/2 for periods greater than 10 ms. This choice of a period-dependent pulse width was used in Cordes et al. (2005, submitted), which describes the ALFA pulsar survey being performed with the Arecibo Observatory. Equation 2.2 was used to produce the plots, using the information in Table 2.2. Pulse broadening due to dispersion smearing is included in the total pulse width W. 2.3 Expectations A survey covering 1.8 sr of the Northern sky was performed at 390 MHz using the 92-m Green Bank telescope (Dewey et al., 1985). We use the results from this survey to obtain a very rough estimte of the number of pulsars that we expect to detect in the portion of sky covered by our drift-scan survey. Their observing frequency of 390 MHz is not significantly different from the frequencies used in our survey (~ 400, 600, 800 MHz), so we have not attempted to correct for changes in flux at different frequencies due to pulsar spectral indices (typically ~ 1.6 (Lorimer et al. 1995)). With a sensitivity of Smin ~ 2 mJy, Dewey et al. detected 83 pulsars in ~ 5909 square degrees. Based on our sky coverage of 591 square degrees, this number of pulsars scales to ~9. However, our sensitivity is ~ 1 mJy. In order to estimate the scaling factor due to this difference in sensitivity, we need to make an assumption about the distribution of pulsars in the sky. Assuming a minimum detectable flux density 5 m i„, and that the number of detectable pulsars is limited by luminosity rather Chapter 2. Pulsar Surveys and the GBT Drift-Scan Survey 19 400 MH7 600 MHz 1 10 100 1000 1 10 100 1000 Figure 2.2: The minimum detectable flux density (in mJy, log scale) versus the pulse period in milliseconds, for 400, 600 and 800 MHz. A range of dispersion measures is plotted for each frequency. A duty cycle of 30% is assumed for periods less than 10ms, and the duty cycle is scaled as W cc P~1/2 for periods P greater than 10 ms. Pulse broadening due to dispersion smearing is included. Chapter 2. Pulsar Surveys and the GBT Drift-Scan Survey 20 than scattering, we make two independent calculations, based on a planar distribution and an isotropic distribution of pulsars. If we assume a planar distribution, the number of pulsars should scale as N oc 5~JN, and we should expect ~ = 2 times the number of pulsars per square degree in our survey than in the 390 MHz survey. If we assume an isotropic distribution of pulsars, the number of pulsars should scale as N oc S^^2. We would therefore expect to detect a factor of ~ (^j^)~ 3 / / 2 = 2.8 times more pulsars than the 390 MHz survey. This is about 18 - 25 pulsars in all. Since we know of 5 pulsars that should be detectable in our survey, this leaves ~ 13 - 20 new pulsars left for us to discover in our data. We expect that one or two of these will be MSPs, based on the overall fraction of MSPs in the population of known pulsars (eg, Hobbs et al. 2004). Note that this is a very rough estimate of the number of pulsars we should see in our survey, and it also does not account for the RFI in our survey, which has made detections more difficult. A full analysis would require a pulsar population model covering all Galactic latitudes, recognizing that pulsars are not really isotropically distributed but instead have different scale heights for young and millisecond pulsars. It would also take into account the fact that we used multiple frequencies in our. survey. 21 Chapter 3 Data Reduction: Searching Here we will describe the process through which the drift-scan data was reduced. The standard search procedure was followed (as described in §3.1), and the specific considerations needed in dealing with the drift-scan data will be addressed. We will discuss the organization of the data as well as specifics about searching using the signal processing software package sigproc. The sigproc package, written by Duncan Lorimer, and incorporating his seek software, is publically available from 3.1 Search Procedure The central problem in searching for pulsars is to find a periodic signal with unknown period, an unknown dispersion measure, and an unknown amplitude and pulse width, within a time series containing seemingly random noise. The two main parameters in our search space are the dispersion measure and the period of the pulsar. There have been many techniques and algorithms developed that aid in at efficiently detecting signals that are buried within a time series. Some of the methods used in the drift-scan survey are discussed here. 3.1.1 Dedispersion The process by which a pulsar signal is dispersed by the interstellar medium is discussed in §1.1.2. The effect is only visible due to the pulsed nature of the pulsar signals, and can be seen in a plot of observing frequency versus the time-of-arrival of a particular pulse (see Figure 3.1). Although it is possible to completely correct for dispersion through a process called coherent dedispersion (Hankins and Rickett 1975), during pulsar searches it is not necessary for a time series to be perfectly dedispersed in order to detect a pulsar signal. We therefore use dedispersion techniques that are more efficient, both in storage and computing time, when searching for pulsars. Chapter 3. Data Reduction: Searching 22 Conceptually, the simplest method of dedispersing a time series Xij (sampled at frequency channel i and time sample j) is to apply the appropriate time delay to each frequency channel. This is the method of dedispersion performed in our survey using sigproc. If we have N frequency channels, and v\ is the centre frequency (in MHz) of the ith channel, then the time delay of a dispersed signal with dispersion measure DM, relative to the first channel is (according to Equation 1.1): A'- = 2 . 4 1 x l O - ^ - X K 2 - " - 2 ) - D M The number of time samples by which the ith channel must be shifted is then ki so we can write the dedispersed time series as (Hankins and Rickett 1975): N Vj = '52xiJ-ki (3-2) i=i Although this method is conceptually simple, it is not always the most computation-ally efficient algorithm. An algorithm that can in some cases be much more efficient was developed by Taylor (1974), and is referred to as tree dedispersion. It requires Nlog2N computations, as opposed to the iV 2 computations required for the simple dedispersion pro-cess described above; however, it is not always optimal for all surveys, due to additional steps involved and assumed linearity of the signal with frequency. Tree dedispersion will not be discussed any further here, as it was not used in the analysis of our survey. 