The North American Nanohertz Observatory for Gravitational Waves
aa r X i v : . [ a s t r o - ph . I M ] O c t The North American Nanohertz Observatory forGravitational Waves
M. A. McLaughlin
Department of Physics, West Virginia University, Morgantown, WV 26506-6315, USAE-mail: [email protected]
Abstract.
The North American Nanohertz Observatory for Gravitational Waves(NANOGrav) is a collaboration of researchers who are actively engaged in usingNorth American radio telescopes to detect and study gravitational waves via pulsartiming. To achieve this goal, we regularly observe millisecond pulsars (MSPs) with theArecibo and Green Bank Telescopes and develop and implement new instrumentationand algorithms for searching for and observing pulsars, calculating arrival times,understanding and correcting for propagation delays and sources of noise in ourdata, and detecting and characterizing a variety of gravitational wave sources. Wecollaborate on these activities with colleagues in the International Pulsar Timing Array(IPTA). We also educate students of all levels and the public about the detection andstudy of gravitational waves via pulsar timing.
1. Introduction
Gravitational waves (GWs), ripples in space-time produced by accelerating massiveobjects, are a fundamental prediction of Einstein’s theory of general relativity.Measurements of orbital decay due to GW emission in double neutron star binarysystems provide convincing evidence for their existence [1, 2]. However, as of yet, wehave not detected the influence of GWs on space-time through a measured change inlight travel time between two objects. This direct detection of GWs will immediatelyprovide spectacular proof of Einstein’s theories and will also usher in a new era ofastronomy in which we can use GWs to study objects which are thus far invisible orinaccessible through electromagnetic observations.Pulsars are rapidly rotating, highly magnetized neutron stars produced in thesupernova explosions of massive stars. Pulsar timing arrays (PTAs) are able to detectGWs through high-precision timing, sensitive to small changes in the light travel timesbetween the pulsars and Earth. The pulsars used for our experiment are millisecondpulsars (MSPs), which have been spun-up to very short periods through accretion ofmass and angular momentum from a companion star. These objects are incrediblystable rotators, with arrival times measurable to microsecond precision and spin periodspredictable to one part in 10 . There are over 200 known Galactic MSPs, of which ANOGrav − − − Hz frequency range, complementary to the much higher frequenciesprobed by ground- or spaced-based GW interferometers. The most likely GW sourcesfor detection by PTAs include supermassive black hole binaries, cosmic strings, and,possibly, early universe inflation. Therefore, PTAs will provide crucial input to galaxyformation and evolution scenarios and cosmology. Additional source classes also mayawait discovery.NANOGrav was formed in October 2007 as a collaboration of researchers at NorthAmerican universities, colleges, national laboratories, and observatories. We use the100-m Green Bank Telescope (GBT) in Green Bank, WV, and the 300-m AreciboObservatory (AO) in Arecibo, Puerto Rico to observe an array of MSPs with the goalof GW detection. It is one of three PTA collaborations and, along with the EuropeanPulsar Timing Array (EPTA) and the Parkes Pulsar Timing Array (PPTA), is a memberof the International Pulsar Timing Array (IPTA). In this article, we will review thecollaboration’s organization and then provide an overview of NANOGrav’s activity inkey areas.
