Detection of pulses from the Vela pulsar at millimeter wavelengths with phased ALMA
Kuo Liu, Andre Young, Robert Wharton, Lindy Blackburn, Roger Cappallo, Shami Chatterjee, James M. Cordes, Geoffrey B. Crew, Gregory Desvignes, Sheperd S. Doeleman, Ralph P. Eatough, Heino Falcke, Ciriaco Goddi, Michael D. Johnson, Simon Johnston, Ramesh Karuppusamy, Michael Kramer, Lynn D. Matthews, Scott M. Ransom, Luciano Rezzolla, Helge Rottmann, Remo P.J. Tilanus, Pablo Torne
DDraft version March 25, 2020
Typeset using L A TEX twocolumn style in AASTeX62
Detection of pulses from the Vela pulsar at millimeter wavelengths with phased ALMA
Kuo Liu, Andr´e Young, Robert Wharton, Lindy Blackburn,
3, 4
Roger Cappallo, Shami Chatterjee, James M. Cordes, Geoffrey B. Crew, Gregory Desvignes,
1, 8
Sheperd S. Doeleman,
3, 4
Ralph P. Eatough, Heino Falcke, Ciriaco Goddi,
2, 9
Michael D. Johnson,
3, 4
Simon Johnston, Ramesh Karuppusamy, Michael Kramer, Lynn D. Matthews, Scott M. Ransom, Luciano Rezzolla, Helge Rottmann, Remo P.J. Tilanus,
9, 2, 13 and Pablo Torne
14, 1 Max-Planck-Institut f¨ur Radioastronomie, Auf dem H¨ugel 69, D-53121 Bonn, Germany Department of Astrophysics, Institute for Mathematics, Astrophysics and Particle Physics (IMAPP), Radboud University, P.O. Box9010, 6500 GL Nijmegen, The Netherlands Center for Astrophysics | Harvard & Smithsonian, 60 Garden Street, Cambridge, MA 02138, USA Black Hole Initiative at Harvard University, 20 Garden Street, Cambridge, MA 02138, USA Massachusetts Institute of Technology Haystack Observatory, 99 Millstone Road, Westford, MA 01886, USA) Cornell Center for Astrophysics and Planetary Science, Cornell University, Ithaca, NY 14853, USA Massachusetts Institute of Technology Haystack Observatory, 99 Millstone Road, Westford, MA 01886, USA LESIA, Observatoire de Paris, Universit´e PSL, CNRS, Sorbonne Universit´e, Universit´e de Paris, 5 place Jules Janssen, 92195 Meudon,France Leiden Observatory - Allegro, Leiden University, P.O. Box 9513, 2300 RA Leiden, The Netherlands CSIRO Astronomy and Space Science, Australia Telescope National Facility, PO Box 76, Epping, NSW 1710, Australia National Radio Astronomy Observatory, Charlottesville, VA 22903, USA Institut f¨ur Theoretische Physik, Goethe-Universit¨at Frankfurt, Max-von-Laue-Straße 1, D-60438 Frankfurt am Main, Germany Netherlands Organisation for Scientific Research (NWO), Postbus 93138, 2509 AC Den Haag , The Netherlands Instituto de Radioastronom´ıa Milim´etrica, IRAM, Avenida Divina Pastora 7, Local 20, 18012, Granada, Spain
ABSTRACTWe report on the first detection of pulsed radio emission from a radio pulsar with the ALMAtelescope. The detection was made in the Band-3 frequency range (85 −
101 GHz) using ALMA in thephased-array mode developed for VLBI observations. A software pipeline has been implemented toenable a regular pulsar observing mode in the future. We describe the pipeline and demonstrate thecapability of ALMA to perform pulsar timing and searching. We also measure the flux density andpolarization properties of the Vela pulsar (PSR J0835 − Keywords: pulsars: individual (PSR J0835 − INTRODUCTIONPulsars are steep-spectrum radio sources (e.g., Lorimer& Kramer 2005). As a result, the vast majority of thepulsar population have been discovered at frequenciesbelow 2 GHz. Correspondingly, the study of the ra-dio emission has been limited to similar frequencies,although successful studies have been conducted up to
