GLIMPSE. I. An SIRTF Legacy Project to Map the Inner Galaxy

The orbit, orientation, and viewing limits on SIRTF conspire to make mapping the Galactic plane a complex process. Furthermore, the spectacular sensitivity of IRAC/SIRTF means that only exceedingly short exposures of the Galactic plane will not be saturated. These issues drive much of the GLIMPSE observing implementation.

Survey strategy.—During standard SIRTF operations, IRAC "campaigns," lasting 10 days to 2 weeks, will be scheduled. A fraction of these IRAC campaigns during the first year will be devoted to the GLIMPSE program, which will observe many IRAC frames tiled together into ∼ 15° × 2° segments of the Galactic plane. Each ∼30 deg² segment will consist of 35–45 "chained" astronomical observing requests (AORs).¹³ Each AOR will cover a narrow rectangular strip 03 × (2°–3°) spanning b≅ -1° to b≅ + 1°, but inclined to the Galactic plane. The inclination angle of the AOR to the Galactic plane depends upon the spacecraft roll angle. For GLIMPSE observations, the roll angle will lie between 15° and 50° from Galactic north. Example AOR sky coverage in the direction of the star formation region W51 is shown in Figure 3.

Each IRAC pointing simultaneously images two adjacent 5 17 × 5 17 fields in two bands. An IRAC frame has 256 × 256 pixels; the 3.6 and 5.8 μm fields coincide on the sky and the 4.5 and 8.0 μm fields coincide on the sky, but the frame edges of the two fields are separated by 1 5. The total integration time per position is 2 s. The observations will be stepped by half‐frames (128 pixels; 2 58) along the long axis of the AOR (which could be the spacecraft X‐ or Y‐axis, depending upon the roll angle). Every sky direction in the GLIMPSE region will be visited at least twice. The time separation between the two visits will range from 20 s (the time between pointings) to 3 hr (the time between AORs). The frame overlap along the short axis of the AOR will be 144 (12 pixels). A typical AOR will take 1.5 hr to complete, yielding (3–4) × (47–70) IRAC frames per band. Each ∼ 15° × 2° segment of the Galactic plane will require 35–45 AORs; the entire survey will require eight segments.

Observing strategy validation.—Time will be scheduled during early SIRTF operations to validate our observing strategy (OSV). The GLIMPSE OSV region will be chosen to sample a range of stellar densities and diffuse background levels characteristic of the inner Galactic plane. Regions of particularly high stellar density and diffuse background will be included to assess our strategy in the most challenging cases. The OSV will view either W51, G284.3−0.3, G305.2+0.2, or NGC 6334, depending upon SIRTF launch date. The goal of these observations is to assure that GLIMPSE data will have the highest reliability point‐source extraction and the greatest possible sky coverage, and to prevent the occurrence of gaps in the survey area. GLIMPSE reliability and completeness calculations, which have been based on simulated data, will be tested by these observations. The OSV will consist of four 1° × 017 AORs nearly identical to normal GLIMPSE AORs. These will be repeated with varying frame offsets (along the strip direction) and overlaps (between strips). Finally, GLIMPSE OSV will perform a single 1° × 017 AOR to allow us to evaluate the pointing reconstruction accuracy for the full half‐hour that it takes to do a single diagonal band across the Galactic plane, testing roll angle uncertainties between AORs, and determining criteria for establishing reliability and completeness. At the end of the OSV period, the outer longitude limits of the survey may be adjusted so that the total survey time does not exceed 400 hr.

Frame details.—The simulated 3.6 μm IRAC frame for the W51 region shown in Figure 3 illustrates both the wealth of detail that GLIMPSE will uncover and the significant challenges for processing and analyzing these data. In this region, there are 400–600 2MASS sources per frame, all of which are detectable by GLIMPSE. (See Table 3 for the 5 σ sensitivities and estimated flux limits for 99.5% reliability.) Examination of 2MASS and MSX data for this region indicates that we should expect two to four saturated sources per frame in the 3.6 μm band, decreasing to an average of one saturated source per frame in the 8.0 μm band.

Timetable and survey tracking.—For our timetable, we give the dates relative to the launch date (L + n months). The OSV data, to validate the survey strategy for GLIMPSE, will be acquired at approximately L + 4. Data acquisition for the full survey will begin after OSV data are analyzed and the survey strategy is validated, and will continue for about 9 months. The observability of a section of the Galactic plane as a function of Galactic longitude and time of year is shown in Figure 4.