2. Organization
The organizational structure of NANOGrav is illustrated in Figure 1. NANOGravconsists of faculty, senior researchers, postdocs, graduate students, and undergraduatestudents at 13 institutions in the United States and Canada. Membership is open toall who share our goal of GW detection and study using pulsars. We currently have31 Full, 14 Associate, and 12 Junior (or undergraduate) members. New membershiprequests can be submitted to the Chair at any time and are accepted from participantsin any country who are willing to contribute to our goals. To be eligible for FullMembership, a participant must be in the collaboration for at least one year. Ourauthorship policy states that all Full Members of NANOGrav will be authors ondetection or upper limit papers. NANOGrav is governed by a Chair and a managementteam. The Chair and the management team (MT) members are elected for two-yearterms. Two NANOGrav members also serve on the IPTA Steering Committee. Moredetails about our by-laws, membership policy, and authorship policy can be found athttp://nanograv.org/governance.html.The detection and study of GWs using pulsars requires a diverse portfolio of workin various observational, analytical, and theoretical areas. NANOGrav accomplishesits goals in these areas through various working groups. We being by describing thework involved in observing and timing an array of MSPs and in applying algorithms todetect and characterize sources, as these are the core activities of NANOGrav. We thendescribe how searching for MSPs, developing techniques to mitigate interstellar medium
ANOGrav §
3. Timing Observations and Analysis
Gravitational wave detection requires observations of a number of pulsars with thehighest precisions possible. A detailed discussion of timing methodology and algorithmsis given by Demorest & Lommen in this issue. We have observed 18 pulsars for morethan five years (see Figure 2), with 17 of them used in the analysis for NANOGrav’sfirst paper reporting an upper limit on the GW background [3]. Over the past five years,three to four pulsars per year have been added to the timing program and a total of36 pulsars are currently being observed. In this section, we first describe the timingobservations used for the first upper limit paper and then the current observing scheme.Preserving the narrow intrinsic pulse profiles which provide the highest timingprecisions requires coherent dedispersion, or the removal of dispersive delays (see § ANOGrav § § µ s (for PSR J1643 − §
7) was detected with high significance in the residuals of two objects and withmarginal significance in the residuals of two others. The remainder of the pulsars haveresiduals consistent with white noise.Our current timing program consists of regular observations of 36 pulsars withArecibo and the GBT (see Table 1). Observations are carried out roughly once everythree weeks at both telescopes, with each pulsar observed for about 10–40 minutes at twowidely separated radio frequencies. At the GBT, data are accumulated and coherentlydedispersed in 1.5625-MHz wide subbands using the Green Bank Ultimate PulsarProcessing Instrument (GUPPI), an FPGA-based spectrometer capable of processingup to 800 MHz of bandwidth [5]. A bandwidth of 200 MHz is used for observations at acenter frequency of 820 MHz and a bandwidth of 800 MHz is used for observations at acenter frequency of 1500 MHz. At Arecibo, the PUPPI backend coherently dedispersesdata, also in 1.5625-MHz wide subbands, over bandwidths of 40/700/600 MHz at centerfrequencies of 430/1410/2310 MHz, respectively. Data are folded in real-time with 15-sintegrations using GUPPI and 10-s subintegrations using PUPPI. A calibration scanis obtained for each pulsar by injecting a 25-Hz noise diode into the signal path forboth polarizations. At each observation epoch and at each frequency, a flux calibrator(B1442+101) is observed.Some properties of the pulsars in our current sample and some observational detailsare provided in Table 1. A “living” version of this table, with updated RMS values andmedian TOA uncertainties, is available at http://nanograv.org/sources.html. Naively,the over 10-fold increase in bandwidth due to GUPPI and PUPPI should result inarrival times measured with over three-fold increased precision at some frequencies,leading to RMS residuals over three times smaller than those listed in Table 1. Actualimprovements (see Figure 3 for an example) in RMS residual range from factors of ∼ ANOGrav
4. Detection and Characterization
NANOGrav develops algorithms for detection of continuous, burst, and stochastic GWsources. Continuous sources emit at a roughly constant GW frequency. Super-massiveblack hole binaries are the most likely source of continuous GWs for PTAs. The reviewarticle by Ellis in this issue describes a Bayesian analysis pipeline that will detect and
ANOGrav
Name Period DM Obs. Frequencies RMS
ASP ˜ σ t, ASP
