Pulsar ALFA Galactic Plane SurveyArecibo Project P2030

Progress Report24 September 2009

Jim Cordes1

For the PALFA Consortium2

SUMMARYThis report summarizes the PALFA Project, namely activity for projects P2030 (survey), P2177,P2283 (general timing) and P2391 (timing on the MSP J1903+0327). from mid-2008 to mid-2009.We discuss progress made toward the goals of our original survey proposal. We report on processingbeing done with pipelines that analyze the data using different algorithms, including the newestpipeline that involves the Einstein@Home community. Starting in Summer 2009, we are acquiringdata using the new spectrometers built by Jeff Mock, which provide up to three times more band-width than the WAPPs and greater immunity to radio-frequency interference. When observationsrecommence in 2009 November when ALFA is returned to service, we will use longer observations (9min instead of 4.5 min) for Galactic latitudes less than about 45 deg in order to reach more volumethat contains pulsars. We will also begin a commensal pulsar survey that acquires data during theALFALFA extragalactic HI survey. This “PALFALFA” survey will be sensitive to millisecond, bi-nary, and high-velocity pulsars that reside at higher Galactic latitudes than covered by our primaryin-plane survey.

Highlights: We have now discovered four millisecond pulsars, all of which are objects ofpotential utility in the pulsar timing array for gravitational wave detection. The first, J1903+0327,has now been timed for two years, demonstrating consistency with GR and providing a determi-nation of the pulsar’s mass as 1.68M to better than 1%. This object is in a highly-eccentricorbit that is itself a mystery, as discussed in the Science paper published last year (Champion etal. 2008). Two other MSP discoveries are also in binaries while the third MSP is isolated. Ayoung, high-E pulsar, J1856+0245, is a likely counterpart to a TeV γ-ray source detected by HESSthat most likely is a pulsar wind nebula. Timing of the young relativistic binary J1906+0746 willcontinue because the secular changes in pulse shape from geodetic precession allow us to samplethe pulsar’s beam shape. Timing will also allow us to determine the masses and test the orbitalperiod decay against GR to reasonably high precision. Intermittent objects have been found andwe have analyzed survey results to put limits on the rate of bursts from extragalactic sources. Onthe outreach side, ARCC students from UT Brownsville visited Arecibo and made presentations atthe 2008 Jan AAS Meeeting; an ARCC Scholars Program has started.

Pulsar yield: We have analyzed the pulsar yield so far taking into account the most recent analysisof the Galactic distribution of pulsars using results of the Parkes multibeam survey, which suggestthat the pulsar luminosity function is slightly shallower than thought before. With this analysis, weconclude that the pulsar yield will increase if we double our integration time from 4.5 to 9 min aswe make the transition to the new Mock spectrometers. This increase will be done only for Galacticlongitudes <∼ 45 deg for which the long line of sight reaches regions relatively dense in pulsars.

Einstein@Home Processing: In 2009, we began processing of PALFA data via the Ein-stein@Home community, which was originally developed to process LIGO data for coherent gravita-tional wave signals. E@H now devotes 1/3 of its processing resources to PALFA data to search forvery compact binary pulsars in circular orbits. A total ∼ 200k clients around the world have signedon to E@H, of which some fraction is active at any epoch. It is 1/3 of the active clients that receivePALFA data. The URL for E@H is given at the end of the main part of this report.

Commensal Pulsar Survey with ALFALFA: Later this year (2009) we will initiate data acqui-sition commensally with the extragalactic HI ALFALFA survey. Drift scans will provide 14 s persky position in directions well outside the Galactic plane. The survey sensitivity is sufficient to findmillisecond pulsars and other objects at high latitudes. The data will also be analyzed to searchfor transient events, including those of cosmological origin that may appear at very high dispersionmeasures.

Commensal Zone of Avoidance, Recombination Line and SETI Surveys: We have coordi-nated the PALFA pointing plan with groups interested in obtaining commensal data to detect ZoA

2

galaxies, Galactic plane recombination lines, and ETI sources The multiplicity of Mock spectrome-ters along with the WAPP spectrometers and a dedicated SETI spectrometer enables this fourfolduse of the PALFA pointings.

Data Storage: Long-term storage is at the Cornell Center for Advanced Computing (CCAC) andup to now has been on a robotic tape archive. We are in the process of migrating the data from thetape archive to a large-volume, RAID6 disk system. While doing the migration, we will reformatthe data from the proprietary WAPP format to PSRFITS format. We will also preprocess the datato remove spectral bandpasses and pack the data as 4-bit samples, a four-fold reduction from thecurrent 16-bit representation. New data from the Mock spectrometers most likely will also be storedlong term as 4-bit samples.

Processing: Data taken with the WAPPs from 2005-2009 have been 90% processed with the Cornellcode (non-acceleration search), 50% processed with the PRESTO-based code (acceleration search),and a few % processed with Einstein@Home (full circular-orbit search).

Survey Completion: By far most PALFA data have been taken with the WAPP spectrometers,which process only 100 MHz out of the 300 MHz total that is available from ALFA. Initial ex-pectations were that the new Mock spectrometers would be available in 2005 but we only beganusing them in 2009 summer. With the WAPPs we have covered about 60% of the total numberof beams needed to cover the Galactic plane in the inner Galaxy. Our plan all along was to re-dothe sky coverage using the full sensitivity provided by the Mock spectrometers. Also, since we willuse longer integration times for Galactic longitudes below 45, the total telescope time needed is2300 hr at 100% observing efficiency. At a nominal yearly rate of 500 hr and 80% efficiency (dueto slewing and spectrometer overheads), this will require another 5.8 yr to complete. Dependingon other programmatic constraints for usage of the telescope, an increase in yearly observing timewould cut this down to below 4 yr. Available options for reducing the total time include shrinkingthe Galactic latitude range from ±5 deg by a factor of two. Recent analyses of the Parkes multibeampulsar survey suggests that the young pulsar population is more closely concentrated toward theGalactic plane than ±5. However, owing to the growing interest in transient pulsars (and othersource classes) and to mitigate against survey incompleteness from astrophysical effects (eclipses andother intermittency) there is impetus to make two passes on the Galactic plane, which pushes thetotal time needed to complete the survey back to ∼ 4 yr if the latitude coverage is reduced from ourcurrent coverage.

Survey Issues: While most aspects of the survey are either in good shape or require attentionfrom the PALFA Consortium, some do require direct support from NAIC. The primary such issue islong-term storage and curation of the raw data and data products. It has been proven from previouspulsar surveys that raw data should remain accessible so that improved and new algorithms can beapplied in the search analysis. RFI excision and new detection algorithms applied to the Parkesmultibeam survey, for example, have increased the number of new discoveries by about 20%. Dataproducts from survey analyses are also of interest, including intermediate products (pulse profiles,dynamic spectra, etc.) and pulsar catalogs. The eventual data volume will be of order 1 Petabyte.Cornell and NAIC have provided the storage so far, but looking to the future, we need supportin developing a plan for migration to new storage media as they become available. We also needto evolve the user interfaces and data protocols as the community moves toward greater usage ofvirtual observatory methods.

Document: The main part of this document focuses primarily on events of the last year havingto do with Consortium structure, notable results, and new developments related to the data ac-quisition, processing and data management. Publications and websites are listed at the end of themain document and before appendices. Details of the survey, including the pulsar yield, processingpipelines, and long-term data management are placed in appendices so that the document is selfcontained, overall. Much of the material in the appendices was originally in the main part of ourprevious reports to NAIC. PALFA participants are listed after the appendices.

Introduction: The Pulsar-ALFA (PALFA) Consortium is conducting a large-scale Galactic plane survey for pulsarsand transients. Precursor work with ALFA began in August 2004 under project P1944, leading to our 2004 Octoberproposal to NAIC for a multi-year survey, which became the P2030 project. An affiliated follow-up timing projectusing Arecibo (P2177, P2283) commenced in 2007 March and routinely acquires full-Stokes arrival-times on some

1 Astronomy Department, Cornell University, Ithaca, NY 14853, cordes@astro.cornell.edu2 See end pages for list

3

PALFA discoveries. A few specific PALFA pulsars have been timed elsewhere, including the GBT, Jodrell, Nancay,Parkes and Westerbork.Our goal is to complete a deep search of the Galactic plane, defined as latitudes within 5 degrees of the midplane forthe two longitude sectors available to Arecibo. So far we have used the WAPP correlation spectrometers. In 2009we began using the new PALFA spectrometers. We intend to extend the survey further out of the Galactic plane asa means for optimizing detection of millisecond pulsars (MSPs) and relativistic binary pulsars.The PALFA Consortium and the PALFA Executive Committee (PEC): The PALFA Consortium wasformed in late 2002 in advance of the commissioning of ALFA in 2004. Since then the Consortium has conductedall aspects of the project including the Arecibo survey observations, data processing and management, timing obser-vations at Arecibo and elsewhere, and multi-wavelength follow-up. Camilo and Cordes are the PIs on the Areciboobserving project (P2030), which has been active since early 2005. P2030 has received ∼ 500 hr per year that includesoverhead and confirming observations. NAIC’s Skeptical Review panel’s latest review (2007 Sept) of all large ALFAprojects recommended continuation of P2030. Other Consortium members have led companion projects (P2177,P2283) for long-term follow-up timing at Arecibo on PALFA discoveries. Others have led timing efforts on affiliatedtelescopes (the GBT, the Lovell telescope at Jodrell Bank, and the Nancay telescope in France). Data are centrallymanaged at Cornell, where they are archived, curated, partially processed and made available to the Consortium,discussed in the Appendix.As the Consortium has evolved, we have instituted the PEC in order to maintain “corporate” memory and toconcentrate decision making among some of the Consortium members who are most active in the project. Thecurrent PEC membership is F. Camilo (Columbia), J. Cordes (Cornell), J. Hessels (ASTRON), D. Lorimer (U WestVirginia), V. Kaspi (McGill; Chair), D. Nice (Bryn Mawr), S. Ransom (NRAO) and I. Stairs (UBC).The PEC has recently defined a two-tier membership scheme for the Consortium, Tier-1 comprising active memberswho are involved in a subset of the activities needed for the project or who play crucial roles in the project andare eligible for (rotating) membership in the PEC. Tier-2 includes less active members who play occasional, ad-hocroles in follow-up on particular objects, such as multiwavelength observations. Our intention is to be as inclusiveas possible while acknowledging the very significant, sustained efforts made by some Consortium members. Tier-2members can propose to become more active and thus join Tier-1. Students’ PhD projects are protected throughvetting of proposed activities through the PEC and the entire Consortium.Database Activities: Please see Appendix for a discussion of our database system. Raw data continue to betransferred to Cornell for archival using portable disk drives. Disk drives are then sent on to another processing siteand then returned to Arecibo. In addition, using our Tracking Database, accessible through a password protectedweb application, 10 TB of data has been selected, restored from tape and securely transferred to remote processingsites via ftp so far.Remote processing sites upload their data products, in return, to a Common Database hosted at the Cornell Centerfor Advanced Computing from where candidate information and plots can be accessed remotely via a viewing programproduced by Patrick Lazarus of McGill and rated according to likelihood of being a pulsar discovery. Informationabout the highest-rated candidates can then be retrieved and incorporated into a confirmation observing schedule.Development of PALFA Processing with Einstein@Home: Exploratory discussions have taken place forprocessing PALFA data on E@H clients in areas of parameter space that cannot be searched by Consortium com-putational facilities owing to throughput issues. In particular, the scheme now being implemented is to dedisperseraw data on a central server and send a single time series to an E@H client. There ∼ 200k E@H clients around theworld, many of which are active. The code running on each client machine searches for circular binary pulsars withorbital periods less than one hour, a task that typically takes about 12 hr. This can be compared with a processingtime ∼ 0.5 hr per time series in the acceleration search done in the PRESTO-based pipelines. Acceleration searchesbecome insensitive to massive binaries for Porb

<∼ 1 hr. Given the merger rate of double neutron star binaries and theorbital lifetime at a given orbital period, there should be one binary in the Galaxy with an orbital period between 5and 10 minutes. There would be larger numbers in going to longer periods. Raw data are served from the CornellCenter for Advanced Computing, sent to the Albert Einstein Insitute (Hannover, Germany), and then disseminatedand managed using the BOINC technology developed for SETI@Home and other grid computing projects. In ad-dition to being an exciting research activity, this joint PALFA/E@H project is an excellent opportunity for furtherpublic outreach.Notable results from the PALFA Survey: We have discovered pulsars in all the target classes that give thegreatest long-term payoffs: MSPs, a relativistic binary pulsar, young pulsars with likely high-energy counterparts,a high-magnetic field object, and intermittent objects/RRATs found with our single-pulse analysis. Our startingassumption is that intermittent objects will turn out to be highly-modulated pulsars, but that need not be the case.New astrophysical source classes are certainly a possibility and only larger source samples and follow-up observationswill tell.