Corresponding author: K. [email protected] much higher frequencies. Before 1990, the highest fre-quency used for successful pulsar studies was 25 GHz,while in the 1990s observations were pushed to 32 GHz(Wielebinski et al. 1993), 43 GHz (Kramer et al. 1997),and finally 87 GHz (Morris et al. 1997). Emission fromnormal pulsars was later observed at 138 GHz (Torne2016), and finally, detections of the radio-emitting mag-netar PSR J1745 − a r X i v : . [ a s t r o - ph . GA ] M a r Liu et al. more effective pulsar searches in highly turbulent envi-ronments like the Galactic Center (Cordes & Lazio 1997;Lorimer & Kramer 2005; Spitler et al. 2014; Dexter et al.2017). Previous pulsar studies have suggested that thecoherent emission seen at lower radio frequencies mayundergo changes at high radio frequencies (Kramer et al.1996). This may be understood as a break-down of thecoherent radiation mechanism, which can be expected tooccur when the observed wavelength becomes compara-ble to the coherence length. The break-down would cor-respond to a transition from the coherent radio emissionto the incoherent infrared or optical emission (Lorimer& Kramer 2005). At the same time, a standard model ofpulsar emission physics interprets observed pulse char-acteristics as the result of a “radius-to-frequency map-ping”, where higher radio frequencies are emitted fromlower emission heights (Cordes 1978). In the contextof this model, performing observations at higher radiofrequencies is equivalent to approaching the polar capregion of the pulsar.All previous studies of pulsars above 30 GHz havebeen conducted with Northern hemisphere radio tele-scopes, especially the 100-m Effelsberg radio telescopenear Bonn, Germany and the 30-m telescope of the Insti-tut de Radioastronomie Millim´etrique (IRAM) on PicoVeleta, Spain (see L¨ohmer et al. 2008 and Torne et al.2017 and references therein). With the advent of the At-acama Large Millimeter/submillimeter Array (ALMA),a large collecting area is now available to study South-ern hemisphere pulsars. Here we report on the estab-lishment of a fast time-domain capability (hereafter a“pulsar observing mode”) for ALMA’s phased-array sys-tem (Matthews et al. 2018). This new pulsar observingmode can be used for observations of compact objectsin the Galactic Center and elsewhere in the Galaxy. Wedemonstrate the capabilities of this system with obser-vations of the Vela pulsar.As one of the brightest radio pulsars in the sky, Velawas one of the first pulsars discovered (Large et al. 1968)despite its relatively fast spin period ( P spin = 89 ms).Recently, an imaging detection of Vela was made at mil-limeter wavelengths using ALMA in its standard inter-ferometry observing mode (Mignani et al. 2017). Asshown below, we can now use phased ALMA (i.e., ALMAin the phased-array mode) to obtain a complementarystudy of Vela’s pulsed and polarised emission at frequen-cies of 90 GHz and above.ALMA can also play a key role in probing neutronstar populations in the Galactic Center and using de-tected objects for the study of spacetime around thecentral black hole, the Sgr A* (Liu et al. 2012; Psaltiset al. 2016; Liu & Eatough 2017). Despite previous ef- forts, no pulsar in a sufficiently close orbit to Sgr A*has yet been detected (Wharton et al. 2012). However,the discovery of a rare radio-emitting magnetar (PSRJ1745 − ≈ . − ALMAPulsar Mode Project , hereafter APMP) involves a se-ries of steps needed to acquire phased ALMA data forpulsar studies: definition and implementation of the ap-propriate phasing mode, providing the signal path fromthe ALMA Phasing Project (APP) system to Mark 6baseband recorders, offline resampling and reformattingof data into PSRFITS format (Hotan et al. 2004), anddevelopment of the backend pulsar/transient process-ing customized to ALMA science contexts. The projectleveraged software development for pulsar phased-arraymodes for the Very Large Array (VLA) and devel-opments for the Event Horizon Telescope (Doelemanet al. 2008; Event Horizon Telescope Collaboration et