Daniel KocevskiKavli Institute for Particle Astrophysics and Cosmology
Stanford University
On behalf of the Fermi collaboration
On The Lack of LAT Detected GRBs
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Fermi GRB Detections
GBM Detected GRBs (until March 1st): 620 - Blue
GRBs in LAT FOV: 288 (46%) - Green
LAT Detected GRBs (>100 MeV): 23 (8%) - Red
LAT LLE Only Detected GRBs: 5 (2%)
Preliminary
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Expected Detection RateTake BATSE spectra, extrapolate and compare to actual detection rate
Predicted: 9.3 GRBs/year > 100 MeV
Observed: 8.0 GRBs/year > 100 MeV
This includes GRBs with extra components
We are seeing fewer GRBs then predicted, especially at GeV energies
Possible explanations
High energy emission is suppressed
Extrapolations are uncertain
Extra components must be rare!
PRELIMINARY
Omodei’s Presentation
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Spectral Fits
Fit NaI+BGO spectrum from 8 keV to 40 MeV in RMFIT
Estimate the expected flux in the 100 MeV to 10 GeV range
Compare upper limits to the expected LAT flux
LATGBM
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Spectroscopic SampleBright BGO Sample:
GRBs with 70 cts/s in BGO in LAT FOV: 92
“LAT Dark GRBs” (i.e. no LAT detection)
“Gold” Sample:
Number of bright BGO GRBs with ΔBeta < 0.5: 30
Expected LAT Flux
Extrapolate β to find expected LAT flux
We use the full covariance matrix to estimate beta error
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Expected Flux Comparisons
15 of the 30 GRBs have expected photon flux that exceed the T90 LAT photon flux upper limit
Same for the expected photon fluence and LAT fluence upper limit
10!5 10!4 10!3
LAT Flux Upper Limits 95% (photons cm!2 s!1) ! T100
10!7
10!6
10!5
10!4
10!3
10!2
10!1
Expe
cted L
AT F
lux (p
hoton
s cm!2
s!1)
0.0001 0.0010 0.0100 0.1000LAT Fluence Upper Limits 95% (photons cm!2)
10!5
10!4
10!3
10!2
10!1
100
Expe
cted
LAT
Fluen
ce (p
hoto
ns cm
!2)
Preliminary Preliminary
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Beta vs Ratio
High Energy Spectral Index ( )
Expecte
d L
AT
Flu
x / L
AT
Upper
Lim
it
GRBs with values of β > -2.2 typically exceed the LAT upper limits
Preliminary
Daniel Kocevski, Annapolis, Nov 1st-4th 2010
Lorentz Factor Distribution3 LAT detected bursts have Γmin > 800
For 6 LAT dark GRBs:
Δt ~ 0.01s and 1 < z < 5
If we assume Ec ~ 100 MeV
Γmax ~ 50-600
LAT bursts may represent the high end of the Γ distribution
LAT dark bursts may represent the low end of the Γ distribution
0 1 2 3 4 5Redshift
0
200
400
600
800
1000
1200
1400
! ("
max =
20 M
eV, t
var =
0.1s
)
090902B
090510
080916C
090926
LAT Detections !minLAT Non!detections !max
Preliminary
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Expected LAT Flux
0.1 1.0 10.0 100.0GBM Flux 20!2000 KeV (photons cm!2 s!1)
10!8
10!6
10!4
10!2
100
Expe
cted
LAT
Flux
0.1
!10
GeV
(pho
tons
cm!2
s!1)
Bright BGO SampleKaneko et al. 2006
Simulated BATSE Sample
LAT Detected Bursts
Simulate a population of GRBs using BATSE Epk, α, and β distributions
Roughly 65-75% of a simulated BATSE sample have expected flux values that exceed the median 30s LAT sensitivity
High energy extrapolations must be misleading in order to explain the number of LAT “dark” bursts
Preliminary
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
Joint GBM+LAT Spectral Fits
Very different beta value if we include LAT limits in the spectral fits.
For bright BGO sample, median β = -2.2 -> -2.5
Which fit is statistically preferred?
β = -1.8Cstat = 408.55
DOF = 370
β = -2.2Cstat = 416.52
DOF = 380
GBM Only Fit GBM + LAT Fit
GBM OnlyBand Fit
GBM+LATBand Fit
We cannot statistically compare these two scenarios using ∆C-Stat because we are using different data sets for the two fits
Model Comparisons?
beta = -2.10 beta = -2.54
GBM+LATBand + Step Fit
GBM+LAT FitBand + Cutoff
We have to compare the ∆C-stat values for the Band only, Band+Step Function, and Band+Cutoff fits to the same GBM+LAT data
beta ~-2.10beta = -2.10
Nested Model Comparisons
∆C-Stat
Band vs. Band + Step Function, change of 1 degree of freedom
Only 6 of 30 (20%) GRBs result in ΔC-Stat > 10
We can reject the null hypothesis (the Band model) only for these bursts
!5 0 5 10 15 20 C!Stat
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Band fit to GBM+LAT data results in softer beta values compared to fits to GBM data alone
Possibly consistent with suggestions by Hascoet (this conference)
!3.0 !2.8 !2.6 !2.4 !2.2 !2.0 !1.8 !1.6High Energy Spectral Index ( ! )
0
2
4
6
8
10Band: GBMBand: GBM+LAT
Preliminary
ΔC-Stat Correlations
Correlation between over-prediction of LAT flux and ΔC-Stat. Likewise, anti-correlation between σβ and ΔC-Stat
Statistical errors on β do not reflect the true, systematic, uncertainty in the parameter estimation
0.001 0.010 0.100 1.000 10.000 100.000 1000.000! C!Stat
0.001
0.010
0.100
1.000
10.000
100.000
1000.000
Expe
cted L
AT F
lux / L
AT U
pper
Limi
t
3" Fit Improvement
0.001 0.010 0.100 1.000 10.000 100.000 1000.000! C!Stat
0.01
0.10
1.00
High
Ene
rgy S
pectr
al Ind
ex U
ncer
tainty
(")
3# Fit Improvement
Preliminary
Preliminary
Daniel Kocevski - Fermi Symposium, May 9th-12th 2011
ConclusionsGBM to LAT extrapolations can be misleading!
Statistical uncertainties may not fully reflect the systematic uncertainties and cross-correlations among the spectral parameters
ΔC-stat for a nested model comparison is the proper method of distinguishing between fits of increasingly complexity
24 (80%) GRBs in our spectroscopic sample are consistent with having a steeper beta value
6 of 30 (20%) prefer a spectral break
Two of these bursts show this break in the LLE selection
Our previous estimates of the β distribution may be biased
Use of future LLE data may help distinguish between cutoffs and softer β
Simulation TestsGRB 101113483
GBM Only: β = -1.8
GBM+LAT: β = -2.2
GBM+LAT+Step: β = -1.8
ΔC-stat ~ 5
Simulated GRB: β = -1.8
GBM Only: β ~ -1.8
GBM+LAT: β ~ -2.2
GBM+LAT+Step: β = -1.8
ΔC-stat ~ 25Nested model comparison can distinguish the difference between the two scenarios, even though the different beta values are statistically excluded