The code and additional files that make up this page and the data reduction are available at github.com/ivh/CRIRES-LM.

CRIRES-LM

Bulk reduction of all public CRIRES+ L-band (2.8--4.2 um) and M-band (4.2--5.5 um) science data from the ESO archive (2021-01 to 2025-02), with improved slit tilt calibration and per-observation telluric correction using vipere (a fork of viper).

Approach

The standard CRIRES+ pipeline (cr2res) does not account for slit tilt in L/M bands, and ships no telluric correction. We use vipere in two stages:

Stage 1: Calibration from telluric standards

For each of the 16 wavelength settings, we reduce one telluric standard star (hot B/O-type) observed at nod positions A and B. Vipere fits a forward model of telluric absorption + wavelength solution to each nod position independently. The A vs B wavelength difference at each pixel directly measures the slit tilt, since the two nod positions sample different spatial positions on the slit.

From these measurements:

• Slit tilt: fit a linear tilt(wavelength) relation per band, interpolating across the CO2 gap at 4.2--4.5 um where no telluric fit is possible. Written into the SlitPolyB column of the tracing tables (*_tw.fits).
• Wavelength calibration: vipere wavelength polynomials replace the pipeline solution for 72% of traces; the remainder keep pipeline wavelengths (mostly orders with weak/no telluric features).
• Aggregate wavelength refinement: after all science reductions, the median vipere wavelength from all observations is compared to the tracing table per (setting, chip, order). A linear velocity offset vs wavelength is fit per setting and applied as a multiplicative correction to all polynomial coefficients, removing systematic offsets of 1--8 km/s.
• Nod throw: actual throw measured from spatial profiles (2--3% larger than commanded).

Stage 2: Science reduction with telluric correction

Before extraction, the trace positions are adjusted per observation to account for instrument flexure: adjust_traces.py cross-correlates the spatial profile of each raw frame against the reference trace boundaries to measure the Y-shift (typically 5--10 pixels, up to ~50 in rare cases) and writes a corrected tracing table into each reduction directory.

The pipeline-provided tracing tables (*_tw.fits) have been cleaned of spurious edge traces that the pipeline's tracing algorithm occasionally creates for partial orders at detector boundaries. These duplicate traces poison the SlitFraction metadata, preventing extraction of valid orders (affected settings: M4211, M4461, M4504).

All 5237 AB pairs and their combined per-template spectra are then reduced with esorex cr2res_obs_nodding using these adjusted tracing tables (with slit tilt and updated wavelengths) and the nearest-in-time flat field and blaze function. Then tellcorr.py runs vipere on each extracted spectrum to fit and divide out the telluric absorption, followed by wavecorr.py which updates the wavelength scale: fitted orders get the vipere wavelength polynomial directly, while unfitted orders (no telluric features, CO2-saturated regions) receive a velocity correction interpolated from a linear fit to the vipere corrections across all three chips.

The reduced and telluric-corrected spectra are browsable and downloadable via a web app.

Web app

The data browser shows all observations that have telluric-corrected spectra. The front page table is filterable by target, setting, and programme, and sortable by clicking column headers.

Each observation page shows the combined per-template spectrum (interactive Plotly plot), diagnostic plots, and download links for the _tellcorr.fits files. The raw frames table groups frames into AB pairs; clicking a pair navigates to the individual pair reduction, which has its own spectrum, plots, and downloads.

Data scale

• 16 wavelength settings (L3244--L3426, M4187--M4519), 5--7 echelle orders per chip, 3 chips
• 11,135 raw science frames from the ESO archive, forming 5237 AB nod pairs across 412 observing sequences
• 233 flat field epochs across 16 settings, each matched to the nearest-in-time science observation
• Combined per-template spectra for all sequences

Repository structure

*.py                   scripts (fetch, reduce, telluric correct, wavecal, compare, webapp)
*_tw.fits              tracing/wavelength tables per setting (calibration products, aggregate-corrected)
cr2res_obs_nodding.rc  esorex recipe config for nodding
cr2res_cal_flat.rc     esorex recipe config for flats
flats/                 flat field data: {setting}_{date}/ subdirs with SOFs and output
tellurics/             telluric standard reductions, tilt + wavecal scripts
templates/             HTML templates for the inspection webapp

Raw data, reduced products, and FITS outputs are not included (too large); they live on disk and can be regenerated from the scripts + ESO archive.

Requirements

• ESO esorex with the cr2res pipeline
• vipere (with L/M band atmospheric models)
• Python 3.10+, managed via uv; all scripts have PEP 723 inline metadata so uv run script.py pulls in the right dependencies automatically