Fermi GBM Status, Results, Plans Colleen Wilson-Hodge GBM PI, NASA/MSFC

Fermi Users Group 18 August 2016

1

5138 triggers as of July 31, 2016 Gamma-ray bursts (GRBs): 1879 (triggered twice on each of four long GRBs) Soft gamma repeaters (SGRs) aka magnetars: 267 (from 6 sources) Terrestrial gamma flashes (TGFs): 686 triggered, ~5x more untriggered Solar Flares: 1121 Particles: 746 Others (galactic XRBs, accidental, uncertain): 435 (169 from V404 Cygni)

168 positive Autonomous Repoint Recommendations 2

3

In response to user requests, GBM GRB catalog is now updated within 1 hour, spectral information ~weekly

•  RoboBA–automa,cproduc,onanddistribu,onoffinalGRBlocaliza,ons(viaGCN)– No,cesdistributedwithin10minfor80%ofGRBs– No,fieshumanBAiffiEngfails– OnlyforGRBs.Futuremodplannedforon-groundtriggerclassifica,on

•  HourlyCTTEdatafiles– Currentlybeingproducedandtested.WillbedeliveredtoFSSCwhentes,ngiscomplete.

– Removes,mingglitchesandimproveslatencyforuntriggeredsearches.– Convenient:simplertopredictwhichfilecontainsanevent

•  PMTGainBalanceTest– PerformedonMarch3,2016,12:20UTto23:50UT– BGOdatawerenon-standardandthereforenotdeliveredtotheFSSC.– PMTsfoundtobebalancedforbothBGOdetectors,sonochangesneeded. 4

•  Under an MOU with the LIGO consortium, GBM has implemented searches of GBM data for short GRB as counterparts of candidate gravitational wave (GW) events –  A seeded search (Blackburn et al 2015, ApJS, 217,8) of

GBM CTIME data for prompt emission •  Uses count rate template matching, weighting 3 spectral models

folded through the detector response –  An unseeded search of CTTE data for sub-threshold short

GRBs •  http://gammaray.nsstc.nasa.gov/gbm/science/sgrb_search.html

–  A search for persistent emission using the Earth Occultation technique with CTIME data

5

SNR=40.7

!me

LIGOtrigger!me

GBMsearchwindow(~1m)

SNR=27.1

SNR=23.6

SNR=17.9

SNR=7.2

knowledgeofdetectorresponse

GBM

sGRBIndividualGBMdetectordata Likelihood-ra,ocharacterizes

eventasorigina,ngfrommodelvsnoisealone

slidingforegroundwindow(~1s)

predictedcountsdependon:amplitude,light-curve,spectrum,sourceposi!on,Earthposi!on

GBMSeededSearch

6

7

• Localization is performed by comparing the relative observed rates from the GRB in each detector to the expected rates from a 1 degree grid

• This requires an assumption of the spectrum, and the sky grid limits to a statistical minimum uncertainty of 1 degree radius.

NaI 0High Expected Rates

Low Rates— Spacecraft Blockage

FermiGBMlocaliza,on

7.44σ

4.60σ

4.31σ

3.18σ

2014-07-09 08:49:56.600 Found in 1.40s time binning 25 - 494 keV energy range P=7.75e-14

INTEGRAL ACS lightcurves

ACS native time bins

GBM timescale

8

GBMUnseededSearchsGRBCandidate

GW150914(Abbotetal.2016a)

•  BH+BHMerger•  36and29Msun•  410Mpc

– 32 –

Fig. 7.— Count rates detected as a function of time relative to the detection time of GW150914,

summed over all 14 GBM detectors. NaI data are summed over 50 - 980 keV and BGO data over

420 keV – 4.7 MeV. Time bins are 1.024 s wide, with the same time binning as in Figure 6, and

the red line indicates the background level. The blue lightcurve was constructed from CTTE data,

rebinned to optimize the signal-to-noise ratio. The signal-to-noise ratio for this lightcurve, summed

over all detectors in the selected energy range, is 6�. In the top panel, the 0.256 s CTIME binning

is overplotted on the 1.024 s lightcurve. The dip before the spike associated with GW150914-GBM

is not significant. Such dips are common in stretches of GBM data, as can be seen in the longer

stretch of data on the bottom panel. A 1600 s stretch of data centered on GW150914-GBM, with

1.024 s binning, shows 100 runs each of positive and negative dips lasting 3 s or longer relative

to a third-order polynomial fit background over the 1600 s time interval, with 55 (38) negative

(positive) excursions lasting 4 s or longer.