MJD T NPSR (ms) (pc cm − ) (MHz) ( µ s) ( µ s)J0023+0923 3.05 14.3 AO 430/1410 – – 55731 1.9 56J0030+0451 4.87 4.3 AO 430/1410 0.148 0.37 50788 15.5 325J0340+4130 3.30 49.6 GBT 820/1500 – – 55972 1.3 41J0613 − − − − − − − − − − − − − σ t from the first NANOGrav upperlimit paper using the ASP and/or GASP instruments [3], MJD at which the timing program began,total time span of the observations, and number of observing epochs. The RMS and median TOAuncertainty for PSR J1713+0747 are for the combined GBT and Arecibo dataset. ANOGrav σ upper limit on characteristic strain at afrequency of 1/yr of 7 × − by computing covariance matrices of the post-fit residuals.In this work, the GW analysis was performed separately from the timing fit, but theproperties of the timing fit were used to determine the amount of GW that may havebeen absorbed in the fit. A single-pulsar upper limit of 1 . × − was also presented,based on the residuals of PSR J1713+0747. Current work involves developing methodsto calculate and interpret the optimal cross-correlation statistic [15] and more efficient,but approximate, maximum likelihood approaches [16]. NANOGrav aims to makedetection pipelines for all source classes available at http://nanosoft.sourceforge.net afterdevelopment and testing.The article by Siemens et al. in this issue discusses the time-to-detection for astochastic background of GWs under various assumptions. They show that we areapproaching the strong-signal regime, where the GW power is larger than the whitenoise, and that in that regime the time-to-detection depends only weakly on thecadence of observations and the white-noise RMS, and much more strongly on thenumber of pulsars. This is independent of any assumptions about red noise (thoughin the case of significant red noise, adding more pulsars to the array becomes evenmore important). Using realistic simulations assuming NANOGrav’s current observingprogram and modest improvements, they show that a detection could occur as earlyas 2016 and will occur by 2023 given reasonable assumptions about the expected GWamplitude range of the SMBH binary background.It is important to note that our goal encompasses not only detection of GWs,but also characterization of sources and GW astrophysics. Models for supermassive ANOGrav
5. Pulsar Searching
The sensitivity of a PTA increases with the number of MSPs included in the array,making searches for MSPs extremely important to our mission. NANOGrav’s searchingworking group provides support for multiple searches with both Arecibo and the GBT.Two pulsars searches are underway using Arecibo: the 327-MHz Arecibo Drift-Scansurvey (AO327) and the 1.4-GHz Arecibo L-band Feed Array Survey (PALFA). AO327is sensitive to nearby MSPs out of the Galactic plane, while PALFA is sensitive todistant MSPs in the plane. Thus far, AO327 has discovered 20 pulsars, including twoMSPs [18], and PALFA has discovered 116 pulsars, including 17 MSPs [19, 20]. Notethat only a fraction of MSPs found in a particular survey will have the narrow, brightprofile and timing stability essential for inclusion in a PTA. Of the PALFA MSPs, twohave thus far have been included in NANOGrav’s regular timing program.Recent GBT searches at 350 MHz include the GBT Drift-Scan Survey (GBTDrift)and the Green Bank Northern Celestial Cap Survey (GBNCC). GBTDrift discovered 35pulsars, including seven MSPs [21, 22], and GBNCC has so far discovered 62 pulsars,including nine MSPs. One pulsar from GBTDrift and one pulsar from GBNCC havealready been included in NANOGrav’s timing program. One additional MSP wasdiscovered in a separate portion of the GBTDrift survey set aside for analysis by high-school students in the Pulsar Search Collaboratory program [23]. In addition, 28 MSPshave been discovered through GBT searches of unidentified
Fermi sources [24, 25], withthree of these pulsars included in NANOGrav’s timing program.Figure 4 illustrates the rate of MSP discovery since the discovery of the first MSPin 1982, and the importance of the GBT and Arecibo surveys in which NANOGravmembers are involved. The review by Stovall, Lorimer, & Lynch in this issue offersmore details about these surveys and projections for the future. Recent MSP populationstudies show that there are a large number (roughly 30,000–80,000) of Galactic MSPs
ANOGrav † vs. year, withdiscoveries by the GBT in green andArecibo in red. Over 60% of allGalactic MSPs have been discoveredsince 2009. The GBT and Arecibotogether have discovered roughlyhalf of all Galactic MSPs. † http://astro.phys.wvu.edu/GalacticMSPs Figure 5 Flux density at 1400 MHz (left) and distance (right) vs. year of discovery forGalactic MSPs † . Recent surveys continue to reveal bright, nearby MSPs.which remain to be detected [26, 27]. The precision with which we can time a pulsar isdirectly proportional to its flux density; current surveys are still detecting bright MSPs,demonstrating that continued searches may yield rich returns (see Fig. 5). In addition,current surveys continue to reveal nearby MSPs for which we are more likely to be ableto measure precise distances through radio interferometry (see Fig. 5).