Binary MSP J1903+0327: This extraordinary binary millisecond pulsar (P = 2.15 ms, Pb = 95d),the first MSP discovered in our survey, was found by the full-resolution pipeline running at McGill and an

4

s s

r r!

.!

.

Fig. 1.— Probability density functions (pdfs) for the orbital inclination, mass of the companion and mass of the pulsar. Left plot:contours for the 2-D pdf displayed in the parameter space used for the Shapiro delay (companion mass, given by ”r”and co-sine of theorbital inclination, given by ”s”). Center-Right: the 2-D pdf in the mass-mass diagram. Top left: 1-D pdf for the cosine of the orbitalinclination, Top-right: 1-D pdf for the pulsar mass, Right: 1-D pdf for the companion mass. The light contours (or 1-D pdfs) arederived assuming only the Shapiro delay (r,s). The heavy contours are derived assuming that the precession of periastron (omega-dot) isrelativistic. These agree very well with the pdfs derived from the Shapiro delay only, but they are much more precise, particularly for thecompanion and pulsar masses. The splitting of the latter pdf into two clumps is caused by the two possible signs of the contribution of theproper motion to the observed precession of periastron, this needs to be taken into account given the great precision of the measurementof the rate of advance of periastron. Countours include 99.72, 95.44 and 68.28probability.

independent pipeline at Cornell. It is the most important discovery of our survey to date for many reasons.First, its DM is the largest for any known MSP, which demonstrates that high frequency resolution ofthe ALFA pulsar survey make it sensitive to MSPs in an unprecedented volume of our Galaxy. This hasbeen confirmed by the two other MSP discoveries to date, which are also found at very high DMs.Second, this highly recycled pulsar is absolutely unusual in having a fairly massive companion (whichis either a massive white dwarf (WD) or possibly a main-sequence star, as suggested by the spatialcoincidence of a Sun-like star with the astrometric position of the pulsar) and an orbital eccentricityof 0.44. These properties are at odds with the very small eccentricities common to all known GalacticMSP-WD systems, and defy our present understanding of the stellar evolution in binary systems; itspossible origin is now under investigation (Champion et al. 2008).The large mass, the eccentric orbit, the fast spin period and the high timing precision that can be achievedat Green Bank and Arecibo allow the determination of three post-Keplerian parameters: the rate ofadvance of periastron (ω) and the “range” (r) and “shape” (s) of the Shapiro delay. The determinationof three such parameters over-constrains the system, i.e., it allows a test of general relativity (alternatively,we can think of this as a verification that ω is relativistic). GR indeed passes the test, all these parametersgive consistent estimates of the mass of the two components, as shown in Figure 1. Even more importantly,the mass of the MSP is 1.68 ± 0.01M. The precision of this measurement is unprecedented for anyMSP, and it is confirmed by the measurement of multiple relativistic effects; furthermore, the mass issignificantly higher than that of any neutron star with a precise mass measurement made until now.This shows beyond doubt that a) MSPs can accrete significant amounts of matter when they are beingrecycled, but more importantly b) has firm, long-term implications for the study of the equation of statefor cold ultra-dense matter: such a large neutron star mass rules out some EOS models.Binary MSP J1949+31: This MSP (P = 13.1 ms, Porb = 1.95d) is in a circular orbit with a likelymassive WD companion. It was discovered in the Cornell pipeline in 2007 Oct and immediately confirmedthrough observations with the GBT and Jodrell Bank, yielding a tentative timing solution. With DM =

5

Fig. 2.— Secular variations of pulse profiles for J1906+0746. The left-hand panel shows the main pulse vs. epoch while the right-handpanel shows the interpulse. The shapes and separation vary in a way that is consistent with general relativistic geodetic precession, whichcauses the orientation of the spin axis to vary and thus also change the way the pulsar beam(s) intersect our line of sight. Color codesfor different instruments: ASP (Arecibo): magenta; GASP (GBT): cyan; WAPPs 1,2 and 4 (Arecibo): red, blue, green.

164 pc cm−3, it is about 6.5 kpc distant using the NE2001 model. This object is also a good candidatefor measuring the Shapiro delay, which allows us to determine the inclination angle.Binary MSP J1900+03: Discovered earlier this year, this 4.9 ms pulsar is in a 12.5 day binary. It iscurrently being timed to determine its orbit and suitability for the timing-array.Relativistic NS-NS Binary J1906+0746: This 144-ms pulsar is in a 3.96-hr orbit (Lorimer et al.2006). After the double pulsar (J0737−3039), it is the second-most relativistic binary system known.The pulsar is the youngest of any in a NS-NS binary (112 kyr), implying a birth rate ∼ 60 Myr−1 and acorresponding inspiral rate for NS-NS binaries of interest for gravitational wave detection and short-periodGRBs. Follow-up at Arecibo, GBT, and Jodrell Bank show no pulsations from the companion (flux limit= 46 µJy), which imply that it is either unfavorably beamed, radio quiet, or not a NS. However, the massdetermination from the measured apsidal advance ω = 7.6 deg yr−1 and gravitational redshift parameterγ shows the companion mass to be consistent with a NS (1.37±0.02M) and larger than the pulsar mass,1.25 ± 0.02M, as expected if it has undergone accretion while the pulsar has not (Kasian et al. 2007).Timing residuals are partly due to secular changes in pulse profile caused by geodetic precession, whichwas suggested in the discovery paper by comparing PALFA profiles with archival data from the Parkesmultibeam survey in which the pulsar was present, but was classified as RFI. We expect to measure theorbital period derivative Pb, determined by gravitational radiation, the rate of geodetic precession, andmap the emission beam of the observed pulsar. The measurements of these effects are more difficultthan for other binary pulsars where the recycled pulsar is timed; the young pulsar in J1906+0746 showssignificant timing noise.Over the past two years data have been collected with Arecibo and the GBT. The profile evolution needsto be modeled in order to obtain finalized TOAs and timing solution. The profiles are modeled at eachepoch with a set of gaussians, one of which is used as a fiducial timing component. Figure 2 showscumulative profiles that are used as template profiles for a given day. The current fit yields stable massestimates for the pulsar and companion, respectively, as quoted above.

High-energy Targets: We have discovered a high E, 82 ms Vela-like pulsar (J1856+0245) that appears

6

Fig. 3.— RIGHT: Single pulse from PALFA object G50.64-00.97 showing the characteristic dispersion sweep ∆t ∝ ν−2DM where thedispersion measure DM is the column density of free electrons to the pulsar. The line in the top spectrum is 21 cm hydrogen. LEFT:Single pulses from PALFA object J1928+15. These consecutive, dedispersed pulses are the only three detected in a total of 1.2 hr. Itappears to be a rotating radio transient (RRAT) with a period of 0.40 sec.

associated with a TeV γ-ray source (Hessels et al. 2008). The high-energy emission is most likely due toa pulsar-wind nebula. A second pulsar, the isolated 68.7 ms J1928+1746, has large E = 1036.3 erg s−1

and short spindown age, 82 kyr (Cordes et al. 2006). It has an unusually flat spectrum from 0.4 to 9 GHz( |spectral index| <∼ 0.3), somewhat reminiscent of the recently discovered flat-spectrum radio emissionfrom two magnetars (Camilo et al. 2006), except that J1928+1746 has an ordinary 1012 G field. It residesin the error box of one unidentified EGRET source and is a top candidate for confirmation with GLAST.Pulsars that are both radio and gamma-emitting objects are important for elucidating the beaming ofthe two kinds of radiation and also for probing the interaction of high-E pulsars with their environment.Transient Events and Pulsar Intermittency: Pulsars have long been known to be intermittentthrough the appearance of bursts of pulses, pulse nulling and other modulations. Recent work hasonly underscored intermittency, including the discovery of “rotating radio transients” (RRATs) by are-analysis of the Parkes Multibeam pulsar survey data using the single-pulse detection modules of theCornell pipeline (McLaughlin et al. 2006, Nature, 439, 817). It is not yet clear to what extent RRATsrepresent a new physical class of radio pulsar as opposed to being merely an empirical extreme, wheresome objects are missed in standard periodicity searches but emit atypically strong pulses detectable insingle-pulse analyses. We remain agnostic on this point because it is clear that the PALFA survey willyield good statistics on this empirical class.Of the previously known pulsars found blindly in our periodicity search analysis, we detect 63% in oursingle-pulse (SP) analysis, somewhat surprising if one assumes pulse amplitudes are fairly steady inthe data streams that typically contain >∼ 102 pulse periods. Obviously, pulse amplitudes are highlymodulated, with RRATs being extreme cases. So far, we have found 8 objects through the Cornell single-pulse analysis that were missed in the periodicity analysis. One has been confirmed through reobservation(J0628+09), one other has been reobserved several times without a redetection (J1928+15), while in allcases, the characteristic differential arrival time ∆tDM ∝ ν−2DM is seen. An example is shown in Figure 3.While transients with an underlying period (like that in the figure) are probably pulsar like and, in theend, may be fairly ordinary pulsars in other respects, their extreme intermittency is a puzzle. Other SPdetections may, of course, be from entirely different source classes, such as the event reported by Lorimeret al. (2007) that may have originated from a cosmological source. PALFA observations are sensitiveto events like the Lorimer event; our recent paper (Deneva et al. 2009) places limits on the rates andamplitudes of such events using PALFA data.Data products from our pipeline are suitable for assessing some aspects of intermittency by defining anintermittency ratio (McLaughlin & Cordes 2003) that is the ratio of S/N from the single-pulse analysis toS/N from the periodicity analysis: r ≡ (S/N)SP/(S/N)FFT. If r > 1, single pulse analysis yields higherS/N in the detection scheme while r < 1 implies the periodicity analysis is better. Figure 4 shows rfor PALFA data, for Parkes Multibeam data, and for analytical results for theoretical pulse-amplitudedistributions. The results indicate that short period pulsars, which provide larger Np in observations ofa fixed duration (e.g. T = 268 s in most PALFA pointings), are likely to have r < 1 while long-periodobjects can have r 1. Truly intermittent objects with low pulse rates will invariably require thesingle-pulse detection analysis in order to be detected.

7

Fig. 4.— Intermittency ratio r plotted against Np, the number of pulses analyzed in a data set as defined in text. Ratios r > 1 signifythat the single-pulse analysis provides larger S/N than a periodicity analysis. Some objects have only a lower bound on r, as do theRRAT objects. Filled red points are RRATs found from the Parkes multibeam survey, open red points are standard Parkes pulsars, blackcircles with arrows are single-pulse detections in our analysis without a corresponding periodicity detection, while black points are caseswith both kinds of detection. Vertical bars in some cases indicate the range of r seen in multiple data sets. The dotted and dashed linesshow r for various models of the pulse amplitude distribution.

Ongoing and New ActivitiesCandidate Selection: We are still working on different schemes in the Consortium for winnowing candidates lists.We are planning a face-to-face meeting in the early Fall to discuss heuristics and algorithms for choosing candidatesthat we will use telescope time for confirmation. Confirmation time will come out of our P2030 survey time andwe are cognizent of the need to have a well chosen pulsar candidate list in order to optimize telescope time usage.However, it is also true that there is an experimental aspect to finding the balance between false positives and falsenegatives.As described in our analysis of the pulsar yield, filtering of the Cornell pipeline’s signal candidates using searchanalysis data products (not pulsar catalog information) yields a pulsar candidate list with a good fraction of pulsars(PALFA, DMB and aliased ATNF-catalog pulsars) and some very good candidate new pulsars. However, thisparticular scheme is too severe and we are still experimenting with other filtering heuristics and thresholds. The planfor the near future is to (a) compare pulsar candidate lists used by different winnowing methods to identify a shortlist of very good candidates that will be re-observed and (b) define a longer list of candidates of gray-area candidatesthat we will spot select for reobservation in order to empirically determine our success rate. Going forward, we willthen optimize the use of telescope time for confirmation and survey time with respect to the rate of new detections,false positives and false negatives.The Common Database of signal candidates from all the pipelines is the enabling aspect of candidate selection. Weare continuing to develop algorithms and scripts for using all the information in this database.New Spectrometer: New spectrometers were delivered in 2008 to the Arecibo Observatory that replace theWideband Arecibo Pulsar Processors (WAPPs). Produced by Jeff Mock, the new spectrometers are based on apolyphase filterbank architecture rather than the correlation-spectrometer basis of the WAPPs. The advantages ofthe “Mock” Spectrometers are twofold: they provide three times more bandwidth (300 MHz instead of 100 MHz) and

8

they are more fault tolerant of narrowband radio-frequency interference (RFI). J. Deneva (Cornell) spent 6 monthsat Arecibo during the rampup period for the new spectrometers in late 2008 - spring 2009 and made test observationsusing the new spectrometers. The test observations on known pulsars probed their detectability amid different levelsof RFI and helped identify the characteristics of the RFI in the wider band. Examples are shown in Figure 5.Data from the Mock spectrometers typically are much cleaner than WAPP data in the presence of RFI, as shownin the dynamic spectra on a known pulsar in the two 170-MHz wide bands processed by the spectrometers (low:1215-1385 MHz, high: 1365-1535 MHz).