3.1.2 Periodicity Searches Once we have a dedispersed time series, it can be searched for periodic signals. This can be done in either the time domain or the frequency domain. Our searches were done in the frequency domain. Within sigproc we used find, which applies a Fast Fourier Transform on the dedispersed time series and performs a periodicity search in the frequency domain. If a Fourier amplitude of a particular spectral bin (of width l/£0fes) -is above a given signal-to-noise threshold, the centre frequency of the bin is recorded as a periodicity. Since the number of samples in each time series is unlikely to be a power of two (which is convenient for performing Discrete Fourier Transforms), the technique of zero padding is used in our survey. This constitutes the addition of zero-valued samples at the end of our time series, and can also be useful in increasing the number of spectral bins and decreasing the bin (3-1) Chapter 3. Data Reduction: Searching 23 1 1 Figure 3.1: Top: Frequency channel versus the time of arrival of two consecutive pulses. The time delay within one frequency channel (of width B) is illustrated. Bottom, left: Our total bandwidth is split up into many smaller frequency channels. Bottom, right: We introduce a time delay to each channel, which is dependent on the centre frequency of the channel, relative to the centre frequency of the lowest frequency channel, as well as the DM of the pulsar. For a total of N frequency channels, this can increase the signal-to-noise of the pulse by a factor of VW. Chapter 3. Data Reduction: Searching 24 size. A feature of zero padding is that it is less likely for a periodicity to fall between two neighboring bins, thereby splitting its power between the two bins and lowering its signal-to-noise. See Lorimer and Kramer (2004) for more details on frequency-domain searching. Harmonic Summing During this stage of the pulsar signal processing, harmonic summing is performed to better detect narrow pulses. For pulses with large duty cycles, their pulse shape is more nearly sinusoidal than those with very narrow duty cycles. A pulse that is nearly sinusoidal will require fewer Fourier components to describe it, and so its Fourier power in the fundamental will be stronger than that of a narrow pulse, which has its Fourier amplitudes split between several components. It is for this reason that we consider five harmonic folds. The process of harmonic summing sees the original spectrum stretched by a factor of two, and then its amplitude added to that of the original spectrum. The same occurs for the second, third and fourth harmonic folds. In the find algorithm, the first, 2nd, 4th, 8th and 16th harmonics are summed together. The idea here is to add up the Fourier components that are strongly harmonically related, so as to reduce any bias towards pulsars with larger duty cycles. Summing up to 16 harmonics suffices for duty cycles as low as 2%; however, for pulses that are narrower summing to 32 harmonics may improve the sensitivity (eg, Ransom et al. 2002; Taylor and Huguenin 1969). Acceleration Searching One other consideration that is often used in searching for binary pulsars in short orbits is an acceleration search. The need for a search in orbital acceleration is clear when searching for short-orbital period binaries within long integrations. The acceleration due to the orbital motion of the binary system often causes the the topocentric period of the pulsar to vary so much that the signal-to-noise of the pulsar is dramatically decreased. This can cause such a pulsar to go undetected, if the signal-to-noise is reduced to a value below the detection limit. Within find there is the option of including an acceleration search. The program searches for signals using a list of trial acceleration values, and outputs a list of signal-to-noise values for each acceleration value in the list. Although we are searching for systems that are potentially short-orbit binaries, the option to search in acceleration space was not used. The first reason for this is that this particular search is so computationally intensive Chapter 3. Data Reduction: Searching 25 that it is best used in long-integration globular cluster searches, in which the dispersion measure is already known, which eliminates the need to search a large range of DMs. A second reason is that for this drift scan, the integration times are typically between one and five minutes, which we-reasonably assume to be short compared to the orbital period of the binary systems we are looking for.' We do not expect a significant decrease in signal-to-noise due to binary motion. 3.2 Data Reduction: Searching 3.2.1 Drift Time Calculation In a drift-scan survey, an important consideration is the fact that the telescope does not track a particular position on the sky, but stays stationary while allowing the sky to move through its beam. The drift time is the amount of time that a point source at a particular declination will remain in the beam of the telescope for a given observing frequency. This is the amount of time we would expect to be able to see a pulsar. Most of the drift times in our survey are on the order of a few minutes, while each scan is a few hours long. If we were to search each entire scan for periodicities, we would not expect to detect a signal from a pulsar that is visible for such a small portion of the time. We therefore need to split the whole observation into smaller pieces of length equal to the drift time. Each small piece can be searched for periodicities. To calculate the drift time, we need to consider first the apparent angular speed of a source on the sky, and also the size of the beam on the sky. In order to calculate the apparent speed of a source at a given declination 8, we must first consider that the beam of the telescope does not sweep out an equal amount of sky solid angle at each declination. This is illustrated in Figure 3.2: the apparent path of a source on the sky throughout one Earth rotation is greater for declinations closer to zero. In fact, the apparent angular distance traveled throughout one day by a source at a declination of zero degrees is 6(6 = 0) = 2TT. (3.3) To obtain the angular distance traveled by a source at any declination 5, we scale Equation Chapter 3. Data Reduction: Searching 26 Ax is of Rotat ion Celest ia l Equator Figure 3.2: An illustration of the path swept out by a source at declination 5 during one Earth rotation. Notice that this apparent path scales as cos 5 with declination. 3.3 by cos 6, so that 9(5) = 2TTCOS<5. Now, since the corresponding period is 24 hours, we can write: 27T cos 5 cos 5 24 hr arcmin/sec (3.4) (3.5) Because the GBT is a radio telescope, its beamwidth is diffraction-limited . The beamwidth is therefore frequency-dependent, and is given on the GBT website ( by B = 740 arcseconds "GHz for a frequency UGHZ in GHz. The drift time is then B 49,600 sec ^drift — W VMHz • C O S 5 (3.6) (3.7) Chapter 3. Data Reduction: Searching 27 for a frequency VMHZ m MHz. The drift time depends on both the observing frequency and the declination of the source. These two factors are constant throughout any single observation, so that the drift time need only be calculated once per scan. For each scan, the drift time was calculated according to equation 3.7. Small files (with integration time equal to the drift time) were created using the f ilterbank task in sigproc, which convertes the raw BCPM data into a standard filterbank format that can be further processed using other sigproc tasks. Each small file was processed individually. Our goal is to find unknown pulsars within our data, located at unknown positions on the sky. Because we have a drift scan survey, each chunk of data (which corresponds to a particular time interval) corresponds to a particular strip of sky. However, we do not know a priori exactly where a pulsar will be on the sky. For this reason, some consideration needed to be made about how to split up each