ProbingtheFastRadioBurstPopulation(s?)withCHIMESHRIHARSH P.TENDULKAR

+CHIME-FRBTEAM

Canadian Hydrogen Intensity Mapping Experiment

2/26/18 SHRIHARSHTENDULKAR– FRB2018 1

2/26/18 SHRIHARSHTENDULKAR– FRB2018 2

CHIME/FRBCollaboration

UBCDr. Davor CubranicDeborah Good, PhD studentProf. Mark HalpernProf. Gary HinshawDr. Kiyo MasuiDr. Cherry NgProf. Kris SigurdsonProf. Ingrid Stairs

DRAODr. Tom LandeckerDr. Paul Scholz

Perimeter InstituteUtkarsh Giri, PhD studentDr. Dustin LangMasoud Ravandi, PhD studentProf. Kendrick Smith

WVUProf. Kevin Bandura

NRAODr. Paul DemorestDr. Scott Ransom

University of TorontoProf. Dick BondNolan Denman, PhD studentProf. Bryan GaenslerProf. Ue-Li PenAndre RecnikProf. Keith Vanderlinde

McGillDr. Michelle BoyceDr. Jojo BoyleShiny BrarProf. Matt DobbsDr. Emmanuel FonsecaProf. David HannaAlex Josephy, MSc studentProf. Vicky KaspiDr. Arun NaiduChitrang PatelZiggy Pleunis, PhD studentDr. Shriharsh Tendulkar

2/26/18 SHRIHARSHTENDULKAR– FRB2018 3

Arewefightingonehorse-sizedduckor

ahundredduck-sizedhorses?

FRBZoo(Repetition)Aretheretwodistinctpopulationsofrepeatersandsinglebursts?

2/26/18 SHRIHARSHTENDULKAR– FRB2018 4

Orisitasmoothdistribution?

Rateparameter

FRBpo

pulatio

n

RepeatersSinglebursts

Rateparameter

FRBpo

pulatio

n

RapidRepeatersRareRepeaters

Temporalclusteringmakesit*very*challengingtodrawconclusions(Opperman &Pen2017)

FRBdesert

Also,propagationaffectsdetectablerepetition.

FRBZoo(Polarization)Unpolarized/Lowpolarization

Stronglinearpolarization

Strongcircularpolarization

FRB140514— ~23%circ polzn(Petroff et al 2017)

FRB150215— 43%Polzn (Petroff et al 2017)

FRB150418— Lin polzn 8.5%(Keane et al 2016)

FRB150807— 80%Polzn (Ravi et al 2017)

FRB110523— 44%Polzn (Masui et al 2017)

FRB121102— 100%Polzn (Michilli etal2018)

2/26/18 SHRIHARSHTENDULKAR– FRB2018

Noinformationabouttherest

5

FRBZoo(RadioFrequency)MostFRBsurveyshavebeenatLband(1.4GHz)◦ Reasonablywell-determinedrate

4detectionsat~800MHz(GBT+UTMOST)

Butnodetectionsat350MHzwiththeGBNCCsurvey(Chawlaetal2017)

Isthisduetofree-freeabsorption,scatteringorisitintrinsic?

2/26/18 SHRIHARSHTENDULKAR– FRB2018 6

FRBZoo— Propagationeffects• Interveningmediumaffects• Intensity,• Scattering,hencedetectability• Alsopolarization

• Boostingduetodiffractivescintillation• Suppressionduetofree-freeabsorption

• Somestatisticalevidenceofhigherratesathighgalacticlatitudes(VanderWeiletal2016butseeMacquart+2018)

2/26/18 SHRIHARSHTENDULKAR– FRB2018 7

21&

10

Figure 10. Simulation of detections vs Galactic coordinates. Thecolor scale shows the number of sources detected out of a total of1.5 ⇥ 1010 sources. The luminosities and distances of the popula-tion are set up so that no sources are detectable in the absence ofISS. The results are for a power-law ↵ = 0 with cutoffs of 100 and750 Jy Mpc2 and a homogeneous population between 50 and 2000Mreshold pc. The survey threshold is 0.3 Jy and bandwidth aver-aging has been included over a 300 MHz bandwidth centered on1.5 GHz. Reduction of survey sensitivity by temporal broadeninghas not been included, so that dearth of detections at low latitudestoward the inner Galaxy is more severe than is shown.