The first installment of the GLIMPSE Point Source Catalog will be delivered to SSC at L + 9; the first installment of the mosaicked data and GLIMPSE Point Source Archive will be delivered at L + 15. Updates to each of these data products will be provided at 6 month intervals after the first release dates. The final version of all GLIMPSE data products will be delivered to SSC at L + 27. Documentation of the GLIMPSE survey will be available shortly after launch (L + 2); updates will be provided with the data releases. A preliminary version of the GLIMPSE Legacy Science Data Products document is already available on our Web site and will be available from the SIRTF Science Center.

The SSC will deliver basic calibrated data (BCD) from the GLIMPSE IRAC observations to the GLIMPSE team. BCD will have gone through the following steps: validation, addition of header keywords, sense of InSb flux flipped, conversion to floating point, correction for dichroic flip, normalization by Fowler number and barrel shifts, corrections of electronic bandwidth limitations, subtraction of dark/bias frames, correction for multiplexer bleed, correction for first‐frame effects, linearization, flattening, detection of radiation hits, subtraction of sky darks, flux calibration, and detection of latents. The positional information will be good to 14, and the photometric accuracy should be better than 10% early in the mission. BCD will also contain several ancillary data files, including the raw data, and several mask images that contain information on cosmic rays, linearity corrections, etc. A complete description of the processing that goes into generating the BCD is given in § 6.3 of the SIRTF Observers' Manual.¹⁴

The GLIMPSE team will further process these data using an automated pipeline (a "post‐BCD" pipeline) to correct for remaining instrumental artifacts, extract and cross‐identify point sources, and mosaic the images. The resulting GLIMPSE Point Source Catalog and Archive and the set of mosaicked images will be released to the astronomical community via the SSC. These released data will be more useful than the original BCD, since they will benefit from the GLIMPSE team's experience in analyzing IRAC data in crowded and confused fields that have bright and positionally varying background emission.

The data reduction will occur at the University of Wisconsin–Madison using a network of LINUX workstations. The GLIMPSE pipeline is parallelized and will simultaneously use multiple processors; data flow will be controlled using OPUS pipeline software (Swade & Rose 1999). Locally, the data will be stored using the commercial database system Oracle 9i, Release 2.

We have divided the Post BCD pipeline into a few key levels with clearly defined steps within each. In order, these levels are as follows:

• 1.  
  Data verification.—Observed AORs are verified and checked to make sure the AOR was properly executed. The data are checked to verify that all SSC pipeline steps were carried out and checked for simple artifacts such as excessive cosmic‐ray hits and instrumental and down‐link problems. The Quick‐Look Validation Tool (QLVT) is used for spot‐checks of frames and to inspect frames with flagged problems.¹⁵
• 2.  
  Basic processing and mask propagation.—The data are corrected for the zodiacal background (Gorjian, Wright, & Chary 2000). Pixels affected by stray light or banding¹⁶ or that are in the wings of saturated sources are flagged and corrected wherever possible. An error mask for these pixels is created.
• 3.  
  Point‐source extraction.—Flux density and position of point sources are determined using DAOPHOT (Stetson 1987). Positions of the brighter sources are checked against the 2MASS Point Source Catalog. Statistics are computed for the residual images and used to assess the extraction process. Flux calibration is also spot‐checked.
• 4.  
  Band merging.—The point source lists obtained in the eight pointed observations (two passes each in four bands) are merged to produce the input for the generation of the GLIMPSE Point Source Catalog and GLIMPSE Point Source Archive.
• 5.  
  Mosaicked image production.—The data are resampled and registered to a Galactic coordinate system, and IRAC frames are background matched, if necessary. The resulting images are turned into tiles of mosaicked images of 20^' × 20^'. The pixel size for these images is not yet finalized but will be approximately 06.
• 6.  
  Point‐source photometry on the mosaicked image.—The mosaicked images are used to perform point‐source photometry. The resulting source list is then band‐merged with source list from single‐frame data.
• 7.  
  GLIMPSE Point‐Source Catalog and Archive generation.—Sources found in mosaicked and single frames are cross‐identified. Appropriate quality and reliability filters are applied to generate the GLIMPSE point‐source products.

