Source-linked AI summary
CERES: A Set of Automated Routines for Echelle Spectra
Rafael Brahm, Andrés Jordán, Néstor Espinoza
TL;DR
CERES addresses the complexity and inconsistency of reducing echelle spectra from different instruments by providing modular routines for fully automated pipelines. It combines calibration, order tracing, extraction, and analysis tools, and achieves author-reported radial-velocity precision ranging from below 10 m s−1 for stabilised fibre-fed instruments to about 400 m s−1 for a non-stabilised spectrograph. The authors report pipelines for thirteen instruments, including cases without dedicated pipelines.
Problem
Echelle reduction is complicated by multiple curved, variably illuminated, and sometimes overlapping orders, while instrument-specific processing and human intervention can yield inconsistent results across datasets.
Method
CERES is a modular collection of routines that automates CCD reduction, order identification and tracing, rectangular or optimal extraction, wavelength calibration, radial-velocity estimation, and atmospheric-parameter estimation.
Results
CERES pipelines were developed for thirteen instruments, with long-term radial-velocity precision below 10 m s−1 for stabilised fibre-fed instruments and about 400 m s−1 for a non-stabilised spectrograph.
Takeaways & Limitations
The recipes can guide pipelines for other instruments, including instruments without dedicated pipelines, while providing reduced spectra, radial velocities, bisector spans, and quick atmospheric-parameter estimates.
Takeaways & Limitations
CERES does not rectify order curvature before extraction, lacks sky-contamination correction for slit spectrographs, and still requires an upgrade for precise iodine-based radial velocities.
Abstract
from arXiv · showhide
We present the Collection of Elemental Routines for Echelle Spectra (CERES). These routines were developed for the construction of automated pipelines for the reduction, extraction and analysis of spectra acquired with different instruments, allowing the obtention of homogeneous and standardised results. This modular code includes tools for handling the different steps of the processing: CCD image reductions, identification and tracing of the echelle orders, optimal and rectangular extraction, computation of the wavelength solution, estimation of radial velocities, and rough and fast estimation of the atmospheric parameters. Currently, CERES has been used to develop automated pipelines for thirteen different spectrographs, namely CORALIE, FEROS, HARPS, ESPaDOnS, FIES, PUCHEROS, FIDEOS, CAFE, DuPont/Echelle, Magellan/Mike, Keck/HIRES, Magellan/PFS and APO/ARCES, but the routines can be easily used in order to deal with data coming from other spectrographs. We show the high precision in radial velocity that CERES achieves for some of these instruments and we briefly summarize some results that have already been obtained using the CERES pipelines.
1. INTRODUCTION
CERES addresses the complexity and inconsistency of reducing echelle spectra by providing modular routines for fully automated pipelines across different instruments. The pipelines target homogeneous processing, low-signal data, and precision radial-velocity measurements.
- Echelle spectra require complex reduction because multiple orders vary in intensity, curve substantially, and can overlap vertically.
- Instrument-specific pipelines and human intervention can produce inconsistent results across instruments and complicate repeated precision radial-velocity measurements.
- Existing automated tools had not been broadly extended beyond a small number of echelle spectrographs.
- CERES is a modular set of computational routines for developing fully automated reduction pipelines for echelle-spectrograph data.
- CERES pipelines were developed for thirteen spectrographs and aim to handle low-SNR data and measure precision radial velocities for extrasolar-planet studies.
2.1. General considerations
CERES pipelines share a common architecture while accommodating instrument-specific tasks and both fibre-fed and slit spectrographs. Their goal is to turn raw echelle images into calibrated, extracted spectra and additional analysis products without human intervention.
- CERES accepts raw echelle images and aims to produce optimally extracted, wavelength-calibrated, instrumentally corrected spectra plus radial velocities, bisector spans, and atmospheric parameters.
- Each pipeline uses a main driver that combines reusable GLOBALutils functions with modules containing instrument-specific tasks.
- The pipeline design accommodates the overlapping spectral orders produced by high-resolution echelle dispersion.
2.2. Pre-processing
CERES preprocessing classifies raw frames, constructs and applies instrument-dependent calibrations, identifies and traces echelle orders, and models scattered light. The routines support both fibre-fed and slit spectrographs, with tracing designed to remain effective in low-SNR regions.
- Master CCD Frames: Raw images are classified by type before master CCD calibration frames are constructed through median combination.
- Master CCD Frames: Slit-spectrograph flat fields use dispersed continuum illumination, while fibre-fed instruments use a different partial correction because the detector borders are difficult to illuminate.