Figure 11. Number of detected sources vs. Galactic latitude, ob-tained by summing the values shown in Figure 10 over longitude.

sampled from the population distribution is ⌘Tng , so themean number of detections is

Nd,s = ⌘TngPL(Ssd2/gng )id (24)

The relevant value of gng (and thus ng) needs to be solvedfor to match Nd,s to the actual number of survey detections.

Figure 12. (Top) Plot of the number of survey trials vs. scintillationgain gng for the ng largest values in a survey. Eq. ?? is evaluated forISS with an exponential PDF for Nd,s = 10 and for different valuesof ru = Su/Ss, the ratio of the largest possible unscintillated fluxdensity to the survey threshold Ss. Burst amplitudes have a power-law PDF with a dynamic range RS = 103 and a power-law index↵ = 5/2. (Bottom) Upper limits on ⌘T from the lack of multipleburst detections in a survey observation that yields one burst on aparticular source. Different curves correspond to different values ofthe single-source ru1 . The single-source ampitude PDF is a powerlaw with ↵1 = 2 and dynamic range RS1 = 100.

Using ng = NgPg(gng ), the number of sampled sources is

n⌦⌦s =Nd,s

⌘TPg(gng )PL(Ssd2/gng )id(25)

which is an implicit equation for n⌦⌦s because gng dependson it. We therefore investigate the dependence of populationparameters on gng . This allows us to identify the minimum

population size required to explain survey results.

6.4. Exponential DISS

Though DISS with an exponential probability Pg(g >gng ) = e�gng is not realistic for surveys with bandwidthaveraging, it illustrates some important trends.

Figure 12 (top panel) shows Nevents,s plotted vs. gng forRS = 103, ↵ = 5/2, and several values of ru. There is aminimum value of Nevents,s at gng given by solving g↵�1

ng�

(↵ � 1)g↵�2ng

= r�(↵�1)u (for ↵ 6= 1). For ↵ = 2, gng =

1 + 1/ru.The curves of N (s)

events increase with gng for large gng be-cause a large population is needed to make large gng likely.The curves also increase at smaller gng when gngru ! 1 be-cause scintillated flux densities are then just above threshold,requiring large ⌘T to make a detectable amplitude probable.Thus the rate factor ⌘T must be large when ISS is small butat large gng , the population size must be large.

The no-ISS case discussed in § 4 corresponds to points onthe curves where gng = 1. Of course for ru < 1 there is

Detec.on&of&FRBs&favored&at&high&la.tudes:&&DISS&is&quenched&by&bandwidth&averaging&and&possibly&sourceRsize&suppression&&Simula.ons&include&luminosity&func.on,&steady&burst&rate,&and&DISS&with&bandwidth&averaging&(but&point&sources,&no&pulse&broadening)&&No&FRBs&detectable&without&DISS&

13/6/17 FRB Scintillations/Lensing: Near and Far

FigurefromJimCordes(Cordes etal,inpreparation)

Whatdowewant?*1. RateofFRBs(withagoodunderstandingofcompletenessandsensitivity)

1. Asafunctionofgalacticlatitude2. Asafunctionoffluence (w/otelescopebeampattern)3. AsafunctionofDMexcess

4. Asafunctionoffrequency(say,400-600MHzvs600-800MHz)

2. Repetition!1. Arethereotherrepeaters?2. Dotheyclusteratlowgalacticlatitudes?3. Ratesandtemporalclusteringproperties

3. Spectral,polarizationandscatteringproperties1. Distributionsandcorrelations

2/26/18 SHRIHARSHTENDULKAR– FRB2018 8

*incompletelist

Howdowegetit?

