A Distant Fast Radio Burst Associated to its Host Galaxy with the Very Large Array
C. J. Law, B. J. Butler, J. X. Prochaska, B. Zackay, S. Burke-Spolaor, A. Mannings, N. Tejos, A. Josephy, B. Andersen, P. Chawla, K. E. Heintz, K. Aggarwal, G. C. Bower, P. B. Demorest, C. D. Kilpatrick, T. J. W. Lazio, J. Linford, R. Mckinven, S. Tendulkar, S. Simha
DDraft version July 7, 2020
Typeset using L A TEX preprint2 style in AASTeX63
A Distant Fast Radio Burst Associated to its Host Galaxy with the Very Large Array
Casey J. Law, Bryan J. Butler, J. Xavier Prochaska,
3, 4
Barak Zackay, Sarah Burke-Spolaor,
6, 7, 8
Alexandra Mannings, Nicolas Tejos, Alexander Josephy,
10, 11
Bridget Andersen,
10, 11
Pragya Chawla,
10, 11
Kasper E. Heintz, Kshitij Aggarwal, Geoffrey C. Bower, Paul B. Demorest, Charles D. Kilpatrick, T. Joseph W. Lazio, Justin Linford, Ryan Mckinven,
15, 16
Shriharsh Tendulkar,
10, 11 and Sunil Simha Cahill Center for Astronomy and Astrophysics, MC 249-17 California Institute of Technology, Pasadena, CA 91125,USA National Radio Astronomy Observatory, Socorro, NM, 87801, USA University of California Observatories-Lick Observatory, University of California, 1156 High Street, Santa Cruz, CA95064, USA Kavli Institute for the Physics and Mathematics of the Universe, 5-1-5 Kashiwanoha, Kashiwa, 277-8583, Japan Institute for Advanced Study, Princeton Center for Gravitational Waves and Cosmology, West Virginia University, Chestnut Ridge Research Building,Morgantown, WV 26505 Department of Physics and Astronomy, West Virginia University, Morgantown, WV 26506 CIFAR Azrieli Global Scholars program, CIFAR, Toronto, Canada Instituto de F´ısica, Pontificia Universidad Cat´olica de Valpara´ıso, Casilla 4059, Valpara´ıso, Chile Department of Physics, McGill University, 3600 University Street, Montr´eal, QC H3A 2T8, Canada McGill Space Institute, McGill University, 3550 University Street, Montr´eal, QC H3A 2A7, Canada Centre for Astrophysics and Cosmology, Science Institute, University of Iceland, Dunhagi 5, 107 Reykjav`ık, Iceland Academia Sinica Institute of Astronomy and Astrophysics, 645 N. A’ohoku Place, Hilo, HI 96720, USA Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr, M/S 67-201, Pasadena, CA91109 USA Department of Astronomy and Astrophysics, University of Toronto, 50 St. George Street, Toronto, ON M5S 3H4,Canada Dunlap Institute for Astronomy and Astrophysics, University of Toronto, 50 St. George Street, Toronto, ON M5S3H4, Canada
ABSTRACTWe present the discovery and subarcsecond localization of a new Fast Radio Burstwith the Karl G. Jansky Very Large Array and realfast search system. The FRB wasdiscovered on 2019 June 14 with a dispersion measure of 959 pc cm − . This is thehighest DM of any localized FRB and its measured burst fluence of 0.6 Jy ms is lessthan nearly all other FRBs. The source is not detected to repeat in 15 hours of VLAobserving and 153 hours of CHIME/FRB observing. We describe a suite of statisticaland data quality tests we used to verify the significance of the event and its localizationprecision. Follow-up optical/infrared photometry with Keck and Gemini associate theFRB to a pair of galaxies with r ∼
23 mag. The false-alarm rate for radio transients of
Corresponding author: Casey J. [email protected] a r X i v : . [ a s t r o - ph . H E ] J u l this significance that are associated with a host galaxy is roughly 3 × − hr − . The twoputative host galaxies have similar photometric redshifts of z phot ∼ .
6, but differentcolors and stellar masses. Comparing the host distance to that implied by the dispersionmeasure suggests a modest ( ∼
50 pc cm − ) electron column density associated with theFRB environment or host galaxy/galaxies. Keywords:
Radio transient sources, radio interferometry INTRODUCTIONFast Radio Bursts (FRBs) are millisecond-timescale radio transients of extremely highbrightness originating at cosmological distances(Petroff et al. 2019; Cordes & Chatterjee 2019).More than hundreds of FRBs are known cur-rently, and the inferred occurrence rate isroughly 10 sky − day − above a fluence limitof 1 Jy ms at frequencies near 1.4 GHz (Cham-pion et al. 2016; Lawrence et al. 2017). FRBdistances can be estimated from the disper-sive delay induced by propagation through ion-ized gas (quantified by a Dispersion Measure,DM, which measures the total electron columndensity along the line of sight to the source);for FRBs, the measured DMs are significantlylarger than those expected due to contributionsfrom our own Galaxy (Cordes & Lazio 2002).By attributing the dispersion induced outsideof our Galaxy to predictions for the intergalac-tic medium (IGM), FRBs are estimated to orig-inate at characteristic distances one to a fewgigaparsecs (Inoue 2004; Lorimer et al. 2007).Several FRBs have been localized by radio in-terferometers and associated with host galaxiesof known distance; their luminosity distancesrange from 149 Mpc to 4 Gpc (Marcote et al.2020; Tendulkar et al. 2017; Bannister et al.2019; Prochaska et al. 2019a; Ravi et al. 2019;Macquart et al. 2020).It is not yet known what causes FRBs orwhether there are multiple formation channels(Lu & Kumar 2016; Ravi 2019a). Identificationsof FRB host galaxies is a critical test of forma-tion models, as it can constrain the age of the stellar populations in FRB environments. Thefirst host galaxy suggested that FRBs are asso-ciated with peculiar star-forming environments(Bassa et al. 2017) but later hosts have a widerrange of environments (Bannister et al. 2019;Ravi et al. 2019; Bhandari et al. 2020).Radio waves are modified as they propagatethrough ionized gas (e.g., dispersion, scattering,lensing, Faraday rotation; Cordes et al. 2017;Vedantham & Ravi 2019). This fact, combinedwith the large distance to FRBs, makes themnovel probes of the IGM and other galaxies(Ginzburg 1973; Masui et al. 2015; Prochaskaet al. 2019a). Furthermore, the fact that disper-sion is an unambiguous tracer of baryonic masshas opened potential for FRBs to study galaxyhalos and cosmology (Ravi 2019b; Prochaska &Zheng 2019). However, most of this science po-tential can only be achieved by measuring dis-tances to FRBs. Multiple radio interferometersfor precision FRB localization are in phases ofconceptual development, construction, or com-missioning (Law et al. 2018; Kocz et al. 2019;Bannister et al. 2019; Caleb et al. 2019; Oost-rum et al. 2017). The goal of all these projects isto localize FRBs to arcsecond precision, whichis required to unambiguously associate it to ahost galaxy (Eftekhari et al. 2018).Many FRBs are seemingly single flashes, andbefore the advent of widespread of use of GPUsto accelerate complex processing, single-dishtelescopes generally led blind searches for newFRBs (Burke-Spolaor et al. 2011; Thorntonet al. 2013; Spitler et al. 2014). However,some FRBs, such as FRB 121102, emit mul-tiple bursts at irregular intervals (Spitler et al.2016; Zhang et al. 2018), which made it pos-sible to target with interferometers (Chatter-jee et al. 2017; Marcote et al. 2017). TheCanadian Hydrogen Intensity Mapping Experi-ment (CHIME) is a transit telescope operatingbetween 400–800 MHz that is rapidly discov-ering both repeating and non-repeating FRBs(CHIME/FRB Collaboration et al. 2018; TheCHIME/FRB Collaboration et al. 2019). TheCHIME/FRB search has a localization precisionof roughly 10 (cid:48) , which is too large to unambigu-ously identify host galaxies for FRBs.Here, we present a new FRB discovery and lo-calization by the Karl G. Jansky Very Large Ar-ray (VLA) using realfast (Law et al. 2018). TheFRB was found coincidentally during a searchfor CHIME/FRB FRB 180814.J0422+73 (here-after FRB 180814, CHIME/FRB Collaborationet al. 2019). This new FRB is associated witha unique host galaxy with a distance that isconsistent with expectations for its DM. Thecombination of radio interferometric data andoptical associations support the conclusion thatit is a new FRB and we refer to it as FRB20190614D. We discuss the FRB environmentand constraints on the distribution of DM inthe IGM and host galaxy. OBSERVATIONS2.1.