scan into its many small (drift time - length) files. We decided to overlap the small files by one-half the drift time, so as to minimize the chance of missing a pulsar due to its signal being split between two files. See Figure 3.3 for a schematic of how the scan was split into small files. file #2 file #3 + file #4 - t -file #5 + file #6 0.5 1.5 2 2.5 Time (in units of the drift time) etc. Figure 3.3: A schematic explaining how each scan is split up into small files. Our choice to overlap the files by half the drift time effectively doubles the amount of processing required, but minimizes the chance of missing a pulsar due to its signal being split between two files. The colours are used only to differentiate between consecutive files. Chapter 3. Data Reduction: Searching 28 3.2.2 RFI Excision Radio-frequency interference (RFI) can be a substantial problem in pulsar surveys. Many Earth-based periodic signals (such as the AC power line at 60 Hz and broadband television stations) can seem very pulsar-like, and can limit the effectiveness of a survey. RFI costs time (by looking at the RFI instead of other, possibly real, candidates), and often causes real pulsars to go undetected if they are weak or have periods very similar to that of the unwanted signals. It can also have an influence on the spectrum. In this survey, RFI excision was performed both before and after a time series is searched for periodicities. When excision is done before the searching, the DM zero timeseries is searched for periodicities, and a list of bad signals (often referred to as a birdies file, or a zap list) is compiled. It is useful to excise RFI before the search for several reasons. First, it reduces the total RMS of the signal, thereby giving any real pulsars a higher total SNR. Second, it can prevent signals that are known a priori to be bad from swamping candidate lists in later stages of data analysis. Looking at the RFI before the searching takes place also gives a good indication of which signals can be expected to appear in the candidate lists. Since the birdies list itself need not contain all bad signals (in some cases, zapping all bad signals could result in a large percentage of the spectrum being removed!), the bad signals not included in the birdies list will likely appear in the final candidate lists. In such a case, it is useful to have a prior list of signals present in the data so as not to waste time further analyzing these signals. RFI filtering after the search process can also be very useful. We take advantage of the fact that we are looking at multiple positions on the sky at different times, and cross-examine all signals detected within one scan. This screening process significantly decreases the number of human hours required in looking at pulsar candidates. Both of these types of RFI filtering are discussed in the following sections. Pre-search Filtering Fortunately, it is possible to identify a large amount of RFI relatively easily. Terrestrial sources of RFI are usually detected by sigproc with a DM of zero. A broadband signal that is emitted from the Earth does not become significantly dispersed by the time it reaches our receiver. The time delay between two frequencies within our bandwidth (Equation 1.1) will be small so that any dispersion effects will not be noticeable. In pulsar searching, it is Chapter 3. Data Reduction: Searching 29 is therefore useful to create a time series that has been dedispersed at a DM of zero and to inspect it for periodic signals. Any periodicities found in the DM zero time series are most likely RFI. The signals found in the DM zero time series are put into a list called a birdies file. This is a list that specifies which spectral bins to zap during the search process. We specify the central frequency to be zapped, the width of the channel to be zapped, as well as the number of harmonics of each signal to be zapped. Careful analysis was performed to enter the cor-rect pulse width and number of harmonics. An analysis of the DM zero time series was performed for all scans using both sigproc and another pulsar signal processing software package, PRESTO (written by Scott Ransom (Ransom, 2001), and publicly available from sigproc was used to obtain a list of period-icities present in the DM zero time series, and PRESTO's explorefft was used to obtain information about the frequency width and the number of harmonics to be zapped.. Not zapping enough of the spectrum can allow the signal from a true pulsar to be swamped by RFI; however, zapping too much spectrum has the danger of ignoring spectral bins that contain the pulse frequency of a true pulsar. For the drift-scan survey, an additional measure was taken to assure,that the signals in the DM zero timeseries are indeed RFI and not nearby pulsars. In a drift scan, the telescope is continuously looking at different pointings throughout an observation. As a result, a time series covering more than a few drift times should not yield any true candidates. A real pulsar should only be visible for one drift time, and over a longer time series its signal should become diluted so as to produce no real candidates. Therefore, a relatively long DM zero timeseries should detect any persistent RFI present during that portion of the observation. The drift-scan survey consisted of 31 scans taken at different declinations and frequencies, for different lengths of time ranging from 2 to 14.3 hours (see Table 4.1). Because each scan was different, this required a different birdies file to be created for each scan. The time variation of RFI within one scan was monitored, and it appears that the RFI does not vary much over the course of a single scan. The assumption was therefore made that the RFI did not change significantly within each observation of this drift-scan survey, so that one birdies file was sufficient for a large portion of a scan. The first half hour of each observation was used to represent the RFI within that entire observation (except for the longest observations, for which two or more birdies files were required). The DM zero timeseries of the first half hour of data is formed and searched for periodicities. The resulting periodicities are placed Chapter 3. Data Reduction: Searching 30 into the birdies file. One last consideration during the pre-search RFI excision is to ignore specific frequency channels from our bandpass. The idea here is to eliminate any specific channels that consis-tently receive strong RFI. A preliminary attempt was made at characterizing the RFI as a function of the bandpass. However, this did not prove to be simple: the bandpass changed drastically with time for the scans that were analyzed, and we were not able to find a clear consistent association between any particular channel(s) and the strong RFI. Further inves-tigation continues, as we hope to create an automatic system for zapping specific channels within our bandwidth. Post-Search Filtering Once the searching process is finished for a given scan, the next step is to look at each candidate plot by eye to determine how likely it is that the candidate is a pulsar. However, there are typically anywhere from hundreds to thousands of candidates that result from a single scan. Not only is this time consuming