6. Interstellar Medium Mitigation
Two frequency-dependent ISM effects - dispersion and scattering - affect pulse TOAs.These are discussed in detail in the review by Stinebring in this issue. Dispersion dueto refraction by free electrons results in delays proportional to DM × ν − , where theDM is the integrated column density of electrons along the line of sight and ν is radiofrequency. Observations at widely separated frequencies (see Table 1) maximize thedifferential dispersive delays and allow accurate DM fitting. Because of the relativemotion between pulsars and the ISM, DMs can change on timescales of ∼ weeks or less,necessitating that the observations at different frequencies occur within, at maximum, ANOGrav ν − . Scattering delays forsome MSPs can be greater than several µ s at our observing frequencies and correctionis crucial for achieve the highest precisions possible. NANOGrav is exploring severalmethods for removal of scattering delays. The first involves estimating the characteristicbandwidth ∆ ν d through auto-correlation of dynamic spectra, which describe how thepulsar flux changes with time and frequency. Then, the pulse broadening time, whichis equal to the scattering delay under certain assumptions [28], can be calculated as(2 π ∆ ν d ) − . This method may result in a modest improvement in RMS for some pulsars[29]. The second method is termed cyclic spectroscopy (CS) and relies on the periodicsignature of the pulsar signal to use the phase information to recover unscattered pulseprofiles and scattering delays [30]. We are exploring the use of CS to calculate delays andcorrect TOAs using both simulated and real data. We are also applying the methodto MSPs with a variety of fluxes, pulse shapes, and DMs to determine the range ofapplicability as the only published CS application is on a very bright, moderatelyscattered pulsar [30]. Because baseband-sampled data is necessary, we are developinga GPU-based real-time CS implementation that can be applied to every NANOGravobservation [31].
7. Noise Budget
A growing focus of NANOGrav’s work is on understanding, characterizing, and exploringmitigation techniques for all sources of noise affecting pulsar TOAs. These sourcesinclude those both extrinsic and intrinsic to the pulsar. Extrinsic sources of noiseinclude radiometer noise, the ISM, and the ionosphere. Intrinsic sources of noise includerotational instabilities and pulse jitter. Most of these sources of noise are “white”, withflat power spectral density, and therefore easily distinguished from the “red” spectrumexpected due to background of GWs. However, spin variations from torque fluctuationsand internal NS activity display a red spectrum [32], similar to that expected due toGWs, and can have a profound effect on detection prospects (see Figure 6).Therefore, determining the contribution of red noise to the residuals is crucial.Methods used include autocorrelating the residuals, measuring the number of residualzero crossings, and testing the residuals for non-Gaussianity. The results from theseanalysis techniques are, so far, similar to those based on spectral analysis presented inthe NANOGrav upper limit paper: the majority of NANOGrav pulsars have residualsconsistent with white noise, with a few exceptions [3]. There are departures from non-Gaussianity but they are consistent with flux modulation due to interstellar scintillation.However, as timing precisions increase, it is possible that red spin noise will becomeapparent in many more pulsars. If red spin noise is shown to be ubiquitous in MSPs,
ANOGrav
8. Education and Outreach
An essential component of the NANOGrav mission is to inform the general publicabout NANOGrav science, inspire the next generation of scientists, and train studentmembers of NANOGrav to perform research and excel in scientific careers. Our websitehosts various materials that inform the general public. These include descriptions ofGWs, pulsars, and GW detection methods, podcasts with NANOGrav astronomers,animations, and several “Kahn-academy” style videos which describe our science ata level appropriate for an advanced high-school or beginning undergraduate student.NANOGrav members at all levels are involved in producing these materials. We alsoinclude information on the “Einstein@Home” project through which citizen scientistscan search for pulsars