Detectability of low energy X-ray spectral components in type 1 AGN

In this paper we examine the percentage of type 1 AGN which require the inclusion of a soft excess component and/or significant cold absorption in the modelling of their X-ray spectra obtained by XMM-Newton. We do this by simulating spectra which mimic typical spectral shapes in order to find the maximum detectability expected at different count levels. We then apply a correction to the observed percentages found for the Scott et al. (2011) sample of 761 sources. We estimate the true percentage of AGN with a soft excess component to be 75+/-23%, suggesting that soft excesses are ubiquitous in the X-ray spectra of type 1 AGN. By carrying out joint fits on groups of low count spectra in narrow z bins in which additional spectral components were not originally detected, we show that the soft excess feature is recovered with a mean temperature kT and blackbody to power-law normalisation ratio consistent with those of components detected in individual high count spectra. Cold absorption with nH values broadly consistent with those reported in individual spectra are also recovered. We suggest such intrinsic cold absorption is found in a minimum of ~5% of type 1 AGN and may be present in up to ~10%.

💡 Research Summary

This paper investigates how common low‑energy X‑ray spectral components—specifically a soft excess and intrinsic cold absorption—are in type 1 active galactic nuclei (AGN). The authors start from the large, serendipitous XMM‑Newton sample compiled by Scott et al. (2011), which contains 761 type 1 AGN with optical identifications from the SDSS DR5 catalogue. In the original analysis only about 8 % of the sources required a soft‑excess component and roughly 4 % required an additional intrinsic neutral absorber, but those figures are lower limits because detection depends strongly on spectral quality (total counts) and redshift.

To quantify the true incidence, the authors performed extensive Monte‑Carlo simulations. They defined a “typical” AGN spectrum based on the 55 sources that originally needed a soft excess: a power‑law with photon index Γ≈1.8, a black‑body soft excess with temperature kT≈0.2 keV, and a black‑body‑to‑power‑law normalisation ratio of ≈0.04 (corresponding to a luminosity ratio of ~0.2). For each of five redshifts (z = 0.4, 0.75, 1.0, 1.3, 1.5) and for a range of total counts (from ~10³ to ~5 × 10⁴), they generated 1 000 fake spectra using XSPEC’s fakeit command, adding Poisson noise appropriate to the chosen exposure. Each simulated spectrum was then fitted with a simple power‑law (po) and with a power‑law plus black‑body (po+bb). An F‑test at the 99 % confidence level decided whether the soft excess was statistically required. The resulting detection fractions as a function of counts and red‑shift form a set of “detectability curves”.

These curves show that detection probability rises steeply with increasing counts and falls with increasing red‑shift, as expected because the soft excess is shifted out of the EPIC band at high z. For sources at z ≤ 1 with >10 000 counts the curves reach 100 % detection, indicating that the soft excess would be seen in every such spectrum if present.

The observed sample was then divided into count bins (75–320, 320–1000, 1000–3000, 3000–10 000, >10 000 counts). Within each bin the fraction of sources at each red‑shift was used as a weight to apply the appropriate detectability curve. The observed detection fraction in each bin was divided by the maximum detection probability (from the curves) to obtain a “corrected” intrinsic fraction. After weighting across red‑shifts, the corrected fractions are roughly constant across all count bins, hovering around 70–80 %. A χ² fit of a constant model yields an intrinsic soft‑excess incidence of 75 % ± 23 % (null‑hypothesis probability 91 %). This suggests that the soft excess is a ubiquitous feature of type 1 AGN, with the low observed percentages in the original work being purely a consequence of limited spectral quality.

A parallel analysis was performed for intrinsic cold absorption. Simulations without a soft excess but with a range of column densities showed that absorption can be detected up to ~25 % of high‑count spectra, but the detection fraction drops sharply at lower counts. After the same correction procedure, the authors infer that at least ~5 % of type 1 AGN harbour intrinsic neutral absorption, and the true fraction could be as high as ~10 %.

The paper discusses the implications: the high prevalence of soft excesses challenges models that treat it as a rare or exotic phenomenon and supports interpretations that link it to a common physical process (e.g., blurred ionised reflection or warm Comptonisation). The detection of intrinsic neutral absorption in a non‑negligible subset of type 1 AGN also calls for revisions to the simplest orientation‑based unified schemes, which predict negligible X‑ray absorption in unobscured objects.

Limitations are acknowledged. The simulations fix the soft‑excess temperature and normalisation ratio to median values, potentially under‑representing the full diversity of real sources. The reliance on the F‑test introduces a ~1 % false‑positive rate, which the authors test by simulating spectra without a soft excess; the spurious detection rate is generally consistent with this level, though it rises to ~5 % at the extreme low‑ and high‑count ends. Future work with higher‑resolution, higher‑throughput missions (e.g., Athena) and more sophisticated spectral models (including ionised absorbers and reflection) will be needed to refine these percentages.

In summary, by combining realistic simulations with a large, heterogeneous X‑ray sample, the authors demonstrate that soft‑excess emission is present in roughly three‑quarters of type 1 AGN, while intrinsic cold absorption is present in at least a few percent. Their methodology provides a robust framework for correcting detection biases in large surveys and highlights the importance of accounting for spectral quality and redshift when interpreting the prevalence of subtle spectral components.