for the individual who must look through each candidate, but it is also not efficient. In looking through each candidate, there are certain things that we look for in determining how interesting the candidate is. The most obvious consideration is a quick glance at the candidate's period. If a signal of exactly the same period is detected in several different regions of the sky, it is probably of terrestrial origin. A script called f i l t e r was written to compare each candidate to all other candidates within a scan. Only the candidates that 'passed' the conditions set by f i l t e r were allowed on to the next stage of the candidate selection process (see §4). f i l t e r allows a candidate to 'pass' if its period is significantly different (i.e. to ~ 4 decimal places) from all but 5 other candidates. We give preference to periods which are detected in adjacent beams. We here define two beams to be adjacent if they are either overlapping (eg, files #1 and #2 in Figure 3.3) or consecutive (eg, files #1 and #3 in Figure 3.3). If the candidate's period is detected in at least one adjacent beam, we then allow for up to 10 other candidates to have the same period. The choice of the number of similar periods allowed for a candidate was somewhat arbitrary, but it is based on the fact that a true pulsar should appear in 3 or 4 (assuming that the pulsar does not happen to fall exactly into one of the arbitrarily chosen drift time segments) beams: two adjacent beams, and one or two overlapping beams. The post-processing period filtering has eliminated many false candidates that appear in Chapter 3. Data Reduction: Searching 31 multiple pointings, and has also been successful in allowing the known (re-detected) pulsars to be easily detected. Another filter that allowed for a significant drop in the number of bad candidates was to place a lower limit on the period of the pulsar candidates. The number of sub-millisecond candidates that appeared in the candidate lists was astounding, and most of these candidates were not convincing and were often barely detections at all. We therefore decided to place a lower cutoff at one millisecond, so as to eliminate any sub-millisecond detections. This adds the danger of eliminating some interesting pulsars, but is much more efficient. A final technique that was used in the drift-scan survey was to eliminate unwanted signals through consideration of the DM smearing across a channel. This is done once the candidate lists have been produced. Pulsars emit radiation at a broad range of wavelengths. Dispersion (see §1.1.2) causes the pulses from higher frequency emission to reach us slightly earlier than those at lower frequencies. Dedispersion corrects for this effect; however, the pulse is still smeared over a certain time interval TDM given by equation 1.2: «i D M A TDM = 8.3 x 10**-3 Au sec vMHz Therefore, any signals with period less than TDM will have a significant decrease in sensitivity due to this smearing effect. Therefore, if the period of each candidate is compared with the value of TDM f° r that particular observation, any candidate having period less than TDM CAN be dismissed as an interesting candidate. Although this particular filter was performed for all candidates, the elimination of all sub-millisecond pulsar candidates (see above) generally made this filter redundant. 3.2.3 D M List Once the observation has been split into small segments as in §3.2.1, the next step is to decide, for each segment, the DM's at which it will be dedispersed and searched. The DM list is defined by two values: the maximum DM and the DM step . The maximum DM is the highest dispersion measure at which we would expect to be able to detect a pulsar. Since the dispersion measure depends on the amount of gas that lies between us and the pulsar, a model of the galaxy must be used for a reasonable estimation of the maximum DM along a given direction. The Galactic coordinates are calculated for each pointing, and Chapter 3. Data Reduction: Searching 32 the maximum DM is determined using the NE2001 Galactic Free Electron Density Model (Cordes and Lazio 2002) given those coordinates. The DM step is the space between two consecutive DMs at which-the timeseries will be dedispersed and searched. It is important to choose a DM step appropriate for the observation. If too large a step is used, then a pulsar with a DM somewhere between two trial values in the list may have its pulse broadened too much to be detected. If too small a step is used there is a danger of performing many redundant calculations, as the dedispersed timeseries of two consecutive DMs may not be significantly different. A compromise can be made between the need for sufficient resolution in DM and the desire to limit computing requirements. This can be found by setting the time delay between the outermost frequency bands (as given by equation 1.2) equal to the sampling time of the survey. The resulting DM step is given by (eg, Lorimer and Kramer with sampling time tsamp, centre frequency v, and bandwidth Av in MHz. Although the values of the maximum DM and DM step described above are reasonable choices, it was desirable in the drift-scan survey to use a larger maximum DM, and a smaller DM step, thereby increasing the total number of trial DMs by a factor of 9. This increased both the range and resolution in DM. The increased range (typically searching DMs .in the range from 0 to 120 pc cm - 3 ) gave us confidence that we were not missing pulsars by not searching to high enough DMs. The higher resolution in DM allowed the SNR versus DM plot to be a better diagnostic tool. A plot of SNR versus DM can help to distinguish whether or not the candidate is a pulsar (see §3.2.5 and §3.2.6). The search algorithm (§3.2.4) included a script called hunt which searched for periodicities at each trial DM, and then recorded each such detection into a file. However, each periodicity found by hunt was not necessarily detected at all DMs: weaker periodicities were often detected only at a few dispersion measures. The SNR versus DM plot could consequently only deliver a limited amount of information. Increasing the total number of trial DMs provides opportunity for more detections and therefore better resolution in the plot. See §3.2.6 for a description of further attempts at improving the diagnostic plots. 