Fig. 5.— Dynamic spectra for the two bands provided by the Mock spectrometers for a pointing toward a bright, known pulsar(B2016+28). The lower band (left) typically shows worse RFI than the high band. The gray scale shows the normalized spectrum vs.time; the top panel shows the average spectrum and the right-hand panel shows the time series of power integrated over frequency.Narrowband interference can be seen in the spectra while pulses from the pulsar can be seen in the time series. Nonetheless, the pulsaris easily visible. The Mock spectrometers are superior to correlators because for the latter, the strong narrowband lines distribute powerover a wide frequency range. The polyphase filters of the Mock spectrometers limit such spectral leakage.

Longer Scans in Selected Parts of the Galactic Plane: As argued earlier, it appears sensible if not necessaryto use longer integration times along with the wider bandwidth of the new PALFA spectrometers to reach moredeeply into the inner Galaxy for the lower longitudes (≤ 60. When ALFA becomes available in 2009 Nov, we willtake this approach.Commensal Drift Scan Observations with the ALFALFA Project: We successfully proposed a commensalpulsar and transients survey (“PALFALFA”) that records data during pointings of the ALFALFA extragalactic, drift-scan survey, which targets HI from galaxies. Using the new Mock spectrometers with all fourteen ALFA receivers,the 300-MHz bandwidth and 14 sec integrations provide a sensitivity to periodic pulsar signals more than a factor oftwo better than recent Parkes surveys, or an increase in survey volume per deg2 of about three. We expect to find 5to 10 millisecond pulsars in the 2100 deg2 remaining to be surveyed by ALFALFA. The survey also targets pulsarswith relativistic companions, high-velocity pulsars and fast transients of Galactic or extragalactic origin. Data willbe processed both as they are obtained using a multi-node cluster at the Observatory and after the fact using alarger grid of possible dispersion measure values. The commensal survey will yield 50 to 100 TB of data (dependingon many bits we store; 50 TB results with 4-bit packing, which we are leaning toward). Quick-look processing willoccur in real time and additional, full-resolution processing off line. Data will be archived at the Cornell Center forAdvanced Computing.Commensal Observing with ZoA and RRL Surveys: We have modified the way we schedule the pointing

9

positions in our grid so that our PALFA data can be further used by the Zone of Avoidance (ZoA) and RecombinationLine (RRL) groups. The primary requirement of these groups is that there is at least an on/off pair of pointingsto use for calibration. The on/off pointing pair must follow, as closely as possible and preferentially to within the∼ 3-arcminute Arecibo beam at 1.4 GHz, the same azimuth and zenith angle track across the Arecibo dish. Thisrequirement was not taken into account back when the PALFA pointing grid was calculated, but is still possibleto achieve with some clever scheduling. We have tested taking commensal data in the last few PALFA observingsessions and an independent analysis by Robert Minchin indicates that these data can be calibrated properly forZoA and RRL studies. Though further tests may still be needed, we will observe in this commensal mode from nowon.Einstein@Home Processing:In 2008 we began exploring a joint collaboration with the group at the Albert Einstein Institute (Hannover, Germany)and at the University of Wisconsin, Milwaukee that leads the Einstein@home project. E@H currently has 100k clientsaround the world that process LIGO data to search for coherent gravitational waves. Discussions between JMC andBruce Allen have led to a plan for processing PALFA data by E@H clients for an analysis that simply is not feasibleusing the small computer clusters available to us. The basic idea is to search for pulsars in circular binary orbits withvery short orbital periods (less than 1 hour), an analysis that complements the acceleration-search analysis in thePALFA PRESTO-based pipeline and the non-acceleration Cornell pipeline. From the merger rate of NS-NS binariesin the Milky Way and the lifetime of a binary at a given orbital period, we calculate that there is of order one binaryin the Galaxy with an orbital period in the range of 5 to 10 minutes. Accordingly there should be a larger numberof objects with longer periods. The analysis went “live” in 2009 March and a press release was issued (see below).The Einstein@home processing uses a template bank for the three orbital parameters searched (orbital period,amplitude and phase of the orbital sinusoid), defined to cover the parameter space efficiently. With reference toFigure C13, data flow involves the following sequence:

1. Acquire data at Arecibo.2. Ship data on portable disks to Cornell for archival at the CCAC.3. Process data through the Cornell pipeline, which searches in dispersion measure (DM), period and pulse duty

cycle.4. Send data via network to several institutions running the PRESTO pipeline, which does an acceleration search

in addition to the DM, period and duty cycle search (i.e. one extra parameter).5. Send data via network to the Albert Einstein Institute for Einstein@home processing. Data are dedispersed

in Hannover using 1300 trial values of DM. An individual time series for a given DM value is sent to an E@Hclient.

6. Each client processes the time series using the template bank. Each template is used to map the time seriesinto what it would be for an inertial frame; then it is Fourier analyzed to search for periodic signals.

7. Candidate signals are returned to Hannover and then Cornell for further processing to identify viable pulsarcandidates. This latter step is now being developed by constructing another database and developing tools toconduct a metaanalysis and provide visual aids.

8. Viable candidates will be reobserved at Arecibo.

Feasibility Study for Longer-dwell Observations Out of the Galactic Plane: In addition to the commensalsearch during ALFALFA time, we are doing simulations to determine optimal parameters for a deeper search atlatitudes out of the Galactic plane. The MSP population extends to higher latitudes, on average, than young,canonical pulsars and there is less longitude dependence because the population is old and has migrated far away fromtheir birth sites. Consequently, a sensible strategy here is to search in LST ranges that are relatively undersubscribed.Finding MSPs in these ranges will also allow follow-up timing to be conducted with Arecibo.Multiple Passes on Pointing Directions: Our original proposal to NAIC discussed the pros and cons of multiplepasses. We will begin making repeats of some sky positions because RFI was severe. Also, we need to assess howmuch better the new spectrometer will perform and whether it is optimal on that basis to reobserve all sky positions.More fundamentally, pulsar intermittency suggests that a multiple-pass strategy is needed to find rare objects thatare potentially of the greatest interest in the overall yield. We can assess all these issues through at least repeatedobservations on a test area of the sky using the new spectrometers.Processing at Cornell: The Cornell pipeline does a standard periodicity search and single-pulse search withoutdoing an acceleration search. As a consequence it runs faster than the other, PRESTO-based pipelines. It has beenrun on all data archived at the CAC and has provided 2.5× 106 “signal candidates.” During the last year we havediscovered two MSPs with this pipeline along with several longer-period objects. As with the PRESTO pipelines,we are applying heuristics to the large signal-candidate list to yield a pulsar candidate list (beyond the very obviouscandidates, like those mentioned above) worthy of confirmation at Arecibo. The code processes one beam on a singleprocessor in 2 hr, corresponding to about 2.4 TB of raw data per week. 9k pointings have been processed out ofthe 13k search pointings made and of the 11k pointings that have been shipped to the CAC. The gap between the

10

9k processed and 11k shipped is due to the fact that in the early days, data were shipped such that only partialpointings were received at the CAC and the Cornell pipeline requires data from all seven beams simultaneously.The remaining pointings are mostly from the P1944 project which used shorter dwell times than in our standardobservation. The backlog has been slow to clear because restoring data from tape back to disk at Arecibo has beenlimited by available personnel availability. Graduate student J. Deneva has spent time at Arecibo to rectify thisbacklog and is making good headway.Processing at Franklin & Marshall: Processing at Frankin and Marshall College: Processing at F&M usesthe PRESTO pipeline and the 8-level candidate classification scheme developed at McGill. Raw survey data aredownloaded from the Cornell data archive via the internet onto a 72-processor cluster and are processed on thiscluster. Resulting candidates are viewed and classified with the help of undergraduate students who have beentrained to do this. The resulting classifications (and the candidates themselves) are uploaded regularly to theCornell database. We also view the single pulse plots that are produced for each beam using a custom viewing script;promising single-pulse candidates are rated and posted on a web page for later confirmation. We have processedabout 5000 survey pointings, generating about 233000 candidates, of which about 60000 have been classified byhumans. We have also detected 23 previously known pulsars through the standard Fourier processing or via singlepulse detection, and we have several candidates ready for confirmation.Processing at McGill University: Approximately 0.3-0.5 TB of raw data are processed per week on the McGillUniversity Pulsar Group’s Beowulf cluster. Currently, all survey data analysed at McGill University are downloaded,over the internet, from the PALFA archive hosted at Cornell University.Scoring schemes for automatically identifying/filtering RFI, and pulsar-like signals, at the database level have beendeveloped by P. Lazarus, S. Bogdanov, and A. Archibald at McGill University, and have been applied to all ”signalcandiates” generated at McGill University. Tools are available to use these scoring schemes to select/rank candidatesin the database. This reduces the number of candidates to examined by eye, at least on a first pass through theresults.A total of 1681 pointings (excluding directed pointings at known pulsars) has been processed. Of the 68366 2 ”signalcandidates”, 206032 are human classified.In March, 2009, J1900+03, a 4.9-ms period pulsar in a circular 12.5 day binary system was discovered in PALFAresults at McGill. J1900+03’s dispersion measure to period ratio (DM/P) is second highest of all known, non-globular cluster pulsars, owing to the fact that the PALFA survey is one of the best surveys for discovering, highdispersion measure, distant millisecond pulsars. (The pulsar with the highest DM/P ratio is J1903+0327, anotherPALFA-discovered millisecond pulsar). Follow-up and timing observations of J1900+03 are on-going.Processing in the Netherlands: Processing at ASTRON (Netherlands Institute for Radio Astronomy) andUniversity of Amsterdam: The recently formed pulsar group at ASTRON/ University of Amsterdam is very interestedin becoming a significant PALFA processing site. To this end, the group has secured funding for dedicated light pathswhich can be used to, e.g., transfer PALFA data from the Cornell database to the Netherlands. For the processingitself, the ASTRON pulsar group has built a mid-sized computer cluster, which is already churning through pulsarsearch data from the GBT. To supplement the available processing power, the group is installing the PALFA searchpipeline on the Dutch national supercomputer Huygens, which will provide the equivalent of many years of computingtime.Processing at Swinburne University (Melbourne): PALFA processing at Swinburne uses the PRESTO-basedpipeline for FFT and acceleration searches and the Co rnell pipeline for single-pulse searching. The processing hasramped up over the past year, covering 12437 beams so far, resulting in 597,443 candidates produced, of which 24,883have been hu man classified. We plan to sustain the current rate of processing - approximately 1000 beams perevery m onth.The Swinburne supercomputer comprises 145 Dell power Edge nodes (dual quad-core Clovertown processors at 2.33GHz), each with 16 GB RAM and 2 x 500 GB drives. Typically 25(256 nodes) resources are available for the PALFAprocessing. Benchmarks on this system are very pleasing – approximately 20 hours to process a single beam atfull-resolution.With an overwhelmingly large number of candidates from our processing, it has become difficult to sift th rough theentire candidate database. We typically view and classify approximately 5% of the top-ranked c andidates and theinformation on any promising candidates is routinely made available for confirmation ob servations. Our processinghas found 26 pulsars so far, including some PALFA discoveries. RFI-generated c andidates typically amount to lessthan 5% of total - this analysis is mostly carried out using the cur rently available filters within MSP finder.Data transfer from CTC to Swinburne is achieved via multiple parallel FTP streams (as Teragrid/NLR are no t viableoptions for Australia) and by running this routinely we are able to transfer as much as 1 TB per week. This datatransfer is covered under University’s AARNET subscription. The PALFA activities at Swin burne are supportedby the grants from the University Researcher Development Scheme and Faculty Grant Sch eme.Processing at University of British Columbia: Raw PALFA data have been transferred to UBC by 10 Gbnetwork from Cornell, using GridFTP. Work on network transfers was initiated by J. van Leeuwen and aided the