Program and Overall Description
In 2018, the VLA and CHIME/FRBteams began collaborating to use the VLAfor follow-up of repeating FRBs found byCHIME/FRB. We have had two approvedprojects: VLA/18B-405 and VLA/19A-331. We targeted FRB180916.J0158+65 andFRB190303.J1353+48 for 40 hours scheduledunder VLA/18B-405 and FRB 180814 for 39hours scheduled under VLA/19A-331; this pa-per focuses on the second project.We observed using the L-band system of theVLA, spanning 1-2 GHz, in twenty separate ob-servations. We observed a field centered at (RA, Dec) [J2000] = (04h22m22s, +73d40m00s), theapproximate position of FRB 180814. The nom-inal field of view of the VLA antennas at L-bandis ∼ (cid:48) (full width half maximum at 1.4 GHz),but the realfast system is configured to imagea field 2 times wider than that. The first sevenobservations were performed in December 2018,in the C-configuration of the VLA, with maxi-mum baselines ∼ ∼ (cid:48)(cid:48) at 1.4 GHz. Thirteen later observations wereperformed in February through July of 2019, inthe B- or BnA-configurations of the VLA, withmaximum baselines ∼
11 km in length and a res-olution of ∼ . (cid:48)(cid:48) at 1.4 GHz. Each observationhad an on-source time of around 1.5 hours thatwas searched by the realfast system. The de-tection reported here is the strongest FRB-likeevent found in this campaign and is the focus ofthe analysis presented.2.2. Search Technique
The observations used a commensal correlatormode that generated visibilities with an inte-gration time of 5 ms to be searched by realfast .The same data also were used to generate andsave the standard visibility data product to theNRAO archive with a sampling time of 3 s, forall observations in June and July 2019 (nine ofthem). Prior to that, all visibilities were savedto the archive at their full time resolution, re-sulting in large datasets (of order 1.5 TB). Bothfast and slow visibilities were made in 16 64-channel spectral windows, with each channel setto a width of 1 MHz. Taking typical interferenceflagging into account, the usable bandwidth is600 MHz.The fast-sampled visibilities were distributedto a dedicated GPU cluster using vysmaw (Poko-rny et al. 2018) and searched with rfpipe (Law2017). After applying available on-line cali-brations, the search pipeline dedispersed andintegrated visibilities in time before formingimages. Calibration solutions derived from ∼ minute-long scans and are stable in time (lessthan 5 deg change from mean value). Imageswere generated with a simple, custom algorithmthat uses natural weighting and a pillbox grid-ding scheme. The search used 215 DM val-ues from 0 to 1000 pc cm − and four temporalwidths from 5 to 40 ms, which is inclusive of theknown properties of FRB180814 (CHIME/FRBCollaboration et al. 2019).For the B-configuration observations, each im-age had 2048 × (cid:48)(cid:48) , covering a field of view of 1 ◦ . The C-configuration images were 512 ×
512 pixels witha pixel size of roughly 6.8 (cid:48)(cid:48) . The nominal1 σ sensitivity in a single 5 ms integration is6 mJy beam − . All candidates detected withsignificance greater than 7.5 σ trigger the record-ing of 2–3 s of fast sampled visibilities and a vi-sualization of the candidate. Each candidate isclassified by fetch , a convolutional neural net-work for radio transients (Agarwal et al. 2019).Finally, realfast team members review the vi-sualizations of the real-time analysis to eitherremove data corrupted by interference or iden-tify candidates for more refined offline analysis.2.3. Discovery
On 2019 June 14 (UT), the realfast system de-tected a candidate transient in the FRB 180814field. The realtime detection system reporteda candidate with image significance of 8.0 σ and DM = 959 pc cm − , far in excess of theexpected DM contribution of the Milky Way(83.5 pc cm − , Cordes & Lazio 2002). However,the DM of FRB 180814 is 189.4 pc cm − ; noFRB has shown changes in DM of more than afew pc cm − (Gajjar et al. 2018), so the can-didate FRB is likely unrelated to the CHIMEFRB.The realtime candidate analysis revealed mul-tiple signatures consistent with an astrophysi-cal source. First, the spectrum (Figure 1, rightpanel) shows emission over a range of frequen-cies spanning at least 50 MHz and the imageshows a compact source. Most sources of in- Time (s; relative) F r e q u e n c y ( G H z ) Figure 1. (Left) Stokes I dynamic spectrum forthe candidate FRB as seen by VLA/ realfast . Thedynamic spectrum was generated by summing cal-ibrated visibilities for all baselines and the two or-thogonal polarizations. The gap and higher noiselevel toward the top left of the dynamic spectrumresults from when the data recording was initiated.(Right) Stokes I spectrum taken from a single 5 msintegration of the dynamic spectrum. terference tend to have circular polarization,narrow spectral extent, or are spatially inco-herent (i.e., radio frequency interference in thenear-field of the array). Second, the fetch
FRB classification system reported an astro-physical probability of 99.9%. Third, there isa weak prior expectation for blindly-detectedastrophysical events to be detected where theantenna sensitivity is highest. The candidatewas detected roughly 9 (cid:48) away from the pointingcenter, where the antenna has roughly 80% ofits nominal sensitivity; only 10% of the imagehas this sensitivity or higher.The realfast search system was starting to re-ceive visibilities from the VLA correlator dur-ing the burst. This is seen in Figure 1, whichshows that the mean of all recorded visibilitiesduring the burst (phased toward the event) isnoisier at early times and at higher frequencies.Visibilities for each baseline, polarization, andspectral window (64 channels) are distributedseparately such that the fraction of data growsto 100% over a few hundred milliseconds as thesystem turns on.2.4.
Verification Tests and SignificanceAnalysis
Traditional fast transient surveys measureevent significance based on a noise estimate thatis local in time (e.g., a standard deviation of atime series). Our interferometric search mea-sures significance in a single image, so the noiseestimate is made simultaneously. Appendix Adescribes how the visibility domain search canbe thought of as a time-domain search that al-lows for more accurate noise estimates.In our initial analysis of the candidate, weconfirmed that the event significance was notaffected by different flagging algorithms or cal-ibration solutions from a calibrator observationa few minutes after the event. We also con-firmed that removing an antenna from the 27-antenna array reduced the detection significanceby roughly 5% ( ≈ /
27 antennas). With con-fidence in the quality of data, we proceeded tomore carefully quantify the event significance.We used the raw, saved visibilities to re-runthe search with a larger image (8192 × .
