Data Set Overview : This dataset contains calibrated, 1.05- to 4.8-micron spectral images of comet 103/P Hartley 2 acquired by the High Resolution Infrared Spectrometer (HRII) from 01 October through 26 November 2010 during the Hartley 2 encounter phase of the EPOXI mission. Initial results based on these data are discussed by A'Hearn, et al. (2011) [AHEARNETAL2011].  Version 3 includes a change in how the dark subtraction was implemented as compared to Version 2. A scaled master dark subtraction was applied to DOY 307 and an in scene dark subtraction was applied to DOY 311-313, as well as the use of an average per scan optical bench temperature in the pipeline processing. These improvements are described in the EPOXI Calibration Pipeline Summary document in this dataset.  Version 3 also includes the calibration enhancements implemented in Version 2 of this dataset: a new per-pixel linearity correction treatment and its propagation through the calibration steps (i.e., bad-pixel maps, flat-field file update, revised spectral calibration curve), new mode-dependent master darks, an optimized scaling factor for the master dark, and a refinement in the absolute spectral calibration curve.  The following list summarizes the comet observations in this dataset. Descriptive text for each activity is included below. Additionally, the HRII Hartley 2 Flyby (E-18 hours to E+2 days) Log in the DOCUMENT directory provides notes about each scan, such as frames containing the comet, that were recorded by the science team as the data arrived on the ground.  -------------------------------------------------------------------------- Mid-Obs Exposure IDs Mission Activity Date/DOY Min Max (E : Encounter) -------------- ------- ------- ---------------------------------------- 2010-10-01/274 4000001 4000063 Approach imaging E-34 to E-8 days; to Coma scans every 30 minutes; 2010-10-27/300 Odd ExpIDs only, most repeated daily*  2010-10-28/301 4000001 4000063 Approach imaging E-8 days to E-18 hours; to Coma scans every hour; 2010-11-03/307 Odd ExpIDs only, most repeated daily**  2010-11-03/307 4000001 4000604 Flyby imaging E-18 to E-3 hours; to Scans every two hours; then reduced to 2010-11-04/308 one hour cadence ExpIDs are not repeated  2010-11-04/308 5000000 5008002 Flyby imaging E-2 to E+1.5 hours; Near-continuous scans of coma/nucleus; Nadir imaging at closest approach; ExpIDs are not repeated  2010-11-04/308 4000700 4010100 Flyby imaging E+2 to hours E+2 days; to Coma scans every 30 minutes; 2010-11-06/310 ExpIDs are not repeated  2010-11-06/310 4100001 4500023 Departure imaging E+2 to E+12 days; to Coma scans every ~15 minutes; 2010-11-16/320 Odd ExpIDs only; repeated daily  2010-11-16/320 4100001 4500011 Departure imaging E+12 to E+21 days; to Coma scans every 30 minutes; 2010-11-26/330 Odd ExpIDs only; repeated daily -------------------------------------------------------------------------- * Data acquired on 06 Oct were never downlinked due to a pointing problem with the Deep Space Network (DSN). ** First cycle of 28 Oct, only six hourly scans were taken before the sequence stopped and restarted at 'dosido' section for 7 hourly scans. Therefore the middle 11 scans were not acquired on purpose. Some frames in a scan may be missing because they were scheduled for transmission after the HGA was turned off, i.e. loss of signal. Any data the DSN received after this were considered extra credit.  Hartley 2 Approach Imaging, E-34 to E-8 Days (Start HRII): From 01 to 28 October 2010, MRI and HRIV imaged 103P/Hartley 2 about every 5 minutes while the HRII spectrometer scanned for outbursts once every 30 minutes. The instruments observed the comet for 16 hours per day allowing for 8 hours of downlinking; the same sequence was repeated daily yielding one full cycle per day. Data from the 6th cycle on 06 October 2010 (DOY 279) were never downlinked because of a pointing problem with the Deep Space Network. Those data had to be erased on board the spacecraft to make room for the next daily cycle and could not be recovered.  Hartley 2 Approach Imaging, E-8 Days to E-18 Hours: From 28 October to 03 November 2010, the MRI and HRIV imaged 103P/Hartley 2 continuously and HRII scanned the comet about every hour for 16 hours per day allowing for 8 hours of downlinking punctuated by hourly maneuvers, called dosido, to observe the comet. During this imaging phase there was only a single downlink of all images with zero margin; thus some images were occasionally lost as expected. The first cycle (DOY 300/301) was abbreviated such that the first comet-imaging session was only 6-hours long, followed by the standard 8-hour dosido.  