- Identification of the echelle orders: Order identification combines central columns into a reference cut, smooths it, detects peaks, and rejects shallow peaks using an adjustable threshold.
- Identification of the echelle orders: CERES traces curved orders across the CCD with fitted centroids and polynomial representations, rejecting anomalous drifts during the process.
- Identification of the echelle orders: For flexing slit spectrographs, CERES retraces orders by cross-correlating reference traces with science images in pixel space.
- Scattered light subtraction: Scattered light is estimated from median fluxes in inter-order regions and linearly interpolated across the cross-dispersion direction before subtraction.
2.3. Extraction
CERES provides rectangular and optimal extraction methods that convert each echelle order from a two-dimensional image into a one-dimensional spectrum. Optimal extraction uses profile-based weights to minimize variance while preserving an unbiased flux estimate, including for distorted echelle traces.
- Extraction sums signal across the trace in the cross-dispersion direction, producing one one-dimensional spectrum for each echelle order after systematic corrections.
- Rectangular extraction sums pixels within a user-defined vertical window, whereas optimal extraction assigns profile-based weights.
- Optimal extraction chooses weights that minimize the expected flux variance while keeping the flux estimator unbiased.
- The weighting model accounts for pixel variance from detector gain and readout noise and uses normalized spatial profiles.
- For highly distorted echelle spectra, CERES models profiles with multiple polynomials aligned along the traces and iteratively identifies cosmic rays.
2.4. Wavelength Calibration
CERES calibrates echelle spectra by identifying ThAr emission lines, fitting a global wavelength solution across pixels and physical orders, and tracking instrumental drift. The procedure accommodates different spectrograph designs while addressing blended lines, outliers, and imperfect calibration lamps.
- ThAr calibration: Arc-lamp spectra provide the pixel-to-wavelength mapping used to calibrate science spectra.The pipeline uses characterized emission lines from a reference gas lamp as wavelength standards.
- ThAr calibration: CERES extracts ThAr spectra with optimal extraction for fibre-fed instruments and rectangular extraction for slit spectrographs.The extraction choice depends on whether a reference profile from flat frames is available.
- Line identification: The pipeline builds a common line list, estimates long-term pixel drift by cross-correlating orders with a binary mask, and refines line positions with Gaussian fits.Blended regions are handled by fitting multiple Gaussians.
- Order numbering: Real echelle orders are assigned by selecting m0 from the grating-equation relation, then setting each physical order to mi = ji + m0.The selected m0 produces the smaller slope in the diagnostic relation.
- Global wavelength solution: A global Chebyshev-polynomial expansion of the grating equation fits wavelength as a function of pixel and order, with iterative 3σ rejection of outlier lines.The polynomial degrees in pixel and order are instrument-dependent and selected by inspecting residual structure.
- Drift correction: Instrumental drift is monitored with additional ThAr exposures or simultaneous calibration fibres, and the drift model fits a single Doppler shift while holding reference coefficients fixed.This approach is used to determine the velocity drift relative to the reference wavelength solution.
- Calibration limitations: ThAr lamps enable precise radial velocities but remain imperfect calibrators because line coverage, saturation, lamp composition, and lamp replacement can introduce systematic errors.Fabry-Perot calibration can supplement the system, although it does not directly provide absolute wavelengths.
2.6. Final output
CERES produces reduced, wavelength-calibrated spectra and homogeneous measurements through modular routines for blaze correction, radial velocities, uncertainty estimation, and rapid stellar classification.
- Reduced spectral products: CERES saves reduced spectra in three-dimensional FITS files indexed by data type, echelle order, and pixel.The wavelength and optimally extracted stellar flux products occupy dedicated data-type entries.
- Instrumental correction: Blaze correction uses normalized fibre flats for fibre-fed spectrographs or polynomial fits to rapidly rotating hot-star spectra for slit instruments.The resulting deblazed spectra can retain slopes because calibrator and target continua differ.
- Radial velocities: Radial velocities are derived by cross-correlating spectra with spectral-type binary masks, combining order-level CCFs by SNR, and fitting a Gaussian to the CCF minimum.CERES provides G2, K5, and M5 masks, with the G2 mask as the default.
- Radial velocities: Moonlight contamination can create a secondary CCF peak, so CERES estimates the contaminant velocity and fits two Gaussians when the peaks overlap.RV uncertainties use empirical relations based on CCF width and continuum SNR, calibrated with Monte Carlo simulations.