19 pc cm − . Us-ing the same refinement procedure on other can-didates typically does not reproduce the initialdetection. Noise-like events are expected to besensitive to the image gridding parameters, sowe ignore all events that cannot be reproducedin larger images. We use these refined proper-ties for visualizations and all further analysis.Figure 2 shows the cumulative distribution ofevent significance for all events seen in this cam-paign. The FRB search pipeline automaticallyapplies flags for bad calibration, antenna state,missing data, and interference. We visually in-spected the 263 candidates detected above 7.5 σ Figure 2.
Circles show the cumulative distribu-tion of candidates in this observing campaign as afunction of image S/N ratio. The solid line showsthe expected cumulative event rate for a Gaus-sian (noise-like) S/N distribution. The yellow crossshows the candidate FRB S/N ratio after refine-ment analysis. in observations of this field and removed thoseaffected by unflagged interference to get a sam-ple of 31 candidates.Figure 2 also shows an independent estimateof the ideal event rate significance distributionfor the array and correlator configuration usedto find this candidate. The ideal cumulativeevent rate assumes that each pixel imaged hasa brightness that is drawn from a stationaryGaussian distribution. The number of inde-pendent pixels searched is (N pix / O pix ) × N int ∗ (N DM / O DM ), where N pix is the width of an im-age in pixels, N int number of integrations (at alltime widths), N DM is the number of DM trials,and O pix / DM are the oversampling of the synthe-sized beam and dispersion sensitivity function,respectively. Both images and DMs are over-sampled to maintain uniform sensitivity to alllocations and DMs. The search run here usesO pix = 2 . DM = 3. In this configuration,we have 8.4 hrs of observing time and 5 × in-dependent pixels. The candidate signal-to-noiseratio (S/N) of 8.27 corresponds to a False AlarmRate (FAR) of once in 250 hr. The measuredand ideal distributions are independent and inrough agreement, which shows that the signifi-cance follows a Gaussian distribution and thatthis candidate is an outlier.The FRB search pipeline also uses spectralbrightness fluctuations to distinguish candi-date events from noise (Law et al. 2017; TheCHIME/FRB Collaboration et al. 2019). TheKalman detector (Zackay, in prep) is a methodto estimate the statistical significance of FRBspectral variations for an assumed noise modeland signal smoothness. For a given noise andsignal model, we can marginalize the detectionstatistic over all matched filters, weighted bytheir prior probability. This prior probability isdefined by a random walk with one free param-eter, the coherence bandwidth. We calculatedthe Kalman score on the candidate FRB, us-ing logarithmic spaced options for the smooth-ing scale, but found no significant change in thetotal confidence for the candidate FRB (otherFRBs do show some improvement; Zackay, inprep). We conclude that the candidate FRBspectrum is consistent with a constant flux den-sity. 2.5. Localization
The realtime FRB search software makes sev-eral assumptions to improve computational effi-ciency, and as a result images which are usedwithin it are not optimal. To address this,we used the stored raw, de-dispersed visibili-ties to re-image the burst data with a combina-tion of CASA (McMullin et al. 2007) and AIPS (Greisen 2003) . Here, we describe a unifiedcalibration and imaging procedure used in bothfast and deep imaging. This procedure allowsus to quantify the systematic error in the FRBlocalization.Prior to re-imaging the burst data, we reducedall of the data taken in June and July 2019 fora deep image of the field. Nine datasets dur-ing B configuration were included in this anal-ysis. We excluded C configuration data, as ithas poorer spatial resolution. We also excludedearly B configuration data recorded at the fastsampling rate, as it was computationally expen-sive to include in the deep imaging analysis.We started by applying the calibration andflagging tables for each observation which wereprovided by the VLA calibration pipeline. Forall observations, the flux density scale was setwith an observation of the calibrator source3C 147, and at these frequencies is accurateto 1-2% (Perley & Butler 2017). Bandpassand delay calibrations were also determined bythe 3C 147 observation. Complex gain (am-plitude and phase) fluctuations over time werecalibrated with observations of the calibratorsource J0410+7656 every 30 minutes. We thenexported the calibrated visibilities from CASAand imported them into AIPS. After furtherRFI flagging, we averaged in time (to 9 sec-onds), and frequency (to 4 MHz channels) toreduce the computational load for the imaging.We used faceted imaging in AIPS to imageto beyond the first null of the antenna primarybeam response (1.1 ◦ width). A total of 73 sep-arate fields, each 1024 × (cid:48)(cid:48) pixel size), and 250 CLEAN boxes were usedto image and clean the area. After cleaning,the 73 images of the fields were combined to-gether, and that result was used to self-calibrate Both CASA and AIPS calibrate with a different algo-rithm from that used by the realtime calibration systemknown as “telcal” (Law et al. 2018). (Cornwell & Fomalont 1999) the visibilities ona 1-minute timescale. The imaging and self-calibration was then repeated using this self-calibrated dataset, on a 9-second timescale - es-sentially self-calibrating every visibility. A finalimage was then made, and a primary beam cor-rection made to it, based on Perley (2016). Thisis the final deep image used for further analysis.The synthesized beam in this final deep imageis 3.6 (cid:48)(cid:48) × (cid:48)(cid:48) at a position angle of 79 ◦ (Norththrough East). The image has a 1 σ sensitivityof 3.6 µ Jy beam − , consistent with expecta-tions for the total on-source time and flagging.For the re-imaging of the burst data, we firstcopied the VLA calibration pipeline tables (cal-ibration and flagging) from the full June 14thobservation, and ran a modified version of theprocedure to re-apply these tables. Calibra-tion tables from the three spectral windows (384MHz of bandwidth) with valid, uncorrupteddata were applied. The synthesized beam inthis final burst image is 10.3 (cid:48)(cid:48) × (cid:48)(cid:48) at a posi-tion angle of 67 ◦ . It is significantly worse thanthe resolution of the deep image because of thedrastically reduced amount of data that wentinto it.The deep and fast radio images were exportedto CASA format for source detection and mod-eling. The source detected by the realfast sys-tem (using rfpipe ) is also detected in the burstimage. We fit an ellipse to that source to mea-sure the centroid location, peak flux density,and their 1 σ uncertainties (see Table 1). Thelocalization precision is approximately 1/10 th ofthe synthesized beam diameter, which is typicalfor sources of this significance observed with theVLA (Becker et al. 1995).We then searched the deep image to deter-mine whether there is persistent radio emissionassociated with the candidate FRB. We find nosuch associated persistent radio emission at thelocation of the candidate FRB, to a 3 σ limit of11 µ Jy (see Figure 3).
Figure 3.