2004): ADM = 1.205 x 10 - 7cm- 3pc t. samp (3.8) Chapter 3. Data Reduction: Searching 33 3.2.4 Searching In §3.2.1, it is explained how one observation (scan) is split up into small sections which are converted into filterbank format. Each of the small filterbank files must go through the searching stage independently, and a master candidate list is compiled from a combination of the results from the individual files. I will here consider just one of the small files, and explain how it is searched for periodicities. In order to search for periodicities in the data, several sigproc tasks are called into action. The process for a single drift-time length filterbank file is as follows: • Given a dispersion measure, a time series is created by introducing the appropriate delay to each frequency channel through dedisperse. See §1.1.2 for more details on dedispersion. • find zero-pads and fourier-transforms the time series created by dedisperse and searches for significant features in the frequency domain. The raw spectrum as well as the first four harmonic folds (see §3.1.2) are searched. For every harmonic fold, the period and SNR is stored for each detected signal. • In order to easily create a list of dispersion measures to search at (see §3.2.3), step is used. Given a DM range and a DM step, a DM list is created. • hunt automates the search process by iterating dedisperse and find for each trial dispersion measure in the DM list, find produces a list of detections for each trial DM, all of which are collected into one file. • A list of the best overall candidates is created using best. It looks at the output from find, and after calculating how the signals are harmonically related, decides which signals appeared the strongest in that particular observation. The candidates are listed in order of decreasing SNR. 3.2.5 Folding Once the time series has been searched for periodicities, each detection can be folded and inspected by eye. The folding algorithm used by fold in sigproc is a basic one. For a Chapter 3. Data Reduction: Searching 34 given time series and pulse period, the.pulse phase is calculated for each time sample, and, the sample is placed in the corresponding phase bin. The result is a time series that is folded modulo a given period. The script which automates the folding of the best signals resulting from a search is called foldsignals. The foldsignals script obtains a list of all candidates produced by best and creates a dedispersed time series at the detected DM. Since it is important to know the period extremely precisely, foldsignals proceeds to fold the time series at 64 periods in a narrow range about the period initially detected by find, and chooses the best period to be the one which gives the highest SNR upon folding. The time series is separated into eight sub-bands, and is also separated into eight sub-integrations. This information is all used by sigproc's plotting program quickplot which gives an output as in Figure 3.4. The quickplot output can then be inspected by eye for specific features that are char-acteristic of pulsars. Some such features are: , • The folded time series should have a clear peak • Ideally, the peak should appear at the same phase for all sub-bands and all sub-integrations. If it appears in only one sub-band, odds are that it is RFI. If the signal appears in only one sub-integration, then it was not on during the entire observation. It is possible that such a signal is from a nulling pulsar, or it could be a pulsar that is just not completely in this beam (considering the transiting nature of this survey). However, if this same signal does not appear in adjacent beams, this could mean that the signal is due to RFI. Signals that are found in overlapping (or adjacent) beams are given some preference over those that appear only once. • There should be a clear peak in the plot of SNR versus DM. This is because the pulse becomes significantly broadened when the DM is not close enough to the true DM of the pulsar, thus lowering the peak SNR of the signal. A signal that peaks at zero DM is likely RFI. The above are guidelines that were followed in inspecting the detections for pulsars. Any interesting signals that met some number of the above criteria were saved as candidates and the search was refined. Chapter 3. Data Reduction: Searching 35 3.2.6 Refining the Search Parameters Finally, once a good list of candidates has been created based on the folded plots, a more intense search can be performed to further refine the search parameters. A search is per-formed in a grid of DM and period, and a plot is created from the period and DM producing the highest SNR. The value of this search is two-fold. First, the search refines the DM and period, thereby allowing for a higher peak SNR in the folded time series, and hence, more confidence in the classification as either a true detection, false detection, or RFI. Second, the two-parameter grid is plotted in greyscale and used as a final diagnostic as to whether or not the signal is pulsar-like. A true pulsar will exhibit a peak in the two-dimensional space, probably close to the centre of the space (depending on how good the original detection was). A false detection will likely show no distinct pattern, and RFI will peak at zero DM. In order to refine the search, a modified quickplot program was written. This modi-fied program was called quickgrid, which has a similar output to quickplot, but with a greyscale DM-period grid replacing the non-folded time series. The gridding algorithm is similar to that used iii the ATNF program pdm, which also performed a two-dimensional search in period and DM. See Figure 3.4 for a sample output of quickgrid. The ranges of period and DM over which to search were chosen in the following way. The search in period follows that used in f oldsignals, using a range determined by the observation time. The range of periodicities searched is (— T ^ ) to (— \- j^-) \Pmid *-obs J \Pmid ^obs / where pmid is the period of the detected signal and t0bs is the length of the observation (drift time). The number of trial periods can be specified, but for most purposes of the drift-scan survey, was kept at 64 trial periods (which is the same number of trial periods used in quickplot. The range for the DM search is determined by the number of DM trials and the resolution in DM step. The DM step is calculated as in §3.2.3, but is made to be even smaller at this stage, in order to further increase the resolution in DM. For the purposes of this survey, the DM step used was equation 3.8 divided by 30 (a factor of 10 smaller than the original DM step used), and the number of trial DM's was kept at 32. The number of DM steps was chosen to be 32 after having experimented in an attempt to balance between good resolution and limiting the number of iterations required from the CPU-intensive program that produces the plots. When less than 32 DM steps were used, Chapter 3. Data Reduction: Searching 36 there was degradation in the resulting plots. However, there did not appear to be much benefit to increasing the number of DM steps above 32. Chapter 3. Data Reduction: Searching 37 Figure 3.4: Left: An example of the output from quickplot. The search information is listed at the top of the plot. The top left plot is the time series. The top right plot is signal-to-noise versus trial DM. The third from the bottom is a plot of signal-to-noise for each of 8 frequency bands versus pulse phase; the second from the bottom is a similar plot for each sub-integration; the bottom plot is the integrated profile. The bottom three plots are shown for the period and DM which maximize the SNR. Right: An example of a plot using quickgrid. This is a similar to the quickplot output in that the signal-to-noise is maximized in period and DM space, but we here include a grid in period and DM (the top left plot) which can help to diagnose whether a candidate is a pulsar. For a real pulsar, we expect to see the signal-to-noise peak up in both period and DM as is seen here. Both plots were created using data from J1518+4904, a pulsar that was re-detected in our survey. 