Hartley 2 Encounter Imaging, E-18 hours to E+2 Days: From 03 to 06 November 2010, the HRII, HRIV, and MRI performed high resolution encounter imaging of 103P/Hartley2. The HRIV and MRI instruments began sampling about once every two hours until one hour before encounter when the cadence changed to once every 15 minutes. At E-30 minutes the instruments began continuously imaging of the comet. At E+30 minutes simultaneous observing and data playback began with samples being taken every 30 minutes. During the encounter imaging period, HRII infrared scans occurred every two hours until four hours prior to encounter when the cadence increased to hourly then more frequently one hour before closest approach. About one hour after closest approach, regular infrared sampling at 30-minute intervals resumed.  During this period, HRII obtained several full spectral maps of the coma with a scale < 250 m/pixel (the exposure IDs are provided):  5001000 (E-14 min, ALTFF/Mode5 scan at ~115 m/pixel) 5001002 (E-7 min, ALTFF/Mode5 scan at ~60 m/pixel 5006000 (E+7 min, ALTFF/Mode5 scan at ~50 m/pixel) 5007000 (E+14 min, ALTFF/Mode5 scan at ~115 m/pixel)  Also HRII obtained a full spectral map of the nucleus with a scale < 100 m/pixel:  5005001 (E+3 min, BINSF2/Mode3 scan at ~30 m/pixel)  Please note the comet is in fewer frames than expected at closest approach because there was an error in how the spacecraft was commanded to point during closest approach. However this unexpected offset enabled serendipitous imaging of cometary debris near the nucleus.  Hartley 2 Departure Imaging, E+2 to E+12 Days: From 06 to 16 November 2010, the HRII spectrometer scanned 103P/Hartley 2 every ~15 minutes while the MRI CCD imaged the comet every 2 minutes and HRIV once every hour.  Hartley 2 Departure Imaging, E+12 to E+21 Days: From 16 to 26 November 2010, the HRII spectrometer scanned 103P/Hartley 2 every 30 minutes, and HRIV performed rotation sampling at the same cadence. MRI performed rotation sampling every 30 minutes and imaging using gas filters every two to four hours.   Data Acquisition Strategy ------------------------- The data acquisition strategy for the IR scans throughout the Hartley 2 encounter had to balance data volume limitations with desired sampling. The goal of the IR observations during approach and departure was to monitor the coma for any changes that occurred from one coma scan to another. Therefore, the field of view covered by a scan was selected such that for a reasonable outflow velocity, the scan would cover the distance traveled by any new material released from the nucleus since the previous observation. In order to meet the sampling and field of view criteria, the cadence of scans was selected and then scan rate of the spectrometer (that is, the slew rate of the spacecraft), perpendicular to the slit length, was set to be either one slit width per exposure frame or two slit widths per exposure frame:  Mission Timeline Scan Rate ------------------------ ------------------------------------------ Approach Imaging E-34 to E-8 days 2 slit widths per exposure frame E-8 days to E-18 hours 1 slit width per exposure frame Flyby Imaging E-18 hrs to E+2 days 2 slit widths per exposure frame for these observations (exposure IDs): 4000200, 4000500, 5008002, and 4000700-4010100 excluding 4002700 and 4002800. All other observations are 1 slit width per exposure frame. Departure Imaging E+2 days to E+12 days 2 slit widths per exposure frame E+12 days to E+21 days 1 slit width per exposure frame   Required Reading --------------- The documents listed below are essential for the understanding and interpretation of this dataset. Although a copy of each document is provided in the DOCUMENT directory of this dataset, the most recent version is archived in the Deep Impact and EPOXI documentation set, DI-C-HRII/HRIV/MRI/ITS-6-DOC-SET-V4.0, available online at http://pds.nasa.gov.  EPOXI_SIS.PDF - The Archive Volume and Data Product Software Interface Specifications document (SIS) describes the EPOXI datasets, the science data products, and defines keywords in the PDS labels.  EPOXI_CAL_PIPELINE_SUMM.PDF - The EPOXI Calibration Pipeline Summary provides an overview of the final version of the calibration pipeline that generated the data products in this dataset. For a thorough discussion of the pipeline, see 'EPOXI Instrument Calibration' by Klaasen, et al. (2013) [KLAASENETAL2011].  INSTRUMENTS_HAMPTON.PDF - The Deep Impact instruments paper by Hampton, et al. (2005) [HAMPTONETAL2005] provides very detailed descriptions of the instruments.  