al.2019), Black Hole Camera (Goddi et al. 2017) projects,and the ALMA Phasing Project (Matthews et al. 2018). OBSERVATIONSFeasibility studies conducted for the APMP usedtest data obtained in conjunction with ALMA Phas-ing Project commissioning runs in 2016 April and 2017January. The former provided data used to test theintegrity of the Mark 6 to PSRFITS transformationwhile the latter provided data on the Vela pulsar (PSRJ0835 − https://science.nrao.edu/facilities/vla/docs/manuals/oss/performance/pulsar ela pulses at mm wavelengths with ALMA − × × . µ s and packed in PSRFITS format in search mode(channelized time series). For the purpose of detect-ing the Vela pulsar, the data were folded to form 10-ssub-integrations, using an ephemeris obtained from lowfrequency observations around the same period of time.Details on the data flow and the pre-processing can befound in Figure 1. RESULTSThe Vela pulsar has been successfully detected af-ter folding the search mode data. Figure 2 shows theintegrated pulse profiles obtained from the lower andupper sideband, respectively. An effective integrationtime of 25 min yields a peak signal-to-noise ratio of thedetection of ∼
50 for the lower and ∼
40 for the uppersideband. The signal was detected at frequencies upto 101.268 GHz, the highest radio frequency that thepulse profile of Vela has ever been seen. Overall, theALMA detection of the Vela pulsar provides a success-ful demonstration of the APMP and provides the basisfor future execution of our primary scientific motivation:searches and follow-up observations of pulsars and tran-sient sources in the Galactic center.3.1.
Timing
The utility of pulsars as precision clocks, the so-calledpulsar timing technique, requires high timing stabilityduring the process of data recording. To examine sucha capability of the phased ALMA data, we carried outa timing analysis with the 10-s sub-integrations fromboth sidebands. For each of them, we averaged the pulseprofile over frequency, and calculated its time-of-arrival(TOA) with the canonical template-matching method
ALMACorrelator
VDIF
Mark VIVLBIRecorderMark VIVLBIPlayback
VDIF
DataTransport P o l P o l Packet Buffer
Metadata vdif2psrfits tools Time-avgPSRFITS
ChannelizedPSRFITS
FullResolution PSRArchive
Figure 1.
The developed ALMA system for offline processing toproduce pulsar/transients data. The phased ALMA voltage dataare recorded on Mark6 VLBI recorders (Whitney et al. 2013). Inour study the data were played back and processed at Cornell Uni-versity, the Massachusetts Institute of Technology (MIT) HaystackObservatory, the Center for Astrophysics | (Taylor 1992). Figure 3 shows the timing residuals ob-tained by subtracting the predictions of the ephemeris(obtained independently from low-frequency observa-tions) from the TOAs. No time offset was seen betweenthe four individual scans. A weighted root-mean-square(rms) timing uncertainty of 134 µ s has been achieved,which is in general consistent with the TOA errors ex-pected from radiometer noise. This shows that a timingprecision of order 100 µ s that was used in the simula-tion of Liu et al. (2012), is in fact possible for pulsarobservations with ALMA at 3-mm wavelengths.3.2. Search capability
To demonstrate the capability of the APMP forsearching for periodic signals in the data, we used theoverall dataset to directly carry out a blind search forthe Vela pulsar. We first averaged the time series fromall four individual sub-bands for each scan, and com-bined the time series from all four scans, with the powermean padded in the gaps among the scans. Then aperiodicity search was performed using the presto software package (Ransom et al. 2002). Figure 4 shows ∼ sransom/presto/ Liu et al.
LowerUpper MPIfR CfA
Figure 2.