Deep 1.4 GHz radio image of theFRB 180814 field with the location of FRB20190614D shown with white cross-hairs. Blackcontours show radio brightness levels of 25 and50 µ Jy. No persistent radio emission brighter than3 σ (11 µ Jy) is seen at the location of the new FRB.The noise level of this image is 3.6 µ Jy beam − , andthe beam shape is (3.6 (cid:48)(cid:48) , 2.8 (cid:48)(cid:48) , 78 ◦ ), marked by a yel-low ellipse in the bottom left corner of the image. We tested the astrometric precision by as-sociating compact radio sources with opticalsources in Pan-STARRS DR2 catalog (Cham-bers et al. 2016). We ran the aegean sourcefinding package (Hancock et al. 2018) and iden-tified 270 compact radio sources with a flux den-sity greater than 100 µ Jy ( > σ ). Of these,102 had optical counterparts within 3 (cid:48)(cid:48) andnDetections = 5. No systematic offset is foundbetween the radio and optical sources; the stan-dard deviation of the radio/optical offsets is0.2 (cid:48)(cid:48) . We note that given the resolution of theradio image (3.6 (cid:48)(cid:48) × (cid:48)(cid:48) ), we expect the astro-metric accuracy to be of the order of 0.1 (cid:48)(cid:48) forthese brighter sources (a few % of the synthe-sized beamwidth).2.6. CHIME/FRB Limits
The CHIME/FRB system, operating in itscommissioning phase, has observed the sky po-sition of FRB 20190614D for a total of 153 hoursduring the interval from 2018 August 28 to 2019
Time (MJD, @2.0 GHz) 58648.05071771R.A. (J2000) 4h20m18.13sDeclination (J2000) +73d42m24.3sR.A. (J2000, deg) 65.07552Declination (J2000, deg) 73.70674Centroid ellipse ( (cid:48)(cid:48) , (cid:48)(cid:48) , ◦ ) 0.8 (cid:48)(cid:48) , 0.4 (cid:48)(cid:48) , 67S/N ratio image obs (pc cm − ) 959.2 ± MW (pc cm − ) 83.5Peak flux density (mJy) 124 ± ± µ Jy/beam) < Table 1.
Measured properties of FRB 20190614Dwith 1 σ errors. The centroid ellipse is defined withthe major and minor axes and orientation (east ofnorth). Deep limit refers to the flux density limiton 1.4 GHz radio counterparts in a deep image ofthe FRB field. The Milky Way DM estimate iscalculated from Cordes & Lazio (2002). September 30. The large exposure is due tothe circumpolar nature of the source and issplit between 88 hours for the upper transitand 65 hours for the lower transit. The aver-age duration of the upper and lower transitsis 17 and 13 min, respectively, during whichthe source is within the FWHM region of thesynthesized beams at 600 MHz. We searchedthrough all low-significance events that weredetected by the CHIME/FRB system in theabove-mentioned observing time. No signifi-cant event or excess event rate was found tobe consistent with the location and DM of FRB20190614D, so there is no evidence for repeti-tion from this FRB.To determine CHIME/FRB sensitivity toFRB 20190614D, we follow the methods de-tailed in Josephy et al. (2019). The sensitivity ofCHIME/FRB varies with observing epoch, po-sition along transit, and a burst spectral shape.We used a Monte Carlo simulation with 10 re-alizations to generate fluence thresholds for dif-ferent detection scenarios within the quoted ex- C D F Figure 4.
Cumulative distribution of fluence de-tection thresholds for the CHIME/FRB instru-ment. Note that the FRB candidate is circumpo-lar and thus transits the CHIME FoV twice a day;thresholds shown here are valid for the upper tran-sit, whereas the lower transit is a factor of ∼ σ in 5 ms at 1.4 GHz). posure. These simulations define a set of rela-tive sensitivities, which are tied to a flux densityscale using beam-formed, bandpass-correctedobservations. As a reference, we use a burstfrom FRB 180814.J0422+73 detected on 2018November 11 with a S/N ratio of 9.8 σ , fluenceof 2 . ± . Optical Associations
We considered the significance of this candi-date high enough to trigger observations de-signed to find an optical counterpart. On UT2019 July 2, we observed the field surround-ing FRB 20190614D with the Gemini Multi-Object Spectrograph (GMOS) on the Gemini-N h m s s s Right Ascension (J2000) D e c li n a t i o n ( J ) Keck V-band
FRB 4 h m s s s Right Ascension (J2000)
Gemini r-band h m s s s Right Ascension (J2000)
Keck I-band AB Figure 5.
Cut-out images from Keck/LRIS and Gemini/GMOS centered on the candidate FRB. Thedashed line shows the 1 σ radio centroid region. Source A (brighter, to south) is red with brightest flux inthe I band. Source B (fainter, to north) is bluer with colors indicative of star formation. telescope. We obtained a series of 8 ×
300 simage exposures in the r -band. These datawere reduced with standard procedures usingthe Gemini’s pyraf package, and the imageswere registered using Pan-STARRS DR1 astro-metric standards (Flewelling et al. 2016). Weperformed photometry on these images using DoPhot (Schechter et al. 1993) and calibratedthe image using Pan-STARRS r -band calibra-tors.On UT 2019 September 25, we obtained aseries of 4 ×
600 s images with the Low Res-olution Imaging Spectrograph (LRIS) on theKeck I telescope in V and I bands. Thesedata were reduced using a custom-built pipelineused for transient searches and based on the photpipe imaging and reduction package (Restet al. 2005). Following standard procedures,we removed bias and flattened our images us-ing bias and dome flat-field exposures obtainedon the same night and in the same instrumen-tal configuration. We registered the imagesusing Pan-STARRS astrometric standards andcombined the individual exposures with SWarp (Bertin et al. 2002). We performed point-spreadfunction photometry on the final stacked im- ages with
DoPhot and calibrated these data us-ing Pan-STARRS grizy calibrators transformedto
V I using the bandpass transformations de-scribed in Tonry et al. (2012).On UT 2019 November 26, we obtained anadditional set of 18 ×
200 s z -band images ofthe FRB field with the Alhambra Faint ObjectSpectrograph and Camera (ALFOSC) on theNordic Optical Telescope (NOT). The imageswere processed with standard procedures andastrometrically-calibrated to the Gaia -DR2 ref-erence frame.On UT 2020 March 09, we also obtained a setof 4 ×
300 s images (each one coming from 5 ×
60 sco-adds) in the near-infrared J-band using theNear InfraRed Imager and spectrograph (NIRI;Hodapp et al. 2003) on the Gemini-N telescope.The images were reduced with standard proce-dures using the dragons package and wereastrometrically-calibrated to the Gaia -DR2 ref-erence frame. A photometric calibration wasderived using 2MASS sources in the image.Figure 5 shows the
V rI images centered onthe radio localization of the candidate FRB. https://dragons.readthedocs.io Å ]10 f ( J y ) A V r I z J
SED templateSED in filtersObserved flux0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 z phot p ( z ) PosteriorPriorz Estimate68% interval
Object A Å ]10 f ( J y ) B V r I z J
SED templateSED in filters3 Upper limitObserved flux0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 z phot p ( z ) PosteriorPriorz Estimate68% interval
Object B
Figure 6. (Top left) The photometric measurements of source A with best-fit model in blue. SED in filtersshows the best-fit template fluxes in each filter, with the black points showing the measured flux in thefilters. (Bottom left) The redshift posterior for source A as estimated by eazy . The red dashed line showsthe expectation value of the redshift over the posterior. With the pink shaded region marking the 16-84thpercentile range. The stated σ associated with z phot is half of the difference between the upper and lowerlimits shown above. (Right) Same left panels, but for source B. All optical images were registered in the Pan-STARRS DR1 astrometric frame, and so theuncertainty in their relative alignment is givenby the precision of the original alignment solu-tions. We estimate a registration precision of ≈ . (cid:48)(cid:48) (1 σ ) for each image.There are two optical sources that are plau-sibly associated with the radio source. Thebrighter is J042017.85+734222.8, referred to asSource A, and approximately 1 (cid:48)(cid:48) north of thatis J042017.86+734224.5, referred to as source B.The 1 σ radio localization region overlaps withsource B, but the 2 σ (90% confidence interval)radio localization region overlaps with source A.We consider both sources as potentially associ-ated with the event. Final V rIzJ photometryof the candidate FRB hosts was obtained usinga 1 (cid:48)(cid:48) aperture centered at the locations describedin Table 2 and corrected for Galactic extinction.With the photometry of the galaxies as in-puts, we have used the software package
Eazy (Brammer et al. 2008) to estimate photomet- ric redshifts for the two sources closest toFRB 20190614D. We find z phot = 0 . ± . z phot = 0 . ± .