HRII_HARTLEY2_FLYBY_LOG.PDF - This log provides notes recorded by the science team as each Flyby exposure (scan) acquired from E-18 to E+48 hours was received on the ground. Annotations include data quality and a list of frames within each scan that appeared to contain the comet.  HRII_3_4_EPOXI_HARTLEY2.TAB - This ASCII table provides image parameters such as the mid-obs Julian date, exposure time, image mode, mission activity type, and description or purpose for each observation (i.e., data product) in this dataset. This file is very useful for determining which data files to work with.   Related Data Sets ----------------- The following PDS datasets are related to this one and may be useful for research:  DIF-C-HRII-2-EPOXI-HARTLEY2-V1.0 - Raw HRII comet Hartley 2 observations  DIF-CAL-HRII-2-EPOXI-CALIBRATIONS-V2.0 - Raw HRII in-flight calibrations from 2007 to 2011  DIF-C/E/X-SPICE-6-V1.0 - EPOXI SPICE kernels  DIF-CAL-HRII/HRIV/MRI-6-EPOXI-TEMPS-V2.0 - HRII, HRIV, and MRI instrument thermal telemetry data for EPOXI which may be useful for determining how temperature fluctuations affect the science instruments, in particular the IR spectrometer. N.B. The pipeline does not use these thermal data to calibrate IR spectra of Hartley 2. Instead it uses instrument temperatures recorded in the FITS headers.  DI-C-HRII/HRIV/MRI/ITS-6-DOC-SET-V4.0 - Deep Impact and EPOXI documentation set   Processing : The calibrated two-dimensional (wavelength and spatial/along-slit) FITS spectral images and PDS labels in this dataset were generated by the Deep Impact/EPOXI calibration pipeline, maintained by the project's Science Data Center (SDC) at Cornell University. The final version of the pipeline for HRII processing, dated January 2014, was used. Known limitations and deficiencies of the pipeline and the resulting data are discussed in the EPOXI Calibration Pipeline Summary document in this dataset and by Klaasen, et al. (2013) [KLAASENETAL2011].  For HRII spectra, the pipeline generates two types of calibrated products:  - Uncleaned radiance data provided in units of Watts/(meter**2 steradian micron) and identified by the mnemonic 'RADREV'. The RADREV data are considered to be reversible because the calibration steps can be backed out to return to the original, raw data numbers.  - Irreversibly cleaned radiance data provided in units of Watts/(meter**2 steradian micron) and identified by the mnemonic 'RAD'. The RAD data are considered to be irreversible because the calibration steps, such as smoothing over bad pixels, cannot easily be backed out to return to the original, raw data numbers.  The calibration pipeline performed the following processes, in the order listed, on the raw HRII FITS data to produce the RADREV and RAD products found in this data set (the process uses the image mode to select the appropriate set of calibration files):  - Calibration of temperatures and voltages in the FITS header - Per-pixel linearization of raw data numbers (version 1.0 used per-quadrant linearization) - Subtraction of dark noise, derived using pixel-by-pixel linearization of either in scene dark frames or optimized mode- dependent master dark frames (the prisms/spectral imaging module and IR focal plane array temperatures, SMOBENT and IRFPAT in the FITS header, are used for scaling if dark modeling is required) - Division by a flat field, derived from pixel-by-pixel linearization - Determine spectral registration and bandwidth for each pixel (using SMOBENT from FITS headers) - Conversion of data numbers to units of radiance for an absolute, radiometric calibration that is reversible (RADREV) and that was derived from pixel-by-pixel linearization - Interpolation over bad and missing pixels identified in the RADREV data to make a partially cleaned, irreversible, radiometric calibration with units of radiance (RAD); Steps for despiking (i.e., cosmic ray removal) and denoising the data which are part of the RAD stream were not performed because the existing routines are not robust. - Set non-image pixels at the left, right, and bottom edges to zero in the RADREV and RAD products. The 'real data' window of an image is given by CALWINDW in the FITS header. If edge pixels need to be analyzed, the original DN values can be found in the raw products located in the PDS dataset, DIF-C-HRII-2-EPOXI-HARTLEY2-V1.0.  As part of the calibration process, the pipeline updated the per-pixel image quality map, the first FITS extension, to identify:  - Pixels where the raw value was saturated, - Pixels where the analog-to-digital converter was saturated, - Pixels that were ultra-compressed and thus contain very little information, and - Pixels considered to be anomalous as indicated by bad pixel maps derived for per-pixel linearity (missing pixels were identified when the raw FITS files were created).  