A unified modeling of X-ray and gamma-ray spectro-polarimetric data: the case of Cygnus X-1

Source: arXiv:2608.06328 · Published 2026-08-06 · By Tristan Bouchet, Thomas Siegert, Victoria Grinberg, Jérôme Rodriguez, Floriane Cangemi, Philippe Laurent et al.

TL;DR

This paper addresses the challenge of jointly modeling X-ray and soft gamma-ray spectro-polarimetric data, particularly for black hole X-ray binaries like Cygnus X-1, where discrepancies between polarization measurements across energy bands complicate physical interpretation. It proposes a robust, unified framework that fits spectral and polarization data simultaneously from multiple independent instruments with differing energy binning and statistical qualities. This approach rigorously accounts for nonlinear mixing of polarization from distinct emission components and enables comprehensive broadband modeling from 2 keV to 2 MeV. Applied to Cygnus X-1 in its hard state, the authors successfully characterize the elusive soft gamma-ray hard tail and constrain the high-energy synchrotron cutoff at (3.9⁺⁰.⁶_−₀.₅)×10² keV. Their results provide insight into non-thermal electron acceleration consistent with Bohm diffusion and synchrotron cooling and offer a novel explanation for the polarization angle misalignment by invoking magnetic field helicity combined with Doppler boosting effects. The method is general and applicable for multi-instrument spectro-polarimetric studies beyond this source.

Key findings

• The synchrotron photon cutoff energy is tightly constrained at (3.9^{+0.6}_{-0.5}) × 10^2 keV for Cygnus X-1's hard tail.
• The joint spectral and polarization fit achieves a reduced χ² of 1.07 on 98 degrees of freedom, indicating high-quality modeling of heterogeneous data across 2 keV–2 MeV.
• The inverse Comptonization (IC) component is constrained with plasma optical depth τ_p=1.1, producing a polarization fraction saturation around 15%, consistent with theoretical opacity and inclination.
• Polarization mixing between components results in significant depolarization or shifts in polarization angle depending on the relative polarization angles and flux weighting.
• Polarization fraction of synchrotron emission shows energy dependence near the cutoff due to the electron energy distribution cutoff, diverging from classical constant PF assumptions.
• The polarization angle misalignment between soft X-rays (jet-aligned) and soft gamma-rays (jet-misaligned by ~40°) can be explained by magnetic field helicity combined with relativistic Doppler boosting effects.
• Independent polarization measurements from IXPE, INTEGRAL/SPI, INTEGRAL/IBIS, AstroSat/CZTI, and others are combined coherently over non-simultaneous observations spanning ~15 years.
• Including polarization data improves physical constraints on electron acceleration mechanisms, favoring Bohm diffusion and synchrotron cooling models.

Methodology — deep read

The authors start by defining a threat-aware framework combining spectral and polarization data for Cygnus X-1 from various missions covering 2 keV to 2 MeV, focusing on the hard state where the non-thermal hard tail is prominent. The adversary analogy is not applicable, but implicit assumptions include stable polarization properties over integrated observations.

They gather spectral data from INTEGRAL instruments JEM-X, ISGRI, and SPI spanning 3–1500 keV, all reduced following consistent pipelines and observation selections. Polarization data are collected from multiple instruments (IXPE, PoGO+, XL-Calibur, AstroSat CZTI, INTEGRAL/IBIS and SPI) with widely differing energy binning, statistical uncertainties, and time coverage but assumed long-term stability in the hard state.

The key modeling innovation is a forward-folding approach within the 3ML Python library to produce predicted counts from spectral component models, paired with a rigorous polarization mixing procedure: input polarization fractions and angles are converted to normalized Stokes parameters (q,u) per component on a fine logarithmic energy grid (300 points from 1 to 2500 keV), summed over components weighted by flux and then integrated over observational bins. This accounts for nonlinear mixing of polarization, which can cause depolarization or angle shifts when emissions of differing polarization superpose.

The spectral model includes two components: a thermal Comptonization component modeled with comptt, characterized by electron temperature, optical depth, and seed photon temperature; and a highly polarized synchrotron hard tail modeled as a stretched exponential cutoff power-law electron distribution producing a synchrotron spectrum with an energy-dependent cutoff and polarization fraction.

The IC polarization is modeled phenomenologically with a log-sigmoid function that starts low and saturates around 15% polarization at higher energies, consistent with Comptonization theory and constrained from spectral opacity. The synchrotron polarization is computed numerically integrating over the electron energy distribution weighted by synchrotron kernel functions to capture energy-dependent polarization near cutoff energies, going beyond the usual power-law constant PF assumption.

The combined spectral plus polarization χ² is minimized using Levenberg-Marquardt implemented in the lmfit Python library, treating spectral count residuals and normalized Stokes residuals with appropriate error propagation. Confidence intervals are estimated by fixing parameters and minimizing χ², and posterior parameter distributions explored via emcee MCMC with 100 walkers and 40,000 steps.