There are four principal data products that will result from the GLIMPSE program:

• 1.  
  The GLIMPSE Point Source Catalog (GPSC, or "the Catalog"). The flux limit for this catalog will be determined by the requirement that the reliability be ≥99.5%. We currently estimate this flux limit to be 1.1 mJy at 3.6 μm and 2.5 mJy at 8.0 μm. The 8 μm channel has a brighter limit due to the increased diffuse background from PAH emission near 7.7 μm in the Galactic plane. The Catalog photometric uncertainty will be less than 0.2 mag. For each IRAC band, the Catalog will provide fluxes (with errors), positions (with errors), the density of local point sources, the local sky brightness, and flags that provide information on source quality and any anomalies present in the data. The Catalog is expected to contain ∼10⁷ objects.
• 2.  
  The GLIMPSE Point Source Archive (GPSA, or "the Archive"), consisting of point sources with signal levels ≥5 σ above the local background, to approximately a flux limit of 0.2–0.4 mJy. The photometric uncertainty is expected to be less than 0.2 mag. The information provided will be the same as for the Catalog. The Archive will contain ∼5 × 10⁷ objects.
• 3.  
  Mosaicked images for each band, each of approximately 20^' × 20^' angular coverage. About 9000 FITS formatted images will be tiled to smoothly cover the entire survey area, using a Galactic coordinate system. The pixel resolution has not been finalized but will be about 06.
• 4.  
  The Web Infrared Tool Shed (WITS), a Web interface to a collection of model infrared spectra of dusty envelopes and PDRs, updated for IRAC and Multiband Imaging Photometer for SIRTF (MIPS) bandpasses. WITS currently resides on servers at IPAC.¹⁷ The interface contains two "toolboxes": the Dust Infrared Toolbox (DIRT) and the Photodissociation Region Toolbox (PDRT), which provide databases of circumstellar shell emission models and PDR emission models. Users can input data and retrieve best‐fit models. DIRT output includes central source and dust shell parameters; PDRT output consists of gas density, temperature, incident UV field, and IR line intensities.

The GLIMPSE survey will uncover for the first time a huge number of stars in the inner Galaxy. As a result, it will be the survey of choice for those interested in the stellar structure of the Galaxy. Moreover, since the preponderance of star formation in the Galaxy is expected to occur in the inner Galaxy, studies using GLIMPSE will be able to characterize star formation in a wide range of environments. These are the two principal goals of the GLIMPSE science team.

3.1.1. Star Formation in the Inner Galaxy

3.1.2. Galactic Structure

The wide variety of objects contained in the GLIMPSE survey region is illustrated in Figure 6, which shows the positions of many types of objects overlaid on an 8 μm MSX map of a section of the GLIMPSE survey region. With the higher sensitivity and angular resolution of GLIMPSE, together with the color‐select possibilities of seven or more photometric bands (GLIMPSE+2MASS+MSX), data from GLIMPSE will allow the astronomical community to evaluate the statistics, spatial distribution, and internal structures of numerous classes of Galactic objects as well as providing new probes of the interstellar medium. These include studies of stellar populations, PDRs, extinction, as well as serendipitous discoveries.

Stellar population studies.—Using a combination of 2MASS and MSX data toward a sample of previously classified objects in the Large Magellanic Cloud, Egan, van Dyk, & Price (2001) showed that the mid‐IR 8 μm band provides an important "lever arm" that allows color separation of many classes of objects. Planetary nebulae, H ii regions, and some classes of C‐ and O‐rich AGB stars have very red mid‐IR colors. Other attempts to develop mid‐IR and near‐IR color selections focus on infrared carbon stars using J−K_s versus K_s (Cole & Weinberg 2002), young stellar objects (YSOs) using ISOGAL [7]−[15] versus [15] (Felli et al. 2002), brown dwarfs (Burrows et al. 1997), and carbon stars, OH/IR stars, planetary nebulae (PNe), Herbig AeBe stars, compact H ii regions, and massive YSOs using a combination of 2MASS J, H, and K and MSX 8 μm (Lumsden et al. 2002).

PDRs.—The near/mid‐IR spectrum of photodissociation regions at the surface of molecular clouds is dominated by emission bands at 3.3, 6.2, 7.7, 8.6, and 11.3 μm, probably arising from polycyclic aromatic hydrocarbon (PAH) molecules (Peeters et al. 2002 and references therein). Figure 6 shows that this emission is a striking characteristic of the MSX maps of the Galactic plane. GLIMPSE data can be used to characterize the spatial distribution of different charge states of PAH to constrain the chemistry and evolution of PDRs. The IRAC 3.6 μm band is sensitive to the 3.3 μm feature from neutral PAHs; the 5.8 and 8.0 μm bands are sensitive to PAH ⁺, while the 4.5 μm band contains no PAH features and thereby monitors the continuum (Bakes et al. 2001).

