Overview of the New Hubble Spectroscopic Legacy Archive

This paper introduces the Hubble Spectroscopic Legacy Archive (HSLA), a new automated system that aggregates and coadds Cosmic Origins Spectrograph and Space Telescope Imaging Spectrograph data from the MAST archive to produce comprehensive, classified spectra and metadata for individual astronomical targets while also providing open-source tools for custom analysis.

The Gist

Imagine the Hubble Space Telescope as a giant, 28-year-long photography session of the universe. Over the decades, it has taken tens of thousands of pictures (spectra) of nearly 7,000 different cosmic objects using two powerful cameras: STIS and COS.

For a long time, these photos were stored in a massive digital warehouse (the MAST archive), but they were organized by the "photographer's" schedule (the specific observing program) rather than by the subject. If you wanted to see every photo ever taken of a specific star, you had to hunt through hundreds of different folders. It was like trying to find every picture of your grandmother in a library where the books were sorted by the date the camera was invented, not by the person in the photo.

The Hubble Spectroscopic Legacy Archive (HSLA) is the new, super-organized photo album that fixes this. Here is how it works, broken down simply:

1. The "Name Tag" Problem (Target Association)

The first challenge was figuring out that "Star A" in Program 1 is the same object as "Star A" in Program 50, even if the astronomers gave them slightly different names or pointed the telescope at slightly different spots.

• The Analogy: Imagine trying to find all the photos of a specific celebrity. Sometimes they are listed as "Brad," sometimes "Brad Pitt," and sometimes the camera was pointed 2 inches to the left.
• The Solution: The HSLA team created a smart matching system. They decided that if two observations are within 2 arcseconds (a tiny angle, roughly the width of a human hair seen from 10 meters away) of each other, they are the same target. They also used a "master address book" (SIMBAD and NED databases) to verify names. This ensures that every observation of a specific object is grouped together, no matter which program took it.

2. The "Labeling" System (Target Classification)

Once the photos are grouped, the archive needs to know what the object is so you can search for it.

• The Analogy: Instead of just having a folder named "Object 123," the archive puts a detailed label on it: "Star," "White Dwarf," "Galaxy," or "Active Galaxy."
• The Solution: The system automatically reads the labels astronomers wrote in their original proposals and cross-references them with the master address books. It uses a three-tiered hierarchy:
  • Tier 1 (Broad): "Star" or "Galaxy."
  • Tier 2 (Medium): "White Dwarf" or "Active Galaxy."
  • Tier 3 (Specific): "O-type Star" or "Quasar."
    This allows you to search for "all stars" or just "all white dwarfs" instantly.

3. The "Mosaic" (Data Products & Coaddition)

This is the magic part. HSLA doesn't just list the photos; it stitches them together.

• The Analogy: Imagine taking 50 blurry, low-light photos of a firefly at night. If you stack them perfectly on top of each other, the result is one incredibly sharp, bright image.
• The Solution: The archive takes all the individual spectra (light readings) for a single object and combines them into one coadded spectrum.
  • Higher Quality: By combining data, the "signal-to-noise ratio" (the clarity of the image) goes up. Faint details that were invisible in a single observation become clear.
  • Wider Coverage: One instrument might see blue light, and another might see red light. HSLA stitches these together to show the full rainbow of light the object emits, from ultraviolet to near-infrared.

4. The "Instruction Manual" (Metadata & Tools)

The archive provides a "human-readable" file for every object.

• The Analogy: It's like a museum placard next to a painting. It tells you the object's name, coordinates, what it is, and exactly which "photos" (programs) were used to make the final image.
• The Tools: The team also released "Jupyter Notebooks" (interactive coding guides). These are like "DIY kits" for scientists who want to build their own custom mosaics if the standard ones don't fit their specific needs (for example, if an object is moving or changing brightness).

5. Quality Control (Testing)

Before releasing this new archive, the team ran rigorous tests.

• The Analogy: Before opening a new restaurant, the chef tastes every dish to ensure the ingredients are fresh and the recipe is correct.
• The Results: They checked that the coordinates were right, the names matched, and the light measurements were accurate. They found that the combined data is accurate to within 5% of the true value, which is excellent for astronomy. They even tested it against "standard stars" (known cosmic lighthouses) to make sure the colors and brightness were correct.

Why Does This Matter?

The HSLA turns a scattered collection of 64,000 individual observations into a unified, searchable library.

• For a single star: You can now see it with higher clarity than ever before, revealing details about its atmosphere or the gas around it.
• For a group of stars: You can instantly pull up data on 800 white dwarfs to study them as a group, rather than looking at them one by one.

In short, the Hubble Spectroscopic Legacy Archive is the ultimate "greatest hits" collection of Hubble's ultraviolet observations, organized so that anyone can find, combine, and study the light of the universe more easily than ever before.

