Gerry Share, Nicola Omodei, Kim Tolbert
A short description of the LAT instrument is provided
here; a detailed description is in
Atwood et al 2009; ApJ, 697, 1071.
We provide plots showing the flux of >100 MeV solar gamma rays since the launch of Fermi in 2008 calculated by two different methods, light-bucket and maximum-likelihood. Both methods use Pass 8 Source-class data which are well screened and are comprised of true external photons with very limited extraneous background. Both approaches accept all photons whose directions are within a given angle from the Sun, with some exceptions discussed below. The 'Light Bucket' method plots total fluxes (and statistical uncertainties) comprised of Galactic, extra-galactic, and solar photons. Thus these fluxes vary significantly during the year as the Sun moves along the Ecliptic. The fluxes are significantly higher around December and June when the Sun passes through the Galactic Plane. Any transient high-energy solar emission will appear above this 'quiescent' flux. In contrast the maximum-likelihood method typically plots upper limits to any high-energy solar flare gamma-ray emission and only plots fluxes and uncertainties when there is high confidence of a transient solar excess. Both methods are described in more detail below.
These plots are primarily meant to identify times when
high-energy solar events appear to have been detected. We strongly recommend
that any scientific studies be based on separate detailed analyses.
LIGHT-BUCKET PLOTS
The light-bucket plots were generated at GSFC (contact: Kim Tolbert) and are based on a procedure developed by Gerry Share to study the LAT >100 MeV time profiles. Gamma-ray background was reduced by restricting the allowable events to zenith angles <100 degrees and further reduced by restricting events to those photons having measured locations within 10 degrees of the Sun. Using Pass 8 Source-class data (V6 prior to 27-Oct-2013, Pass 7 Reprocessed prior to 30-Jun-2015, entire data set reprocessed with Pass 8 data in September 2015) and effective areas, the plots show the total flux in the selected region assuming the spectrum follows a power law with index -2. The >100 MeV flux varies significantly during the year as the Sun moves along the Ecliptic and passes into and out of the Galactic plane or passes a transient.
An automated search was done for high-energy solar transients by obtaining a mean flux over a four-day period and then removing intervals with fluxes differing from the mean by 3 standard deviations to create a modified mean flux. This modified mean flux is then compared with the observations, and intervals are identified where the fluxes deviate by more than 3 standard deviations. These points are shown in red on the plots, and are tabulated in a text file.
Key to light-bucket plots:
Blue: Fluxes and uncertainties
Red: Flux and uncertainty for points > 3 standard deviations above mean (many plots have no red points)
Cyan: 3 sigma uncertainties from the weighted mean
Black: Weighted mean (mean flux over 4-day period after rejecting outliers > 3 sigma from mean of all points)
Green: Weighted mean +- RMS
Magenta: Effective area averaged over the solar exposure for each flux value plotted, scale on the right axisThe width of the horizontal bar plotted for the flux and effective area shows the duration of the solar exposure used for each point.
Note: Any 1-minute data accumulation with a solar effective area less than 650 cm2 (500 cm2 prior to Pass 8) is not included in the plotted data points.. In addition, data points are not plotted for any solar observation period for which the total exposure (effective area x number of minutes in the exposure) is less than a threshold (shown on plots, usually set to 2.6 m2 min), or where the relative error (error/flux) is greater than a threshold (shown on plots, usually set to .8). These limits were applied to reduce scatter and uncertainties in the plots. For extremely large events (e.g. 7-mar-2012), this results in missing valid data points.
MAXIMUM-LIKELIHOOD PLOTS
The maximum-likelihood plots are based on data generated by the LAT team (contact: Nicola Omodei) and use well-studied techniques applied to LAT celestial studies. Pass 8 (Reprocessed Pass 7 prior to June 2015) Source-class data are accumulated during times with sufficient solar exposure. These are typically for 20-40 minutes every other orbit. Gamma-ray background from the Earth`s atmosphere is reduced by restricting the allowable events to zenith angles <100 degrees. A point source is placed at the location of the Sun and a power-law with an exponential cut-off is assumed as the best fit spectral model. All events with energies above 100 MeV with directions within 12 degrees of the Sun are included in the analysis. A solar flux (Test Statistic* > 20, plotted in red) or upper limit (95% Confidence Level, plotted in blue) is obtained using a maximum-likelihood analysis that compares the likelihood obtained by fitting the data with the solar source included with the likelihood of the null hypothesis (no solar source). Included in the study is a model with all the known LAT point sources (from the 2 FGL catalog (P. L. Nolan et al. 2012 ApJS 199 31, a model of the Galactic diffuse emission, and an isotropic background whose amplitude is allowed to vary (to account for any residual cosmic-ray background).
Key to maximum-likelihood plots:
Blue: Upper limit of flux
Red: Flux and uncertainty for points with TS>20 (many plots have no red points)
Magenta: Overall exposure (effective area x time), scale on the right axis
Green: Flux of the quiet Sun, 5.6 X 10-7 ph cm-2 s-1 (A. A. Abdo et al. 2011 ApJ 734 116)The width of the horizontal bar plotted for the flux and exposure shows the duration of the solar exposure used for each point.
* Test Statistic is the difference of the (-2) log likelihood of the two fits (null hypothesis and the model with the solar source). It derives from the Wilks' theorem which holds the (approximate in our case) relation that sqrt(TS) is distributed as a chi square with N degrees of freedom, where N is the number of additional parameters in the fit. Briefly: sqrt(TS=20) ~ 4 sigma
CAVEATS
The analysis is performed in a variable time window, depending on the apparent motion of the Sun in the LAT field of view. Because of the precession of the orbit and of the variation of the latitude of the Sun, the exposure in each time window can vary with time. This causes the value of the upper limits to oscillate in the maximum-likelihood plots.
During its operation, the Fermi satellite can execute Autonomous Repoint Requests (ARR) and Target Of Opportunity observations that can also impact the exposure of the Sun.
During bright impulsive flares, the high flux of X-rays can cause pile-up in the Anti Coincidence Detector (ACD), resulting in a high probability of misclassifying good photons as background. This effect has been described in detail in Ackermann et al. 2012 ApJ 745, 144. These data are flagged by the LAT team but are used by the SunMonitor (maximum-likelihood) in its analysis. The SunMonitor analysis is indicative in case of detection, however we discourage the use of the flux values for final analysis. We therefore strongly encourage to repeat the analysis in an optimized time window, excluding the data taken during bright X-ray impulsive flares.
Last updated September 29, 2015 by Kim Tolbert