1. Dave Tierney S. McBreen, R. Preece, G. Fitzpatrick and the GBM Team Low-Energy Spectral Deviations in a Sample of GBM GRBs DT acknowledges support from SFI under grant No. 09-RFP-AST-2400

2. Introduction Previous Work An X-ray excess of greater than 5 sigma above the Band model (~ 5 – 20 keV) was reported for ~14% of an 86 burst sample observed by BATSE (Preece et al 1996). Band Model Spectral model for fitting prompt GRB emission Consistent across GRBs Parameterised by , , E peak Empirical Model Additional components using Fermi Band+PLAbdo et al. 2009, Ackermann et al. 2010 Band+BBGuiriec et al. 2011, McGlynn et al. in prep Band+PL+BBGuiriec et al. in prep,

3. Fermi – GBM Key Advantages of GBM over BATSE Much higher data resolution Single detector from 8 – 1000 keV

4. Sample Selection Good In Sample Bad Not In Sample BlockagesFluence > 2x10 -5 erg / cm 2 10 -1000 keV (Paciesas et al. 2012) Detector Angle < 60 o

5. Analysing the Sample (Initial) 45 GRBs from the first 2 years Performed for time-integrated fitting only. Single Fit Select all good NaIs Select at least 1 BGO Perform a Band fit from 8 keV - 40 MeV Sum Low-Energy Residuals below 15, 20, 25, 30, 50, 100 keV

6. Select all good NaIs Select at least 1 BGO Perform a Band fit from LET* keV - 40 MeV Extrapolate function downwards to 8 keV Compare data to extrapolated function Sum Low-Energy Residuals between 8 keV and the LET Extrapolated Fit Analysing the Sample (Extended) 45 GRBs from the first 2 years Blind search using time-integrated, time-resolved spectral analysis. *LET = L ow- E nergy T hreshold. Selected to be 15, 20, 25, 30, 50 & 100 keV

7. Some cuts are applied to the results… E peak > 100 keV for LET = 15, 20, 25, 30 E peak > 300 keV for LET = 50, 100 Alpha_Err < 0.2 E peak _Err/E peak < 0.45 Time-Resolved (SN 50  ) Time-Integrated Distributions of Low-Energy Residuals GRB090902B

8. How to quantify the uncertainty... Simulations… Left: Combined distribution of 5 GRBs simulated with perfect Band model Right: Time-Integrated Data Distribution Simulations (Perfect Band)Sample Data

9. How to quantify the uncertainty... Simulations…Simulations… GRB080817.161 – No Deviations in the Time-Integrated GRB090902.462 – Strong Deviations in the Time-Integrated Distributions (Blue) showing when no deviations are present and a line (Red) showing the deviations in the data.

10. How to quantify the uncertainty... Simulations…Simulations…Simulations… GRB090926.181 – Strong Excess in the Time-Resolved (9 – 10 s) GRB090424.592 – Strong Deficit in the Time-Resolved (2 – 3 s) Distributions (Blue) showing when no deviations are present and a line (Red) showing the deviations in the data.

11. Low-Energy Excesses GRB090902B - PL (TI)GRB090926A – PL (TR)GRB090323 – BB (TR) Low-Energy Deficits GRB090424 – BB (TR)GRB090820 – BB (TR) Closer Analysis This method demonstrates the requirement for an extra component without any prior knowledge of the nature of the extra component.

12. Conclusions Excesses and Deficits can mean additional components… Excess tend to be an additional component dominant at low energies. Deficits tend to be an additional component dominant between the LET and E peak, forcing alpha higher. Systematic blind search shows that low-energy deviations are rarer than previously thought. 2% of my sample compared to 14% from BATSE (Time- Integrated). Additional components can become washed out with time-integrated spectral fitting. Time-resolved analysis is a must. There is more spectral curvature in some bursts than expected. This gives hints of extra-components BB, PL, other / new models.