1. Status of GRB Scaler Analysis David Williams Taylor Aune

2. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting2Contributors Work of Eitan Anzenberg –Senior Thesis (on Memo’s page) –Memo on pressure and temperature corrections Monte Carlo data for low energy (500MeV – 5GeV) generated by Vlasios Vasileiou GRB Catalog from Pablo Saz Parkinson

3. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting3 Scaler Analysis Nothing fancy - no individual events are reconstructed Record aggregate counts from each PMT each second Increased sensitivity to lower energy events (< 100 GeV) Spatial and spectral resolution of a source is impossible Check for an increase in counting rates above background coincident with satellite-detected GRBs

4. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting4 Time Resolution Whether two hits on same channel overlap depends on ToT –At threshold: ≥35 ns separation needed for low; ≥53 ns separation for high –Longer separations needed for larger pulses E.g at high threshold, ≥188 ns separation needed for two low threshold hits –Multiple hits on same channel not a bit effect –Only consider 0 or 1 hits per channel at each threshold Small underestimate of A eff –> higher flux limit Whether two hits on different channels in an OR’d group overlap is pulse height independent –≥25 ns separation needed for low –≥46 ns separation needed for high –Hits with less separation overlap and extend OR’d pulse –Sort list of hits within each group and determine number of resolved counts based on minimum separation

5. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting5 Effective Area Calculation Two Monte Carlo data sets: –Standard 5 GeV – 500 TeV, E -2 gamma-ray files –Special 500 MeV – 5 GeV, E -2 gamma-ray files Basically just counting hits –Modified version of CalibrateMC2.c Calculate effective area assuming all PMTs are live and used Calculate effective area in bins of  and interpolate linearly in log A eff vs. sec  for each burst Scale effective area down by fraction of PMTs which actually are live and used for a given burst Tubes excluded are both those known to be dead from the EMS data, and those that are members of the OR groups that were excluded earlier because of bad behavior

6. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting6 Fluence Upper Limits Assume an unabsorbed spectrum of dN/dE ~ E -2 from the source Calculate limits on the fluence (5 - 50 GeV): –For sources with estimated z, use that. –Otherwise use four generic redshifts: 0.1, 0.5, 1, 2 Take into account the absorption from the EBL (Primack ‘05) unabsorbed z = 0.1 z = 0.5 z = 1.0 z = 2.0

7. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting7 Fluence Upper Limits In the current GRB catalog, we have 17 bursts with “known” z. Of those, 11 have z<3. Fluence upper limits range from 6.9 x 10 -6 to 0.2 ergs/cm 2 These limits are comparable to those of ARGO-YBJ and those predicted to be obtained with Auger

8. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting8 Work to be done… Incorporate the low energy (500 MeV - 5 GeV) monte carlo data Clean up the code, potentially add a scaler specific histogram function to milinda Add the GRBs from the last few months Recalculate fluences using the low threshold muon layer - currently seems to have comparable sensitivity to the air shower

9. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting9 Lesson for HAWC #2 Muon layer effective area for gamma- rays is same as air shower layer –~2x effective area per PMT HAWC scaler sensitivity should not suffer from the tubes being deeper Preliminary

10. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting10 Flux Limits Still working on the details of the effective area calculations –Use EBL absorption for burst z, when known –Use EBL absorption for z=0.1, 0.5, 1.0 for bursts with no z determined –Need to stitch together results from the low energy and high energy MC samples Rough estimates suggest limits will be interesting: –~10 -6 ergs cm -2 s -1 –Good enough to have seen GRB 940217 –Better than recent ARGO-YBJ limits (for a different set of bursts) in astro-ph/0609317

11. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting11 PMT Groupings Broadly, the PMT counts are classified into six categories or arrays –Air Shower [High & Low] –Muon [High & Low] –Outrigger [High & Low] In each array, PMTs are further organized into logical OR groups –16 PMTs per group for ASHigh, Muon, and Outrigger arrays –ASLow: A,D are OR combined as are B,C 8 PMTs per group

12. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting12 Hardware Readout During normal operation, each OR-group is read out every second There are 133 OR-groups –AS High: 29AS Low: 58 –Mu High: 18Mu Low: 17 (last group just 737) –Out High: 11Out Low: 11 Data is collected in 16 second clock cycles within the OR groups To cut down on disk usage, individual tubes are read out once every 16 seconds, using a bit mask system –The individual tube rates from the bitmask are not used in this analysis

13. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting13 Example of Rate Change with Pressure Apply linear correction to rate — see Eitan Anzenberg’s July 4, 2006 memo Uncoupled linear corrections for any number of variables straightforward Corrections normalized to keep average rate unchanged

14. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting14 Pressure, Temperature curves

15. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting15 Lesson for HAWC #1 Air conditioning output air blows directly into racks –Large rapid temperature excursions –Seen in rates Don’t do this in HAWC Did not succeed in correcting for this with a linear correction (Did not need Milagro to teach us this lesson — objections to this design were ignored during Milagro construction)

16. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting16 Effect of P and T (outside) Correction

17. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting17 Burst Analysis, continued Bootstrap the significance and upper limits from the data Use 11 UT days of data centered on the burst Divide the data into test intervals, each with a Test Burst and corresponding background region –Determine the frequency of a test region yielding a given raw significance or more –Assign probability and significance based on that frequency For upper limits, determine how much extra signal (if any) must be added to the Test Burst to yield the same raw significance as the actual burst –99% c.l. upper limit is the amount of signal which pushes 99% of the test intervals to higher raw significance than the actual burst GRB BKG BKG i Test i BKG i+1 Test i+1...

18. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting18 Correlations & GRB 000607 GRB 000607 is the most significant burst in all four groups –0.12 s short burst –  = 42 0 The air shower and muon layer turn out to be highly correlated Cannot simply combine significances Need to look at this particular burst a bit more closely In general, calculate flux limit using each group and choose the best limit Raw signifance of test intervals Muon layer low vs. air shower low

19. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting19 Quality checks on test regions Repair some corrupted files by truncating corrupted part Exclude regions with bad GPS clock info Exclude from 1 minute before shutdown/missing data to 5 minutes after startup Lightcurves examined by hand for all bursts: exclude ~5 regions Require background rate before and after burst to be roughly consistent –Looking for gross outliers –Cut 1 per 1000 or less Same tests applied as well to GRB intervals, but all passed GRB BKG BKG i Test i BKG i+1 Test i+1...

20. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting20 Significance Distributions GRB 000607 GRB 000607 GRB 000607 GRB 000607

21. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting21 OR Channel Exclusion Faulty PMTs are excluded on a group by group and an event by event basis Steps for finding which groups to exclude: 1.Calculate the threshold RMS value for the entire array (the largest RMS that, when added still improves the signal-to-noise ratio) 2.Throw out all OR groups that have RMS values above the threshold RMS for the array 3.Repeat steps 1 and 2 until there are no OR groups with RMS values above threshold This excludes the noisiest tubes and increases the sensitivity

22. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting22 Results of OR channel exclusion

23. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting23 Burst Analysis Use 11 UT days of data centered on the burst Use background interval 10x the length of the burst, centered on burst –Calculate mean rate s and error on mean  s  during burst –Calculate mean rate b and error on mean  b in background region –Poisson statistics are not assumed –“Raw significance” = (s - b)/sqrt(  s 2 +  b 2 ) –Raw significance not normally distributed (see next slide) GRB BKG BKG i Test i BKG i+1 Test i+1...

24. April 21, 2007GRB Scaler Analysis — Milagro Collaboration Meeting24 Determining Significance To determine the significance: –The average rates and RMS of the signal and background intervals are calculated –The difference of the rates and error of the difference are then compared to the null hypothesis The significance of 38,000 random intervals was calculated and plotted –The resulting distribution should be Gaussian –There are systematic effects that prevent unbiased measurements