3 8 Chapter 4 Results: Candidates A s o f t h e w r i t i n g o f t h i s t h e s i s , 2 4 o f t h e 3 1 s c a n s h a v e b e e n p r o c e s s e d . T a b l e 4 . 1 i s a l i s t o f t h e s c a n s b y o b s e r v a t i o n d a t e , a n d s h o w s e a c h s c a n ' s d e c l i n a t i o n a n d r i g h t a s c e n s i o n r a n g e . T h e f o c u s o f t h i s s e c t i o n i s t h e o v e r a l l l i s t o f c a n d i d a t e s t h a t w a s p r o d u c e d b y t h e s e a r c h c o d e . T h e c l a s s i f i c a t i o n s c h e m e u s e d i n r a n k i n g t h e c a n d i d a t e s w i l l b e d i s c u s s e d , a n d t a b l e s o f b o t h n e w c a n d i d a t e s a n d r e - d e t e c t e d p u l s a r s w i l l b e p r e s e n t e d . H o w e v e r , a r e c e n t l y -d i s c o v e r e d b u g i n t h e f o l d s i g n a l s c o d e t h a t p r o d u c e s t h e d i a g n o s t i c p l o t s h a s m a d e i t n e c e s s a r y f o r u s t o r e p r o d u c e t h e s e p l o t s u s i n g a c o r r e c t e d v e r s i o n o f f o l d s i g n a l s f o r a l l c a n d i d a t e s . T h i s w i l l s i g n i f i c a n t l y c h a n g e t h e c l a s s i f i c a t i o n o f m a n y c a n d i d a t e s , r e q u i r i n g t h a t w e r e - e v a l u a t e o u r c l a s s i f i c a t i o n s y s t e m . W e h a v e r e p r o d u c e d t h e p l o t s f o r a s u b s e t o f t h e d a t a t h a t w a s i n i t i a l l y p r o c e s s e d ( t h e n o n - b o l d s c a n s i n F i g u r e 4 . 1 ) , a n d w i l l p r e s e n t s o m e C l a s s 1 c a n d i d a t e s t h a t r e s u l t e d . 4.1 Classification System I n o r d e r t o o r g a n i z e t h e t e n s o f t h o u s a n d s o f d e t e c t i o n s t h a t r e s u l t e d f r o m t h e d r i f t - s c a n s u r -v e y , a c l a s s i f i c a t i o n s y s t e m w a s r e q u i r e d . I t w a s n e c e s s a r y t o k e e p t r a c k o f w h i c h w e r e g o o d p u l s a r c a n d i d a t e s t h a t s h o u l d b e f o l l o w e d u p , a n d w h i c h a r e p r o b a b l y n o t w o r t h l o o k i n g a t a g a i n . W e h a v e s u c c e s s f u l l y p r o p o s e d t o r e - o b s e r v e g o o d c a n d i d a t e s f o r c o n f i r m a t i o n , w h i l e C l a s s 4 c a n d i d a t e s a r e c o n s i d e r e d f a l s e d e t e c t i o n s , a n d w i l l n o t b e s t u d i e d a n y f u r t h e r . T h e f o l l o w i n g s e c t i o n s d e s c r i b e t h e c l a s s i f i c a t i o n s y s t e m i n m o r e d e t a i l , a n d g i v e s o m e e x a m p l e s o f C l a s s 1 c a n d i d a t e s . 4.1.1 Class 1 Candidates C l a s s 1 o b j e c t s a r e t h o s e t h a t l o o k c o n v i n c i n g l y l i k e p u l s a r s . T h e s e o b j e c t s a p p e a r i n t w o t o f o u r a d j a c e n t d r i f t t i m e s , a n d d o n o t h a v e a m a x i m u m i n S N R a t a D M o f z e r o . T o Chapter 4. Results: Candidates 39 Date Frequency Declination RA start RA end Scan length Drift time mm-dd-yyyy MHz dd:mm:ss hh:mm:ss hh:mm:ss hr sec 05-14-2002 590.00 68:24:10 08:15:20 13:31:12 5.3 232 05-15-2002 590.00 49:04:34 14:16:19 22:02:36 7.8 130 06-04-2002 575.00 70:00:00 13:54:06 22:55:35 9.0 256 06-05-2002 575.00 69:40:00 13:58:02 00:30:24 9.5 252 06-08-2002 820.00 69:23:00 08:48:55 12:11:59 3.4 174 06-09-2002 820.00 69:23:00 19:02:36 06:34:29 11.5 174 06-13-2002 820.00 69:15:00 17:37:18 05:39:16 12.0 173 06-29-2002 570.00 69:20:00 03:05:45 08:21:37 5.3 250 07-04-2002 570.00 69:00:00 17:33:27 03:20:03 9.8 246 07-04-2002 590.00 69:00:00 05:35:25 11:59:48 6.4 238 07-07-2002 590.00 46:33:29 05:12:18 12:13:27 7.0 124 07-08-2002 590.00 68:40:00 10:34:42 12:35:01 2.0 234 07-09-2002 590.00 68:40:00 20:16:59 01:32:41 5.3 234 07-10-2002 590.00 70:00:00 10:52:22 13:52:52 3.0 249 07-18-2002 410.00 68:10:00 19:05:34 01:38:18 6.5 330 07-29-2002 590.00 67:40:00 21:38:15 02:39:04 5.0 224 08-06-2002 590.00 67:20:00 20:36:57 03:08:01 6.5 221 08-06-2002 590.00 67:40:00 12:55:06 15:10:28 2.3 224 . 08-07-2002 590.00 39:00:04 13:10:44 01:57:49 12.8 109 10-12-2002 820.00 54:59:10 14:27:27 18:28:06 4.0 107 10-13-2002 820.00 54:59:10 18:42:35 22:43:15 4.0 107 10-13-2002 820.00 54:59:10 00:45:25 03:15:15 4.0 107 10-15-2002 820.00 54:59:10 21:53:52 04:24:56 6.5 107 10-17-2002 570.00 54:58:55 19:49:56 05:28:21 9.6 154 10-17-2002 570.00 54:38:46 15:17:10 05:36:11 14.3 152 10-21-2002 570.00 54:19:23 20:59:56 05:53:53 8.9 151 10-22-2002 820.00 54:04:17 22:08:20 06:02:07 7.9 104 03-20-2003 800.00 52:12:00 07:24:49 15:26:08 8.0 102 03-21-2003 820.00 53:50:00 06:36:29 16:12:42 9.6 104 06-03-2003 800.00 36:45:00 16:25:00 22:40:00 ~6.3 76.7 06-06-2003 800.00 36:00:00 16:40:00 23:40:00 ~7.0 76.0 Table 4.1: A list of all observations in the drift-scan survey. The first column shows the date of the observation; the second column is the observing frequency; the third column is the declination of the strip; the fourth and fifth columns are the right ascensions at the beginning and end of the scan; next is the length of the scan, and the last column is the drift time for each scan's frequency and declination, as calculated by Equation 3.7. Those in bold font are scans that are yet to be processed. Chapter 4. Results: Candidates 40 Period DM Detected SNR Coordinates Drift Time Flux Density (ms) (cm -3pc) RA+Dec (sec) (mJy) 1083.12898 5.30 8.1 (13.2) 22:19:00+69:00:00 246 1 620.739771 22.40 9.2 (36.3) 07:40:09+69:00:00 246 2 Table 4.2: A list of two Class 1 candidates from the survey. The detected period, dispersion measure, and coordinates are shown. The flux densities are estimated from the radiometer equation, and are very rough estimates because we do not know if the pulsars were centered in the telescope beam. be in this category, the candidate should have enough signal for it to be clear that they are true signals and not random noise added in such a way as to produce a signal-to-noise ratio above the threshold. Figure 3.4 shows the foldsignals and quickgrid outputs for the test pulsar, PSR 1518+4904, and shows a clear, strong pulsar-like folded profile as well as a peak in both DM and period. The next step for a Class 1 candidate is a follow-up observation using the GBT. We have thus far detected two Class 1 candidates, which are listed in Table 4.2 and shown in 4.1. It will be necessary to rerun foldsignals on all Class 3 and Class 4 candidates. This will likely increase the number of Class 1 and Class 2 candidates. 4.1.2 Class 2 Candidates Class 2 candidates are those that appear like pulsars, but which are either too weak to be able to confidently classify as a Class 1, or for which there is some reason to cast doubt on its pulsar status. The next step for Class 2 candidates is to refine the search as in §3.2.6 using quickgrid. If the refined search increases confidence that it is a pulsar, the candidate will be promoted to Class 1. If it is still not clear that it is a pulsar but it still looks promising, the candidate will remain in Class 2. If the candidate looks worse upon refining the search, the candidate may move down to either Class 3 or Class 4. 