1. Singlelargesurvey

2. Carefulinstrument+pipelinedesign

3. Sensitivitytestingandmonitoring

2/26/18 SHRIHARSHTENDULKAR– FRB2018 9

CHIME:CanadianHydrogenIntensityMappingExperimentTransittelescopedesignedtostudyBaryonAcousticOscillationsatz=0.8-2.5

Four20mx100mcylinders

256dual-polfeedsoneachcylinder

400-800MHz

FOV:E-W~2.5o–1.3o,N-S~120o

Beamsize0.5o–0.3o

CFIfundedFRBbackendforreal-timedetection:“CHIME/FRB”

2/26/18 SHRIHARSHTENDULKAR– FRB2018 10

≈220sq.deg.

CHIMEFRBRates

2/26/18 SHRIHARSHTENDULKAR– FRB2018 11

FromP.Chawla

BasedonLawrenceetal2017rates(5.9× 102 sky−1 day−1 @1.4GHz,1Jy)

CHIMEisexpectedtodetect0.6-11FRBsperday (above10-sigma)

(andmoreat8-sigma)

FRBsandCHIMEWhatwewant CanCHIMEdeliver?

Thousandsofeventsforeventrate,fluxdistribution,angulardistribution,DMdistribution,scatteringvs DM,…

Yes

Repeatedobservations Yes

Real-time triggers Yesà GCN,VO, ATel Digest

Sensitivity topolzn vsfreq,vs time Yes

Localization:

Absolutelynecessaryfordistinguishing models

Arcminutes:WithinCHIME(SNRdependent)

Arcseconds: Maybe,ifoptical/X-raycounterparts exist,arelong-lived&brightORVLBI(downtheroad,outriggers?)

2/26/18 SHRIHARSHTENDULKAR– FRB2018 12

2/26/18 SHRIHARSHTENDULKAR– FRB2018 13

CHIME

F-Engine F-Engine

Analogsignalchain

Digitizers+FPGAs

X-Engine X-EngineGPUcorrelator

CHIMECosmology CHIME-FRB CHIME-Pulsar

Eastandwestreceiverhuts

NorthandsouthGPUcans

1024freq channelsn2 visibilities30sec

16kfreq channels1024stationaryintensitybeams1-mssampling

1024freq channels10trackingvoltagebeams2.5-ussampling

13Tbps

130Gbps

• Cylinder focuses light only in EW direction

• Gives us large FOV2/26/18 SHRIHARSHTENDULKAR– FRB2018 14

• FFT telescope in NS direction

• 256 beams per cylinder2/26/18 SHRIHARSHTENDULKAR– FRB2018 15

\

• 1024 beams from full 4-cylinder CHIME

2/26/18 SHRIHARSHTENDULKAR– FRB2018 16

2/26/18 SHRIHARSHTENDULKAR– FRB2018 17

CHIME-FRBPipeline

2/26/18 SHRIHARSHTENDULKAR– FRB2018 18

L0node

L0 L1 L2 L3 L4FFTBeamforming1024intensitybeams

RFIremovalDedispersionMoreRFIrejection

Multi-beamgroupingMulti-beamRFIremovalLocalization

IdentificationActionchoice

DatabaseforheadersIntensitydataBasebanddataOfflineanalysis

L0node

L0node

L1Node

L1Node

L1Node

L2/L3Node Archiver+DB

Basebandcallback

Intensitycallback

128streams

64freq chanx8beams

Pulsartrackingbeam

Realtime (dispersionsweep+2-3seconds)AdaptedfromadiagrambyCherryNg

WorkbyC.Ng,K.Smith,M.Rafaiei,U.Giri,A.Josephy,SPT,P.Scholz,C.Patel,Z.Pleunis,E.Fonseca,S.Brar,P.Boyle,M.Boyce,V.Kaspi

Buffer Buffer

RFIexcision&filtering

ThreestagesofRFIdetection:Intensitydata,singlebeameventinformationandmulti-beaminformation

2/26/18 SHRIHARSHTENDULKAR– FRB2018 19

Example

• Next step is “sifting”: we decide whether an event is astrophysical or RFI, based on its profile in the (DM, time) plane.