Turbulence and structure in SFRs.—A comparison of the infrared brightness fluctuations in SFRs with the spectral line information from CO and H i observations can yield information about the source of ISM turbulence by using the velocity channel analysis technique of Lazarian & Pogosyan (2000).

Interstellar extinction.—Data from GLIMPSE will allow studies of interstellar reddening in dense dusty regions and diffuse environments, using the colors of stars that lie behind dark clouds. This will allow testing near/mid‐infrared extinction models, two of which are given in Table 3 (Li & Draine 2001; Lutz et al. 1996). Since extensive grain coagulation occurs in the inner regions of dense clouds, the IR extinction law could be quite different between very dense clouds such as the MSX dark clouds (Egan et al. 1998) and presently observable regions. GLIMPSE data will allow a vital characterization of the variation of extinction properties with environment.

Serendipity.—Since the inner Galaxy is the region of the sky with the greatest extinction, it is also the direction in which one is most likely to make serendipitous discoveries. The recent 2MASS discoveries of new globular clusters in the Galactic plane (Hurt et al. 2000) and galaxies in the "zone of avoidance" (Jarrett et al. 2000) hint at the possibilities for GLIMPSE.

The value of the GLIMPSE data products will be enhanced by the availability of complementary IR and radio data sets. The basic characteristics of the IR surveys are given in Table 4. Data sets that we anticipate will be the most useful and that will play important roles in the GLIMPSE team science studies are as follows:

• 1.  
  2MASS.—This survey (Cutri et al. 2001) provides an ideal companion data set to GLIMPSE, with an excellent match in both sensitivity and angular resolution for many types of objects. Many objects will have GLIMPSE+2MASS magnitudes in a total of seven near‐IR and mid‐IR bands! This will allow for a wide variety of different possible color selections and spectral energy distributions (SEDs) from ∼1 to 8 μm. The GLIMPSE bands provide crucial mid‐infrared information.
• 2.  
  MSX.—The MSX (Price et al. 2001) survey provides a good match to the GLIMPSE/IRAC 8 μm band. The MSX data set will be particularly useful for studies of diffuse emission. It provides a good complement to the GLIMPSE data for bright sources, since the saturation limit of GLIMPSE is only slightly brighter than the faint detection limit for MSX.
• 3.  
  Arecibo/Green Bank Telescope/Australian Telescope Compact Array Surveys of GLIMPSEH ii regions.—This data set resolves the distance ambiguities to many massive star formation regions. The data include more than 100 objects with resolved distance ambiguities that will be published and made available on the GLIMPSE Web site.
• 4.  
  Milky Way Galactic Ring Survey (GRS).—A Boston University and Five College Radio Astronomy Observatory collaboration, this is a large‐scale¹³CO J = 1→0 molecular line survey of the inner Galaxy between latitudes −1° to +1° and longitudes 18°–52°, with an angular resolution of 22^'' and a velocity resolution of 0.3 km s⁻¹ (Simon et al. 2001). It is available through the GRS Web site.¹⁸ GRS will be completed by winter 2003.
• 5.  
  The International Galactic Plane Survey.—This survey will map the Milky Way disk in the H i 21 cm line with a resolution of 1^' and 1 km s⁻¹ over the entire GLIMPSE survey area. The data cubes will be available on‐line.¹⁹ The Southern Galactic Plane Survey is also available on‐line.²⁰

There are several other surveys that will provide a useful complement to GLIMPSE. These include the ISOGAL survey at 7 and 15 μm (A. Omont et al. 2003, in preparation; Felli et al. 2002), which provides complementary data for the inner Galaxy region not covered by the GLIMPSE survey, and the planned all‐sky Astro‐F survey at 8.5–175 μm, which will provide an extension to large Galactic latitudes and longitudes, although at lower resolution and sensitivity than GLIMPSE (see Table 4). High angular resolution X‐ray surveys of the Galactic plane using Chandra (Grindlay et al. 2003)²¹ and XMM (Helfand et al. 2002) will also provide a useful comparison to GLIMPSE data in selected regions of the Galactic plane.

Support for this work, part of the Space Infrared Telescope Facility (SIRTF) Legacy Program, was provided by NASA through an award issued by the Jet Propulsion Laboratory, California Institute of Technology, under NASA contract 1407. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France, the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA, and data products from the Midcourse Space Experiment. Processing of the Midcourse Space Experiment data was funded by the Ballistic Missile Defense Organization with additional support from NASA Office of Space Science.