Left panel: Pulse profiles detected in the lower (bottom, 85 . .
268 GHz) and upper sideband (top, 96 . .
268 GHz) ofALMA Band-3, using the MPIfR pipeline. Middle and right panel: Comparison of pulse profile achieved from the MPIfR and CfA pipeline(folded using the presto software package, Ransom et al. 2002). In short, the MPIfR pipeline makes power detection of all four Stokesparameters in frequency domain, while the CfA pipeline derives total intensity power directly from the state counts. With the same sectionof data, the detection from these two pipelines shows highly consistent measures of detection significance in total intensity and the shapeof the pulse profile of the Vela pulsar. The product from the MPIfR pipeline is used for the data analysis in the rest of this paper. R e s i d u a l ( m s ) Wrms = 134.2 µ s Figure 3.
Timing residuals of Vela pulse profiles derived from10-s sub-integrations. The open circles and squares representresiduals from the lower and upper sideband, respectively. the power spectrum of the combined time series. Thelow frequency noise, mainly caused by fluctuations of the power level in the time series, starts to becomesignificant for frequencies below 5 Hz. Meanwhile, theoverall power level of the rest of the spectrum is mostlyflat. The fundamental spinning frequency of the Velapulsar (around 11.2 Hz) and its higher order harmonicsare clearly seen in the Figure 4 inset. The periodicitysearch resulted in 28 candidates with detection signif-icance above 3 σ . The fundamental spin frequency ofthe Vela pulsar was the top candidate and most of therest were higher harmonics. At frequencies of 50 and1 Hz, periodic signals were detected and associated withthe electricity power cycle and cryogenic pump’s cycle,respectively. Another signal was detected at 2.6 Hz, andits origin is unknown but is likely to be terrestrial, as itwas also seen in scans of the calibrator . The ALMA baseline correlator is clocked (precisely) at 125MHz with microprocessor interrupts at 16 ms and timing signalsat 48 ms—so signals commensurate with these are likely to havebeen produced in the correlator. ela pulses at mm wavelengths with ALMA -3 -2 -1 Frequency (Hz) P o w e r
10 20 30 Figure 4.
Fourier power spectrum of the overall time seriesformed by combining data from all four individual Vela scans.For each scan, the time series was first averaged in frequency.The inset shows a zoomed-in region of the power spectrum forfrequencies between 5-35 Hz, with linear scale on the x-axis. Notethat the spinning frequency of the Vela pulsar is approximately11.2 Hz.
Pulsar properties
We can use the obtained data to study the proper-ties of the Vela pulsar at frequencies between 80 and100 GHz and compare those to Vela’s properties at lowerfrequencies, and to those of PSR B0355+54 (to be pre-sented in future work), the only other normal pulsar de-tected at similar frequencies (Morris et al. 1997; Torne2016). This will allow us to gauge the prospects of fu-ture pulsar observations with ALMA and further ourunderstanding of pulsar emission physics.3.3.1.
Flux Density
The recorded data of the phase calibrator J0828 − . Then for each individualVela scan, we used the closest calibrator scan to calibratethe flux density of the pulsed emission, by using the stan-dard flux calibration formula (Lorimer & Kramer 2005).This gave us a mean flux density of 0 . ± .
17 mJy and0 . ± .
12 mJy at 87.268 and 99.268 GHz, respectively,where the error bars represent the actual standard de-viation of all four individual measurements. Using theordinary ALMA interferometry data recorded in parallel https://almascience.nrao.edu/sc/ during the observation and calibrated following the ded-icated procedures developed for phased ALMA (Goddiet al. 2019), we also produced image detections of theVela pulsar from all individual scans, which deliveredmean flux density measurements of 0 . ± .
09 mJy and0 . ± .
05 mJy at 87.268 and 99.268 GHz, respectively.These are fully consistent with the measurements de-rived from the phased ALMA data.3.3.2.