11

system of network transfers for other insitutions in the PALFA Consortium. As with the other PRESTO pipelinesites, signal candidates are uploaded to the Common Database at Cornell for global assessment. UBC processing isgoing again after a hiatus due in large part to lack of personnel. The UBC processing cluster is being upgraded anda graduate student is being recruited to work on the project.Processing at University of Texas, Brownsville: The Presto pipeline has been updated to use MPI on theTACC machines. We are now able to process 64 beams per day (9.1 pointing per day.) on the Lonestar cluster. Withthe newly installed cluster, Ranger, we expect to get a factor of at least 2-3 times more per day. Ranger is the thirdlargest cluster in world. Locally at UTB, we are processing about 10 beams per day. To date we have processed atotal of about 286 pointings. We expect this number to grow considerably in the next year using the MPI version ofPRESTO and the new cluster.Outreach: Our primary outreach activity involves the Arecibo Remote Command Center (ARCC) at the Universityof Texas, Brownsville, which involves students from the high school, undergraduate, and graduate levels in actual re-search at Arecibo. Using Arecibo’s remote observing capabilities together with a dedicated center at UT Brownsville,student teams performed PALFA observations in early 2007 and will continue when observations resume. The ob-serving itself is conducted with groups of three students. One of the three students has observing experience ands/he mentors the other two students. As students progress in their confidence and abilities, they earn the right tolead an observing run on their own. ARCC students also look through PALFA candidate files produced at Texas. Aspart of a three week summer academy, students learn the basics of Astronomy and Astrophysics together with signalprocessing and data analysis. This prepares them to evaluate the results of the PRESTO and Sigproc pipelines. Ascandidate viewing system has been developed in order for the students to look through the data both in the ARCCroom and on their home computers (http:arcc.phys.utb.edu/viewer).Four of the ARCC students had the opportunity to visit the Arecibo observatory this year and perform followupobservations for the PALFA survey on site.This year, UTB launched the ARCC scholars program. Funded by an NSF PAARE grant, this project providesfull four year undergraduate scholarships for up to ten students per year. The students must work on researchprojects associated to the ARCC program for the first two years of their undergraduate career, while maintaininggood academic standing as a student in the physics department of UT Brownsville. We recruited five undergraduatestudents for fall 2008. They were actively involved in the PALFA survey by leading observations, searching throughcandidates, and training the next generate of high school ARCC students.Also, in 2009, we launched the Einstein@Home project for PALFA processing, involving thousands of worldwide userclients covering a wide range of backgrounds, from high school students to scientists working in many different fields.Tools for Candidate Selection: The Cornell pipeline produces a large volume of “signal candidates,” in both theperiodicity search path and the single-pulse path. Signal candidates are identified for values of a test statistic thatexceeds some noise-defined threshold. For the periodicity search we use an 8σ threshold on the harmonic sum andwe reject 60 Hz and its harmonics, along with a few other prevalent interfering signals. For the single-pulse searchwe use a 5σ threshold. Everything that passes these tests is termed a signal. Tools used to identify astrophysicalsignals include: (1) The pipeline produces graphical output that goes into the data products database and can beviewed through the web. We scan these visually to identify the most obvious periodic and single-pulse celestialsignals. An example of single-pulse output showing a pulsar amid RFI is given in Figure A9. (2) We have writtenscript driven C programs (winnow1, winnow2) that flag periodicity candidates by using sky position information,data quality, candidate parameters, identify signals that are almost certainly RFI, signals that are known pulsars(through comparison with the ATNF pulsar catalog), signals that are plausible new pulsars, and signals that are lessplausible but still viable celestial candidates. Figure A8 shows a histogram of periods of all signal candidates (minus60Hz and its harmonics), with periods of detected known pulsars indicated. As can be seen, there are large peaksin the period histogram corresponding to interfering signals. The histogram is used, in turn, to winnow many of thesignal candidates in the pipeline, so as to reduce pipeline output that goes into the data products database. Thepipeline does, however, output the full signal list prior to this winnowing so that we can keep track of interferingsignals, some of which are highly episodic.The PRESTO pipeline produces similar candidate types but in different formats. A viewer tool has been written atMcGill by graduate student P. Lazarus that allows rapid perusal of candidates and navigation to related diagnosticoutput.The data-products DB includes diagnostic plots for the single-pulse and periodicity analyses that canbe perused via SQL queries. Example plots are available publicly (for the purpose of this report) athttp://arecibo.tc.cornell.edu/PALFA/graphics.html.Tools for RFI Characterization and Mitigation: The Cornell pipeline routinely analyzes the dynamic spectrum(DS) of each beam for each pointing on data degraded in time resolution to 0.1 s (from 64 µs) while maintainingfull frequency resolution. In the dynspec code, we calculate the mean spectrum after filtering out large events thatdeviate from the mean spectrum (an iterative process). Output from the tool includes the mean spectrum, a scalefactor that measures changes in total power, and an events file that tabulates all frequency-time cells in the DS thatdepart by 3σ from the mean spectrum. From the events file, we calculate the occupancy fraction of non-noise signals

12

in the frequency-time plane, which can be used to characterize RFI vs. epoch and as a metric for indicating whichpointings may require re-observation.The worst interference is from the FAA airport radar, the Punta Salinas air defense radar, and the aerostat radar inLajas used for drug interdiction. These have 12 s periods and are modulated so as to produce spectral features up toand beyond 1 kHz, and thus appear in initial candidate lists even in dedispersed time series.3 Radar usually entersthe telescope optics through scattering off telescope substructures and thus interferes with itself across the ALFAaperture plane. Consequently, RFI is not uniform across the seven ALFA feeds, requiring careful consideration whenrejecting particular signals as RFI (c.f. Figure A9). Interfering signals are mitigated in several ways through thepipeline. Up to now, we have rejected only 60 Hz and its harmonics within the periodicity search part of the pipelineand we filter candidates in a post-analysis of the global output (as discussed above with the winnow1,2 codes).We have developed modules that can mask the frequency-time data prior to dedispersion in order to remove the mostoffending radar pulses. The mask is formed from the dynspec events files described above. Event lists are combinedcarefully so as to flag RFI that appears in more than three beams (though not necessarily in all seven) yet not rejectstrong pulsars that can appear in multiple beams. The mask is inputed to a new version of the Cornell dedispersionmodule. Tests on data sets with a variety of RFI levels indicates that output candidate lists are reduced significantlywhen the mask is applied. We have not yet invoked the mask code into the routine pipeline simply because RFIhas not been so bad as to hinder pulsar detections in the vast majority of our data and we want to be sure thatthe mask adapts properly to the episodic aspects of the RFI. Because we will be changing soon to the new PALFAspectrometer with much wider bandwidth with different RFI, we will make the decision to deploy the mask codeonce we have explored performance of our pipelines on the new data format. We expect the new spectrometer tobe more resistant to RFI than the WAPPs because it is a polyphase filter bank rather than correlator. However,the wider processed bandwidth with the new machines will contain more RFI than the current WAPP band that wehave processed.

3 Some of the radars use frequency hopping, so the dedispersion algorithm will sum non-simultaneous time-frequency patches that canproduce large S/N at non-zero values of DM, especially in the single-pulse search.

13

Publications: published, submitted and in preparation:

1. J. Deneva 2009, “Elusive Neutron Star Populations: Galactic Center and Intermittent Pulsars,” PhD Thesis,to be completed Fall 2009.

2. J. Deneva et al. 2008, “Arecibo Pulsar Survey Using ALFA. Probing Radio Pulsar Intermittency and Tran-sients,” ApJ, 703, 2259

3. J. Hessels et al. “PSR J1856+0245: Arecibo Discovery of a Young, Energetic Pulsar Coincident with the TeVγ-ray Source HESS J1857+026,” 2008, ApJ, 682, L41

4. L. Kasian et al. 2008, “Timing and Precession of the Relativistic Binary Pulsar PSR J1906+0746,” in prepa-ration

5. D. Nice et al. 2008, “Arecibo Pulsar Survey Using ALFA. IV. Timing of Twenty Five PALFA Pulsars,” inpreparation.

6. D. Champion et al., 2008, ‘An Eccentric Binary Millisecond Pulsar in the Galactic Plane,” Science, 320, 13097. D. Champion, 2008, “The Discovery of an Eccentric Millisecond Pulsar in the Galactic Plane,” in 40 Years of

Pulsars: Millisecond Pulsars, Magnetars and More, AIPC Proceedings, 983, 4488. J. Cordes, 2008, “Arecibo Pulsar and Transient Surveys Using ALFA,” in 40 Years of Pulsars: Millisecond

Pulsars, Magnetars and More, AIPC Proceedings, 983, 5679. L. Kasian, 2008, “Timing and Precession of the Young, Relativistic Binary Pulsar PSR J1906+0746,” in 40

Years of Pulsars: Millisecond Pulsars, Magnetars and More, AIPC Proceedings, 983, 48510. J. Cordes + 22 other authors “Arecibo Pulsar Survey Using ALFA. I. Survey Strategy and First Discoveries,”

The Astrophysical Journal, 637, 446, 2006 (Paper I)11. D. Lorimer + 35 other authors “Arecibo Pulsar Survey Using ALFA. II. The Young, Highly Relativistic Binary

Pulsar J1906+0746,” The Astrophysical Journal, 640, 428, 2006 (Paper II)12. M. Calimlim, J. Cordes, A. Demers, J. Deneva, J. Gehrke, D. Kifer, M. Riedewald, and J. Shanmugasun-

daram “A Vision for PetaByte Data Management and Analysis Services for the Arecibo Telescope” IEEE DataEngineering Bulletin, 27, 12

Talks and Posters:

1. 2009 June, Poster at the 8th Amaldi Conference on Gravitational Waves, Columbia University, NY (B. Knispel,B. Allen, O. Bock, J. M. Cordes, B. Machenschalk, C. Messenger, H. Pletsch & R. Prix), “The Search for TightBinary Radio Pulsars in Arecibo Radio Data with Einstein@Home”

2. 2009 June, Talk at Penn State Workshop, Probing Neutron Stars with Gravitational Waves (J. Cordes), “RadioPulsars and Gravity”

3. 2009 April, Colloquium at UC Berkeley (J. Cordes), “The Right Place and the Right Time: Pulsars, ALFAand Gravity”

4. A. Miller, F. Jenet, Zermeno, A., & Stovall, K., “The Arecibo Remote Command Center: Creating an InspiringEnvironment for Astrophysics” AAS, 213rd Meeting, BAAS, 41, 264

5. J. Cordes, “The Arecbo ALFA Survey for Pulsars and Transients,” at joint FAST/Cornell/NAIC meeting, 2008June, Beijing

6. K. Stovall et al., “Pulsar Search Results from the Arecibo Remote Command Center,” 2007, BAAS, 38, 9187. F. Jenet et al., ”The Arecibo Remote Command Center: Inspiring the Next Generation of Astrophysicists,”

BAAS, 38, 7358. J. Deneva, “Pulsar surveys present and future: The Arecibo pulsar-ALFA survey and projected SKA survey”,

ArXiv Astrophysics e-prints, arXiv:astro-ph/0701181, 20079. J. Cordes, “The Pulsar ALFA Survey,” presentation to the National Astronomy and Ionosphere Center Visiting

Committee, Arecibo, PR 19 Feb 2007 (via videocon)10. J. Cordes, “The Pulsar ALFA Survey,” presentation to the NSF’s Management Review Committee of the

National Astronomy and Ionosphere Center, Arecibo, PR 14 Mar 200711. A. Miller, A. Rodriguez-Zermeno, and F. Jenet, “The Arecibo Remote Command Center: Involving Students

in Major Astronomical Research,” , 2006 BAAS, 38, 99312. J. Hessels, “Imaging the Environment of the Newly Discovered Young Pulsar J1856+0245”, XMM-Newton

Proposal ID #05059201, 213, 200613. J. Deneva, “Assessment of Pulsar and Rotating Radio Transient Detection Rates for the Arecibo Pulsar-ALFA

Survey”, BAAS, 38, 82, 200614. J. van Leeuwen et al., “Arecibo and the ALFA Pulsar Survey”, Chinese Journal of Astronomy and Astrophysics

Supplement, 6, 311, 200615. J. van Leeuwen, “Big, Smart Dishes to Find Thousands of New Radio Pulsars”, Bulletin of the American

Astronomical Society, 38, 1065, 2006

14

16. L. Kasian and PALFA Consortium, “New Observations of the Young Relativistic Binary PSR J1906+0746”,BAAS, 38, 81, 2006

17. L. Kasian and PALFA collaboration, “Discoveries from the ALFA Pulsar Survey,” CASCA meeting, displaypaper (2006 June)

18. P. Freire, “Pulsar Surveys with ALFA”, 36th COSPAR Scientific Assembly, 36, 1431, 200619. W. Arms et al. “Three Case Studies of Large-Scale Data Flows,” ICDEW, p. 66, 22nd International Conference

on Data Engineering Workshops (ICDEW’06), 200620. J. Deneva, “Assessment of Pulsar and Rotating Radio Transient Detection Rates for the Arecibo Pulsar-ALFA

