Fermi Gamma-ray Space Telescope

Cicerone: Data — LAT Data Products

Overview

The LAT astrophysical data analysis that you will perform with the Fermitools begins with a list of counts that have been identified as resulting from astrophysical photons. The analysis requires information about where the LAT was pointing and what was the observing efficiency. Therefore you will extract and manipulate two types of FITS files: the 'photon file' and the 'spacecraft file.'

An additional 'extended file' is also available that provides additional photons with looser cuts as well as a few additional parameters that are not required for analysis with the Fermitools.

Event Reconstruction

These filetypes result from the processing of the data downlinked from the Fermi spacecraft (considered to be 'Level 0' data), and are therefore regarded as 'Level 1' data. For the LAT the Level 1 processing involves reconstructing the interaction of the event in the LAT from the 'hits' in the various parts of the LAT, identifying the type of event (e.g., astrophysical photon), and characterizing the event's relevant physical parameters (e.g., direction, energy). 'Hits' are the signals that result from the interaction of the event or its products with the various components of the LAT.

Over the course of the mission the event-level analysis software has been periodically updated to take advantage of improvements in the understanding of the LAT and its orbital environment. Since launch, there have been four major data releases (Passes): Pass 6, Pass 7, Pass 7 Reprocessed, and Pass 8. The latest Pass 8 release is the first since launch to introduce significant changes to the event-level reconstruction. Whereas previous passes were primarily focused on reducing systematic uncertainties in the instrument response functions, Pass 8 is a comprehensive revision of the entire analysis chain that yields substantial gains in instrument performance.

One of the the primary motivations for Pass 8 was to mitigate the effect of 'ghost' events, instrumental pile-up away from the gamma-ray shower that introduced errors in the measurement of the energy, and shower center and direction. Although originally motivated by the issue of ghost events, the new event reconstruction features improvements in many key areas beyond that. These include:

  • A new pattern recognition algorithm in the Tracker reconstruction that does not depend on Calorimeter information and is less sensitive to track confusion induced by the backsplash.
  • A clustering stage in the Calorimeter reconstruction aimed at finding and discarding the charge deposition due to ghost events.
  • A better energy reconstruction that improves the handling of energy leakage and crystal saturation.
  • A better algorithm for the association between tracks and ACD tiles for the rejection of charged particles.
  • A new event classification analysis based on boosted decision trees that improves the separation power between photon and cosmic-ray background events.

The instrument Monte Carlo has been revised based on the improved understanding of the instrument. From this, the LAT team has generated new event selection cuts that increase the acceptance over the entire LAT energy band. An important change to the event selection with respect to Pass 7 was the removal of a cut on the minimum energy deposited in the Calorimeter. This change significantly increases the acceptance at lower energies (< 100 MeV).

As a result of the improvements made to the event reconstruction and analysis, Pass 8 contains many more events than Pass 7 for a given time span. This increased effective area also corresponds to an increase in absolute background levels, but the signal-to-background ratios have been improved so that the point source sensitivity is improved over the whole LAT energy range. Details are available on the LAT Performance page.

The characterization of an event results in a set of ~1000 parameter values. Most of the events are not astrophysical photons, and most of the parameters describing an event are not relevant for the data analysis carried out by the Fermitools. Therefore, a small set of parameters for the counts considered astrophysical photons have been extracted from the event data to form the event file you will use for most astrophysical analysis. This optimal screened event list based on our current best knowledge of the instrument comprises our basic archived LAT data product.

Photon Classification

As part of the event reconstruction process, the LAT instrument team makes cuts that classify the events based on their photon probability and the quality of their reconstruction. These cuts are used to separate events into event classes with each class characterized by its own set of instrument response functions. Events within a class are subdivided into event types that use selections based on individual event topologies (for instance whether an event converted in the Front or Back section of the Tracker). The reconstruction methodology and the event class cuts have evolved and are likely to continue to do so.

The current event classes are a nested hierarchy in which the higher probability photon selections are subsets of the less restrictive selections. Higher probability photon selections have smaller effective areas, narrower point spread functions (PSF), and lower contamination of background events. In this set of Pass 8 data (called P8R3 by the LAT team) the background contamination in each class is calibrated to the best-fit power-law parameterization of the Isotropic Diffuse Gamma-Ray Background (IGRB) emission from Abdo et al. 2010 (hereafter A10). The loosest selection criteria (the TRANSIENT classes in Pass 8) are designed for short duration events, such as gamma-ray bursts, and timing studies that benefit from increased photon statistics while tolerating a higher background fraction and broader PSF. These classes have background fluxes that are generally equal to or greater than the IGRB and are denoted by their background level relative to the A10 reference spectrum (e.g. the background in TRANSIENT020 is 2x higher than the A10 reference spectrum). The cleaner photon selections (SOURCE through ULTRACLEANVETO/SOURCEVETO classes in Pass 8) provide lower background contamination at the expense of lower effective areas (particularly at low energies) and have background fluxes that are generally equal to or lower than the IGRB. An intermediate selection (SOURCE class in Pass 8) is most favorable for analysis of moderately extended sources and point sources on medium to long timescales. The most restrictive selections (in Pass 8: the ULTRACLEANVETO class at all energies or the SOURCEVETO class above a couple of GeV) are ideal for analysis of large regions that are more sensitive to spectral features caused by instrumental backgrounds.