- Additional analysis: CERES also computes barycentric corrections and rapidly estimates stellar atmospheric parameters for reconnaissance spectroscopy.The atmospheric-parameter module is intended to identify false positives, fast rotators, and giants rather than deliver publication-level precision.
3. CURRENTLY SUPPORTED INSTRUMENTS
CERES currently supports automated reduction and analysis pipelines for thirteen echelle spectrographs spanning low-to-mid and high spectral resolutions. The supported instruments include fibre-fed and slit spectrographs, with instrument-specific capabilities and boundaries.
- Instrument capabilities: CAFE provides R ≈70000 spectra across 84 orders, and CERES achieves σRV ≈30 m s−1 for this instrument.CAFE lacks a simultaneous calibration system and divides the complete optical spectrum across a 2Kx2K CCD.
- Instrument capabilities: For PUCHEROS, optimal extraction is important because the small telescope aperture produces mostly low-SNR observations.PUCHEROS has R = 20000 and no simultaneous wavelength calibration system.
- Instrument-specific boundaries: FIES processing currently excludes simultaneous calibration data, whereas DuPont/Echelle can retrace orders to handle telescope-flexure-related position changes.FIDEOS was still in commissioning, so its CERES pipeline was expected to be updated.
- Instrument-specific boundaries: Several pipelines have explicit scope limits: MIKE is currently processed only in its red arm, while HIRES is limited to its green CCD chip.CERES also does not handle I2-cell data for PFS or HIRES; PFS currently relies on ThAr wavelength calibration.
4. PERFORMANCE AND INTERESTING RESULTS
CERES pipelines deliver homogeneous radial-velocity measurements across multiple spectrographs, with stabilised fibre-fed instruments reaching planetary-companion sensitivity and non-stabilised instruments supporting false-positive screening. The pipelines have also enabled diverse follow-up studies, including planet discoveries and stellar-system characterization.
- Radial-velocity performance: Below 10 m s−1 precision is achieved for stabilised fibre-fed instruments, while the non-stabilised DuPont spectrograph reaches approximately 400 m s−1.The DuPont precision is insufficient for planetary-mass companions but supports identification of stellar or brown-dwarf companions and other false positives.
- Radial-velocity performance: ≈7 m/s precision is achieved by the CORALIE and FEROS CERES pipelines, compared with ≈30 m s−1 for the dedicated FEROS pipeline.The CERES FEROS pipeline also handles SNR <30 data with uncertainties governed by Poisson errors rather than reduction-systematics-dominated errors.
- Radial-velocity performance: The CERES pipelines for CORALIE, FEROS, and HARPS are precise enough to detect planetary-mass companions.This capability underpins their use in multiple exoplanet follow-up studies.
- Scientific applications: CERES reductions supported K2 planet discoveries and validation, including a dense Neptune-mass planet, two hot Jupiters, and 104 validated planets.These studies used combinations of CORALIE, FEROS, and HARPS data reduced with CERES.
- Scientific applications: CERES-enabled analyses also covered giant-planet systems, updated parameters for transiting hot Jupiters, eclipsing binaries, supernova progenitors, young associations, and novae.The examples span data from FEROS, CORALIE, HARPS, and PUCHEROS.
5. CONCLUSIONS
CERES provides robust, fully automated echelle-spectra pipelines for thirteen instruments, including instruments without dedicated pipelines, while producing radial velocities, bisector spans, and rapid atmospheric-parameter estimates. Its main stated limitations concern order curvature, sky contamination, and the absence of automated iodine-cell radial-velocity routines.
- Contributions: CERES supports robust, fully automated reduction, processing, and analysis pipelines for thirteen instruments with differing specifications.The recipes can guide pipeline construction for other instruments and produce homogeneous results across datasets.
- Performance and outputs: The FEROS pipeline reaches σRV =7.5 ms−1 at high signal-to-noise and remains effective on low-signal-to-noise data, enabling planet discoveries around stars fainter than V = 14.CERES also provides a fully automated DuPont pipeline with σRV =400 ms−1 despite instrument instability.
- Performance and outputs: CERES additionally estimates rough, fast atmospheric parameters and computes bisector spans for rapid target vetting at the telescope.The pipelines also provide reduced spectra and radial velocities.
- Limitations: CERES does not rectify echelle-order curvature before extraction, lacks sky-contamination correction for slit spectrographs, and has no automated iodine-cell radial-velocity routine.These are identified as limitations or future upgrades rather than reported failures of the existing pipelines.