4.1.3 Class 3 Candidates Class 3 candidates are those that are probably either false detections or RFI, but which may still be of interest. One or more of the criteria in §3.2.5 must have been met; usually, either the integrated profile or the DM plot has a peak, but some of the other criteria are Chapter 4. Results: Candidates 41 F i g u r e 4.1: quickplot / f o l d s i g n a l s o u t p u t f o r t w o C l a s s 1 c a n d i d a t e s . S e e T a b l e 4.2 f o r m o r e i n f o r m a t i o n a b o u t t h e c a n d i d a t e s . Chapter 4. Results: Candidates 42 Name Observing Known Known Known Coords Flux Frequency Period DM (J2000) Density (MHz) (ms) (cm~3pc) RA+Dec (mJy) B0105+68 590 1071.07252 61.092 01:08:30+69:05:53 < 2.1 B0329+54 570 714.48373 26.833 03:32:59+54:34:44 1337 B0329+54 820 714.48373 26.833 03:32:59+54:34:44 550 B0355+54 820 156.38242 57.153 03:58:54+54:13:14 20 J1518+490 590 40.94456 11.6109 15:18:17+49:04:34 8.7 B2241+69 575 1664.46403 40.74 22:42:56+69:50:52 1.6 Table 4.3: A list of the known pulsars that were in the region of sky processed to date in our survey. The topocentric period, DM, J2000 coordinates, flux density and predicted SNR of the pulsars are listed here. The observing frequency of the scan in which each pulsar would have been observable is listed as well. The flux densities listed correspond to these frequencies. The DMs and coordinates were obtained from the ATNF pulsar catalogue. The periods were estimated using the known ephemeris of each pulsar with TEMPO. Flux density measurements were obtained from Stairs et al. 1999 (J1518+4904), Lorimer et al. 1995 (B0105+68,B0329+54, B0355+54, B2241+69), interpolating between measured values when necessary. The predicted SNR was calculated using Equation 2.2 along with the fluxes listed in this table. Pulsars discovered by Dewey et al. 1985 (B0105+68, B2241+69), Cole and Pilkington 1968 (B0329+54), Manchester et al. 1972 (B0355+54) and Nice et al. 1996 (J1518+4904). B2241+69 was not detected as it lies at the edge of two of the observed declination ranges. Chapter 4. Results: Candidates 43 Name Observing Detected Detected Detected Coords SNR Flux Frequency Period DM (J2000) Density MHz (ms) (cm~3pc) RA+Dec (mJy) B0105+68 590 1071.004851 67.42 01:08:04+69:00:00 9.4 1.8 B0329+54 570 714.46733 40.63 03:33:01+54:19:23 247.2 7.9 820 714.53173 31.61 03:31:58+54:04:17 19.6 0.54 B0355+54 820 156.37017 57.44 003:58:14+54:04:17 84.46 11.6 J1518+4904 590 40.94431 11.7 15:17:42+49:04:34 121.47 12.3 Table 4.4: Previously known pulsars, re-detected in the drift-scan survey. Shown here are the period, DM, coordinates, SNR and predicted flux density of the pulsars as found by this survey. The SNR is the signal-to-noise ratio calculated from profiles dedispersed and folded at the known DM and topocentric period of the pulsar. Note that B0329+54 was detected in two separate scans of this survey, and so two sets of search parameters are given here. Both of these detections were made while B0329+54 was outside of the beam of the telescope, which explains the decreased flux as compared with that listed in Table 4.3. 1 B0105+68 was detected at its second harmonic, with period 535.50243 ms. 2The offset between the DM listed here and the known DM in Table 4.3 is due to the fact that the observing frequency in the header file being used was incorrect by 20 MHz. 3We had multiple detections of B0329+54 with a range of DMs. The plot in Figure 4.2 has a clear slope in sub-bands plot, indicating that the detected DM is incorrect. Chapter 4. Results: Candidates 44 Figure 4.2: Detection plots for the re-detected pulsars. Top left: B0105+68; Top right: B0329+43; Bottom left: B0355+54; Bottom right: J1518+4904. There was a second detec-tion of B0329+54 with a lower SNR at 820 MHz. Note that the SNR-DM plot is strange in the B0329+54 plot, since it has two peaks in DM. This probably occurred due to saturation by the strong signal, and also by the fact that there appears to be significantly more signal in some sub-bands than others. Chapter 4. Results: Candidates 45 not met. Often, interesting candidates that appear to be RFI are initially classified as Class 3, just in case they turn out to be pulsars. As with Class 2 objects, the search for a Class 3 candidates is refined in DM and period. Typical results from refining the search for a Class 3 candidate have either proven the candidate to be uninteresting, or to be a source of terrestrial RFI. 4.1.4 Class 4 Candidates Class 4 objects are unlikely to be pulsars. In the survey results, many of the 'candidates' found by sigproc have been either persistent RFI (often accompanied by many harmonics), or random noise that has been folded in such a way that it appears to have a periodicity. Once a signal is decided to belong to Class 4, it is no longer considered to be a pulsar candidate, and it is no longer studied. The search is not refined on these candidates. A list of Class 4 objects was not compiled, since there are too many such candidates and because they are not considered interesting. However, we will take a fresh look at these candidates with the revised version of foldsignals. 4.2 Redetections There are five previously-known pulsars in the sky area processed to date (listed in Table 4.3). A good check of whether our software and method are working properly is to determine whether or not these were redetected. Four of the previously-known pulsars (B0105+68, two separate detections of B0329+54, B0355+54, and J1518+4904) were re-detected in the survey. Table 4.4 provides the period, DM and SNR of the pulsars as found by the search software. Also included here are the SNR of the detected pulsars, folded at the known DM and topocentric period, and the flux density in mJy, derived from the integrated profiles and the radiometer equation. We performed a blind search with the drift-scan survey (i.e., the candidate viewing soft-ware did not identify previously-known pulsars), which resulted in three of the four detected pulsars having been classified as Class 1 candidates (shown in Figure 4.2). The initial classification of B0105+68 as a Class 3 candidate led to the identification of the error in foldsignals. We will reclassify all Class 3 and Class 4 candidates using foldsignals now that the software problem is solved. Chapter 4. Results: Candidates 46 The pulsar B2241+69 was not detected in the survey; however, this was probably because it was at a declination that lies at the edge of two scans. 