Euclid preparation. First investigation of the impact of cross-contamination on spectroscopic redshift measurements with pixel-level simulations

Technical Summary: First Investigation of the Impact of Cross-Contamination on Spectroscopic Redshift Measurements with Pixel-Level Simulations

Problem Statement
The Euclid mission aims to construct one of the largest three-dimensional maps of the Universe using the Near-Infrared Spectrometer and Photometer (NISP). A primary challenge for Euclid's slitless spectroscopy is "cross-contamination," where the dispersed spectra of nearby sources overlap. Unlike slit-based spectroscopy, slitless observations disperse light from all objects in the field of view (FOV) without pre-selection. This leads to three main drawbacks: high background light, "self-contamination" (degradation of spectral resolution due to object size), and cross-contamination from neighboring sources. While background and self-contamination reduce the signal-to-noise ratio (S/N), cross-contamination is particularly complex as it depends on source density, contaminant intensity, and dispersion angles. This contamination reduces the S/N of emission lines, introduces residuals after decontamination, and lowers the fraction of usable pixels, potentially leading to incorrect redshift measurements (interlopers). These errors can distort the three-dimensional distribution of galaxies, affecting cosmological analyses such as the correlation function and power spectrum. Previous studies assessed Euclid spectroscopy performance in the absence of cross-contamination; this paper addresses the lack of simulation-based analysis regarding the impact of cross-contamination on redshift measurements.

Methodology
The authors present a study based on pixel-level simulations processed through the official Euclid spectroscopic pipeline. The analysis utilizes the SPRING (Spectroscopic Pipeline Runner and Input Generator) framework, which integrates the TIPS simulator (for image generation), the SIR pipeline (for spectral extraction and decontamination), and the SPE pipeline (for redshift measurement).

• Input Catalogues: The simulations use the "Flagship" galaxy mock, a catalogue of 3.4 billion galaxies derived from an N-body dark matter simulation. The study distinguishes between two samples:
  • H$\alpha$ Sample: Galaxies with redshift $0.9 \le z \le 1.8$ and H$\alpha$ + [N II] flux $F_{H\alpha+[N II]} \ge 1$ (in units of $10^{-16}$ erg s$^{-1}$ cm$^{-2}$).
  • Contaminant Sample: All other galaxies not in the H$\alpha$ sample.
• Simulation Setup: Ten field observations were simulated, selected from a 36 deg$^2$ region of the Flagship mock to cover a range of galaxy densities (from voids to clusters).
  • Idealized Case: A "grid" simulation was performed where H$\alpha$ galaxies were arranged to avoid overlap, serving as a benchmark for intrinsic redshift accuracy.
  • Contamination Scenarios: Realistic scenarios were created by adding contaminant galaxies with increasing magnitude limits ($H_E \le 19, 20, 21, 22, 23, 24$).
• Simplifications: To manage computational costs, the study adopted an idealized photometric bypass. Instead of simulating full photometric observations and processing, SPRING generated idealized photometric cutouts directly from the Flagship morphological parameters, assuming perfect photometric and morphological measurements. The analysis was restricted to the first-order spectra, ignoring zeroth and second orders.
• Metrics: The performance was evaluated using the Redshift Success Rate (SR), defined as the fraction of galaxies with accurate ($|z_{meas} - z_{true}| \le 0.005$) and reliable ($spe\_z\_prob \ge 0.99$) redshift measurements.

Key Contributions

1. Extension of Simulation Framework: The authors extended the existing Euclid simulation framework to include full end-to-end spectral decontamination and redshift measurement, moving beyond previous studies that excluded cross-contamination.
2. Quantification of Cross-Contamination Impact: The study provides the first simulation-based assessment of how cross-contamination from galaxies of varying brightness affects redshift success rates.
3. Differentiation of Contaminant Types: The analysis distinguishes between contamination from H$\alpha$ emitters (targets) and all other galaxies, and further separates contaminants based on their redshift relative to the target sample.
4. Preliminary Estimate of Same-Redshift Contamination: The paper offers a preliminary estimate of the specific degradation caused by contaminants lying within the same redshift range ($0.9 < z < 1.8$) as the target sample, a scenario that creates selection biases difficult to mitigate with standard random catalogues.

Results

• Baseline Accuracy (No Contamination): In the idealized "on-grid" simulation, 97% of H$\alpha$ galaxies with flux $F_{H\alpha+[N II]} \ge 2$ and reliable probability ($spe\_z\_prob \ge 0.99$) were measured accurately ($|z_{meas} - z_{true}| \le 0.005$). A small systematic bias of $2.5 \times 10^{-4}$ was observed, attributed to the H$\alpha$ + [N II] complex.
• Dependence on Flux and Size: The redshift success rate is strongly dependent on H$\alpha$ flux and galaxy size ($R_{50}$). For compact sources ($R_{50} < 0.25''$), the success rate rises from ~10% at low fluxes ($F \in [1.4, 1.9]$) to 100% at high fluxes ($F \ge 10$). Larger galaxies suffer from lower success rates due to flux loss in the extraction aperture.
• Impact of H$\alpha$ Contamination: Contamination from other H$\alpha$ galaxies (which are generally faint, $H_E \approx 21.5$) was found to have a negligible impact on the redshift success rate compared to the grid simulation.
• Impact of General Contamination: When contaminant galaxies were added:
  • Adding sources brighter than magnitude 19 caused a significant drop in the success rate (approximately 25% decrease).
  • Adding contaminants in the magnitude range [19, 20] caused a further, smaller decrease.
  • Adding sources fainter than magnitude 20 resulted in no statistically significant further degradation; the success rate remained constant within uncertainties.
• Same-Redshift Contamination: The authors estimated that contamination from galaxies within the target redshift range ($0.9 < z < 1.8$) contributes to approximately 4% of the total degradation in success rate caused by all cross-contamination. This is a critical finding because such contamination creates a density-dependent selection effect that cannot be mitigated by standard visibility masks.

Significance and Claims
The paper positions itself as a qualitative investigation and a preliminary estimate rather than a definitive assessment of Euclid's redshift performance. The authors explicitly state that the results are likely optimistic due to the idealized photometric bypass and the exclusion of photometric errors.

The significance of the work lies in:

1. Validating Mitigation Strategies: The results support the development of mitigation strategies for slitless spectroscopy, confirming that while bright contaminants are the primary source of degradation, fainter sources ($H_E > 20$) have a limited impact.
2. Highlighting Systematic Risks: The study identifies that contamination from sources at the same redshift as the targets is a specific systematic effect requiring dedicated mitigation, as it biases the density field measurement.
3. Pipeline Testing: The work demonstrates the capability of the official Euclid pipeline (TIPS, SIR, SPE) to handle spectral decontamination and redshift measurement in a simulated environment, providing a baseline for future refinements.

The authors conclude that while the small magnitude of the same-redshift contamination effect (4%) is reassuring, further investigations with more statistics and observed data are required to fully quantify its impact on cosmological analyses.