Profile evolution
The lack of a drift during the folding of the data (seeFig. 2) and the apparent timing stability (Section 3.1),imply that the data timestamps are reliable. Going fur-ther, we can compare the phase of the pulse arrival timeat 1.4 GHz (obtained at the Parkes Radio Telescope),defined by that of the main pulse peak (identified withphase zero), with that of the pulse peak seen with ALMAat mm-wavelengths. Figure 5 demonstrates that the lat-ter is delayed by about 1.8 ms with respect to the mainpulse peak at 1.4 GHz. This offset is explained by com-paring the profiles to those at intermediate frequencies.We use those presented by Keith et al. (2011), follow-ing their alignment based on a separation of the pro-file into individual components . As can be seen fromFigure 5 the main component seen at 1.4 GHz (and be-low) becomes progressively weaker at high frequenciesand is undetectable at ALMA frequencies. The com-ponent remaining is the second, weaker component ofthe 1.4 GHz profile. This is consistent with the iden-tification of the dominant 1.4-GHz component seen inthe top profile of Figure 5, with a so-called “core com-ponent” (see e.g. Johnston et al. 2001) which tends tohave steeper spectra. Indeed, Keith et al. (2011) mea-sure a spectral index of − . ± . − . ± . . Extrapolating from the 24 GHz flux density(3.4 mJy, Keith et al. 2011), we therefore expect a fluxdensity of 0 . ± .
14 mJy at 87 GHz and 0 . ± .
31 mJyat 99 GHz. This is in perfect agreement with the fluxdensity measurement we obtained (see Section 3.3.1).3.3.3.
Polarization properties Keith et al. (2011) modelled the profiles at different frequen-cies as a sum of a number of symmetric components representedby scaled von Mises functions. They found that a set of the samefour component fits to all frequencies whilst keeping the width andseparation of each component fixed. We aligned their solution rel-ative to our ALMA and Parkes observations by eye. Note that we identify Keith et al. (2011)’s component Cand/or D with the ALMA component, for which a spectral in-dex of − . ± .
06 and − . ± . − . Liu et al.
Figure 5.
Vela profiles measured as a function of frequency. Thebottom profile (red) is the result of the addition of the two ALMAsidebands. It is aligned in time, using the data’s timestamps, withthe 1.4 GHz profile shown on top (red), which resulted from anobservation with the Parkes Radio Telescope at 1.4 GHz made on2017 January 8. Note that using nearly contemporaneous obser-vations with Parkes, we minimize any confusing time delay poten-tially caused by rotational instabilities known as “timing noise.”Indeed, the apparent delay of 1.8 ms of the ALMA profile withrespect of the pulse peak seen at 1.4 GHz can be explained bya distinct profile evolution. This becomes clear when adding theprofiles (black) observed and aligned by Keith et al. (2011). Seetext for details.
The MPIfR pipeline allows for the extraction of full-Stokes information from ALMA baseband data. We findthat Vela still shows some significant linear and circularpolarisation. Assuming that the observed polarisationcharacteristics are similar to those observed at 24 GHz,we can use the data by Keith et al. (2011) as a polarisa-tion template and perform a system calibration withoutneeding to make any assumptions on “ideal feeds” oradditional constraints on degeneracy in the system pa-rameters (van Straten 2006; Smits et al. 2017). Theresult is shown in Figure 6, where we show an expandedregion around the pulse. Here, the degree of linear po-larisation ( ∼ Here we chose only to fit for differential gain and phase, asthe leakage in ALMA Band-3 was shown to be no more than afew percentage (Goddi et al. 2019). degree of polarisation, while maintaining a constant PAswing (believed to be tied to the magnetic field and theviewing geometry of the pulsar), is perfectly consistentwith overall trends in other pulsars (Lorimer & Kramer2005). The sign of circular polarisation is identical tothat at lower frequencies and its fraction relative to to-tal intensity is similar to at 24 GHz. I n t e n s it y ( a r b . un it s ) -15 -10 -5 0 5 10 15 20 25Pulse longitude (deg)-150-100-50050100150 P o s iti on a ng l e ( d e g ) Figure 6.