Survey” poster presented at AAS 208, Calgary, Canada, 2006 June21. J. Cordes “Pulsar Surveys,” talk presented at the NRAO Legacy Projects Workshop, 2006 May, Socorro, NM

(invited)22. J. Cordes “Massive Radio Astronomy Surveys at Arecibo Using ALFA: Data Mining and Management,” talk

presented at URSI/National Radio Science Meeting, Boulder, 2006 Jan23. F. Camilo, “The Pulsar ALFA Survey,” talk presented at special ALFA session at AAS meeting, Washington,

DC, 2006 Jan

Websites:

http://arecibo.tc.cornell.edu/PALFA/: the primary web site for the PALFA project, which has both publicly accessible and pass-word protected sections.

http://arecibo.tc.cornell.edu/PALFA/graphics.html: example graphical output for both the periodicity and single-pulse searches,including PALFA discoveries.

http://einstein.phys.uwm.edu/: Web site for Einstein@Home:

http://www.brynmawr.edu/physics/dnice/timing.pdf: web site to obtain timing results on PALFA pulsars.

http://arecibo.tc.cornell.edu/PALFAData/default.aspx: web page for accessing raw PALFA data (requires account)

http://arecibo.tc.cornell.edu/legacypulsardata/: the Legacy Pulsars Database that is publicly accessible.

http://mingus.astro.cornell.edu/∼deneva/palfa/tablesearch.php#: A PHP-based webtool written by Cornell graduate studentJulia Deneva.

http://astro.cornell.edu/∼cordes/PALFA: constains a collection of memos relevant to the PALFA survey, including notes on thepulsar yield, etc.

http://arcc.phys.utb.edu: main website for the ARCC project.

http://arccview.phys.utb.edu: Live webcam view of the ARCC room.

http://arcc.phys.utb.edu/viewer/index.php: Candidate viewing system used by the ARCC students to search for pulsars.

Press Release for Einstein@Home Processing of PALFA Data

NEW EINSTEIN@HOME EFFORT LAUNCHED: THOUSANDS OF HOME COMPUTERS TOSEARCH ARECIBO DATA FOR NEW RADIO PULSARShttp://einstein.phys.uwm.edu/Einstein@Home, based at the University of Wisconsin-Milwaukee (UWM) and the Albert Einstein In-stitute (AEI) in Germany, is one of the world’s largest public volunteer distributed computing projects.More than 200,000 people have signed up for the project and donated time on their computers to searchgravitational wave data for signals from unknown pulsars. Today, Prof. Bruce Allen, Director of theEinstein@Home project, and Prof. Jim Cordes, of Cornell University and Chair of the Arecibo PALFAConsortium, announced that the Einstein@Home project is beginning to analyze data taken by thePALFA Consortium at the Arecibo Observatory in Puerto Rico. The Arecibo Observatory is the largestsingle-aperture radio telescope on the planet and is used for studies of pulsars, galaxies, solar systemobjects, and the Earth’s atmosphere. Using new methods developed at the AEI, Einstein@Home willsearch Arecibo radio data to find binary systems consisting of the most extreme objects in the universe:a spinning neutron star orbiting another neutron star or a black hole. Current searches of radio data losesensitivity for orbital periods shorter than about 50 minutes. But the enormous computational capabil-ities of the Einstein@Home project (equivalent to tens of thousands of computers) make it possible todetect pulsars in binary systems with orbital periods as short as 11 minutes.“Discovery of a pulsar orbiting a neutron star or black hole, with a sub-hour orbital period, would providetremendous opportunities to test General Relativity and to estimate how often such binaries merge,” saidCordes. The mergers of such systems are among the rarest and most spectacular events in the universe.They emit bursts of gravitational waves that current detectors might be able to detect, and they are alsothought to emit bursts of gamma rays just before the merged stars collapse to form a black hole. Cordesadded: “The Einstein@Home computing resources are a perfect complement to the data managementsystems at the Cornell Center for Advanced Computing and the other PALFA institutions.”

15

“While our long-term goal is to detect gravitational waves, in the shorter-term we hope to discover atleast a few new radio pulsars per year, which should be a lot of fun for Einstein@Home participants andshould also be very interesting for astronomers. We expect that most of the project’s participants will beeager to do both types of searches,” said Allen. Einstein@Home participants will automatically receivework for both the radio and gravitational-wave searches.The large data sets from the Arecibo survey are archived and processed initially at Cornell and otherPALFA institutions. For the Einstein@Home project, data are sent to the Albert Einstein Institute inHannover via high-bandwidth internet links, pre-processed and then distributed to computers around theworld. The results are returned to AEI, Cornell, and UWM for further investigation.Additional Background Material:

Gravitational waves were first predicted by Einstein in 1916 as a consequence of his general theoryof relativity, but have not yet been directly detected. For the past four years, Einstein@Home has beensearching for gravitational waves in data from the US LIGO detectors.Radio pulsars, first discovered in the 1960s, are rapidly spinning neutron stars that emit a lighthouse-likebeam of radio waves that sweeps past the earth as frequently as 600 times per second. Radio pulsars inshort-period binary systems are especially interesting because the effects of general relativity can be verystrong. Systems that have already been discovered have been used to verify that Einstein’s predictionsabout gravitational wave emission are correct to better than 1%.The discovery of new pulsars in much shorter-period binaries would improve estimates of the rates atwhich binary star systems form and disappear in our Galaxy, and also provide new targets to search forwith gravitational wave detectors.The Arecibo Observatory is the largest single-aperture radio telescope on the planet and is used forstudies of pulsars, galaxies, solar system objects, and the Earth’s atmosphere. The first binary pulsarwas discovered at Arecibo in 1974 and led to Hulse and Taylor’s 1993 Nobel Prize in Physics, because ofits stringent test of general relativity. The new pulsar survey uses a specialized radio camera, the AreciboL-band Feed Array, and is conducted by the PALFA Consortium.The Max Planck Institute for Gravitational Physics (Albert Einstein Institute) is the largestresearch institute in the world devoted to the study of general relativity. Its two branches in Potsdam andHannover support research in astrophysics, theoretical physics, mathematics, and experimental physics. Itoperates the GEO600 gravitational wave detector near Hannover, Germany, is a partner in the AmericanLIGO project, and plays a major role in the analysis of the data from all existing gravitational wavedetectors, including the VIRGO detector in Italy. The software that will be used in the Einstein@Homeradio searches was developed by the AEI in Hannover.The University of Wisconsin Milwaukee hosts the Einstein@Home project and plays a major rolein the data analysis activities of the LIGO Scientific Collaboration. It also carries out Arecibo radioobservations as an Arecibo Remote Control Center (ARCC).Funding The U.S. National Science Foundation supports this work through grants to the Einstein@Homeproject, to the PALFA project, to the BOINC project at the University of California at Berkeley, andthrough a cooperative agreement with Cornell University to operate the Arecibo Observatory. The AlbertEinstein Institute for Gravitational Physics is supported by the Max Planck Society and the Universityof Hannover.The Einstein@Home project, launched in 2005, is an undertaking of the LIGO Scientific Collaboration,and was primarily developed by UWM and the AEI. Einstein@Home is built using the Berkeley OpenInfrastructure for Network Computing (BOINC) developed at the University of California at Berkeley’sSpace Sciences Laboratory.

16

APPENDIX

Pulsar Detection Rate: Issues and Assessment

Diagnosis of the pulsar yield: The PALFA survey has blindly detected 162 pulsars of which 46 are new (found inthe periodicity analysis, the single-pulse analysis or both) and 11 were also found in the unpublished and uncatalogedParkes Deep Multibeam Survey (DMB) that found 14 pulsars in a portion of Arecibo’s declination range. Forcomparison, there is a total of 201 pulsars in the ATNF catalog in our inner-Galaxy survey region and 11 pulsars inour outer-Galaxy survey region. Given that we have surveyed only about 1/3 of the directions in the inner-Galaxyregion, it is clear that we will rediscover all the known pulsars and we will find additional new pulsars. However,the yield-rate (e.g. per square degree) is not as high as we originally predicted. We have spent considerable timeanalyzing the situation and expect that the yield rate will go up considerably through a number of measures.We conclude that

1. A significant number of new pulsars is contained in candidate signals in our pipeline outputs thathave not yet been distinguished from the overall huge number of signal candidates that are mostlyfrom RFI; we are confident that we will be able to winnow the large number of signal candidatesinto a short list of pulsar candidates with a high probability of being confirmed.

2. The shallower luminosity function identified in a recent analysis of the Parkes Multibeam surveyanalysis implies that we need to search more deeply to detect a larger fraction of the pulsars in theinner Galaxy.

3. Though we expect larger numbers of new pulsars to emerge from the PALFA survey, we em-phasize that predicted numbers depend on the Galactic structure of the pulsar population (and ofthe electron density), which we will know well enough only after we have finished our search and itsanalysis.

Prescription for the PALFA Survey: We also conclude that

1. Survey observations with the new PALFA spectrometers should be made to increase the overallproduct of bandwidth and integration time (BT) by a factor ∼ 6, particularly for Galactic longitudes32 ≤ ` ≤ 60 where a significant amount of volume is filled with pulsars inside the solar circle.

2. The pipelines should search to larger dispersion measures than currently for these low longitudes.3. We intend to observe commensally with the extragalactic drift-scan survey, ALFALFA, in order

to find millisecond pulsars, relativistic binaries and other objects in directions out of the Galacticplane.

4. We are looking at the feasibility of an out-of-plane survey using deeper integrations than thoseprovided by the ALFALFA drift scans.

Payoffs: The return will obviously include a larger number of pulsars of all types, but especially those of greatestinterest (binaries, millisecond pulsars, GLAST targets, intermittent pulsars and high-velocity pulsars). We will alsoprobe the pulsar population to sufficient distances that we can understand its Galactic structure, particularly whetherthere is any clustering toward spiral arms, as we would expect for young objects.So far we have observed 15k distinct sky positions with 7 beams per pointing (excluding test observations on knownpulsars that we routinely make), about 1/3 of those needed to cover the inner and outer-Galaxy regions at lowlatitudes. Figure A6 shows PALFA detections vs. DM and Galactic longitude along with known pulsars from theATNF Pulsar Catalog (small dots). PALFA has detected ∼ 150 pulsars so far of which 46 are new objects. Forcomparison, there is a total of ∼ 200 pulsars in the ATNF pulsar catalog with periods greater than 10 ms. Thedistribution of PALFA detections implies that we reach the boundary of the free-electron disk in the 268-s integrationtime we use, at least for the most luminous pulsars. Other conclusions include: (1) For small longitudes, |`| ≤ 30,the density of detected pulsars falls off for DMs larger than that for the tangent point distance. For these longitudes,the actual pulsar density should remain high for these larger DMs, indicating that surveys of this part of the sky(primarily the Parkes multibeam survey) do not reach much past the tangent point. This makes sense given thatthe electron density model predicts very high DMs out to twice the tangent-point distance; (2) At longitudes of−30 to −90, where DM is not so high and selection effects are not as severe, there does seem to be a falloff inpulsar density beyond the solar circle, though the density is fairly high between the tangent-point line and the solar-circle line, particularly for longitudes −60 to −90; and (3) On the opposite side of the Galaxy (positive longitudes)where the PALFA survey samples the population, detections to date reach as far as the limiting DM and quite afew of the detections are near or beyond the solar circle. This suggests that the PALFA survey reaches as far as itshould for sufficiently luminous pulsars.The eventual pulsar yield will depend on quantities we do not yet know well. For disk populations, a scale heightdescription would generally be used. However, of order 50% of pulsars escape the Galaxy owing to their birthvelocity kicks (Lyne & Lorimer 1994; Cordes & Chernoff 1998; Faucher-Giguere & Kaspi 2006). At low latitudes,the pulsar yield depends on the velocity distribution of the bound population, also not well known and certainly not

17

Fig. A6.— DM vs Galactic longitude. Thetopmost curve is the NE2001 model integrated to“infinite” (50 kpc) distance from the Sun. Regionscorresponding to spiral-arm tangent points and otherfeatures are labeled as described in the NE2001model (Cordes & Lazio 2002). The middle line showsDM for lines of sight that reach the solar circleat galactocentric radius a = 8.5 kpc, a distancedss = 2a cos `. The lowest line is DM integratedto the tangent point, dt = a cos ` for the innerGalaxy. Both dt and dss vanish for |`| ≥ 90. Small(blue) points show pulsars from the ATNF pulsarcatalog (Manchester et al. 2005). PALFA periodicitydetections are shown as large (red) filled circles.RRAT detections are shown as open (green) squares.

well normalized with respect to the high-V objects. For canonical numbers (20% beaming fraction, 10 Myr radiolifetime, birth rate of 1.4 per 100 yr, galactocentric radial scale of 8 kpc, effective scale height of 0.5 kpc, and afiducial maximum detection distance Dmax = 5 kpc, on average), about one pulsar should be detected in 30 ALFApointings (i.e. 7×30 beams):

Npointing/psr ≈6πR2

gH

7fbNpsr,gD3maxΩb

≈ 30(

0.2fb

)(1.4× 105

Npsr,g

)(5 kpcDmax

)3(Rg

8 kpc

)2(H

0.5 kpc

). (A1)

This number corresponds to about 0.6 deg2 per pulsar for nominal values of other parameters. For the inner Galaxy,we will survey about 430 deg2, implying a total yield of 700 pulsars, subject to the caveat that not all directions inthe inner Galaxy will be equally rich in pulsars, owing to the galactocentric radial scale of the pulsar population. Inthe ATNF pulsar catalog there are 200 previously known pulsars in the same Galactic longitude and latitude region,indicating that ∼ 500 new pulsars can be expected. While the eventual number could be lower, it could also behigher if we extend our reach (Dmax) further than 5 kpc, as we advocate below.