In the Pass 8 data release the event classes are organized in three nested hierarchies: "Standard", "Extended", and "No-ACD". The Standard hierarchy contains all classes currently recommended for LAT analysis (see the Data Preparation page for the current data selection recommendations). The Extended hierarchy contains three TRANSIENT classes that are each supersets of the TRANSIENT classes in the Standard hierarchy (e.g. TRANSIENT020E is a supserset of TRANSIENT020). The Extended classes are defined with a less restrictive fiducial selection that accepts events with projected trajectories that do not pass through the Calorimeter. This selection improves the LAT effective area at low energies and high incidence angles (< 100 MeV and θ > 45 deg) but also slightly worsens the energy resolution. The No-ACD classes are defined using selections that exclude variables associated with the Anti-Coincidence Detector and are therefore less susceptible to X-ray pile-up activity which can occur during the impulsive phase of solar flares. However these classes generally have worse performance than the standard TRANSIENT selections when no pileup activity is present.

Table of event classes and event class hierarchies defined in the Pass 8 data release. Classes within a given hierarchy are nested, with one exception (SOURCEVETO is nested to SOURCE, but not to CLEAN, ULTRACLEAN and ULTRACLEANVETO). The "evclass" column indicates the parameter value that is used to apply the event class selection with gtselect. The columns "Photon File" and "Extended File" indicate whether a given class is available in the corresponding data file type. Event classes that are not in either the Extended or Photon files are not part of the standard public data products but will be made available for specific transient events on a case-by-case basis.

Standard Hierarchy for LAT Event Classes
Event Class evclass Photon File Extended File Description
P8R3_TRANSIENT020 16 X Transient event class with background rate equal to two times the A10 IGRB reference spectrum.
P8R3_TRANSIENT010 64 X Transient event class with background rate equal to one times the A10 IGRB reference spectrum.
P8R3_SOURCE 128 X X This event class has a residual background rate that is comparable to P7REP_SOURCE. This is the recommended class for most analyses and provides good sensitivity for analysis of point sources and moderately extended sources.
P8R3_CLEAN 256 X X This class is identical to SOURCE below 3 GeV. Above 3 GeV it has a 1.3-2 times lower background rate than SOURCE and is slightly more sensitive to hard spectrum sources at high galactic latitudes.
P8R3_ULTRACLEAN 512 X X This class has a background rate very similar to ULTRACLEANVETO.
P8R3_ULTRACLEANVETO 1024 X X This is the cleanest Pass 8 event class. Its background rate is 15-20% lower than the background rate of SOURCE class below 10 GeV, and 50% lower at 200 GeV. This class is recommended to check for CR-induced systematics as well as for studies of diffuse emission that require low levels of CR contamination.
P8R3_SOURCEVETO 2048 X X This class has the same background rate than the SOURCE class background rate up to 10 GeV but, above 50 GeV, its background rate is the same as the ULTRACLEANVETO one while having 15% more acceptance.
Extended Hierarchy
Event Class evclass Photon File Extended File Description
P8R3_TRANSIENT020E 8 X Extended version of the P8R3_TRANSIENT020 event class with a less restrictive fiducial cut on projected track length through the Calorimeter.
P8R3_TRANSIENT010E 32 X Extended version of the P8R3_TRANSIENT010 event class with a less restrictive fiducial cut on projected track length through the Calorimeter.
NON-ACD Hierarchy
Event Class evclass Photon File Extended File Description
P8R3_TRANSIENT015S 65536 X Transient event class designed for analysis of prompt solar flares in which pileup activity may be present. This class has a background rate equal to 1.5 times the A10 reference spectrum.

In previous LAT data releases, each event class was partitioned in two conversion event types (front and back), depending on the location of the Tracker layer where the photon-to-pair conversion occurred. Starting from the top of the instrument, the Tracker consists of 12 layers of 3% radiation length tungsten converters (front or thin section), followed by 4 layers of 18% radiation length tungsten converters (back or thick section). Photons that convert in the front section have intrinsically better angular resolution than those that convert in the back section. This is due to the fact that multiple-scattering is more likely to occur in thicker material. The conversion type partition was introduced to allow the front and back events to be treated separately each with their appropriate instrument response functions.