• Currently a placeholder until initial filtering is finalized.

• We expect that the computational cost will be very small, of order 0.01 cores/beam.

L1B code

M.Rafaiei,U.Giri,K.Smith A.Josephy SPT

Verynon-analyticselectioneffects!

KeepingtabsoncompletenessAcarefulpopulationstudyneedsanunderstandingofsensitivityandcompleteness.

Wewantto:

1. Maintaindetailed configurationhistorydatabase(re-runnable)◦ Whichsoftwareversion,whichconfiguration(bothversionedwithgit)

2. Runinjectiontestsinrealdatafortestingcompleteness

3. Logallinstrument&dataqualityparameters◦ Trackexposure,sensitivity

2/26/18 SHRIHARSHTENDULKAR– FRB2018 20

LiveSensitivityTestingGoal

2/26/18 SHRIHARSHTENDULKAR– FRB2018 21

L1’ L2’ L3’

L1’Node

L1’NodeL2’/L3’Node

InjectionServer

Copyof16beams

InjectFRBsintotheintensitydatawithfullbeammodel.Includebeamsensitivitywithfrequency,positionetc.Varyfluence,width,specidx

Exactlythesameconfiguration

Detections

Injections

Getsensitivity+completenessmeasurementEvery~10min

L0node

L0 L1 L2 L3 L4

L0node

L1Node

L1NodeL2/L3Node Archiver+DB

1024beams

LoggingDataQuality&Health

2/26/18 SHRIHARSHTENDULKAR– FRB2018 22

L0node

L0 L1 L2 L3 L4

L0node

L1Node

L1NodeL2/L3Node Archiver+DB

1024beams

CentralLogger

#activefrequencychannels(sensitivity)#activefeeds(beamshape)ADCscaling(sensitivity)

#nodesalive(searchskyarea)L0-L1packetloss(sensitivity)RFImetrics(sensitivity,falsealarmrate)Spectrum,MaskFractionEvent/reject/acceptrate(sensitivity,sanitycheck)Processingdelays,latencies

L1nodedelays (sanitycheck)Eventrate (sanitycheck)Knownsourcesinthepasthour/day (sanitycheck)

L3/L4EventrateIntensitydatacallbackrateSuccessful/unsuccessfulactionsDiagnosticplots

MonitoringDashboards

Prometheus+Grafanabasedmonitoringdashboards• Fordataquality• Sanitychecks• Hardwarehealthchecks

2/26/18 SHRIHARSHTENDULKAR– FRB2018 23

CHIME-FRBPilot

• Commissioninga8-nodepilotclusterwith64beamsonsky

• Thoroughlytestthesystemarchitectureandsoftwarepipeline

• 64beamsà 0.06– 0.6FRBsperday(1-10per2weeks,atLawrenceetalrates)

• Nodes+storageetc installedonsite.• Datapathtestedwithincoherentbeam.

2/26/18 SHRIHARSHTENDULKAR– FRB2018 24

2/26/18 SHRIHARSHTENDULKAR– FRB2018 25

Crab

B0329+54

B0823+26

B1508+55

B2021+51

PRELIMINARY

52hourrunwiththeincoherentbeam~halftheCHIMEdesignbandwidth

B0329+54

PreliminaryRFIcleaning

Missingchannelsfilledwithmedianvalues

2/26/18 SHRIHARSHTENDULKAR– FRB2018 26

PRELIMINARY

CHIMEStatus

•Workinghardtodebugandcommissionphase-coherentbeams

• Phasecoherentsolutionsbeingtestedaswespeak

• Acquiringfull128-nodesystem

Staytuned!

2/26/18 SHRIHARSHTENDULKAR– FRB2018 27