Understanding the Pulsar Yield: In the following we analyze various observational and astrophysical factorsbehind the current low yield of new pulsars but re-emphasize that the total yield (new + old) so far is not unreasonable.We conclude that a significant number of new pulsars is already contained in our pipeline output but we have not yetsettled on a scheme for winnowing the list from the huge number of signal candidates that is efficient with respectto follow-up telescope confirmation. Several approaches are being explored by different Consortium groups that willbe compared with respect to outcome (commonality of candidate pulsars) and used to select the most promisingcandidates for re-observation. We are confident that we will be able to do so. We also conclude, on astrophysicalgrounds, that observations with the new PALFA spectrometers should use a longer integration time so that we reachto greater distances at low Galactic longitudes.Potential explanations for the low yield achieved to date include:

1. Receiver system and data acquisition problems,2. Analysis pipeline errors,3. Contamination by RFI and associated problems in identifying new pulsars,4. Astrophysical reasons, such as Galactic structure of the pulsar population and of the electron density.

18

Fig. A7.— LEFT: Plot of predicted vs. measured S/N from the periodicity search pipeline that yielded blind detections of known pulsarsthat had cataloged values of 1.4 GHz flux density. The largest S/N is for the millisecond pulsar B1937+21 and the lowest predicted S/N isfor a pulsar detected far off axis, for which the lack of sidelobes in the ALFA beam model causes the predicted S/N to be underestimated.RIGHT: Histogram of S/N for PALFA pulsar discoveries above a threshold S/N of 7. Given realistic spatial distributions in the Galaxycombined with the luminosity function, we expect the number of detections to increase in going to lower S/N. The decrease of thehistogram signifies incompleteness that results from the fact that we have too many low S/N candidates (i.e. S/N <10) that include RFIas well as real pulsars and we have not yet separated and confirmed the best pulsar candidates.

The first two of these potential explanations can be dismissed because we detect known pulsars as expected andwith S/N consistent with catalogued flux densities. Furthermore, we have re-discovered 11 of 14 pulsars found in the“DMB” (Deep Multibeam Survey) conducted by several of us (Camilo, Lorimer and McLaughlin, unpublished) usingone-hour integrations with the Parkes telescope. That we have not yet covered all the sky region of the DMB surveysuggests that PALFA will be complete with respect to DMB objects. Moreover, we detect the DMB pulsars at highS/N in on-axis detections using scans that are more than 10 times shorter. Some of the DMB objects are detectedwell off the boresight of some of the telescope beams in some pointings. The DMB investigators predicted that theywould find about three times as many pulsars as they eventually found (http://www.jb.man.ac.uk/∼mclaughl/dmb/),which may bear an explanation in common with that for the PALFA yield, to some extent.The third and fourth explanations listed above both appear to be relevant for the low yield and we expect that theyield can be increased substantially, as we now discuss:Candidate Winnowing and RFI: Detection statistics on known pulsars show that the pipeline-achieved S/Nfrom our periodicity analysis is consistent with that predicted from catalogued flux densities (left panel of Fig. A7).However, the histogram of S/N (right panel) indicates that detections of new pulsars tail off above our nominalthreshold of 7σ. For any reasonable luminosity function and spatial distribution, the S/N histogram should increasemonotonically in going to lower S/N. This reflects contamination from RFI that is manifested in the very large numberof “signal candidates” above our nominal threshold. Figure A8 shows the histogram of raw “signal candidates” fromthe Cornell pipeline. Nearly every signal above threshold is shown, except for candidates found at 60 Hz or itsharmonics. The figure demonstrates the forest among which we need to sift for high-quality candidates. Similarnumbers result from the PRESTO pipelines. Candidates from both pipleines are examined visually, though cuts aremade to lessen the number required for this. It is a research and development task in itself to develop automatedfiltering algorithms that can further decrease the number of candidates that need to be viewed. This R&D is takingplace at several institutions in the Consortium, in particular McGill and Cornell.At Cornell, we have developed automated winnowing software that identifies known pulsars in the ATNF pulsarcatalog by using location, period and DM tests. The remaining number of candidate signals is ∼ 2M. The numberof candidate signals can be reduced from 2M down to thousands through additional simple filters, such as rejectingcommon candidates that appear in disparate sky positions and other reality checks. But getting from thousandsdown to a few hundred to 1000 is more challenging with respect to minimizing both false-positive and false-negativedetections. We have applied additional filters to Cornell pipeline output based on a number of flags that useassessments of the RFI load on a given day or scan, the S/N, and the statistics on the number of harmonics seen. Astringent application of these filters yields a short pulsar candidate list that has a high proportion (10%) of pulsars(i.e. those already found in the PALFA survey in the quick look analysis or found “by eye” looking through data

19

Fig. A8.— Global period histogram from the Cornell pipeline that includes any signal above 8σ in the harmonic sum of the periodicitysearch; 60 Hz and a few of its harmonics have been excluded. The blue vertical lines indicate the periods of known pulsars that wereblindly detected while the red lines are confirmed PALFA discoveries. The location of known and discovered pulsars near period binswith large counts indicates that pulsar detections are possible even in the presence of severe RFI. Period bins with large numbers ofcounts correspond to RFI signals that are seen in widely spaced pointings, allowing us to reject some candidate signals on the basis thatthey are widespread. However, some RFI is sporadic if not rare, making it less clear on a case by case basis how to discriminate pulsarsfrom RFI. The period bin width is ∆P = P 2∆f , corresponding to 0.5 cycles of smearing over a ∼ 300 s observation, approximately theprecision to which the period can be determined in the FFT + harmonic sum analysis. The “forest” of counts below about 2 ms is fromencoded radar signals.

products; those from the uncataloged DMB survey; and a few pulsars in the ATNF catalog that leak through thecatalog comparison at a high multiple of the spin period or at a discrepant dispersion measure.4) This approachmisses pulsars we have found by visual inspection of output while at the same time finds good candidate objects thathave not been identified in the visual inspection. Moreover, if we allow the filters to run on candidate lists that stillinclude known pulsars in the ATNF catalog, we find that even some of them would not make it into the winnowedlist owing, largely, to a low RFI score. A similar procedure, but with different heuristics, is being applied to thesimilarly large number of signal candidates from PRESTO pipelines and the results will be consolidated with thosefrom the Cornell pipeline. After more experimentation and a less-severe cut, we will take a pulsar candidate list(generated through joint filtering of PRESTO-pipeline results) back to the telescope for confirmation shortly afterthis report is submitted (Aug-Sep 2008).The role of RFI in the single-pulse analysis is shown in Figure. A9. The figure displays output for the 7 ALFA beams

4 Very strong pulsars can be detected at DM values far from the true value.

20

Fig. A9.— Demonstration of RFI excision in the single-pulse analysis. Shown is the graphical output for a known pulsar showingbefore (LEFT) and after (RIGHT) excision of severe 12-s radar, which produces the dark vertical stripes. In both frames, each ALFAbeam corresponds to one of the rows, which contain five panels from left to right: events above 5σ vs. trial DM channel and time (268 stotal); histogram of events per DM channel, scatter plot of S/N vs. DM channel, mapping of DM channel to physical DM value, andhistogram of S/N. The central beam is at the top, beam 6 at the bottom. The excision algorithm first identifies and removes the 12sradar interference using a Fourier method; then it removes isolated impulsive events that occur in more than three of the beams and thatpeak at low DM. The number of counts in the DM vs. S/N scatter plot (third panel in each row) is greatly reduced.

before and after excision of RFI using the single-pulse candidate lists alone to create a mask in the DM-time plane foreach beam. Most of the RFI is radar with a 12s period modulated by up to kHz frequencies, which produces manycandidates in our periodicity analysis. Work in the last year has included development of a similar frequency-timemask based on joint event statistics across the 7 ALFA beams that is applied before dedispersion. We are nowapplying this in our pipeline and consequently are getting far fewer signal candidates in our periodicity analysis.Astrophysical factors: Properties of the pulsar population that influence the rate of detection of new pulsarsinclude:

1. Luminosity function: the pseudo-luminosity Lp = SD2 at 1.4 GHz ranges from approximately L1 = 0.1 toL2 = 104 mJy kpc2. Often the differential distribution is characterized as a power law, fL ∝ L−x

p , where xhas been found to range between 1.6 and 1.8 in a recent study of the Parkes Multibeam survey (Lorimer et al.2006, MNRAS, 372, 777). Many previous results in the literature identify a value of x = 2 (where the slope−d log fL/d logL = x − 1 usually is reported.) As we show, lower values of x favor a strategy that reaches tolarger distances.

2. Period distribution: periods range from 1.4 ms to 8 s with most clustered around 0.7 s.

3. Spatial distribution: many pulsars are distributed in a disk with thickness of ∼ 0.5 kpc and radial scale∼ 4 kpc but extending well past the solar circle. The asymmetry of spiral arms with respect to ` = 0 mayaccount for some differences between the pulsar yields of the Parkes multibeam survey and the PALFA survey.

4. Maximum detectable distance: Dmax corresponds to the volume sampled, Vmax ∝ D3max, where Dmax =√

Lp/Smin, with Smin the minimum detectable flux density in a periodicity search. Smin is very stronglydependent on direction, distance and period.

In Fig A10 theoretical values of Dmax are plotted against spin period P for four different pseudo-luminosities Lp.The curves shown are for the PALFA survey using the current WAPP spectrometers and for the Parkes Multibeamsurvey. The upper and lower boundaries of the bands correspond to full, on-axis gain and 50% gain, respectively.The curves are direction dependent, but the direction used for the figure is representative of low latitude directionstoward the inner Galaxy.From the figure and other considerations, we make the following observations about completeness of the PALFAsurvey to date:

1. Period coverage is complete for P > 10 ms because for periods longer than 10 ms, the Dmax curves flatten aspulses become much less affected by propagation effects (dispersion and scattering).

21

0.1

1

10

0.001 0.01 0.1 1 10

0.1

1

10

0.001 0.01 0.1 1 10

Fig. A10.— LEFT: Maximum detection distance, Dmax, vs. spin period for the direction `, b = 35, 0 (at lower range of thedeclination coverage for Arecibo with b = 0) for our current PALFA observing parameters and for the Parkes Multibeam Survey.We have assumed a fixed intrinsic pulse duty cycle of 0.05 (independent of period). The four frames correspond to different valuesof the ‘pseudo-luminosity’ Lp, which is the period-averaged flux density ×D2. The distribution of Lp for pulsars is broad, cover-

ing >∼ 5 orders of magnitude, because the emission is beamed and because the true beam luminosity is a strong function of spinparameters. The top and bottom boundaries of each shaded region are for full- and half-gain, respectively. The PALFA curves ap-ply to the current observing parameters using the WAPP spectrometers (viz. 268 s integration and 100 MHz bandwith with 256channels and 64 µs time sampling). Propagation effects, which limit Dmax at large distances, are calculated using the NE2001electron density model (Cordes & Lazio 2002). For distances > 5 kpc, Dmax is limited by pulse broadening from scattering.RIGHT: Galactic plane showing spiral arms (blue) as in the NE2001 model, the solar circle (dashed line), solid lines of constantdistance from the Sun at 1, 5, 10 and 15 kpc and lines showing the Galactic longitude coverage (red) for the PALFA survey. Smallgreen points show pulsars in the ATNF catalog with |b| ≤ 5 and large red points show new PALFA discoveries. Note that most of thepulsars within 1 kpc of the Sun in the PALFA search range have been re-detected with PALFA, though this is not indicated in the figure.Distances are calculated using DM and the NE2001 model and are subject to large errors, particularly at the larger nominal distances.With current parameters the survey is complete out to ∼ 1 kpc, i.e. the entire luminosity function of steady pulsars is detectable out to1 kpc. However, the survey samples only about 25% of the LF for pulsars at a distance of 5 kpc (but with considerable uncertainty dueto characterization of the luminosity function and spatial distribution of pulsars). The spatial distribution of pulsars appears to peak ata galactocentric radius ∼ 4 kpc, which is equal to the galactocentric radius of the tangent point of the ` = 32 limit of Arecibo’s coverageof the Galactic plane. The asymmetry of spiral arms with respect to ` = 0 may account for some differences between the pulsar yields ofthe Parkes multibeam survey and the PALFA survey.