The number event class partitions has been expanded and generalized into the concept of event types. On top of the conversion type partition, Pass 8 introduces two new event type partitions:

  • PSF event type: using an event-level quantity indicating the quality of the reconstructed direction, the data is divided into quartiles, from the lowest quality quartile (PSF0) to the best quality quartile (PSF3)
  • EDISP event type: same as PSF except that the event-level quantity used to partition the data indicates the quality of the energy reconstruction.

It should be noted that the subdivision of the data into quartiles is only approximate and the relative fraction of events in each event type will depend on the instrument observing profile of a given source (the distribution of inclination angles at which a source is observed). The PSF and EDISP event type selections were specifically optimized to partition the acceptance of SOURCE class. The constant relative fraction of acceptance across the four types is thus only strictly maintained for this class.

In summary, each event class is partitioned in 3 ways:

  • FRONT/BACK
  • PSF0/PSF1/PSF2/PSF3
  • EDISP0/EDISP1/EDISP2/EDISP3

and the IRFs have been determined for only these event types. The LAT IRFs do not contain information regarding the cross-membership of events between partitions and therefore it is not possible to mix the event type selections. (I.e. there is no IRF corresponding to events that are only FRONT and PSF3, or to events that are only PSF3 and EDISP3). Event type selections that combine event types from more than one partition will generate an error in gtselect.

Table of event types and event type partitions defined in the P8R3 data release. The "evtype" column indicates the parameter value that is used to apply the event type selection with gtselect. Event types within a partition are mutually exclusive and therefore an event can only belong to one of the types in each partition.

Conversion Type Partition
Event Type evtype Description
FRONT 1 Events converting in the Front-section of the Tracker. Equivalent to convtype=0.
BACK 2 Events converting in the Back-section of the Tracker. Equivalent to convtype=1.
PSF Type Partition
Event Type evtype Description
PSF0 4 First (worst) quartile in the quality of the reconstructed direction.
PSF1 8 Second quartile in the quality of the reconstructed direction.
PSF2 16 Third quartile in the quality of the reconstructed direction.
PSF3 32 Fourth (best) quartile in the quality of the reconstructed direction.
EDISP Type Partition
Event Type evtype Description
EDISP0 64 First (worst) quartile in the quality of the reconstructed energy.
EDISP1 128 Second quartile in the quality of the reconstructed energy.
EDISP2 256 Third quartile in the quality of the reconstructed energy.
EDISP3 512 Fourth (best) quartile in the quality of the reconstructed energy.

Data File Contents

For each astrophysical photon the LAT photon file contains the information necessary for science analysis. This includes the energy of the event, the position, as well as information about the quality of the event reconstruction. The extended data file contains the same photon data, plus information helpful for understanding the results of the Level 1 analysis as well as additional photons that are in the TRANSIENT classes. However, this information is unlikely to be useful for the typical analysis. Finally, there is the spacecraft file that contains spacecraft position and orientation information for 30 second intervals. (Some intervals at the ends of a file may be shorter than 30 seconds.) A description of each column in these files has been provided.

Note that during the analysis process, additional information may be added to a FITS file, either as a keyword or an additional column, as long as its name differs from that of an existing type of data. Thus you might find additional information in some event files.

Other Files

You will need to use other files when analyzing the LAT data, some of which the tools access without your intervention, some of which are intermediate products of the analysis. Some examples are:

  • Instrument Response Functions (IRFs) — The LAT's response is characterized by a number of functions with empirically determined parameters. The Fermitools access files with the parameters appropriate for your analysis.
  • Diffuse Emission Models — The LAT detects point sources on top of the bright diffuse emission from the Galaxy and extragalactic sources. The LAT team has provided models of this diffuse emission.
  • Pulsar Ephemerides — The LAT team has provided a list of pulsar ephemerides that have been used in various publications.
  • Livetime Cubes — Over a specified time range, the livetime cube file provides the livetime for which the LAT observed a region of the sky at a given inclination angle (see the discussion in 'Precomputation of Likelihood Quantities').
  • Source Definition — Trial and fitted source models are stored in XML files (one of the few non-FITS files).
  • Binned Spectra — The LAT photon data can be binned into spectra stored in the common 'PHA' format. PHA files are most frequently used for gamma-ray burst analysis.
  • Bin Definitions — You can use FITS files to input the grids used to bin spectra in time and energy.
  • Response Matrices — To analyze binned spectra, the IRFs must be integrated over space and stored in the common 'RSP' format.

» Forward to GBM Data Products
» Back to Data Files
» Back to the beginning of the Data
» Back to the beginning of the Cicerone