Technical Summary

Technical Summary: Instrument Science Report COS 2025-18(v1)

Problem Statement
The Hubble Space Telescope's (HST) primary spectroscopic instruments, the Space Telescope Imaging Spectrograph (STIS) and the Cosmic Origins Spectrograph (COS), have accumulated over 64,000 spectra of nearly 7,000 unique astronomical objects since their operational inception in 1997 and 2009, respectively. While these data are hosted in the Mikulski Archive for Space Telescopes (MAST), they are primarily organized by observing programs rather than by target. This structure makes it difficult for astronomers to efficiently locate, select, and combine data for specific objects or source types, particularly when observations span multiple programs, instruments, or epochs. Previous efforts, such as the original Hubble Spectroscopic Legacy Archive (o-HSLA), provided a static template but lacked STIS data, were limited to pre-2018 COS data, and did not incorporate improved calibrations or automation.

Methodology
The new Hubble Spectroscopic Legacy Archive (HSLA) is designed as an automated, dynamic system to generate coadded spectra for individual targets across the entire operational history of STIS and COS. The methodology involves three primary stages:

1. Target Association: The system identifies unique sources by cross-matching observations based on sky coordinates and target names. To determine the optimal angular tolerance for matching, the authors analyzed the spatial scatter of exposures associated with the same target name against SIMBAD coordinates (treated as ground truth). They found that a cross-match radius of 2 arcseconds maximizes completeness and accuracy (>95%) while minimizing contamination from nearby distinct sources. The process accounts for proper motions by calculating coordinates at a common epoch (January 1, 2016) and groups observations into "pointings" based on visit and target name.
2. Target Classification: A three-tiered hierarchical classification scheme was developed to facilitate broad and narrow searches. Classifications are assigned automatically by querying SIMBAD, NED, and Phase II proposal keywords in a specific order of precedence. If a target is an exoplanet host, the NASA Exoplanet Archive is queried for the host star's spectral type. The hierarchy includes Tier 1 (e.g., Star, Galaxy, ISM), Tier 2 (e.g., Stellar Remnant, Active Galaxy), and Tier 3 (e.g., White Dwarf, Quasar). These classifications are mapped to Unified Astronomy Thesaurus (UAT) concepts.
3. Data Coaddition and Abutment: Coadded spectra are produced for each observing mode (grating) and, for COS FUV, for each Lifetime Position (LP). A "quicklook" spectrum is also generated by abutting (joining at transition wavelengths) spectra from different gratings to cover the full wavelength range (approx. 900–10,000 Å). The abutment prioritization is determined by a JSON table that optimizes line detection sensitivity based on grating resolution, sensitivity, and median exposure time. The process excludes spectra with anomalously low flux to avoid errors from slit/aperture drifts but does not perform flux checking across different gratings.

Key Contributions and Data Products
The HSLA project releases several distinct data products for each target:

• Coadded Spectra: Single-grating coadds (cspec.fits) for specific modes and LPs, and abutted quicklook spectra (aspec.fits) covering the full wavelength range.
• Metadata Files: Human-readable text files containing target names, coordinates, classification tiers, UAT mappings, radial velocities/redshifts, and a summary of constituent modes and proposal IDs.
• Provenance and Trailer Files: FITS extensions and text files detailing the input datasets, pipeline versions, and reasons for any data rejection during coaddition.
• Tools: Publicly available Python code and Jupyter notebooks (hosted in the ULLYSES and HASP repositories) that allow users to perform custom coadditions, handle specific science cases (e.g., flux scaling for variable sources, managing Line Spread Function variations across LPs), and inspect data quality.

Results and Validation
The authors validated the HSLA products through several tests:

• Association and Classification: A random sample of 100 targets showed >94% accuracy for coordinate association and >92% accuracy for naming and classification.
• Flux and Wavelength Accuracy: Comparisons between input spectra and coadds for >50,000 datasets revealed that >93% of spectra matched within 5% in flux and >97% matched within 0.5 resolution elements in wavelength, exceeding the 75% success rate benchmark.
• Flux Calibration: Using CALSPEC standard stars (G 191-B2B, GD 71, WD0308-565), the coadded spectra were found to retain the absolute flux calibration accuracy of individual observations (5%) and achieve a relative accuracy of approximately 2.3%.
• Coverage: The initial release (based on data through June 2025) includes 6,712 unique targets, with the majority classified as Stars (3,894) or Galaxies (2,459).

Significance
The paper positions the HSLA as a critical evolution in HST data delivery, leveraging 28 years of STIS and 16 years of COS observations to enable studies with higher signal-to-noise ratios and extended wavelength coverage than individual programs allow. By automating the association and classification of legacy data, HSLA facilitates large-scale spectroscopic surveys of specific object populations (e.g., White Dwarfs, Active Galaxies). The authors note that the automated association and classification framework developed for HSLA could serve as a template for organizing spectroscopic data from other observatories. The report emphasizes that while the "quicklook" abutted spectra are useful for overview, the single-grating coadds are the primary products for scientific analysis, and custom tools are provided for cases requiring specialized handling of variable sources or complex line spread functions.