Fermi Gamma-ray Space Telescope

Using LAT's New Pass 7 Data

The FSSC is now serving Pass 7 LAT data for analysis. The new version of LAT data provides a number of improvements over Pass 6, and is considered the best dataset for all types of LAT analysis. Pass 6 data will no longer be served. However, pre-existing Pass 6 data are archived. Here we discuss the changes and how they affect your analysis.

Changes to Data files

  • New data set
    As a result of the updated event classification, many (~30%) of the events in the Pass 7 data are new events that were not present in Pass 6. Some of the Pass 6 events have been reclassified as background, and are no longer in the event data. Pass 7 should be treated as a new LAT data set. (Related Cicerone discussion)
  • Photon vs. Extended Data
    In Pass 6, both photon and extended data files contained all events. However, with the increase in effective area at low energies, the Pass 7 Transient class is now significantly larger than the Pass 6 Transient class. As a result, Pass 7 Transient class events are only present in the Extended data files. Pass 7 Source class events and better (those used for most non-transient (>200 s) analyses) continue to be present in both file types.
  • Event Class now a bitfield
    In order to allow events to belong to multiple classes, the EVENT_CLASS data field is now a bitfield. The four main classes are: Transient (bit 0), Source (bit 2), Clean (bit 3), and Ultraclean (bit 4). Bit 1 is unused. As before, the analysis classes (Transient through Ultraclean) are hierarchical. (Related Cicerone discussion)
  • New diffuse responses
    The precalculated difuse response columns have been populated with values for the Pass 7 Source and Pass 7 Clean event classes. These precalculated diffuse responses were generated using the new Pass 7 Galactic diffuse model and class-specific isotropic models (described below). If you choose to use any other diffuse models, you must use gtdiffrsp to calculate the diffuse reponse of those model prior to performing an unbinned likelihood analysis. This includes when you are using one of the 12 individual extended sources included in the 2FGL catalog.
  • New weekly files
    Weekly files for the Pass 7 data contain the Source, Clean, and dataclean event classes. Transient events are only available from the LAT Data Server. Pass 7 weekly data and spacecraft files are available at the ftp site, and can be retrieved with wget:

    wget -m -P . -nH --cut-dirs=4 -np -e robots=off http://fermi.gsfc.nasa.gov/FTP/fermi/data/lat/weekly/photon/
    wget -m -P . -nH --cut-dirs=4 -np -e robots=off http://fermi.gsfc.nasa.gov/FTP/fermi/data/lat/weekly/spacecraft/

    Pass 6 weekly files created prior to August 1, 2011 have been archived and are available via ftp:
    Pass 6 Version 11: http://heasarc.gsfc.nasa.gov/FTP/fermi/data/lat/weekly/p6v11/
    Pass 6 Version 3: http://heasarc.gsfc.nasa.gov/FTP/fermi/data/lat/weekly/p6v3/

    No new data (acquired since August 1, 2011) will be processed using the Pass 6 analysis.

Changes to the Instrument Response Functions

  • Increased Effective Area at low energies
    Recovering effective area at low energies was a priority in the Pass 7 event classification. Biases, for example in the exposure calculation, caused by the energy dispersion in Pass 6 became very apparent in energy ranges where the effective area is changing rapidly (below a few hundred MeV). For Pass 7, the increase in effective area helped, but did not eliminate the behavior. Energy dispersion effects still should be considered at low energies. (Related Cicerone discussion)
  • Introduction of phi-dependence
    Dependence of the effective area on the photon azimuthal angle (phi) was first introduced in the P6_V11 IRFs. When the livetime cube is generated using a specified number of phi bins, the exposure calculation (e.g., in gtexpmap or gtexcube2) will reflect this. Otherwise, the average aziumuthal response will be used. (Related Cicerone discussion)
  • Use of on-orbit determined point spread function
    The instrument PSF was characterized post-launch by using a population of AGN and is included in the Pass 7 IRFs. To accommodate low statistics in the measurement of the PSF, there is no longer dependence of the PSF on inclination angle (theta) and instead a description representing the two-year average is given. This does not apply to Transient class, which still uses the pre-launch PSF formulation. The use of the theta-averaged PSF is only a concern in analyses for which the source is not observed over a large range of inclination angles. The worst cases are generally GRBs. (Related Cicerone discussion)

Changes to Background Models

  • Galactic diffuse model registration
    The pixel size is now 0.125° and has been adjusted to place a row centered at |b|=0°.
  • Galactic diffuse model components
    A new HI map that does not include the Magellanic stream was used to generate the model. In addition, several non-pointlike structures seen in the all-sky residuals were removed through the used of dedicated templates. These structures include: Loop 1 north polar spur, Loop 1 excess, high-energy galactic lobes, and three prominent Galactic plane excesses. As a result, the described features may disappear from the dataset when this model is used. (Full description of Galactic diffuse model)
  • Isotropic diffuse model
    The isotropic diffuse model is always paired with a specific Galactic diffuse model and a specific data set. Once the Galactic model changed, it was necessary to redefine the isotropic model for the Pass 7 class selections using that new Galactic model. You should not use this model with Pass 6 data, nor with a different Galactic diffuse model. And you should not use Pass 6 models with Pass 7 data.
  • As before, these models are designed specifically for point and small-scale extended source analysis. They should not be used for large-scale or all-sky analysis.