Polarisation properties of the Vela pulsar as observedwith ALMA, when adding the Stokes vectors measured for thelower and upper sideband. The top panel shows the total power(black), the linearly polarised emission component (red), and thecircularly polarised emission (blue). The lower panel shows theposition angle of the linearly polarised component as a functionof those pulse phases, where the linear intensity exceeds 1 . σ ofthe off-pulse region. We also indicated the position angle swingmeasured at 1.4 GHz, corrected for Faraday rotation to infinitefrequency. DISCUSSIONWe have demonstrated the capability of phasedALMA for the study of pulsars. The use of phasedALMA in combination with a passive phasing modewill enable future pulsar searches, as well as timing,polarisation, and emission studies. This opens up newscience possibilities, especially in the southern hemi-sphere, which is completely unexplored for pulsars atfrequencies above 30 GHz.Given the typical steep spectrum of pulsars, they willgenerally be far too weak at ALMA frequencies to al-low active phasing of the ALMA array on the pulsar ela pulses at mm wavelengths with ALMA . − , Johnston et al. 2005), guarantees that thecontamination by Faraday rotation during calibrationprocess is negligible at 3-mm wavelengths. To carry outthis experiment, a long track of the Vela pulsar needsto be conducted in order to cover a wide range of par-allactic angle. Then a calibration can be performed byfollowing the approach described in van Straten (2004,2006), after which a polarisation template of the Velapulsar at the given frequency will be constructed. For calibration afterwards, one would only need a short scanon the Vela pulsar and match it to the template.We thank A. Evans, T. Remijan and F. Stoehr for thehelp to stage our data on the ALMA science portal .KL, RW, GD, RPE, HF, CG, MK, LR, PT acknowl-edge the financial support by the European ResearchCouncil for the ERC Synergy Grant BlackHoleCam un-der contract no. 610058. SC and JMC acknowledgesupport from the National Science Foundation (AAG1815242). Work on this project at the Smithsonian As-trophysical Observatory was funded by NSF Grant AST-1440254. This paper makes use of the following ALMAdata: ADS/JAO.ALMA Facility:
ALMA
Software:
PRESTOREFERENCES
Cordes, J. M. 1978, ApJ, 222, 1006Cordes, J. M., & Lazio, J. T. W. 1997, ApJ, 475, 557Dexter, J., Deller, A., Bower, G. C., et al. 2017, MNRAS,471, 3563Doeleman, S. S., Weintroub, J., Rogers, A. E. E., et al.2008, Nature, 455, 78Eatough, R. P., Falcke, H., Karuppusamy, R., et al. 2013,Nature, 501, 391Event Horizon Telescope Collaboration, Akiyama, K.,Alberdi, A., et al. 2019, ApJL, 875, L2Goddi, C., Falcke, H., Kramer, M., et al. 2017,International Journal of Modern Physics D, 26, 1730001 The data are available via: https://almascience.eso.org/alma-data/enhanced-data-products/vela-pulsar-j0835-4510.
Goddi, C., Mart´ı-Vidal, I., Messias, H., et al. 2019, PASP,131, 075003Hotan, A. W., van Straten, W., & Manchester, R. N. 2004,Publications of the Astronomical Society of Australia, 21,302309Johnston, S., Hobbs, G., Vigeland, S., et al. 2005, MNRAS,364, 1397Johnston, S., van Straten, W., Kramer, M., & Bailes, M.2001, ApJL, 549, L101Keith, M. J., Johnston, S., Levin, L., & Bailes, M. 2011,MNRAS, 416, 346Kramer, M., Jessner, A., Doroshenko, O., & Wielebinski, R.1997, ApJ, 488, 364Kramer, M., Xilouris, K. M., Jessner, A., Wielebinski, R.,& Timofeev, M. 1996, A&A, 306, 867