17 (68% confidence interval) forsource B. Figure 6 shows the redshift posteriordistributions for sources A and B and their best-fitting templates. The best-fitting template forsource A is a relatively quiescent galaxy withweak emission features whereas source B, whichexhibits a bluer color, is best-fit with a star-forming template. The SED templates thatwere fitted to the data, agree well with the colordifference that we observe in the source. Withtesting multiple sets of SED templates, we con-sistently find source A to be quiescent, and sim-ilar in shape to what is shown above, as well assource B consistently being fit to star-forming,bluer templates. We also note that the u − r restframe colors from the CIGALE analysis de-tailed below, are consistent with the eazy out-puts.1On UT 2019 September 29, we obtained aseries of long-slit spectra (1 (cid:48)(cid:48) wide) of sourceA and B with LRIS configured to cover wave-lengths λ ≈ − λ ≈ − PypeIt software package (Prochaskaet al. 2019b). While we detect a very faint traceof continuum emission from source A, there isno obvious emission or absorption feature to es-tablish a spectroscopic redshift. This is con-sistent with it being an early-type galaxy withlow or negligible star-formation and correspond-ingly weak nebular emission. We did not iden-tify any significant flux from source B.To roughly estimate the stellar mass andrest-frame color of each candidate host galaxy,we performed a spectral energy distribution(SED) analysis of the measured photometry(Table 2). This analysis, using the CIGALEsoftware package (Noll et al. 2009), also requiresthe source redshift; we adopted the posterior-weighted photometric redshift from the eazy analysis (Table 2). For the SEDs constructedby CIGALE, we adopt a delayed-exponentialstar-formation history model with no late burstpopulation, a Chabrier initial mass function,and the Calzetti et al. (2000) dust extinctionmodel. Because of the applied extinction correc-tions, changes in these assumptions would pro-duce similar results. Consistent with the eazy analysis, the best-fitting SEDs were quiescentfor source A and star-forming for source B. InTable 2, we report estimates for the stellar massand rest-frame u − r color with the latter reflec-tive of the inferred star-forming properties ofeach galaxy. We caution that the stellar mass,especially, bears great uncertainty due to theuncertain redshifts of each source. DISCUSSION 3.1.
Joint Probability of Radio Candidate withOptical Association
The chance of randomly associating a point onthe sky with a galaxy has previously been stud-ied in the context of gamma-ray bursts. Thechance association has an empirically-definedfunctional form parameterized by a associa-tion tolerance and survey depth (Bloom et al.2002) . Following the same approach we esti-mate chance association probabilities of P ch , A =7 . × − and P ch , B = 6 . × − for the twogalaxies to be unrelated to FRB 20190614Dbased on the r -band detections of Galaxy A( r = 23 .
24 mag) and B ( r = 23 .
93 mag). Themaximum “search radius” r ch used for these es-timates take into account the half-light radii R / of the host candidates and the distanceto the galaxy centroids ( d gal , A = 2 . (cid:48)(cid:48) and d gal , B = 1 . (cid:48)(cid:48) ) as r ch = (cid:113) d + 4 R / . We alsouse the background flux as a limit on the pres-ence of a galaxy below the detection limit of r > . P ch , undet = 1 . × − . The probabil-ity that either source A and source B are unre-lated to the FRB is therefore small and the ex-pectation for an even fainter host galaxy coun-terpart within the error region is even smaller.We also used the methods of Eftekhari &Berger (2017) to calculate the chance associ-ation probabilities . Using this approach, weestimate the chance coincidence probability of P ch , A = 2 . × − and P ch , B = 2 . × − for Galaxy A and B, respectively. Although,Eftekhari & Berger (2017) followed a similarprocedure to Bloom et al. (2002), the estimatesusing their methods are smaller because they Eqs. 1-3 of Bloom et al. (2002) are implemented in https://github.com/FRBs/FRB. Implemented in https://github.com/KshitijAggarwal/casp (Aggarwal et al, in prep). Table 2.
Optical Candidates
Source A Source BQuantity Unit Value Error Value ErrorRA (J2000) deg 65.07380 0.00005 65.0745 0.0001Dec (J2000) deg 73.70636 0.00005 73.7068 0.0001 V mag 25.42 0.25 24.58 0.16 r mag 23.25 0.15 23.94 0.24 I mag 22.83 0.10 23.74 0.18 z mag 23.18 0.30 22.53 999. J mag 22.56 0.20 23.66 999. z phot M ∗ M (cid:12) u − r mag 2.1 0.8 Note —This AB photometry has been corrected for Galacticextinction. A 999.9 value for photometric error indicates a3 σ upper limit. Estimates for M ∗ and u − r are based onthe photometric redshift and bear great uncertainty. used a more recent estimate of r -band num-ber counts of galaxies (Driver et al. 2016) tocalculate the number density of galaxies aboveany given limiting magnitude. We use the moreconservative and more widely used estimatesobtained using the formalism of Bloom et al.(2002) for calculating the significance of theFRB candidate.Under the assumption that an FRB shouldreside in a galaxy, we can use the host galaxyassociation to improve the confidence in the sig-nificance of the candidate event. The radiosignal alone has been characterized by a SNRof 8.27 and a FAR of 4 × − hr − . If weassume that false positives are randomly dis-tributed in the field, then the association of theradio source to a host galaxy improves the con-fidence in the significance of the FRB candidateas FAR assoc = FAR · P ch , det . According to thisrelation, we find FAR assoc = 3 × − hr − .Given that the association of a false positivewith a host galaxy is unlikely, we conclude thatthe FRB candidate is an astrophysical event. We used the Transient Name Server to namethe event FRB 20190614D. This naming con-vention is consistent with a new standard de-veloped by several groups in the FRB commu-nity. The common convention used prior to thischange is suitable as a shorthand and is ”FRB190614”.3.2. FRB Host Galaxy and DM
The identification of a specific FRB hostgalaxy can be critical for both estimating thelikely host DM contribution to the total ob-served DM, and for identifying trends in FRBhost galaxy types and environments, which canin turn help discriminate between FRB originmodels. FRB 20190614D is offset from opticalcounterpart(s), the components of which appearto be galaxies that differ in their mass, color,and type.Given the total extent of the optical counter-parts and the color differences, it is most likelythat they are two distinct galaxies. However,the photometric redshifts of the two galaxies are See https://wis-tns.weizmann.ac.il/. z = 0 . (cid:48)(cid:48) ), the galaxy centers are separated by13.6 kpc in projection. Assuming that the twogalaxies are located at the same redshift, theyare likely an interacting pair, in which source Bmay be a star-forming dwarf satellite of sourceA. If instead we do not assume they are inter-acting, this projected separation can occurs bychance in galaxies of this magnitude about 10%of the time. In this case, one galaxy is the fore-ground object to the other.While the optical data do not directly indicatewhich galaxy might be in the foreground, ananalysis of dispersion does provide some hints.The net observed DM is a sum of contributionsfrom: the Milky Way’s interstellar medium andits halo, diffuse contributions from IGM plasma,any intervening galaxies and their related cir-cumgalactic media, any cluster plasma, and anyhost galaxy or FRB-engine contribution (Simhaet al. 2020) . Any distant contribution is cosmo-logically redshifted, causing the rest-frame DMcontribution to scale by (1+ z ) − (e. g. Yao et al.2017).Regarding local contributors to DM, for con-tribution from the interstellar medium of theMilky Way, we adopt the value of 83.5 pc cm − predicted by Cordes & Lazio (2002). Generallycontributions from Milky Way’s ionized haloare taken to 30-80 pc cm − ; here we adopt themodel of Prochaska & Zheng (2019), which pre-dicts a halo contribution of 64 pc cm − for thissightline. Given these local contributions, wearrive at a representative extragalactic contri-bution of DM x = 812 ±
25 pc cm − , which en-compasses all non-local contributions.We can use the scaling of DM with redshift(known as the Macquart relation; Macquartet al. 2020) to estimate a maximum possible red-shift for our FRB. To do this, we attribute allof DM x to an IGM that is devoid of cluster andgalaxy group halos. There are various models Figure 7.