47 Chapter 5 Discussion and Conclusions 5.1 Discussion of Results The processing of the drift-scan data until now has yielded several re-detections of known pulsars, and has also produced some Class 1 and Class 2 candidates. The small number of convincing new class 1 candidates can be attributed mainly to a recently-corrected software problem within sigproc and will increase as more sources are reclassified. The problem in the sigproc software had been distorting the integrated profiles of the candidates so that real pulsar signals did not appear as prominent as they should have. This effect has caused real pulsars to receive too low a classification, and has most likely caused real pulsars to go unnoticed. In addition, the RFI that corrupted the data was strongest at 600 MHz, the observing frequency of at least half of our data. Distinguishing real candidates from fake candidates became more difficult due to this prominent RFI. 5.2 Future Work and Conclusions There are still several prospects for the data. Now that we have identified and fixed the software problem within sigproc, a reanalysis of all processed (and unprocessed) scans using an upgraded version of sigproc will certainly yield a higher detection rate. The next step is to re-observe the most convincing Class 1 and Class 2 candidates to confirm them as pulsars. We will re-observe some of these candidates with the GBT in December 2005. If a candidate is re-detected at the same location on the sky with a similar period and DM, this is confirmation that the pulsar really exists and was neither a random signal nor some source of RFI. In the case that a candidate is confirmed during the December observations, we will further propose timing observations for the pulsar. A third consideration is to re-examine the data using more sophisticated techniques for RFI excision. For example, if bad channels could be removed systematically, this would improve our chances of detecting Chapter 5. Discussion and Conclusions 48 any unknown pulsars. W i t h these future directions, we are optimistic that this drift-scan survey is due to discover some exciting new pulsars. 49 Bibl iography Alpar, M. A., Cheng, A. F., Ruderman, M. A., and Shaham, J.: 1982, Nature 300, 728 Baade, W. and Zwicky, F.: 1934, Proc. Nat. Acad. Sci. 20, 254 Backer, D. C : 1970, Nature 228, 42 Backer, D. C : 1976, ApJ 209, 895 Backer, D. C , Dexter, M. R., Zepka, A., D., N., Wertheimer, D. J., Ray, P. S., and Foster, R. S.: 1997, PASP 109, 61 Backer, D. C , Kulkarni, S. R., Heiles, C , Davis, M. M., and Goss, W. M.: 1982, Nature 300, 615 Callanan, P. J., Garnavich, P. M., and Koester, D.: 1998, MNRAS 298, 207 Camilo, F., Nice, D. J., Shrauner, J. A., and Taylor, J. H.: 1996a, ApJ 469, 819 Camilo, F., Nice, D. J., and Taylor, J. H.: 1996b, ApJ 461 , 812 Camilo, F., Stairs, I. H., Lorimer, D. R., Backer, D. C , Ransom, S. M. , Klein, B., Wielebin-ski, R., Kramer, M. , McLaughlin, M. A., Arzoumanian, Z., and Miiller, P.: 2002, ApJ 571, L41 Camilo, F., Thorsett, S. E., and Kulkarni, S. R.: 1994, ApJ 421, L15 Chakrabarty, D., Morgan, E. H., Muno, M. P., Galloway, D. K., Wijnands, R., van der Klis, M. , and Markwardt, C. B.: 2003, Nature 424, 42 Chen, K. and Ruderman, M.: 1993, ApJ 402, 264 Cole, T. W. and Pilkington, J. D. H.: 1968, Nature 219, 574 Bibliography 50 Cordes, J. M. and Lazio, T. J. W.: 2002, preprint, astro-ph/0207156 Dewey, R. J., Taylor, J. FL, Weisberg, J. M., and Stokes, G. FL: 1985, ApJ 294, L25 Dicke, R. FL: 1946, Rev. Sci. lustrum. 17, 268 Drake, F. D. and Craft, H. D.: 1968, Nature 220, 231 Gold, T.: 1968, Nature 218, 731 Gould, D. M. and Lyne, A. G.: 1998, MNRAS 301, 235 Hankins, T. H. and Rickett, B. J.: 1975, in Methods in Computational Physics Volume 14 — Radio Astronomy, pp 55-129, Academic Press, New York Haslam, C. G. T., Stoffel, H., Salter, C. J., and Wilson, W. E.: 1982, A&AS 47, 1 Helfand, D. J., Manchester, R. N., and Taylor, J. H.: 1975, ApJ 198, 661 Hewish, A., Bell, S. J., Pilkington, J. D. H., Scott, P. F., and Collins, R. A.: 1968, Nature 217, 709 Hobbs, G., Manchester, R., Teoh, A., and Hobbs, M.: 2004, in IAUSymposium, pp 139-+ Johnston, S. and Bailes, M.: 1991, MNRAS 252, 277 Komesaroff, M. M.: 1970, Nature 225, 612 Kramer, M. , Xilouris, K. M., Lorimer, D. R., Doroshenko, O., Jessner, A., Wielebinski, R., Wolszczan, A., and Camilo, F.: 1998, ApJ 501, 270 Lamb, F. K. and Yu, W., in F. A. Rasio and I. H. Stairs (eds.), Binary Radio Pulsars Linares, M. , van der Klis, M. , Altamirano, D., and Markwardt, C. B.: 2005, Discovery of kHz QPOs and shifted frequency correlations in the accreting millisecond pulsar XTE J1807-294, astro-ph/0509011 Lorimer, D. R.: 2001, Living Reviews in Relativity, Bibliography 51 Lorimer, D. R. and Kramer, M.: 2004, Handbook of pulsar astronomy, Handbook of pulsar astronomy, by D.R. Lorimer and M. Kramer. Cambridge observing handbooks for research astronomers, Vol. 4. Cambridge, UK: Cambridge University Press, 2004 Lorimer, D. R., Xilouris, K. M. , Fruchter, A. S., Stairs, I. H., Camilo; F., Vazquez, A. M. , Eder, J. A., McLaughlin, M. A., Roberts, M. S. E., Hessels, J. W. T., and Ransom, S. M.: 2005, MNRAS 359, 1524 Lorimer, D. R., Yates, J. A., Lyne, A. G., and Gould, D. M.: 1995, MNRAS 273, 411 Lyne, A. G. and Graham-Smith, F.: 1998, Pulsar astronomy, Pulsar astronomy / Andrew G. Lyne and Francis Graham-Smith. Cambridge, U.K.; New York : Cambridge University Press, 1998. (Cambridge astrophysics series ; 31) ISBN 0521594138 Lyne, A. G. and Manchester, R. N.: 1988, MNRAS 234, 477 Lyne, A. G., Manchester, R. N., Lorimer, D. R., Bailes, M., D'Amico, N., Tauris, T. M., Johnston, S., Bell, J. F., and Nicastro, L.: 1998, MNRAS 295, 743 Manchester, R. N., Taylor, J. H., and Huguenin, G. R.: 1972, Nature Phys. Sci. 240, 74 Nicastro, L., Lyne, A. G., Lorimer, D. R., Harrison, P. A., Bailes, M., and Skidmore, B. D.: 1995, MNRAS 273, L68+ Nice, D. J., Sayer, R. W., and Taylor, J. H.: 1996, ,4^7 466, L87 Nice, D. J., Splaver, E. M., Stairs, I. H., Loehmer, O., Jessner, A., Kramer, M. , and Cordes, J. M.: 2005, A 2.1 Solar Mass Pulsar Measured by Relativistic Orbital Decay, astro-ph/0508050 Pacini, F.: 1967, Nature 216, 567 Pacini, F.: 1968, Nature 219, 145 Phinney, E. S.: 1992, Phil. Trans. Roy. Soc. A 341, 39 Radhakrishnan, V. and Cooke, D. J.: 1969, Astrophys. Lett. 3, 225 Radhakrishnan, V. and Srinivasan, G.: 1982, Curr. Sci. 51, 1096 Bibliography 52 Rankin, J. M.: 1993, ApJ 405, 285 Ransom, S. M.: 2001, Ph.D. thesis, Harvard University Ransom, S. M., Eikenberry, S. S., and Middleditch, J.: 2002, Astronomical Journal 124, 1788 Richards, D. W. and Cornelia, J. M.: 1969, Nature 222, 551 Sayer, R. W., Nice, D. J., and Taylor, J. H.: 1997, ,4^7 474, 426 Shrauner, J. A., Taylor, J. H., and Woan, G.: 1998, ApJ 509, 785 Slee, O. B., Dulk, G. A., and Otrupek, R. E.: 1980, Proc. Astr. Soc. Aust. 4, 100 Staelin, D. H. and Reifenstein, III, E. C : 1968, Science 162, 1481 Stairs, I. H.: 2003, Living Reviews in Relativity 6, 5 Stairs, I. H., Faulkner, A. J. , Lyne, A. G., Kramer, M. , Lorimer, D. R., McLaughlin, M. A., Manchester, R. N., Hobbs, G. B., Camilo, F., Possenti, A., Burgay, M., D'Amico, N., Freire, P. C , and Gregory, P. C : 2005, Discovery of Three Wide-orbit Binary Pulsars: Implications for Binary Evolution and Equivalence Principles, astro-ph/0506188 Stairs, I. H., Lyne, A. G., and Shemar, S.: 2000, Nature 406, 484 Stairs, I. H., Thorsett, S. E., and Camilo, F.: 1999, ApJS 123, 627 Strohmayer, T. E., Zhang, W., Swank, J. H., Smale, A., Titarchuk, L., Day, C , and Lee, U.: 1996, ,4pJ469, L9 Taylor, J. H.: 1974, A&AS 15, 367 Taylor, J. H. and Huguenin, G. R.: 1969, Nature 221, 816 Van Der Klis, M. , Swank, J. H., Zhang, W., Jahoda, K., Morgan, E. H., Lewin, W. H. G., Vaughan, B., and Van Paradijs, J.: 1996, ApJ 469, L l van Kerkwijk, M. H., Bergeron, P., and Kulkarni, S. R.: 1996, ApJ 467, L89 Weisberg, J. M., Romani, R. W., and Taylor, J. H.: 1989, ApJ 347, 1030 Bibliography 53 Wijnands, R., van der Klis, M. , Homan, J., Chakrabarty, D., Markwardt, C. B., and Morgan, E. H.: 2003, Nature 424, 44 Wolszczan, A.: 1990, IAU Circ. 5073, 1 Young, M. D., Manchester, R. N., and Johnston, S.: 1999, Nature 400, 848 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items