LVT151012(Abbocetal2016a)

•  CandidateBH+BH•  23and13Msun•  1100Mpc

GW151226(Abbocetal.2016b)

•  BH+BHMerger•  14and7.5Msun•  440Mpc

•  GW150914-GBM,a2.9σeventconsistentwithashortGRB•  Notpredictedbytheore,calmodels

•  Nogamma-raydetec,onsforLVT151012orGW151226–notconstraining•  32%and17%ofLIGOlocaliza,onregionblockedbyEarthforGBM•  Backgroundswere18%and3%higherinGBM•  DistanceforLVT151012was3xlarger•  Ifgamma-rayemissionisinajet,only15-30%wouldbepointedtowardEarth

•  Needmoreeventsbeforewecansaymore!

(Connaughtonetal2016) (Racusinetal2016)

GBMGBM GBM

9

•  LIGO’snextobservingrun(O2)expectedtobegininlateSeptember.Crea,ngautomatedsearchpipelinesforGBM.– Seededsearch

•  NowusesCTTEdatatoenableshorter,mescales•  Improvedbackgroundes,mate(unbinnedPoissonmaximumlikelihood)•  Jointsta,s,ctoaccountforspa,alcoincidence•  ReplacinghardtemplatewithComptonizedmodelwithindex-0.5andEpeak=1.5MeV

– Unseededsearch•  Automa,cevalua,onofbackgroundqualitytoreducelatency•  Produceno,cesforcommunica,onwithLIGO•  Inter-comparisonswithseededsearchtointernallyvalidatecandidates•  Addi,onalalgorithmstoincreasesensi,vity

– Tooltoinjectsimulatedsignalsintodata– DocumentourtechniquesonarXivbeforethestartofO2

10

•  TheFermi-GBMThree-yearX-rayBurstCatalog,P.Jenkeetal.2016,ApJ,826,22– 1084events,including752thermonuclearX-raybursts,267eventsfromaccre,onflaresandX-raypulses,and65untriggeredGRBs

•  The3rdFermiGBMGamma-rayBurstCatalog:TheFirstSixYears:N.Bhatetal2016,ApJS:1403bursts

•  TheFermiGBMGRB,me-resolvedspectralcatalog:thebrightestburstsinthefirst4years,H.Yuetal2016,A&A,588,A135:81bursts

•  FirstGBMTGFcatalog:includesGBMandWWLLNdata– hcp://fermi.gsfc.nasa.gov/ssc/data/access/gbm/tgf/– 3356TGFsfrom2008Jul11–2015June23;>80%untriggered

11

12

GW150914-GBM Localization

13

• GBM highest probability region coincides with LIGO highest probability region

• Some probability in the “North” - localization mirror point resulting from event underneath spacecraft

14

Spectrum & Fluence • Only used detector NaI 5 & BGO 0: smallest source angles - standard

GBM analysis • Background fit with a 1st-order polynomial using a 2-pass linear least-

squares minimization (background livetime is ~30 s for 1 s signal) • Spectral fitting is performed using a forward-folding Levenberg-

Marquardt algorithm, minimizing the Castor C-statistic (Poisson likelihood), assuming the background model variance is negligible compared to the Poisson rate variance.

• For each of the 10 points, perform a joint fit of a PL to the signal in the two detectors

• At each point, simulated 1e4 deviates of the fitted spectrum using the background livetime, signal livetime, and responses. These synthetic spectra were then fit, producing an estimate of the spectral PDF at each point.