The range of extragalactic DM contri-butions (predominantly from the IGM and galaxygroup halos) predicted by the model of Prochaska& Zheng (2019). The mean and 68% range of thedistribution is shown in red. The nominal value ofDM x = 812 pc cm − inferred for FRB 20190614Din Section 3.2 is shown here as a dotted black line. that predict the ionization and elemental make-up of the IGM as a function of redshift; mostof these provide results in the same range (e. g.Yao et al. 2017; Pol et al. 2019; Prochaska &Zheng 2019), predicting maximum redshifts inthe z = 1 . − . z = 0 . . Inthis formulation, the mean and 68% range ofthis distribution gives DM = 734 +42 − pc cm − .The estimated DM x for FRB 20190614D is not Model implemented in Prochaska et al. (2019) with cos-mological parameters described in Planck Collaborationet al. (2016) DM x probabilitylies at a higher value than the mean, howevereven if we use that as a reference point for in-dicating potential host contributions, this min-imal difference does not change the conclusionsor analysis presented below.There are a few uncertainties in the compar-ison of measured to expected DM. First, thisestimate ignores the host DM, both from thegalaxy halo and interstellar medium. Any suchcomponent would push the predicted distribu-tion to higher DM, making it more consistentwith the observed value. However, this cor-rection term is diminished from the rest framevalue by 1+ z c , where z c is the host redshift. For z = 0 .
6, a rest frame host contribution would beroughly 57 pc cm − to make the 68% intervalof the prediction consistent with the observedvalue. Second, the DM estimate for the MilkyWay tends to be underestimated because themodels do not include all small-scale contribu-tions (e. g. H α features) of roughly ∼ CONCLUSIONSWe present the discovery of FRB 20190614Dwith VLA/ realfast , the first FRB discoveredblindly via interferometric imaging. The real-fast system has a relatively high sensitivity andlocalization precision, which makes it possibleto identify distant FRBs and associate them tohost galaxies. We describe how the use of in-terferometric images enable simultaneous noiseestimates that are more robust than the tra-ditional time-local noise estimates. The radioevent significance is low, but we argue that thenature of the radio measurement, consideredwith its association to a pair of host galaxies,is consistent with an astrophysical origin.FRB 20190614D has the highest DM of anywell-localized FRB ( DM x ≈
812 pc cm − ) andis likely associated with a pair of host galaxiesthat are among the most distant hosts identified( z ∼ . ∼ erg Hz − ; the fluence, distance, and en-ergy make this a faint version of the populationtypically seen by the Parkes Observatory (Shan-non et al. 2018).The DM is somewhat larger than predictedat the distance of the host galaxies, which im-plies a modest contribution from the FRB en-vironment or intervening galaxy. The two as-sociated galaxies differ in their colors and stel-lar masses, which implies different environmentsfor the FRB. However, they are broadly consis-tent with Milky Way-like stellar masses and starformation rates, as has been identified in otherFRB associations (Bhandari et al. 2020).The realfast system continues to commensallysearch for FRBs and other fast transients dur-ing VLA observations. In the future, the systemwill transition to a community service mode, inwhich real-time alerts are distributed automat-ically.5ACKNOWLEDGMENTSWe thank the CHIME collaboration for sup-porting the analysis for this FRB and theNRAO for supporting realfast development andoperations. We recognize and acknowledge thevery significant cultural role and reverence thatthe summit of Maunakea has always had withinthe indigenous Hawaiian community. We aremost fortunate to have the opportunity to con-duct observations from this mountain.CJL acknowledges support under NSF grant2022546. BZ acknowledges the support of Frankand Peggy Taplin Membership Fund. On be-half of the F team, JXP, AM, NT, KEH,and SS acknowledge support under NSF grantsAST-1911140 and AST-1910471. SBS and KAacknowledge support by NSF grant 1714897.SBS is a CIFAR Azrieli Global Scholar inthe Gravity and the Extreme Universe pro-gram. NT acknowledges support by FONDE-CYT grant 11191217. The NANOGrav projectreceives support from National Science Founda-tion (NSF) Physics Frontier Center award num-ber 1430284. PC is supported by an FRQNTDoctoral Research Award.The National Radio Astronomy Observatoryis a facility of the National Science Foundationoperated under cooperative agreement by Asso-ciated Universities, Inc. Part of this researchwas carried out at the Jet Propulsion Labora-tory, California Institute of Technology, undera contract with the National Aeronautics andSpace Administration. The Pan-STARRS1 Sur-veys (PS1) and the PS1 public science archivehave been made possible through contributionsby the Institute for Astronomy, the Univer-sity of Hawaii, the Pan-STARRS Project Of-fice, the Max-Planck Society and its participat-ing institutes, the Max Planck Institute for As-tronomy, Heidelberg and the Max Planck Insti-tute for Extraterrestrial Physics, Garching, TheJohns Hopkins University, Durham University,the University of Edinburgh, the Queen’s Uni- versity Belfast, the Harvard-Smithsonian Cen-ter for Astrophysics, the Las Cumbres Obser-vatory Global Telescope Network Incorporated,the National Central University of Taiwan, theSpace Telescope Science Institute, the NationalAeronautics and Space Administration underGrant No. NNX08AR22G issued through thePlanetary Science Division of the NASA Sci-ence Mission Directorate, the National ScienceFoundation Grant No. AST-1238877, the Uni-versity of Maryland, Eotvos Lorand University(ELTE), the Los Alamos National Laboratory,and the Gordon and Betty Moore Foundation.Some/all of the data presented in this paperwere obtained from the Mikulski Archive forSpace Telescopes (MAST). STScI is operated bythe Association of Universities for Research inAstronomy, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data isprovided by the NASA Office of Space Sciencevia grant NNX13AC07G and by other grantsand contracts. This