The pipeline also created FITS image extensions for a spectral registration (wavelength) map, a spectral resolution (bandwidth) map, and a signal-to-noise ratio map, which are briefly described in the next section. The calibration steps and files applied to each raw image are listed in the PROCESSING_HISTORY_TEXT keyword in the PDS data label.   Data :  FITS Images and PDS Labels -------------------------- Each calibrated spectral image is stored as FITS. The primary data unit contains the two-dimensional spectral image, with the fastest varying axis corresponding to increasing wavelengths from about 1.05 to 4.8 microns and the slowest varying axis corresponding to the spatial or along-slit dimension. The primary image is followed by four image extensions that are two-dimensional pixel-by-pixel maps providing additional information about the spectral image:  - The first extension uses one byte consisting of eight, single-bit flags to describe the quality of each pixel in the primary image. The PDS data label defines the purpose of each single-bit flag.  - The second extension provides the spectral registration or wavelength for each pixel in the primary image. This extension is required because the wavelength for each pixel changes as the temperature of the instrument increased or decreased.  - The third extension provides the spectral bandwidth for each pixel in the primary image. This extension is required because the bandwidth for each pixel changes as the temperature of the instrument increased or decreased.  - The fourth extension provides a signal-to-noise ratio for each pixel in the primary image.  Each FITS file is accompanied by a detached PDS data label. The EPOXI SIS document provides definitions for the keywords found in a data label and provides more information about the FITS primary image and the extensions. Many values in a data label were extracted from FITS image header keywords which are defined in the document EPOXI_FITS_KEYWORD_DESC.ASC found in the Deep Impact and EPOXI documentation dataset, DI-C-HRII/HRIV/MRI/ITS-6-DOC-SET-V4.0.   File Naming Convention ---------------------- The naming convention for calibrated data labels and FITS files is HIyymmddhh_eeeeeee_nnn_rr.LBL or FIT where 'HI' identifies the HRII instrument, yymmddhh provides the UTC year, month, day, and hour at the mid-point of the observation, eeeeeee is the exposure ID (OBSERVATION_ID in data labels), nnn provides the image number (IMAGE_NUMBER in the data labels) within the exposure ID, and rr identifies the type of reduction:  RR for RADREV data (reversibly calibrated, radiance units) R for RAD data (partially cleaned RADREV data, radiance units)  Up to 999 individual images can be commanded for one exposure ID. Spectral scans often had 8 or more frames for one specific exposure. Therefore, nnn in the file name provides the sequentially increasing frame number within an exposure ID and corresponds to IMAGE_NUMBER in the data labels. For example, if 30 frames were commanded for a scan with an exposure ID of 1000002, the first FITS file name would be HI10110412_5000000_001_RR.FIT and the last would be HI10110412_5000000_030_RR.FIT.   Image Compression ----------------- All data products in this dataset are uncompressed. Specifically all raw Hartley 2 spectral images were never compressed.   Image Orientation ----------------- A true-sky 'as seen by the observer' view is achieved by displaying the image using the standard FITS convention: the fastest-varying axis (samples or wavelength) increasing to the right in the display window and the slowest-varying axis (lines or spatial/along-slit) increasing to the top. This convention is identified in the data labels: the SAMPLE_DISPLAY_DIRECTION keyword is set to RIGHT and LINE_DISPLAY_DIRECTION to UP.  The direction to celestial north, ecliptic north, and the Sun is provided in data labels by CELESTIAL_NORTH_CLOCK_ANGLE, ECLIPTIC_NORTH_CLOCK_ANGLE, and SUN_DIRECTION_CLOCK_ANGLE keywords and are measured clockwise from the top of the image when it is displayed in the correct orientation as defined by SAMPLE_DISPLAY_DIRECTION and LINE_DISPLAY_DIRECTION. Please note the aspect of the North celestial pole in an image can be computed by adding 90 degrees to the boresight declination given by DECLINATION in the data labels.  For a comparison of the orientation FITS image data from the three science instruments, see the quadrant nomenclature section of the the EPOXI SIS document.   