Reflection-Based Spin Inference

X-ray reflection spectroscopy uses fluorescent lines, absorption edges, and the Compton-scattered continuum produced when a hard coronal spectrum illuminates an optically thick disk. Relativistic energy shifts encode the inner emitting radii, but spin is one step farther downstream: it follows only if the fitted inner edge is the Kerr ISCO. Continuum curvature, absorption, density and ionization gradients, elemental abundance, illumination geometry, disk thickness, and calibration can imitate part of the same broadening. The robust target is therefore a response-folded inner-radius conclusion that survives competing physical models.

Required background. Probability and Inference for Gravity supplies likelihood construction, nuisance parameters, predictive checking, and model comparison.

Helpful background. Black-Hole Geometry and Mechanics supplies Kerr photon transfer and the ISCO; Relativistic Matter and Compact Objects supplies disk atmospheres and radiative microphysics; and Black-Hole Astrophysics supplies accretion states, coronae, and absorption.

Reflection before relativistic broadening

At disk radius $r$, an illuminating flux $F_X$ and electron density $n_e$ define the ionization parameter

$$\xi(r)=\frac{4\pi F_X(r)}{n_e(r)}.$$

A rest-frame calculation returns an angle-dependent specific intensity

$$I_{E_{\rm em}}^{\rm refl} =I^{\rm refl}\!\left( E_{\rm em},\mu_{\rm em}; \Gamma,E_{\rm cut},\xi,n_e,A_{\rm Fe},\eta_{\rm atom} \right),$$

where $\Gamma$ and $E_{\rm cut}$ describe the incident continuum, $A_{\rm Fe}$ is iron abundance, and $\eta_{\rm atom}$ labels atomic data and radiative-transfer choices. Photoelectric absorption followed by fluorescence creates the Fe K complex; Compton scattering creates a broad excess at tens of keV. These features must be calculated together. Treating a red wing as an isolated line discards the broadband energy balance that helps separate reflection from absorption or continuum curvature.

Relativistic transfer maps the local spectrum to infinity:

$$F_{E_{\rm obs}}^{\rm refl} =\frac{1}{D^2} \int_{\rm image} g^3\, I_{E_{\rm obs}/g}^{\rm refl} \!\left(r,\mu_{\rm em}\right)d\alpha\,d\beta .$$

The image-plane coordinates $(\alpha,\beta)$ are physical impact lengths, so $d\alpha\,d\beta/D^2$ is a solid-angle element. The ratio $g=E_{\rm obs}/E_{\rm em}$ includes gravitational and Doppler shifts. The local intensity here includes the illuminating flux normalization as well as its effect on ionization; it must not be multiplied by that same flux a second time. A library of unit-normalized spectral shapes instead needs one explicit radial illumination weight. Spin affects both $g$ and the allowed disk edge, while inclination also changes Doppler asymmetry and the local emission angle. Angle averaging can therefore bias a relativistically blurred fit; the construction and its limits are explicit in García et al. 2014, §§ 2–4.

The response-space forward model

For detector channel $j$ and observing interval $k$, a minimally adequate expectation is

$$\lambda_{jk}=t_k\int \frac{R_{jk}(E)}{E} A_k^{\rm Gal}(E) \left\{ A_k^{\rm src}(E) \left[F_k^{\rm cor}(E)+F_k^{\rm relrefl}(E)\right] +F_k^{\rm distant}(E) \right\}dE+B_{jk}, \qquad C_{jk}\sim{\rm Poisson}(\lambda_{jk}).$$

Here $R_{jk}$ is the redistribution times effective area, $A_k^{\rm Gal}$ attenuates every source component, $A_k^{\rm src}$ represents source-frame absorption or partial covering of the compact corona and disk, $F^{\rm cor}$ is the intrinsic coronal continuum, and $B_{jk}$ is the expected background count. All $F(E)$ terms are energy flux densities, consistently with the transfer equation. The factor $1/E$ converts them to photon number flux, with $E$ in the same energy units used by $F(E)$; remove that factor when using spectral-model outputs already in photon units. Source and background counting uncertainty must both be propagated rather than treating a subtracted spectrum as exact.

If the distant reflector is covered by the source absorber in the proposed geometry, its term belongs inside $A_k^{\rm src}$ as well. The fit should share spin, inclination, abundance, and a physically stable inner edge across intervals while permitting absorption, coronal normalization, and ionization to vary on their own timescales. The disk density, radial ionization profile, emissivity or coronal geometry, and returning radiation must be either fitted or declared fixed.

The four coordinated 2012–2013 NuSTAR and XMM-Newton observations of NGC 1365 form a useful adversarial data set because line-of-sight absorption changes dramatically while the reported inner-disk parameters remain stable. The historical reduction used NuSTARDAS 0.11.1 with NuSTAR CALDB 20130509 and SAS 13.0.0; extraction and response details are given in Walton et al. 2014, §§ 2.1–2.2. A modern reproduction should preserve those products for comparison and separately rerun the public data with HEASoft 6.36, XSPEC 12.15.1, the resolved NuSTAR CALDB index, and a declared SAS calibration release. It must not mix response versions within a nominally shared spectrum.

Selection includes flux and absorption state, simultaneous time intervals, detector good-time intervals, source and background regions, pile-up and flare cuts, energy band, grouping or Poisson likelihood, and any decision to include a spectrum because it visibly has a broad residual. Selecting on the desired feature exaggerates its apparent prevalence. Gain, effective-area edges, redistribution, background, cross-normalization, and pile-up belong in the calibration model, especially where astrophysical absorption edges compete with detector structure.

