🔭 astrophysics

An archive of reduced and telluric-corrected CRIRES+ L- and M-band spectra with slit-tilt and wavelength calibrations

This paper presents a newly created, uniformly reprocessed archive of high-resolution CRIRES+ L- and M-band spectra, featuring improved wavelength calibrations and slit-tilt corrections derived from telluric modeling to provide a standardized resource for the astronomical community.

Original authors: Thomas Marquart, Alexis Lavail

Published 2026-04-28

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Thomas Marquart, Alexis Lavail

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Problem: The "Blurry Glasses" of Space Observation

Imagine you are trying to take a high-definition photo of a distant, glowing firefly through a window during a heavy rainstorm. Even if you have the world’s most expensive camera, two things will ruin your shot:

The Tilt: The window is slightly crooked, so everything looks slanted.
The Fog: The raindrops and condensation on the glass blur the light, making it impossible to see the firefly’s true color or shape.

This is exactly what astronomers were facing with the CRIRES+ instrument at the Very Large Telescope (VLT). CRIRES+ is a super-powerful "camera" (a spectrograph) that breaks light into rainbows to study the chemistry of stars. It works perfectly for most colors, but when it tries to look at "thermal infrared" light (the L and M bands, which detect heat), it hits a wall.

Because the instrument lacks certain calibration tools for these specific "colors," the data comes out "crooked" (the slit tilt) and "foggy" (atmospheric interference from Earth's own air). Because the data was so messy, most scientists simply ignored it.

The Solution: The Ultimate Digital Restoration

Two researchers, Thomas Marquart and Alexis Lavail, decided to act like digital forensic experts. They took a massive mountain of "messy" data—over 11,000 raw images—and performed a massive, automated cleanup.

Here is how they did it, using three main "restoration" steps:

1. Straightening the Frame (Fixing the Slit Tilt)

Imagine you’re looking at a striped shirt through a tilted lens; the stripes look diagonal. In the telescope, the "stripes" of light were hitting the detector at an angle. The researchers looked at "standard stars" (stars that are known to be very simple and featureless) to see how the tilt was affecting the light. They then calculated a mathematical formula to "straighten" the image, ensuring that every pixel was exactly where it was supposed to be.

2. Cleaning the Window (Telluric Correction)

Earth’s atmosphere is like a thick, dirty window. It is full of molecules (like water vapor and CO2) that absorb specific parts of the light before it reaches the telescope. To a scientist, this looks like "missing" pieces of the rainbow.
The researchers used a clever tool called viper. Think of viper as an AI-powered "smart filter" that knows exactly what Earth's "dirt" looks like. It models the atmospheric mess and mathematically subtracts it, leaving behind the pure, clean light from the distant star.

3. The AI Assistant (The Secret Ingredient)

This wasn't a job for humans alone. The researchers used an advanced AI (Claude) to act as their "super-intern." The AI wrote the complex code, organized the massive files, and ran the thousands of calculations required to process the data. What would have taken humans weeks of tedious clicking and typing was finished in just a few days.

The Result: A Digital Library for the Stars

The researchers didn't just fix the data; they built a public library (a web archive).

Now, any astronomer in the world can go to their website, search for a star, and instead of seeing a "blurry, crooked" mess, they see a crystal-clear spectrum. This allows them to see things they couldn't see before—like the "breath" of a star (carbon monoxide gas) or the heat signatures of planets forming around distant suns.

In short: They took a pile of "unusable" cosmic static and turned it into a high-definition map of the heat in our universe.

Technical Summary: An Archive of Reduced and Telluric-Corrected CRIRES+ L- and M-band Spectra

Authors: Thomas Marquart (Uppsala University) and Alexis Lavail (Univ. de Toulouse)
Date: March/April 2026

1. The Problem: Calibration Gaps in the Thermal Infrared

The CRIRES+ spectrograph at the ESO Very Large Telescope (VLT) is a high-resolution instrument capable of observing from the Y to the M bands. However, a significant technical limitation exists for the thermal infrared L-band (2.8–4.2 µm) and M-band (4.4–5.5 µm). Unlike the shorter wavelength bands, these bands lack access to standard U-Ne and Fabry-Pérot calibration light sources.