Spectral Scans -------------- Each HRII scan of Hartley 2 consists of multiple frames within one exposure ID (OBSERVATION_ID in the data labels). To work with these spectral scans, it is recommended that all frames for one exposure ID be stacked into a three-dimensional cube. Then, a spatial-spatial map can be produced for a specific wavelength by selecting the appropriate spectral column from the image cube. Spectral wavelengths are provided by the second FITS extension, the spectral registration (wavelength) map.   IR Slit Location ---------------- For a comparison of the relative locations of the IR slit with respect to the fields of view of the Medium Resolution Instrument Visible CCD (MRI) and the High Resolution Instrument Visible CCD (HRIV), see the instrument alignment section of the EPOXI SIS document or Klaasen, et al. (2013) [KLAASENETAL2011].  There are no visible CCD context images provided in this dataset to aid in orienting the IR slit location with the nucleus during a particular observation. In many cases, nearly simultaneous MRI frames, located in the dataset DIF-C-MRI-3/4-EPOXI-HARTLEY2-V1.0, were acquired during the IR scans and may provide field of view context for the slit location. Utilizing the infrared data scans themselves is currently the best way to determine the slit location on the nucleus. The user can create a three-dimensional cube as described above then use the ~2.4-micron spatial-spatial map to determine where the slit was pointed during that particular scan. Geometry values, such as right ascension and declination, given in the data labels are for the instrument boresight and do not easily give positional information along the slit for tying a pixel to a specific point on the nucleus.   Timing for Spectra ------------------ It is important to note that the readout order of the IR detector affects the timing of the spectra. When a HRII spectral image is displayed using the true-sky convention, the wavelength increases horizontally to the right and the spatial or along-slit direction is vertical. In this orientation, the IR detector was read out from the left and right edges and toward the center and starting with the first row at the bottom and ending with the last row at the top of the display. Since the detector is reset and read out on a pixel-by-pixel basis, the read out order affects the time at which each pixel is exposed although each pixel has the same exposure duration -- except for the ALTFF mode that has different read and reset causing the effective exposure time to vary with line number, i.e., along the slit in the spatial direction. Additionally, the end of the spectrometer slit that always points roughly towards the sun is the first line to be readout and the last line to be read out is furthest from the sun, assuming the spacecraft is in its usual orientation with the solar panels pointing roughly toward the sun. For more information about the timing of the spectra, see the zero exposure background section of the EPOXI instrument calibration paper by Klaasen, et al. (2013) [KLAASENETAL2011]. A brief discussion about how the calibration pipeline handles the ALTFF mode is included in the EPOXI Calibration Pipeline Summary document.   Parameters :  Data Units ---------- The calibrated RADREV and RAD image data have units of radiance, W/(m**2 steradian micron).   Imaging Modes ------------- A summary of the imaging modes is provided here. For more information see Hampton, et al. (2005) [HAMPTONETAL2005], Klaasen, et al. (2008) [KLAASENETAL2006] and Klaasen, et al. (2013) [KLAASENETAL2011]. In the table below, X-Size is the spectral dimension and Y-Size is the spatial dimension along the slit.  X-Size Y-Size Bin Mode Name (pix) (pix) Type Comments ---- ------ ------ ----- ----- ------------------------------------ 1 BINFF 512 256 2x2 Binned full frame 2 BINSF1 512 126 2x2 Binned sub-frame 3 BINSF2 512 64 2x2 Binned sub-frame 4 UBFF 1024 512 1x1 Unbinned full frame 5 ALTFF 512 256 2x2 Alternate mode 1 (min. exposure time is 1/2 of mode 1) 6 DIAG 1024 512 1x1 Diagnostic, one reset frame followed by a separate read frame such that odd IMAGE_NUMBERs are reset frames and even IMAGE_NUMBERs are read frames 7 MEMCK 1024 512 1x1 Memory Check  By utilizing the different imaging modes of the HRII instrument, the observational requirements for desired exposure times were met. Note, of the 7 modes, only modes 1-6 were used for the encounter with comet Hartley 2. Subframe modes are binned (2x2), reduce the spatial (LINE) extent of the image FOV, and have a shorter readout time which reduces the exposure time for bright objects and keeps the detector from saturating.   