research has made use ofNASAs Astrophysics Data System.Based on observations obtained at the interna-tional Gemini Observatory, a program of NSFsOIR Lab, which is managed by the Associ-ation of Universities for Research in Astron-omy (AURA) under a cooperative agreementwith the National Science Foundation, on be-half of the Gemini Observatory partnership: theNational Science Foundation (United States),National Research Council (Canada), AgenciaNacional de Investigaci´on y Desarrollo (Chile),Ministerio de Ciencia, Tecnolog´ıa e Innovaci´on(Argentina), Minist´erio da Ciˆencia, Tecnologia,Inova¸c˜oes e Comunica¸c˜oes (Brazil), and KoreaAstronomy and Space Science Institute (Repub-lic of Korea). The Gemini GMOS and NIRIdata were obtained from programs GN-2019A-Q-107 and GN-2020A-FT-201 (PI Spolaor) re-spectively, and it were processed using the Gem-6ini’s pyraf and dragons packages respec-tively.Data were obtained at the W. M. Keck Obser-vatory, which is operated as a scientific partner-ship among Caltech, the University of Califor-nia, and the National Aeronautics and SpaceAdministration (NASA). The Keck Observa- tory was made possible by the generous financialsupport of the W. M. Keck Foundation. Facilities:
EVLA, Pan-STARRS, MAST,Keck, Gemini, NOT
Software: rfpipe (Law 2017), astropy (As-tropy Collaboration et al. 2013; Price-Whelanet al. 2018), aegean (Hancock et al. 2018), drag-ons , pypeit(Prochaska et al. 2019b), (Prochaskaet al. 2019)REFERENCES
Agarwal, D., Aggarwal, K., Burke-Spolaor, S.,Lorimer, D. R., & Garver-Daniels, N. 2019,arXiv e-prints, arXiv:1902.06343.https://arxiv.org/abs/1902.06343Astropy Collaboration, Robitaille, T. P., Tollerud,E. J., et al. 2013, A&A, 558, A33,doi: 10.1051/0004-6361/201322068Bannister, K. W., Deller, A. T., Phillips, C., et al.2019, Science, 365, 565,doi: 10.1126/science.aaw5903Bassa, C. G., Tendulkar, S. P., Adams, E. A. K.,et al. 2017, ApJ, 843, L8,doi: 10.3847/2041-8213/aa7a0cBecker, R. H., White, R. L., & Helfand, D. J.1995, ApJ, 450, 559, doi: 10.1086/176166Bertin, E., Mellier, Y., Radovich, M., et al. 2002,Astronomical Society of the Pacific ConferenceSeries, Vol. 281, The TERAPIX Pipeline, ed.D. A. Bohlender, D. Durand, & T. H. Handley,228Bhandari, S., Sadler, E. M., Prochaska, J. X.,et al. 2020, arXiv e-prints, arXiv:2005.13160.https://arxiv.org/abs/2005.13160Bloom, J. S., Kulkarni, S. R., & Djorgovski, S. G.2002, AJ, 123, 1111, doi: 10.1086/338893Brammer, G. B., van Dokkum, P. G., & Coppi, P.2008, ApJ, 686, 1503, doi: 10.1086/591786Burke-Spolaor, S., Bailes, M., Johnston, S., et al.2011, MNRAS, 416, 2465,doi: 10.1111/j.1365-2966.2011.18521.x Caleb, M., Stappers, B., Barr, E., et al. 2019, TheAstronomer’s Telegram, 13098, 1Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000,ApJ, 533, 682Chambers, K. C., Magnier, E. A., Metcalfe, N.,et al. 2016, ArXiv e-prints.https://arxiv.org/abs/1612.05560Champion, D. J., Petroff, E., Kramer, M., et al.2016, MNRAS, 460, L30,doi: 10.1093/mnrasl/slw069Chatterjee, S., Law, C. J., Wharton, R. S., et al.2017, Nature, 541, 58, doi: 10.1038/nature20797CHIME/FRB Collaboration, Amiri, M., Bandura,K., et al. 2018, ApJ, 863, 48,doi: 10.3847/1538-4357/aad188—. 2019, Nature, 566, 235,doi: 10.1038/s41586-018-0864-xCordes, J. M., & Chatterjee, S. 2019, ARA&A, 57,417, doi: 10.1146/annurev-astro-091918-104501Cordes, J. M., & Lazio, T. J. W. 2002, ArXivAstrophysics e-printsCordes, J. M., Wasserman, I., Hessels, J. W. T.,et al. 2017, ApJ, 842, 35,doi: 10.3847/1538-4357/aa74daCornwell, T., & Fomalont, E. B. 1999, inAstronomical Society of the Pacific ConferenceSeries, Vol. 180, Synthesis Imaging in RadioAstronomy II, ed. G. B. Taylor, C. L. Carilli, &R. A. Perley, 187–199Driver, S. P., Andrews, S. K., Davies, L. J., et al.2016, ApJ, 827, 108,doi: 10.3847/0004-637X/827/2/108Eftekhari, T., & Berger, E. 2017, ApJ, 849, 162,doi: 10.3847/1538-4357/aa90b9 Eftekhari, T., Berger, E., Williams, P. K. G., &Blanchard, P. K. 2018, ApJ, 860, 73,doi: 10.3847/1538-4357/aac270Flewelling, H. A., Magnier, E. A., Chambers,K. C., et al. 2016, arXiv e-prints,arXiv:1612.05243.https://arxiv.org/abs/1612.05243Gajjar, V., Siemion, A. P. V., Price, D. C., et al.2018, ApJ, 863, 2,doi: 10.3847/1538-4357/aad005Ginzburg, V. L. 1973, Nature, 246, 415,doi: 10.1038/246415a0Greisen, E. W. 2003, Astrophysics and SpaceScience Library, Vol. 285, AIPS, the VLA, andthe VLBA, ed. A. Heck, 109,doi: 10.1007/0-306-48080-8 7Hancock, P. J., Trott, C. M., & Hurley-Walker, N.2018, PASA, 35, e011, doi: 10.1017/pasa.2018.3Hodapp, K. W., Jensen, J. B., Irwin, E. M., et al.2003, PASP, 115, 1388, doi: 10.1086/379669Inoue, S. 2004, MNRAS, 348, 999,doi: 10.1111/j.1365-2966.2004.07359.xJosephy, A., Chawla, P., Fonseca, E., et al. 2019,ApJL, 882, L18, doi: 10.3847/2041-8213/ab2c00Kocz, J., Ravi, V., Catha, M., et al. 2019,MNRAS, 489, 919, doi: 10.1093/mnras/stz2219Law, C. J. 2017, rfpipe: Radio interferometrictransient search pipeline.http://ascl.net/1710.002Law, C. J., Abruzzo, M. W., Bassa, C. G., et al.2017, ApJ, 850, 76,doi: 10.3847/1538-4357/aa9700Law, C. J., Bower, G. C., Burke-Spolaor, S., et al.2018, The Astrophysical Journal SupplementSeries, 236, 8, doi: 10.3847/1538-4365/aab77bLawrence, E., Vander Wiel, S., Law, C., BurkeSpolaor, S., & Bower, G. C. 2017, AJ, 154, 117,doi: 10.3847/1538-3881/aa844eLorimer, D. R., Bailes, M., McLaughlin, M. A.,Narkevic, D. J., & Crawford, F. 2007, Science,318, 777, doi: 10.1126/science.1147532Lu, W., & Kumar, P. 2016, MNRAS, 461, L122,doi: 10.1093/mnrasl/slw113Macquart, J. P., Prochaska, J. X., McQuinn, M.,et al. 2020, arXiv e-prints, arXiv:2005.13161.https://arxiv.org/abs/2005.13161Marcote, B., Paragi, Z., Hessels, J. W. T., et al.2017, ApJL, 834, L8,doi: 10.3847/2041-8213/834/2/L8 Marcote, B., Nimmo, K., Hessels, J. W. T., et al.2020, Nature, 577, 190,doi: 10.1038/s41586-019-1866-zMasui, K., Lin, H.