This absence results in two critical issues:

Uncalibrated Slit Tilt: The inclination of the long slit relative to the detector columns is unknown. Because this tilt is wavelength-dependent, it causes spectral features to shift in wavelength depending on their position along the slit, degrading spectral resolution and introducing systematic errors.
Lack of Wavelength Calibration: Without calibration lamps, there is no standard way to establish an absolute wavelength scale for L/M observations.

Consequently, these data have been excluded from ESO’s Phase 3 science products, leading to a low publication fraction for L/M-band CRIRES+ observations.

2. Methodology: A Multi-Stage Reprocessing Pipeline

The authors performed a bulk reprocessing of all public CRIRES+ L/M science data obtained between September 2021 and March 2025. The methodology follows a four-stage approach:

Slit Tilt Calibration: The authors used observations of "telluric standard stars" (hot, featureless B- and O-type stars). By using the tool vipère to fit the atmospheric absorption lines in both the A and B nodding positions of the slit, they measured the wavelength difference caused by the tilt. They derived linear relations for the tilt in both the L and M bands, which were then used to update the ESO Data Reduction System (DRS) tracing tables.
Wavelength Scale Refinement: Instead of relying on static inputs, the authors derived new wavelength scales for each observation by fitting telluric absorption features directly to the science spectra using viper. For orders where telluric features were weak or absent (e.g., the $\text{CO}_2$ gap), they used a global velocity offset correction ( $\Delta v$ ) derived from well-fitted orders across all three detector chips.
Advanced Science Extraction: The pipeline corrects for vertical offsets in the spectral orders caused by the non-repeatable movement of the cross-disperser wheel. It also utilizes deep flat-fields to ensure high signal-to-noise ratios (SNR) and correct the blaze function.
Telluric Correction: Using custom-generated atmospheric transmission templates (based on molecfit data), the authors modeled the molecular absorption (including $\text{H}_2\text{O}$ , $\text{CH}_4$ , $\text{CO}_2$ , $\text{N}_2\text{O}$ , $\text{CO}$ , and $\text{O}_3$ ) to produce telluric-corrected spectra.

3. Key Contributions

The Archive: A uniformly reprocessed archive of 5,649 AB nod-pair spectra spanning 156 unique targets and 68 ESO programmes.
Interactive Web Tool: An interactive web archive (https://www.astro.uu.se/crires-lm) that allows users to browse, plot, and download data, including telluric models and residuals.
Open Source Code: A complete repository of scripts, code, and improved DRS input FITS files (https://github.com/ivh/CRIRES-LM) to facilitate future custom reductions.
AI-Assisted Workflow: The authors highlight a highly efficient development process using Claude Opus 4.6, which was used to write the data retrieval, processing, and web-application code, turning weeks of manual work into days.

4. Results and Precision

The reprocessing significantly improved the data quality:

Wavelength Precision: The A–B difference scatter (the tightest metric) showed a median RMS of only 253 m s⁻¹. The absolute wavelength repeatability has a median RMS of approximately 2.2 km s⁻¹.
Scientific Utility: The authors demonstrate the ability to clearly resolve key astrophysical features, such as the $^{12}\text{CO}$ fundamental band in Herbig Ae stars and T Tauri stars, and the $\text{Br}\alpha$ line in both young stars and supergiants.
Improved Calibration Tables: The project produced updated reference tracing tables with improved default wavelength solutions that can be used as "drop-in" replacements for future ESO pipeline releases.

5. Significance

This work effectively "unlocks" a massive volume of previously underutilized high-resolution infrared data. By providing a standardized, calibrated, and telluric-corrected archive, the authors have lowered the barrier to entry for researchers studying star and planet formation, circumstellar disks, and stellar atmospheres in the thermal infrared. The paper also serves as a modern case study for the integration of Large Language Models (LLMs) in high-throughput astronomical data reduction.