in PALFA data and on three major outreach projects run byNANOGrav members. The Arecibo Remote Command Center (ARCC), based atthe University of Texas at Brownsville, and its satellite program at the University ofWisconsin at Madison, involve undergraduate students in pulsar searches. The PulsarSearch Collaboratory (PSC) is run by West Virginia University and the National RadioAstronomy Observatory and has involved over 800 high-school students from 15 statesin pulsar searching. Finally, the Mid-Atlantic Relativistic Initiative for Education(MARIE) program, based at Franklin & Marshall, brings NANOGrav students intopublic high schools and hosts astronomy open houses at F&M. NANOGrav’s goal isto more directly integrate these programs into NANOGrav over the next several years,building a pipeline for GW studies into the next generation and beyond. We also aimto integrate our efforts with IPTA-wide outreach.
ANOGrav T (shaded region) for a range of plausible values for the GW amplitude RMS dueto supermassive black hole binaries. The GW spectrum is ∝ f − / , while red spin noiseis ∝ f − , though as discussed in the review by Cordes, there is a large amount of scatterin this scaling law. Also shown are 3 σ thresholds based on the RMS timing residualin the absence of GWs. The heavy black curves indicate the thresholds expected whenthere are only white-noise measurement errors. The light (red) curves show thresholdswhen red “timing noise” adds to the white noise. Actual noise curves will likely fallsomewhere between the light and heavy curves depending on presently uncertain levelsof spin noise in MSPs. Figure credit: Jim Cordes.
9. Partnerships for International Research and Education
Since August 2010, many NANOGrav members have received funding through anaward from the National Science Foundation’s Partnerships for International Researchand Education (PIRE) program. The primary goal of PIRE is to “support highquality projects in which advances in research and education could not occur withoutinternational collaboration”. The PIRE funding supports NANOGrav personnel towork on IPTA-related research. It also provides funding for IPTA student workshopsand science meetings and for research-abroad experiences for U.S. students. Thus far,ten NANOGrav undergraduate students and two graduate students have performedNANOGrav-related research abroad supported by this award. Evaluations of the PIREprogram demonstrate that the meetings and research abroad experiences are crucial forencouraging students to stay in the field and for preparing them to perform research in
ANOGrav
10. Strategic Planning
GW astrophysics was named in the National Academy of Sciences “New Worlds andNew Horizons” Decadal Report as one of five key discovery areas. Pulsar timing isa critical capability for GW detection and study, as it is the only means to probesources in the 10 − − − Hz frequency band. However, despite broad support forGW astrophysics, National Science Foundation funding constraints imply uncertainfutures for the GBT and Arecibo, with the GBT recommended for divestment by 2017and Arecibo operational costs secured for only the next four years. A reduction intime on either of these two telescopes would dramatically impact the sensitivity ofour experiment and our time to detection. Carrying out an identical analysis as forNANOGrav’s first upper-limit paper but with only Arecibo or only GBT data increasesthe upper limit by a factor of two. Calculating time-to-detection estimates as in theSiemens review article in this issue shows that our sensitivity would be roughly halvedand time-to-detection roughly doubled if we had access to only one of these telescopes.Furthermore, Figure 7 illustrates the importance of both telescopes to a well-sampledcorrelation curve and the dramatic increases that will follow from additional pulsars.Therefore, we consider continued access to these telescopes, and the scientific resourcesrequired for our observing and analysis programs, to be our most critical strategicplanning task. A secondary goal is to secure funding for ultra-broadband receiverson both Arecibo and the GBT. Receivers with frequency coverage of 700 MHz – 3 GHzwould dramatically increase sensitivity by boosting overall signal-to-noise ratios andalso allowing for more precise correction of frequency-dependent ISM effects.