The fitting procedure jointly constrains 11 free parameters including IC electron temperature, optical depth, synchrotron cutoff energy, normalization, PF parameters, and polarization angles. Final fits achieve reduced χ² close to unity, with parameter correlations and uncertainties characterized. The method explicitly addresses mixing effects and differences in instrument energy binning, statistical weight, and polarization representation. The full workflow is reproducible using released Python libraries 3ML, lmfit, and emcee.

One concrete example: for each energy grid point, the model predicts component spectra and polarization (Q,U), integrates over instrument bins using rebinning matrices to match measured polarization bins, normalizes by total intensity, and compares to observed normalized Stokes parameters to compute χ² residuals — enabling consistent joint fit of spectral and polarization data from multiple instruments with different resolutions and sensitivities.

Technical innovations

• Development of a unified forward-folding approach combining spectral and polarization data from multiple independent instruments with different energy binning and statistical quality.
• Rigorous polarization mixing algorithm that converts polarization fractions and angles into additive normalized Stokes parameters summed over all emission components before integration over detector bins.
• Modeling synchrotron polarization fraction as energy-dependent near the electron distribution cutoff using numerical integration over synchrotron kernels, extending beyond classical constant-PF power-law approximations.
• Phenomenological log-sigmoid function to model energy-dependent Comptonization polarization saturation constrained by spectral opacity parameters.
• Application of combined spectro-polarimetric fitting with MCMC to constrain both spectral and polarization parameters simultaneously, enabling physically meaningful parameter uncertainties and correlations.

Datasets

• Cygnus X-1 spectral data — over a decade of INTEGRAL observations (JEM-X, ISGRI, SPI) — public archive
• Cygnus X-1 polarization data — compiled from multiple instruments (INTEGRAL/IBIS, INTEGRAL/SPI, AstroSat/CZTI, IXPE, PoGO+, XL-Calibur) — various published observational datasets

Baselines vs proposed

• Spectrum-only fit: χ²/Nbins ~1.0 compared to joint spectral+polarization fit χ²/Nbins ~1.07 (98 DOF), illustrating improved constraints but no major fit degradation.
• Polarization fraction saturation for IC constrained at ΠM=15% based on plasma opacity τp=1.1, compared to assumed lower PF near 0% below 10 keV.
• Synchrotron cutoff energy fixed by combined fit at (3.9⁺⁰.⁶_−₀.₅)×10² keV, improving over previous broad estimates without polarization constraints.
• Polarization angle (PA) misalignment of ~40° between synchrotron hard tail and jet axis identified and modeled, not explained by simpler models ignoring polarization mixing.

Limitations

• The spectral model used is relatively simple, excluding reflection and detailed disk components which could improve spectral residual balance.
• Polarization data from different instruments are not strictly simultaneous but treated as stable integrated properties over years, which may hide temporal variations.
• The phenomenological modeling of IC polarization excludes complex energy-dependent variations above ∼100 keV due to lack of robust polarization data there.
• Synchrotron electron super-exponential index β is fixed to 2 due to insufficient high-energy data to constrain it from observations.
• No explicit modeling of circular polarization (V parameter) as it is assumed negligible; this assumption is common but could miss some physics.
• The method depends on accurate instrument response matrices and assumes Gaussian uncertainties in normalized Stokes parameters, which might not capture all systematics.

Open questions / follow-ons

• How would inclusion of reflection and more complex accretion disk spectral components affect combined spectro-polarimetric fits and parameter inference?
• Can contemporaneous or time-resolved spectro-polarimetric observations confirm the assumption of stable polarization properties over years?
• What is the impact of circular polarization measurements, if available, on the interpretation of magnetic field geometry and acceleration mechanisms?
• How does the proposed explanation of polarization angle misalignment scale to other black hole X-ray binaries or accreting compact objects?

Why it matters for bot defense

Although this work is astrophysics-focused and does not directly address bot defense or CAPTCHA technologies, its rigorous approach to combining heterogeneous, multi-instrument measurements with differing resolutions and statistical qualities may inform data fusion methodologies in bot detection pipelines. The polarization mixing algorithm illustrates careful treatment of asynchronous and differing-precision data sources, a challenge analogous to integrating diverse behavioral signals in bot detection. Additionally, the joint fitting framework using forward-modeling and MCMC could inspire analogous approaches to simultaneously model multiple modalities of bot/fraud signals with mixed uncertainties. However, the domain-specific physical modeling limits direct applicability, so practitioners should focus on methodological parallels in multi-source data fusion rather than content.

Cite

@article{arxiv2608_06328,
  title={ A unified modeling of X-ray and gamma-ray spectro-polarimetric data: the case of Cygnus X-1 },
  author={ Tristan Bouchet and Thomas Siegert and Victoria Grinberg and Jérôme Rodriguez and Floriane Cangemi and Philippe Laurent and Joern Wilms },
  journal={arXiv preprint arXiv:2608.06328},
  year={ 2026 },
  url={https://arxiv.org/abs/2608.06328}
}

Read the full paper

• arXiv:2608.06328 (abstract)
• arXiv:2608.06328 (PDF)