Low-frequency spectra of neutron star + OB supergiant binaries: Does wind density drive persistent and flaring modes of accretion?

TL;DR

This paper presents the first coordinated millimeter and radio survey of twelve neutron star high-mass X-ray binaries (SgXBs, SFXTs, and intermediates) using ALMA, NOEMA, and the VLA to study the origin of their low-frequency emission. The results indicate that the detected low-frequency spectra are dominated by thermal free-free emission from the OB donor winds, with no compelling evidence for jet-dominated emission at these frequencies, and show that SFXTs are systematically underluminous at 100 GHz compared to classical SgXBs, likely due to less dense winds. By comparing mm-derived wind constraints with bow-shock measurements and literature wind parameters, the study finds plausible wind-acceleration effects and radial clumping as explanations for observed deviations from simple wind models, and identifies long-term wind variability on multi-day timescales. The findings imply that wind density and velocity structure, rather than accretion physics alone, play a key role in differentiating accretion modes between SgXBs and SFXTs, with important implications for wind physics and binary evolution in HMXBs.

Abstract

Neutron star high-mass X-ray binaries are well-studied in wavebands between the infrared and hard X-rays. Their low-frequency millimeter and radio properties, on the other hand, remain poorly understood. We present observations of the millimeter and radio emission of binaries where a neutron star accretes from an OB supergiant. We report ALMA and NOEMA millimeter observations of twelve systems, supplemented by VLA radio observations of six of those targets. Our targets include six Supergiant X-ray Binaries (SgXBs), four Supergiant Fast X-ray Transients (SFXTs), and two intermediate systems. Nine out of twelve targets, including all SFXTs, are detected in at least one millimeter band, while in the radio, only two targets are detected. All detected targets display inverted radio/millimeter spectra, with spectral indices in the range $α=0.6-0.8$ for those systems where accurate SED fits could be performed. We conclude, firstly, that the low-frequency SEDs of neutron star SFXTs and SgXBs are dominated by free-free emission from the OB supergiant's stellar wind, and that jet emission is unlikely to be observed unless the systems can be detected at sub-GHz frequencies. Secondly, we find that SFXTs are fainter at 100 GHz than prototypical SgXBs, probably due to systematically less dense winds in the former, as supported further by the differences in their fluorescence Fe K$α$ lines. We furthermore compare the stellar wind constraints obtained from our millimeter observations with those from IR/optical/UV studies and bow shock detections, and present evidence for long-term stellar wind variability visible in the thermal emission.

Figures (10)

• Figure 1: The low-frequency spectral energy distributions of all twelve targets, separated into SgXBs (left) and SFXTs / intermediates (right). For each source, we plot all available ALMA and NOEMA data, including those from Van den Eijnden et al. (2023). We also show the coordinated VLA flux density (limits) for the six ALMA targets. At several frequencies, upper limits overlap; in all those cases (particularly the 6/10 GHz bands in the right-hand panel) the overlapping points all represent upper limits. In the left-hand panel at 100 GHz, above the two overlapping limits, two detections are plotted at overlapping flux density.
• Figure 2: The two ALMA images at 40 GHz (left) and 150 GHz (middle), and the NOEMA image at 100 GHz (right) of the field of 4U 1907+09. The green circle in each panel is centred on the position of the SgXB, which is used as the plotting center of each panel. With increasing frequency, the field of view shrinks, causing the pointing centre to shift to the upper right in the ALMA images. While formally not $>3\sigma$ significant in both ALMA images, we measure its flux and report it as a detection due to its consistent presence in all images covering the position; in particular, the NOEMA 100-GHz image shows a millimeter counterpart at $\gtrsim 4\sigma$. The ALMA 300-GHz image, where the source position is not covered, is shown in Figure \ref{['fig:all2']}.
• Figure 3: Fits of a power-law radio / millimeter spectrum $S_\nu = \xi S_0 (\nu / \nu_0)^{\alpha}$ to the six ALMA targets. The left and right panels show the three SgXBs and SFXTs, respectively, where we stress the difference in vertical extent. The shaded regions and dashed lines indicate the $1\sigma$ confidence regions of the MCMC fits, including only the ALMA data and both the ALMA and VLA data, respectively. Below each spectral fit, the residuals are shown as data divided by model, for the fit to the ALMA and VLA data. Note that we plot the 10-GHz point for 4U 1700-37 but do not use it in the MCMC fit.
• Figure 4: An order of magnitude comparison of the jet and wind luminosity assuming wind capture in a circular orbit and the measured Be/X-ray binary coupling between accretion rate and jet power. The legend indicates deviations from the standard input, shown as the black line ($v_\infty = 1000$ km/s, $\nu = 10$ GHz, $M_{\rm donor} = 20$$M_\odot$, $P_{\rm orb} = 10$ days, $\alpha_{\rm jet} = 0$). Despite the assumptions underlying this calculation, the jet is unlikely to be visible above $1$ GHz unless interactions with the wind brighten it significantly.
• Figure 5: Left: The 100-GHz millimeter luminosities versus orbital period. SgXBs are shown as squares; SFXTs as circles; intermediate sources as pentagons. The uncertainties on the millimeter luminosity include both flux density and distance uncertaintes. Right: the millimeter luminosity plotted versus the Equivalent Width of the Fe K$\alpha$ line at $6.4$ keV. Archival points are shown based on vandeneijnden2021, under the assumption that these extrapolated radio luminosities are dominated by stellar wind emission. In both panels, the grey band indicated the range covered by SFXTs.