11. Conclusions
In the five years since its formation, NANOGrav has evolved into a coherent organizationwhich provides a framework for researches to share ideas and resources and totrain students in a collaborative environment. Our work is broad and involves allaspects of GW detection with pulsars from searching for new pulsars to detecting andcharacterizing different source classes. Over the past several years, our sensitivity hasincreased through new pulsars and wider bandwidth instruments. We expect furtherimprovements with new algorithms to increase timing precision, mitigate interstellarscattering, characterize noise, and optimally and efficiently detect and characterizevarious types of sources. Over the next five years, we will increase our focus on GWsources and multi-wavelength characterization to prepare for the post-detection era ofGW astrophysics and build links with the broader astronomy community.
ANOGrav
12. Acknowledgements
MAM is supported through the NSF PIRE program and the Research Corporation.She also gratefully acknowledges the work of and many discussions with NANOGravmembers that have improved this review. She also thanks the anonymous referees,whose comments significantly improved this paper.
ANOGrav References [1] Taylor, J. H., & Weisberg, J. M. 1982,
ApJ , , 908[2] Kramer, M. et al. et al. 2006, Science , , 97[3] Demorest, P. et al. 2013, ApJ , , 94.[4] Lorimer, D. R. & Kramer, M. 2005, Handbook of pulsar astronomy, Cambridge University Press.[5] DuPlain, R. et al. 2008, Proceedings of the SPIE , .[6] Pennucci, T. et al. in preparation.[7] Zhu, W. et al. in preparation.[8] Swiggum, J. et al. in preparation[9] Ellis, J. A. et al. 2012, ApJ , , 175.[10] Ellis, J. A. et al. 2012, ApJ , , 96.[11] Damour, T., & Vilenkin, A. 2000, Phys Rev L , , 3761.[12] Finn, L. S. & Lommen, A. N. 2010, ApJ , , 1400.[13] Cordes, J. M. & Jenet, F. A. 2012, ApJ , , 54.[14] Hellings, R. W. & Downs, G. S. 1983, ApJ , , 39.[15] Anholm, M. et al., 2009, Phys Rev D , , 084030.[16] Ellis, J. A. et al. 2013, ApJ , , 63.[17] Jenet, F. A., Lommen, A., Larson, S. L., & Wen, L. 2004, ApJ , , 799[18] Deneva, J., et al. 2013, in preparation[19] Crawford, F. et al. 2012, ApJ , , 90[20] Deneva, J. S. et al. 2012, ApJ , , 89[21] Boyles, J. et al. 2013, ApJ , , 80[22] Lynch, R. S. et al. 2013, ApJ , , 81[23] Rosen, R. et al. 2013, ApJ , , 85[24] Ransom, S. M. et al. 2011, ApJL , , L16[25] Bangale, P. et al., in preparation.[26] Lorimer, D. R. 2013, IAU Symposium, , 237.[27] Levin, L., Bailes, M., Barsdell, B. R., et al. 2013, MNRAS , 434, 1387.[28] Cordes, J. M. & Rickett, B. J. 1998,
ApJ , , 846.[29] Levin, L. et al. in preparation.[30] Demorest, P. 2011, MNRAS , , 2821.[31] Jones, G.et al., in preparation.[32] Shannon, R. M. & Cordes, J. M. 2010, ApJ , , 1607.[33] Cordes, J. M. & Shannon, R. M. 2012, ApJ , , 89.[34] Shannon, R. M. & Cordes, J. M. 2012, ApJ , , 64.[35] Lyne, A. G. et al. 2010, Science ,329