-H., Sievers, J., et al. 2015,Nature, 528, 523, doi: 10.1038/nature15769McMullin, J. P., Waters, B., Schiebel, D., Young,W., & Golap, K. 2007, in Astronomical Societyof the Pacific Conference Series, Vol. 376,Astronomical Data Analysis Software andSystems XVI, ed. R. A. Shaw, F. Hill, & D. J.Bell, 127Noll, S., Burgarella, D., Giovannoli, E., et al.2009, A&A, 507, 1793Oostrum, L. C., van Leeuwen, J., Attema, J., et al.2017, The Astronomer’s Telegram, 10693, 1Perley, R. 2016, Jansky Very Large ArrayPrimary Beam CharacteristicsPerley, R. A., & Butler, B. J. 2017, ApJS, 230, 7,doi: 10.3847/1538-4365/aa6df9Petroff, E., Hessels, J. W. T., & Lorimer, D. R.2019, A&A Rv, 27, 4,doi: 10.1007/s00159-019-0116-6Planck Collaboration, Ade, P. A. R., Aghanim,N., et al. 2016, A&A, 594, A13,doi: 10.1051/0004-6361/201525830Pokorny, M., Law, C. J., Bower, G. C., et al.2018, Journal of Astronomical Instrumentation,7, 1850005, doi: 10.1142/S2251171718500058Pol, N., Lam, M. T., McLaughlin, M. A., Lazio,T. J. W., & Cordes, J. M. 2019, ApJ, 886, 135,doi: 10.3847/1538-4357/ab4c2fPrice-Whelan, A. M., Sip˝ocz, B. M., G¨unther,H. M., et al. 2018, AJ, 156, 123,doi: 10.3847/1538-3881/aabc4fProchaska, J. X., Simha, S., Law, C., Tejos, N., &mneeleman. 2019, FRBs/FRB: First DOIrelease of this repository, v1.0.0, Zenodo,doi: 10.5281/zenodo.3403651Prochaska, J. X., & Zheng, Y. 2019, MNRAS, 485,648, doi: 10.1093/mnras/stz261Prochaska, J. X., Macquart, J.-P., McQuinn, M.,et al. 2019a, Science, 366, 231,doi: 10.1126/science.aay0073Prochaska, J. X., Hennawi, J., Cooke, R., et al.2019b, PypeIt: Python spectroscopic datareduction pipeline. http://ascl.net/1911.004Ravi, V. 2019a, Nature Astronomy, 3, 928,doi: 10.1038/s41550-019-0831-y—. 2019b, ApJ, 872, 88,doi: 10.3847/1538-4357/aafb30 Ravi, V., Catha, M., D’Addario, L., et al. 2019,Nature, 572, 352,doi: 10.1038/s41586-019-1389-7Rest, A., Stubbs, C., Becker, A. C., et al. 2005,ApJ, 634, 1103, doi: 10.1086/497060Schechter, P. L., Mateo, M., & Saha, A. 1993,PASP, 105, 1342, doi: 10.1086/133316Shannon, R. M., Macquart, J. P., Bannister,K. W., et al. 2018, Nature, 562, 386,doi: 10.1038/s41586-018-0588-ySimha, S., Burchett, J. N., Prochaska, J. X., et al.2020, arXiv e-prints, arXiv:2005.13157.https://arxiv.org/abs/2005.13157Spitler, L. G., Cordes, J. M., Hessels, J. W. T.,et al. 2014, ApJ, 790, 101,doi: 10.1088/0004-637X/790/2/101Spitler, L. G., Scholz, P., Hessels, J. W. T., et al.2016, Nature, 531, 202,doi: 10.1038/nature17168Tendulkar, S. P., Bassa, C. G., Cordes, J. M.,et al. 2017, ApJL, 834, L7,doi: 10.3847/2041-8213/834/2/L7 The CHIME/FRB Collaboration, :, Andersen,B. C., et al. 2019, arXiv e-prints,arXiv:1908.03507.https://arxiv.org/abs/1908.03507Thornton, D., Stappers, B., Bailes, M., et al. 2013,Science, 341, 53, doi: 10.1126/science.1236789Tonry, J. L., Stubbs, C. W., Lykke, K. R., et al.2012, ApJ, 750, 99,doi: 10.1088/0004-637X/750/2/99Vedantham, H. K., & Ravi, V. 2019, MNRAS,485, L78, doi: 10.1093/mnrasl/slz038Yao, J. M., Manchester, R. N., & Wang, N. 2017,ApJ, 835, 29, doi: 10.3847/1538-4357/835/1/29Zackay, B., Venumadhav, T., Roulet, J., Dai, L.,& Zaldarriaga, M. 2019, arXiv e-prints,arXiv:1908.05644.https://arxiv.org/abs/1908.05644Zhang, Y. G., Gajjar, V., Foster, G., et al. 2018,ApJ, 866, 149, doi: 10.3847/1538-4357/aadf31 A. ROBUST ESTIMATE OF EVENT SIGNIFICANCE WITH INTERFEROMETRIC IMAGESWe describe the detection of an event with SNR estimate on the border between statisticallysignificant and not. Therefore, we have to be very careful in assessing both it’s power and theexpected noise floor. A small change in the noise standard deviation estimate (and therefore theSNR) will dramatically change it’s probability of chance occurrence. This type of error, althoughrarely considered, can push the detection threshold for FRBs and pulsars up by as much as 10% intypical radio time-domain surveys. Correcting this type of error was shown to provide a substantialsensitivity improvement when applied to gravitational wave data in Zackay et al. (2019).The SNR is derived from a detection score that is a particular linear combination of the data: S ( α, δ ) = (cid:88) f,i,j,p G ( i, j ) V f,i,j,p e iφ f,i,j ( α,δ ) (A1)Where V f,i,j,p are the visibilities of antennae i, j at frequency f and polarization p , G i,j are theempirically measured gains and φ are the calculated phases using the position α, δ . To a very goodapproximation, under the noise hypothesis and assuming no significant RFI, the score follows aGaussian distribution. The tail of the Gaussian distribution is approximately proportional to P ( S > x ) ∝ e − x / (A2)For a Gaussian distribution, a 5% change in the noise estimate translates to a factor of 30 in thechance occurrence probability of a tail event above a threshold of S/N = 8. We therefore musthave a noise standard deviation estimate that is good to ≈ × independent measurements to average over.Obtaining such an accurate estimate is non-trivial in general. A common approach is to producea time-series of detection scores at a single direction in the sky (either a single-dish beam or phasedarray of an interferometer), and computing a running standard deviation to use. However, in orderto obtain 10 independent samples few seconds of data are required, and slow gain fluctuations wouldtypically bias the measurement. Our case was further complicated by the fact that the candidateFRB was discovered while the realfast system was turning on, so the number of recorded visibilitieschanges as a function of frequency/baseline/time. This precludes simple local noise estimates basedon neighboring time or frequency samples.We chose to use the image standard deviation as our noise estimate. To provide justification inusing this estimate and to assess it’s biases, we think of the phases in Eq. A1 as random, uncorrelatedvariables with uniform distribution in [0 , π ]. From symmetry arguments, this standard deviationwould depend only on the sum of the squared visibility amplitudes, or the total momentary powerregistered by all antennae . Since the FRB contributes (cid:28)
1% of the total power, it can be ignored.Therefore, the relative statistical error in the standard deviation is proportional to the inverse squareroot of the numbers of visibilities, which is smaller than 1% . According to Parseval’s theorem, the standard deviation of the image is exactly equal to the standard deviationestimate using the absolute value squared of the visibilities. N vis = N bl ∗ N ch ∗ N pol ∗ f recorded = 351 ∗ ∗ ∗ .
38 = 4 . e4