Changes to Science Tools

  • Science Tools Version v9r23p1
    A new version of the Fermi Science Tools has been released and is available for download. You will need these tools to be able to properly analyze Pass 7 data.
  • gtselect accepts bit value for event class
    gtselect now includes a new parameter, evclass, which is used to specify the bit number corresponding to the event class for the data you wish to analyze. gtselect is also backwards compatible and will accept evclsmin and evclsmax for use when analyzing previously downloaded Pass 6 data.
  • Binned Likelihood process
    A new tool, gtexpcube2 has been introduced to generate binned exposure maps for use in binned likelihood analysis. This tool replaces the on-the-fly generation of an exposure map by gtsrcmaps. In addition, gtsrcmaps now requires that you account for exposure out to 10° beyond your source region (due to the large PSF at low energies). Practically speaking, this means generating an exposure map with gtexpcube2 that exceeds the boundaries of the counts cube used in your analysis by 10° on all sides. (see the Binned Likelihood analysis thread)

Changes to Recommended Data Selections

  • Recommended event class for source analysis has changed to "Pass 7 Source" class (evclass=2). (Related LAT Caveat)

  • Recommended zenith angle cut to eliminate Earth limb events ("zmax") has changed to 100. (Related LAT Caveat)

  • Recommended source list for analysis is now the 2FGL Catalog. The new source list has several important new features that could affect your analysis:
    1. Fit to Pass 7:
      This catalog was generated from LAT Pass 7 data, and so will better represent the sources seen in that data set. An updated tool (make2GFLxml.py) to generate XML model files from this source list is available at the User-contributed Tools page.
    2. Extended sources:
      Twelve of the sources in the catalog are not point sources, but rather are represented by an extended spatial template. In 1FGL, some these sources were represented by multiple point sources. This effect has been reduced by fitting an extended template to the data. If your model includes an extended source, you will need to run gtdiffrsp on it before preforming an unbinned likelihood analysis.
    3. Multiple spectral models:
      The 2FGL catalog employs three spectral models to represent the various spectral forms seen in the sources. Known gamma-ray pulsars are defined as havng a power-law with exponential cutoff spectral shape. Other sources that show spectral curvature use a log-parabola spectral form. The remainder of the sources are fitted with a power-law.

Concerns for Analysis

  • All LAT data have been reprocessed using the new Pass 7 event classification. This means that if you downloaded data prior to the Pass 7 release, it is no longer the recommended data set for LAT analysis. If you are starting a new analysis, we recommend that you download new data prior to starting your analysis. However, this change does not invalidate work already completed on Pass 6 data.

  • The change to the event class designation means that all scripts using gtselect will need to be modified to remove the evclsmin/evclsmax parameters and replace them with evclass. In addition, you should verify that your parameter file does not hold old values for evclsmin/evclsmax by running the command:

    > punlearn gtselect

    This will reset the parameter file values to the original defaults that were provided with your Fermi Science Tools installation.

  • The change to the event selection means the recommended event class for most analyses is now Pass 7 Source (evclass=2). Previously, the Pass 6 Diffuse class (event classes 3 & 4) was the recommendation.

  • Changing the PSF has had several effects. First, the on-orbit PSF has tails at high energy that are significantly larger than the pre-launch Monte Carlo-based PSF. This may impact source localization. However, if there is an effect it is small and hard to measure. Second: while removing the theta dependence for the PSF will have little effect for long integration times (where the theta value for a given sources averages out over time), small integration times (such as those used when generating light curves) may be affected. To determine the level of the effect, it is advisable to check the distribution of your events in bins of theta.

  • Changing the isotropic and Galactic diffuse models means you will need to change the information contained in any XML model files before using them with Pass 7 data. The diffuse models for Pass 7 are available for download.

  • As a result of the change to gtsrcmaps, your exposure cube must cover a larger area than was previously required. Please review the Binned Likelihood analysis thread for details on how to determine the proper inputs for both gtexpcube2 and gtsrcmaps.

Still analyzing Pass 6?

If you need access to the previous set of documentation and analysis threads appropriate for Pass 6 data analysis, click here.