Time- and Geometry-Related Keywords ----------------------------------- All time-related keywords in the data labels, except EARTH_OBSERVER_MID_TIME, are based on the clock on board the flyby spacecraft. EARTH_OBSERVER_MID_TIME provides the UTC when an Earth-based observer should have been able to see an event recorded by the instrument.  The SDC pipeline was not able to automatically determine the proper geometric information for the target of choice in some cases. When these parameters could not be computed, the corresponding keywords in the data labels are set to a value of unknown, 'UNK'. Also if GEOMETRY_QUALITY_FLAG is set to 'BAD' or GEOMETRY_TYPE is set to 'PREDICTED' in the PDS labels, then this indicates the geometry values may not be accurate and should be used with caution. The value 'N/A' is used for some geometry-related keywords in the data labels because these parameters are not applicable.  Observational geometry parameters provided in the data labels were computed at the epoch specified by the mid-obs UTC, IMAGE_MID_TIME, in the data labels. The exceptions are the target-to-sun values evaluated at the time light left the target that reached the spacecraft at mid-obs time, and the earth-observer-to-target values evaluated at the time the light that left the target, which reached the spacecraft at mid-obs time, reached Earth.   Ancillary Data : The geometric parameters included in the data labels and FITS headers were computed using the best available SPICE kernels at the time the data products were generated. The kernels are archived in the EPOXI SPICE dataset, DIF-C/E/X-SPICE-6-V1.0.   Coordinate System : Earth Mean Equator and Vernal Equinox of J2000 (EME J2000) is the inertial reference system used to specify observational geometry parameters in the data labels.   Software : The observations in this dataset are in standard FITS format with PDS labels, and can be viewed by a number of PDS-provided and commercial programs. For this reason no special software is provided with this dataset.

Confidence Level Overview : The FITS files in this dataset were reviewed internally by the EPOXI project and were used extensively to verify the calibration of the instrument.   Review : This dataset was peer reviewed and certified for scientific use on 21 March 2014.   Data Coverage and Quality : There are no unexpected gaps in this dataset, except those that occurred on 06 and 28 October 2011 as noted in the dataset overview. All science frames received on the ground were processed and included in this dataset.  Any horizontal striping through some images indicates missing data. The image quality map extension identifies where pixels are missing. If the second most-significant bit of a pixel in the image quality map is turned on, then data for the corresponding image pixel is missing. For more information, refer to the EPOXI SIS document. Additionally, any missing pixels in a frame (within an exposure ID) that is used as the in-scene dark will contaminate the calibrated observations of the same exposure ID, and those contaminated pixels will not be identified as missing in the image quality map extension but will appear as horizontal stripes in the calibrated spectra.  Transient anomalous pixels may be present in the data.   Limitations :  Timing ------ The EPOXI project plans to generate a complete and highly accurate set of UTC correlations since launch. This will ultimately result in a future version of a SCLK that will retroactively change correlation for **all** Deep Impact and EPOXI data. When this kernel is available, it will be added to the SPICE data sets for the two missions and posted on the NAIF/SPICE web site at http://naif.jpl.nasa.gov/naif/. If time and funding permit, the EPOXI project will provide more precise times for archived data.   HRI Telescope Focus ------------------- Images of stars acquired early during the Deep Impact mission in 2005 indicated the HRI telescope was out of focus. However, this focus problem does not significantly affect the HRII instrument. For more details please see the Deep Impact instrument calibration paper by Klaasen, et al. (2008) [KLAASENETAL2006].   Displaying Images ----------------- Flight software writes an image header over the first 100 bytes of quadrant A. These image header pixels are included in the calibrated FITS images. Since the values in these pixels vary dramatically, it is recommended that the values of the EPOXI:MINIMUM and EPOXI:MAXIMUM keywords in the data label (or the MINPVAL and MAXPVAL in the FITS header) be used to scale an image for display because these values exclude the header bytes as well as the reference rows and columns located at the edges of the image. For more information, see the quadrant nomenclature section of the EPOXI SIS document.