2. At large periods, the PALFA/WAPP data obtained so far reach a distance about 1.5 times further than theParkes MB survey.

3. The luminosity function is sampled completely out to a distance of only 1 kpc for P >∼ 10 ms and shorter periodobjects are less completely sampled. (This statement assumes that the luminosity function cuts off at L1

rather than rolling off. Nonetheless, L1 is a milestone type luminosity.) The figure shows that Dmax = 1 kpcfor a luminosity L1 = 0.1 mJy kpc2 and P > 10 ms (and beamed toward us). Conversely, the fraction ofthe luminosity function sampled is increasingly smaller for pulsar samples distances greater than 1 kpc. Forcomparison, the Parkes Multibeam survey becomes incomplete beyond ∼ 0.7 kpc. (In both cases we are readingoff from the curves in Figure A10 that correspond to 50% beam gain.) Also, the Parkes survey, with its widerchannel bandwidths and longer sampling time, was less sensitive to the faster spin periods than is the PALFAsurvey.

4. A luminosity completeness fraction is area of the luminosity function above Lp, which is Fc(Lp) ≈ (L1/Lp)x−1

for L1 ≤ Lp L2. Rewriting in terms of D and ηSmin, where η allows us to scale the threshold flux density,we have

Fc(D) ≈ η−(x−1)D−2(x−1), (A2)

where D is in kpc and we have used the fact that, for nominal Smin (for current WAPPs and integration timeof 268 s) at long P , the survey is complete out to 1 kpc. If our threshold is η = 2 times worse than nominal,we sample between 1/4 and 1/2 of the luminosity function.

22

5. While the luminosity function of pulsars further than 1 kpc is sampled increasingly incompletely with greaterdistance, the volume surveyed Vs ∝ D3 increases. The product FcVs is what really matters and scales as

FcVs ∝ D5−2x ∝ D to D1.8 (A3)

for x = 2 and x = 1.6, respectively. If the volume is uniformly filled with pulsars (which it isn’t of course), thisscaling law indicates how detected pulsar numbers will grow with greater D. Eq. A3 shows that the growth ofthe detection rate depends strongly on x. The lowest x identified in analysis of the PMB survey indicates thatreaching greater distances will enhance the pulsar yield substantially because FcVs ∝ V 1.8.

6. RFI: if our effective threshold is higher than we think because of RFI contamination, then η > 1 and thenumber of detections will be lower. For η = 2, for example, FcVmax ∝ 2−(x−1) ∝ 0.5 to 0.7.

7. One of our conclusions is that we should increase Dmax by increasing the product of bandwidth with integrationtime, BT . We have η ∝ (BT )−1/2, so an increase in BT yields an increase in detected numbers ∝ (BT )−(x−1)/2.

8. It is not practical to increase the “completeness distance” (that to which we reach luminosities down to L1) to10 kpc because that requires an increase of BT by a factor of 104. Such completeness requires an SKA typeinstrument that provides a huge increase in telescope gain in addition to a modest increase in BT . However, wecan increase the completeness distance to about 1.7 kpc through the increased B of the PALFA spectrometersand a doubling of the integration time. We propose that this be done where it is sensible to do so, i.e. for thelongitude range 32 ≤ ` ≤ 60 where the pulsar density is high out to distances of ∼ 15 kpc (c.f. Fig. A10).High luminosity pulsars will be detectable to well beyond 20 kpc and thus will probe the outer boundary ofthe pulsar population in those directions.

Description of the PALFA Survey and Processing

Sky Coverage: Our nominal survey domains are Galactic latitudes |b| ≤ 5 for the longitude ranges 32 ≤ ` ≤ 75and 168 ≤ ` ≤ 214. Through early 2007 our telescope time was split roughly equally between the inner and outerGalaxy longitude ranges. The pulsar population declines with Galactocentric radius sufficiently that we have foundthe pulsar yield to be about ten times less in the outer Galaxy. Therefore, even though we have found pulsarsin anti-center directions, we have curtailed — for the time being — our search of the outer Galaxy in order toconcentrate on the more densely populated inner Galaxy. We are about 1/3 of the way through a first pass of theGalactic plane. Our goal is to complete the Galactic plane survey using the new PALFA spectrometers starting latesummer 2008, and make multiple passes on at least some of the sky positions. We will extend the survey further outof the Galactic plane as a means for optimizing detection of millisecond pulsars and relativistic binary pulsars.Data Acquisition: We have used the Wideband Arecibo Pulsar Processors (WAPPs) for data acquisition, whichallow a 100 MHz passband to be sampled with 256 channels every 64 µs. In a given observing session (3 hr),approximately 0.4 TB of data are obtained.During an observation session, data are transferred to the Arecibo Signal Processor (ASP) and subjected to astandard periodicity search accompanied by a single-pulse search analysis. This quicklook analysis is with degradedtime-and-frequency resolution yet has yielded many of our pulsar discoveries. The quicklook analysis informs us thatsignal levels and data acquisition are in proper states. We observe at least one known pulsar in an observing session,positioning ALFA so that each of the seven beams, in turn, points at the pulsar. In this way, we confirm that eachbeam is working, that pointing is accurate, etc.Within 24-hr of the observations, data are transferred to an Observatory RAID system and on a time scale of weeksare copied to portable IDE disks for shipment to Cornell and to other processing sites.Shipping and Archival of Raw Data: Data disks are shipped to the Cornell Center for Advanced Computing(CCAC) where they are archived to a Tivoli Storage Manager (TSM) based tape system. Data disks are theneither shipped to one of the other processing sites (currently McGill, University of British Columbia and U. Texas,Brownsville) or directly back to Arecibo. To track data integrity, we compute MD5 checksums at three stages:when portable disks are written at Arecibo, when they are archived at CCAC, and when the data are restoredfrom the TSM system for processing with the Cornell pipeline. The MD5 checksums are usually consistent but wehave identified instances of data corruption. In some cases, we have transferred the original, uncorrupted data fromArecibo over the network to replace the bad data. The checksums are included in the “Tracking Database” (seebelow) that we have developed for listing all PALFA data.Relevant Statistics: Including data from the precursor PALFA survey (program P1944) and our long-termprogram (P2030), we have the following

1. Observed 13k distinct sky pointings with 7 sky positions per pointing (excluding test observations on knownpulsars that we routinely make). This can be compared to the total of 47547 distinct pointings we havepredefined for the combined inner and outer-Galaxy survey (about equally split between the two regions).

23

2. Shipped 11k pointings worth of data or 150 TB to the Cornell CAC, all of which have been archived to theTSM robotic tape system and verified.

3. Processed 9k pointings at the Cornell CAC and 738 pointings at other sites using the PRESTO code, whichdoes the more challenging acceleration search on the data.

4. With our scheduled time of approximately 500 hr/yr the current WAPP spectrometers produce approximately7 TB/month. The Cornell pipeline processes data at 7 TB/month when we use two sets of seven nodes toprocess two pointings at a time. The aggregate PRESTO processing rate sums to about 8 TB/month. We arecontinuing to add processing capacity through porting of PRESTO-pipeline code to a new cluster at Cornelland at Franklin & Marshall. The TACC and the Cornell CAC have a strategic relationship under the NSF’sCyberinfrastructure Initiative and are developing the means for transporting data via 10 GB network ratherthan by portable disks. UVa is also on the National Lambda Rail and can receive data at the same rate.

5. Of the previously known pulsars that we detect in the periodicity search analysis, we detect 63% of them inour single-pulse analysis.

Processing from the Cornell pipeline has yielded ∼ 2 × 106 signals above threshold (see definition below) in theperiodicity analysis from the pointings processed to date. The vast majority of these are interference, of course. Wehave blindly (re)discovered 120 known pulsars by taking our candidates and comparing them with the ATNF PulsarCatalog. We have identified 20 of the new pulsars also found in the quicklook analysis, five new objects from thesingle pulse analysis. We have also discovered eight objects in the single-pulse search not detected in the periodicitysearch.Pulsar candidates: From the pipelines (Cornelll and PRESTO codes) we have a huge number (several million)signal candidates above threshold. From noise only, we would expect only a small number of false-alarm candidates.The large number is due to RFI amid which are an expected number of 100 to 200 real pulsars. So far, we have notconverged on a uniform method for winnowing from signal → pulsar candidates but we have tried several methods.These involve human review of profiles and other data products in one approach. In another approach, we use anautomated flagging system that tests the extent to which a candidate signal (characterized as P , DM and numberof harmonics detected) is repeated over disparate sky positions. Ranking also makes use of the RFI extent duringthe particular pointing and day in which the signal candidate was obtained. Although we have not converged on auniform method, we expect to do so and provide an agreed upon list of candidates for reobservation.Pulsar Confirmations: Confirmations occur in several ways. First, particularly good pulsar candidates can be“self” confirmed through retroactive analysis of the data in hand. For example, single-pulse candidates can bechecked by looking at the frequency-time plane and identifying the characteristic plasma dispersion law. The samecan be done for periodicity candidates by folding the radio spectrum at the candidate period. The best periodicitycandidates show sharp pulses with a large number of harmonics and will be consistent with a unique position on thesky. This can be checked through statistical tests using the data bases, as we have done. Trickier cases follow whenpulses are broadened either by interstellar scattering or by orbital smearing in the Cornell code; the PRESTO codeaddresses moderate orbital smearing via the acceleration search. True confirmation follows from re-observation. Wereobserve candidates using our nominal survey time and thus must keep the reobservation list to a minimum.Follow-up Timing: When a pulsar is confirmed, we obtain a few timing points using our allocated survey time.Eventually, we migrate timing to separate observational programs, both at Arecibo and elsewhere.In March, 2007, we began systematic observations of all our confirmed pulsars (Arecibo project code P2177). To date,we have had eleven sessions in the inner Galaxy and ten in the Galactic anti-center. We observe any given pulsaronce or twice per session, at two to three minutes per observation. We have established phase-connected pulse timingsolutions for all of new pulsars with established pulse periods, thirty-two new isolated pulsars in all. In all cases,timing residuals are a few percent of the pulse period or less, demonstrating that the timing solutions are robust. Atable of timing solutions can be found on the web site: http://www.brynmawr.edu/physics/dnice/timing.pdf. Weare in the process of writing these results up for publication; we anticipate submitting them to the AstrophysicalJournal by the end of the summer.The Arecibo timing observations were taken in full-Stokes mode, and polarimetric analysis is underway. We havepromising rotation measure values for several pulsars; since PALFA pulsars are typically much more distant thanpreviously known sources, they allow us to probe the magnetic field more deeply in the Galaxy, particularly in thevicinity of the outer spiral arm.Many of the PALFA pulsars are strong enough to be timed using other telescopes, including the GBT, Jodrell Bank,Nancay, Parkes and Westerbork. Most of the timing data on J1903+0327 used in the Champion et al. timing paperwas from the GBT. Recently we have obtained higher-quality data from Arecibo at 3 GHz that has yielded theprecise mass determinations presented in the main text. Although the timing precision achievable with telescopessmaller than Arecibo is less, it is important to maintain phase connection on the pulsar and it is the totality ofmeasurements along with the rms error in each TOA that determines the fitting errors to relevant parameters.The PALFA pulsars tend to have somewhat shorter periods and higher period derivatives than the full population,indicative of their young ages and strong magnetic fields.

24

Fig. B11.— P − P plane showing known pulsars and pulsars discovered in the PALFA survey. Not all PALFA pulsars are shownbecause P has not been measured for all of them.

Follow-up Timing of conventional pulsars. Once a pulsar is confirmed, systematic follow-up observations are madein order to measure its age and magnetic field strength, determine its position to high precision, characterizing itspolarimetry, measure its flux density and spectral index, and determine its dispersion and scattering properties.These observations are made periodically (nominally at six-week intervals) over the course of a year or longer so thattime-of-arrival measurements extracted from them can decouple the pulsar rotation period and spin-down rate fromits position. Observations are made in full-stokes “search mode,” so that single-pulse and polarimetric properties canbe inferred. Observations use the L-wide receiver and four WAPPs in “single pixel” mode, allowing maximal gainand bandwidth compared to the ALFA receiver. Presently thirty-six pulsars with confirmed periods are observed atleast occasionally in this program at Arecibo; some are observed at Jodrell Bank as well. As the data sets mature,pulsars with well-established characteristics will be dropped from the Arecibo schedule.