Results and limitations through 2026-08-18

The evidence does not establish a model-independent spin for every broad spectrum. The unresolved issue is how much of an apparently small inner radius is stable under credible disk, corona, absorption, and calibration families in the relevant signal-to-noise and ionization regime.

Evidence or result	Discriminating axis	Scope or limitation
Atomic reflection predicts linked Fe K emission, edges, and a Compton hump	Fit the full broadband spectrum with one abundance, ionization structure, and illuminating continuum	Atomic tables assume a density, composition, slab structure, and incident spectrum
NGC 1365 shows correlated broad-line and hard-hump behavior across variable absorption states	Joint NuSTAR plus XMM-Newton intervals; require stable inner edge and inclination	The inference remains conditional on partial-covering, continuum, emissivity, and disk models
Monte Carlo circumnuclear reprocessing can reproduce substantial hard curvature	Compare geometry-aware absorption and scattering, not slab attenuation alone	A flexible distant medium can be weakly identifiable and needs line and variability predictions
Blind simulations recover some prograde spins from high-quality reflection spectra	Vary corona height, ionization, spin, response, and starting point	Recovery degrades for other parameter regimes and tests only the simulation family used
High-density, radially ionized, thick, or truncated disks can shift the inferred edge	Refit with alternative density, ionization-gradient, emissivity, and thickness families	Present grids cover finite atomic and geometric domains; null shifts in one source do not generalize

The NGC 1365 debate is instructive rather than embarrassing. Walton et al. found consistent inner parameters across four absorption states in Walton et al. 2014, §§ 3–4, whereas Miller and Turner showed that geometry-aware transmission and scattering can challenge simplified absorption comparisons in Miller and Turner 2013, §§ 2–4.3. The competing positions now include disk reflection reaching the ISCO; truncation; complex absorption or continuum curvature; density and ionization gradients; nonstandard or extended illumination; returning radiation; and calibration error. Broadband coverage and variability constrain these alternatives, but they do not make the source model unique by assertion.

The interpretation could change after a response or atomic-data revision moves the inner-radius posterior; after higher-resolution broadband spectra resolve the absorption or line complex; after reverberation constrains a different coronal geometry; or after evidence for disk thickness, truncation, or returning radiation invalidates the fitted parameter regime.

Model challenges and reproduction boundary

Infer radius before spin. Fit $r_{\rm in}/(GM/c^2)$ without enforcing $r_{\rm in}=r_{\rm ISCO}(a_*)$. Report whether alternative continua, absorbers, densities, ionization gradients, and emissivity families agree on the radius. Only then apply the Kerr mapping.

Predict held-out structure. Fit below 10 keV and predict the hard Compton hump, or fit selected intervals and predict a held-out absorption state. A model that gains spin sensitivity only by absorbing out-of-band failure into cross-normalization is not validated.

Test coverage. Simulate counts from each serious competing family through the same responses and refit them with the baseline model. A nominal $90\%$ interval that contains the injected radius much less often than $90\%$ is overconfident even if the real-data residuals look smooth.

A truthful reproduction publishes the event identifiers and checksums, good-time and extraction selections, source and background spectra, response matrices, calibration indices, atomic-table and relativistic-transfer versions, exact model expression, priors, posterior samples, and count-space predictive residuals. Proprietary table models or unrecorded manual interval choices are explicit barriers to exact recovery. Statistical uncertainty is the width conditional on a model family; the spread across plausible physical and calibration families is systematic model dependence.

Exercises

A fit reports $r_{\rm in}=2.0^{+0.3}_{-0.2}GM/c^2$ after imposing a single ionization zone. An alternative radial-ionization model gives $r_{\rm in}=3.1^{+0.7}_{-0.5}GM/c^2$ with equally acceptable predictive residuals. Is it defensible to combine only the first radius with the Kerr ISCO relation and quote its narrow spin interval?

Solution

No. The two fits answer different formulation choices, and the data have not selected one strongly enough to treat the other as excluded. The appropriate intermediate result is a model-conditional or model-averaged inner-radius inference that exposes the radial-ionization choice. Because prograde $r_{\rm ISCO}$ decreases with increasing spin, the second model maps to a lower spin and a wider interval. Quoting only the first result would convert systematic model dependence into false precision. A stronger diagnostic is to identify energy bands or time intervals where the models make different count predictions and test those predictions on held-out data.

References

• García, Javier A., et al. 2014. “Improved Reflection Models of Black-Hole Accretion Disks: Treating the Angular Distribution of X-Rays.” The Astrophysical Journal 782:76. DOI.
• HEASARC. n.d. “HEASoft: A Unified Release of the FTOOLS and X-Ray Analysis Software Packages,” version 6.36. Official software page.
• Miller, L., and T. J. Turner. 2013. “The Hard X-Ray Spectrum of NGC 1365: Scattered Light, Not Black Hole Spin.” The Astrophysical Journal Letters 773:L5. DOI.
• Reynolds, Christopher S. 2021. “Observational Constraints on Black Hole Spin.” Annual Review of Astronomy and Astrophysics 59:117–154. DOI.
• Walton, Dominic J., et al. 2014. “NuSTAR and XMM-Newton Observations of NGC 1365: Extreme Absorption Variability and a Constant Inner Accretion Disk.” The Astrophysical Journal 788:76. DOI.