• The spectral PDFs were marginalized over the sky points to produce the spectrum estimate over the LIGO arc.

15

GW150914-GBM Spectral Comparison

• All short GBM-triggered short GRBs fit with a PL compared to best fit with a PL

• Weaker short GRBs can only be fit by a PL because curvature cannot be constrained

• Almost all short GRBs where we can fit curvature have an exponential cutoff near the peak spectral density.

16

•  IndependentanalysisofGBMdataforGW150914-GBMfromwhichtheauthorsconcludetheeventwereportinVC+2016ismoreconsistentwithbackgroundthanwiththepresenceofasource.

•  Useronly2detectorstodeterminethesta,s,calsignificanceusingspectralanalysis,whichdoesnotchallengethesta,s,calsignificancereportedbyVC+2016,foundincountspacebycombiningcoherentlythedatafrom14detectorsinaseededsearch,basedonanempirically-derivedFAR

•  Incorrectandexcludedsinglesourceposi,onresultsinamuchhighereffec,veareainNaI5thanusedinVC+2016,resul,nginanoveres,mateofcountspredictedusingthefluencereportedinVC+2016.

•  Use128-channelTTEdata;VC+use8-channelCTIMEdata,coarserbinningwhichdoesnotsufferfromlow-countsta,s,calproblems.

•  LowerfluencewithMLEfitimpliesconsistencywithnon-detec,onbySPI-ACS

17

Asingleposi,onalongtheLIGOarcisusedtotestthedeconvolu,onfromrmfitandMLEfitagainsttheobservedcountratesinasingledetector.

18

Table1ofGreineretal.

Table2ofVC+

19

Purplepointassumessinglesourceposi,onexcludedbyGBMandjointGBM/LIGOlocaliza,on,with27degangletoNaI5vs~70degforfavored

sourceposi,on(i.e.,~3xmoreexpectedcounts).

Thiserrorappearstobepropagatedthroughouttherestoftheanalysis-usingtheamplitudefromVC+butforasourceatasingle

loca,on.

rmfitfitwith128channeldataconvolvedwithresponseforsame

incorrectsourceposi,on.

Fig5ofGreineretal.

However,thisposi,oninFig5isexcludedbytheGBMlocaliza,onandcontainsonly2%oftheLIGOprobability-completelyexcludedbyjoint

GBM-LIGOlocaliza,on

20

Thiserrorappearstobepropagatedthroughouttherestoftheanalysis-usingtheamplitudefromVC+butforasourceatasingleloca,on,excludedbytheGBMdetectorrates.

chi2minimumofGBMlocaliza,on

posi,onusedinGreineretal.

21

ExploringINTEGRALSPI-ACSnon-detec,on:power-lawfit(unphysical)

Fig12ofGreineretal.

MLEfitconsistentwithnon-detec,onofGW150914-GBMbyINTEGRALSPI-ACS(greencrosses)

rmfitfitto128-channeldataintensionwithnon-detec,onofGW150914-GBMby

INTEGRALSPI-ACS(redcrosses)(Valuesdon’tmatchTable1ofJG+)

(Ignoreboxes:theyareforunrelatedGRBs)

GBMandINTEGRALSPI-ACSteamsarecollabora,ngonjointanalysisofshortGRBstounderstandhowtheinstrumentscomplementoneanother.

•  GBMopera,onsandperformancearenominal– RoboBAhasimprovedlatencyfordistribu,onoflocaliza,onsformostGRBs

– TheBGOPMTbalancetestshowsthatthePMTsares,llverystableaver8years.

•  SearchesforLIGOEMcounterparts– AutomatedpipelinesandimprovementstoourseededandunseededsearchesarebeingimplementedanddocumentedforO2

– VCetal.paperaboutGW150914-GBMhasbeenpublishedinApJLecers.

– TwootherGWeventsinO1shownodetectablegamma-raysignal,butdonotconstrainGW150914-GBM

22