We have measured period derivatives of thirty-three of the new pulsars. A P − P diagram of these pulsars is shownin Figure B11. These pulsars span the phase space occupied by previously known pulsars, but there is a tendencyfor them to have relatively short periods and high period derivatives, i.e., they are a relatively young population.Observations over the wide bandwidth provided by the L-wide receiver allow us to detect and measure the effectof interstellar scattering. Pulse shapes at lower frequencies are broadened relatively to higher frequencies. SeeFigure B12 . PSRs J1855+02 and J1856+02, for example, show clear evidence of broadening.

25

Fig. B12.— Pulse profiles obtained for PALFA-discovered pulsars that are timed in the Arecibo timing program. This does not includethe MSP objects or the relativistic binary discovered in PALFA.

26

Processing Pipelines

The quicklook pipeline at Arecibo analyzes PALFA data after degrading the time and frequency resolutions in orderto obtain quasi-real time throughput. Two full-resolution pipelines are active for processing PALFA data at the fullresolution of 0.4 MHz channels and 64 µs time resolution: Both search 1272 trial DMs between 0−1003 pc cm−3,spaced such that the pulse smearing due to interstellar dispersion <1 ms for all DMs <600 pc cm−3.

1. Cornell Pipeline: The Cornell pipeline has been operating on PALFA data in full-resolution mode since mid2005. This pipeline evolved from pulsar surveys of the 1990s and was extended to handle 7-beam ALFA data.It uses the sigproc package (D. Lorimer) as a front-end for unpacking correlation functions and FFT-ing theminto spectra. After dedispersing using a brute-force, post-detection method for 1272 trial values of dispersionmeasure, the resulting time series are searched for periodic signals using a standard harmonic-sum analysis andfor aperiodic signals using two algorithms (matched filtering and friends-of-friends). Diagnostic plots for bothsignal classes are included with flat files of candidates for inclusion in the package of data products for eachALFA pointing. It takes 30 to 36 times real time to process each pointing (all seven beams) using seven nodesof an Itanium cluster. When possible we process two pointings simultaneously, yielding 15 to 18 times realtime throughput.

2. PRESTO Pipeline: This pipeline evolved from the PRESTO single-beam package developed by S. Ran-som (http://www.cv.nrao.edu/∼sransom/presto/). Like the other pipelines, the PRESTO pipeline conductsperiodicity and single-pulse searches. Important features include:

• Extensive automatic terrestrial RFI identification and excision is conducted in both the time and frequencydomains to remove short-duration broadband interference and long-duration narrowband interference.• Computationally intensive linear “acceleration” searches are conducted which give greatly improved sen-

sitivities to fast pulsars in compact binary orbits with massive companions (i.e. PSR-NS or PSR-BHsystems).• Raw periodicity candidates are “sifted” using several heuristics beyond just overall signal-to-noise (such

as the signal-to-noise and/or phases of the individual Fourier harmonics) so that typically “pulsar-like”signals are ranked higher than non-pulsar-like signals.• All single-pulse and periodicity candidates are stored in MySQL databases to allow offline multi-

dimensional browsing, display, sifting, identification, and selection of the best pulsar candidates. A largenumber of parameters for each of the periodicity candidates are saved in the database in order to makethis possible. In addition, one-page graphical diagnostic plots are provided for the top ∼30 candidatesfrom each beam. These plots can be viewed during interactive exploration of the database.• Tests are underway to include either or both a Fast-Folding Algorithm or Coherent Harmonic Summing

search code for slow pulsars with narrow pulse profiles.• The PRESTO pipeline takes 10−20 hrs to search a single 5-min pointing (all seven beams) on a cluster

with one beam per CPU. This is 4 to 7 times longer than the Cornell code owing to the fact that thePRESTO code does an acceleration search.

Databases

MySQL DBs are used at Arecibo for tracking of pointings and at PRESTO processing sites. At Cornell, we havethree SQL Server-based DBs for managing the data through the pipeline and an ancillary DB for dispensing publicly-available data on the Teragrid.1. Tracking Database: The tracking database contains information about which observations have been made,what datafiles came from that observation, whether they have been archived, and whether and with what pipelineand location they have been processed. Some of this information is also stored in the MySQL database at Arecibo.Information in the Tracking Database comes from the following sources:

• CIMA logs written during observations at Arecibo.• Logs written at Arecibo about disk contents as they are shipped from Arecibo to the CTC.• Content listings of files archived to the Tivoli Storage Manager system at the CTC.• Processing centers and one field from the Products database.

2. Cornell Pipeline Data Products Database: This is the largest database, which contains all the dataproducts from the processing pipeline. All data products can be retrieved from this database via the web interfaceat http://arecibo.tc.cornell.edu, which includes web services serving VOTable. The website now has a password-protected interface which allows access to the proprietary dataproducts

27

ALFA Spectrometers

RAID Storage System

Portable Disks

TSM Robotic Tape System

Periodicity/Single-Pulse Processing (full res)

(Unisys Itanium Cluster)

PALFA Common DBMS Server SQL

Current: WAPPsFall 2007: PALFA Spectrometers

FedEx

Arecibo Observatory

Cornell Center for Advanced Computing

Disks Shipped to other processing sites

(UBC, McGill, UVA, UWVA)

Ship back to AO

Periodicity/Single-Pulse Processing (full res)

Parallel ATA disks400 to 750 GB

PALFA Web Site/GUI Web Client

Teragrid10 GBE link

Space Sciences BuildingLinux Cluster for

Processing

Quicklook Analysis(degraded t-f resolution)

AO-based Web Site(informal)

Fig. C13.— Data flow for the PALFA project showing elements at Arecibo (green) and at the Cornell Center for Advanced Computing(yellow). Data are processed at Arecibo (quicklook with degraded resolutions), and with full resolutions at Cornell McGill, Texas(Brownsville), and UBC. Data products are available to PALFA Consortium members over the standard internet. Raw data — currentlyin modest amounts — are available over the Teragrid using Cornell’s National Lambda Rail connection. In large volumes, data aretransported on portable disks.

3. Common Candidate Database: To meld the output from the two pipelines and to al-low for future reprocessing, we have constructed a database that accomodates high-quality pulsar can-didates (periodicity and single-pulse) from both processing pipelines. The schema for this data base(http://arecibo.tc.cornell.edu/PALFA/central/schema.aspx) includes data quality measures, including occupancy inthe time-frequency plane, and graphical output for each candidate.The common database contains high-quality candidates from both processing pipelines and from all processing sites,taking account of the differences between the exact dataproducts from each pipeline. The candidates from the Cornellpipeline are significantly winnowed in number. External access is through python scripts which will run on linux orwindows clients, allowing processing sites to query the database or upload dataproducts to it.4. Legacy Pulsar Database: The Legacy Pulsar Database is a spinoff of the work done on the first twodatabases, with the intent of providing the means for archival of non-ALFA pulsar data, such as the manyWAPP data sets that have been taken for a variety of purposes. The legacy pulsar database and its website(http://arecibo.tc.cornell.edu/legacypulsardata) allow retrieval of basic dataproducts and raw data files for observa-tions of known pulsars. We will use this database and website for distribution of raw data and data products forPALFA pulsars after they have become non-proprietary. The website is data-driven (the website polls the databasewhen loaded, to retrieve lists of what is available for display).The Legacy Pulsars Database and website are listed as a science gateway on the Teragrid. Data can bedownloaded through anonymous ftp and plots can be viewed online. The Teragrid Science Gateway link is:http://www.teragrid.org/programs/sci gateways/projects.php?id=54

28

Data Transport

Using secure access (through encrypted password accounts) consortium members can query the tracking databasethrough a web application to identify and select datafiles for restoration from the tape archive. When the files are re-stored, an email is sent with the details required to transfer those files to the consortium member’s location via authen-ticated ftp/sftp. The URL for this file restoring capability is http://arecibo.tc.cornell.edu/PALFAData/default.aspx.A typical end-to-end restore time for the 4 GB of data from a 268s observation with one WAPP (two beams) is aboutfour minutes. Most of this time is latency for finding and mounting the particular tape and spooling to the file loca-tion. It takes about 1.5 min to transfer the data over the network across the Cornell campus on a 1 GB backbone.The restore/ftp tool has been available to the PALFA Consortium for the last year. For PALFA institutions on theNational Lambda Rail (NLR) or otherwise on the Teragrid, transfer costs should be nil. As PALFA data becomepublic, we will similarly make the transfer tool generally available.Sidelobes and Scattering into the Antenna Optics: Much of the RFI is nonuniform across the seven ALFAbeams. For radar signals in particular, the nonuniformity is extreme in that, in a given data set, some beams aredevoid of RFI in the pipeline output while others show strong RFI. Radar signals evidently enter the telescope’soptical path via scattering off of the platform and other blocking structures and on multiple paths. This yieldsconstructive and destructive interference across ALFA’s aperture, thus explaining the spatial variability.If terrestrial RFI can enter the telescope optics through multiple scattering, so too can celestial signals. Giant pulsesfrom the Crab pulsar are strong enough to be seen in the near sidelobes and it is plausible that they could bedetectable in the far-out sidelobes where the antenna gain is much smaller than the on-axis gain. This raises theinsidious possibility that some single pulse events seen in PALFA data may arise from objects that are very far offaxis and may be either known pulsars that emit giant pulses or previously undiscovered objects. Confirmation ofsingle-pulse objects clearly must take this possibility into account. We are currently trying to understand far-outsidelobes for the ALFA system by looking at the statistics of radar pulses vs. azimuth and elevation angles, time ofday, day of week, etc.

PALFA Participants

High School Students (All at UT Brownsville): Irina Azcona, Maggie Bice, Raquel Castaneda, Jonathon Castillo,Joseph Claudio, Edwin Ferrer, Anthony Freitas, Ramon Garcia, Wendy Garza, Marlen Guerrero, Jessica Gutierrez,Anthony Guerra, Brian Leal, Magaly Lopez,Nick Lopez, Joey Martinez, Chris Martinez, Sebastian Gonzalez Molina,Alan Ponce, Mariela Rivera, Astrid Perez Roman, Hannah Upton.

Undergraduate Students:

Drew Fleckenstien ARCC Scholar, UT BrownsvilleAnthony Ford ARCC Scholar, UT BrownsvilleAlejandro Garcia ARCC Scholar, UT BrownsvilleRossina Miller ARCC Scholar, UT BrownsvilleJesus Rivera ARCC Scholar, UT BrownsvillePeter Cox Melbourne University

Brian Devour Franklin & MarshallJames McBride UC Berkeley (REU student at Cornell, 2007)Chase Morgan Franklin & Marshall

Laura Popa Bryn MawrKyle Story Cornell (academic years 2005-6-7)Yingzi Wang Bryn Mawr

Graduate Students:

Julia Deneva CornellLaura Kasian UBCPatrick Lazarus McGillK.J. Lee UT Brownsville, Peking U.Kevin Stovall UT BrownsvilleAdrienne Rodriguez-Zermeno UT Brownsville

29

Cornell Center for Advanced Computing (CCAC; formerly Cornell Theory Center):

John Zollweg Programming, data integrity, and processing

Ph.D. Scientists:

Bruce Allen Albert Einstein Institute (Hannover, Germany)Zaven Arzoumanian Goddard Space Flight CenterDon Backer UC BerkeleyAdam Brazier CornellFernando Camilo ColumbiaRamesh Bhat University of Swinburne, AUDavid Champion Australia Telescope National FacilityShami Chatterjee CornellIsmael Cognard NancayJim Cordes CornellFronefield Crawford Franklin & MarshallAvinash Deshpande Raman Research InstitutePaulo Freire AreciboBryan Gaensler The University of SydneyJin-Lin Han BeijingJason Hessels AmsterdamFrederick Jenet UT BrownsvilleVicki Kaspi McGillMichael Kramer MPIfRJoseph Lazio NRLAndrea Lommen Franklin & MarshallDuncan Lorimer West Virginia UniversityAndrew Lyne Manchester, UKMaura Mclaughlin West Virginia UniversityDavid Nice Bryn MawrScott Ransom NRAOXavier Siemens University of Wisconsin, MilwaukeeIngrid Stairs UBCBen Stappers U. Manchester (UK)Steve Thorsett UC Santa CruzArun Venkataraman NAICJoel Weisberg Carleton College