Time lags and their association with the Boundary Layer structure in a Z source GX 349+2

Studying the cross-correlation function between the soft and hard X-ray emission in Neutron Star Low Mass X-ray Binaries provides crucial insight into the structure and dynamics of the innermost accretion regions. In this work, we investigate the CCF of the Z-source GX 349+2 using an XMM-Newton observation. We noted that asymmetric CCFs with lags of a few hundred secondsbetween soft and hard band LCs in the horizontal branch, whereas CCFs remained symmetric in normal and flaring branches. We also performed a CCF study during the flux transition duration and observed lags of the order of a few tens to hundreds of seconds. Monte Carlo simulations were performed to assess the robustness of these CCFs, confirming their significance at a 95% confidence level. Spectral analysis during the flux transitions further suggests that the inner accretion disk extends close to the last stable orbit. We propose that the observed hard lags arise from the readjustment of the boundary layer/coronal region located near the inner edge of the accretion disk. From the measured lags, we estimate the characteristic size of the boundary layer. We show that the observed lags could also be associated with the depletion timescale of the boundary layer with low viscosity.

💡 Research Summary

This paper presents a detailed timing and spectral investigation of the Z‑type low‑mass X‑ray binary GX 349+2 (also known as Sco X‑2) using a single XMM‑Newton EPIC‑pn observation performed on 19 March 2008 (ObsID 0506110101, total exposure 22.5 ks). The authors focus on the cross‑correlation function (CCF) between soft (0.8–2 keV) and hard (2–10 keV) X‑ray light curves, aiming to probe the physical connection between the accretion disk, the neutron‑star boundary layer (BL), and any surrounding corona.

Data preparation and HID classification
Because the source is very bright, only the EPIC‑pn camera was used in Timing mode, with central columns excluded to mitigate pile‑up. Light curves were extracted with 50 s bins, corrected for dead time and exposure variations using the epiclccor task, and divided into eleven ≈2000 s segments. A hardness‑intensity diagram (HID) was constructed using the ratio of 2–10 keV to 0.8–2 keV count rates, allowing the identification of three canonical Z‑track branches: the horizontal branch (HB), the normal branch (NB), and the flaring branch (FB).

Cross‑correlation analysis
For each segment the CCF between the soft and hard light curves was computed with the XANADU crosscorr tool. The HB segments (designated Sections 1 and 2) displayed markedly asymmetric CCFs, with the hard band lagging the soft band by several hundred seconds and relatively low peak correlation coefficients. In contrast, the NB and FB segments (Sections 3–6) produced symmetric CCFs that peaked at zero lag, indicating near‑simultaneous variations in both bands. Additional “transition” windows (Sections A, B, C), each about 2200 s long and covering rapid changes between HID branches, also showed asymmetric CCFs with statistically significant lags.

Statistical validation
To confirm that the observed lags are not artefacts of stochastic variability, two Monte‑Carlo approaches were employed. First, synthetic soft and hard light curves were generated by adding Gaussian noise (σ equal to the measurement uncertainties) to each observed data point, preserving the intrinsic variability pattern. Ten thousand such pairs were produced; each pair’s CCF was fitted with a linear baseline plus a Gaussian peak, and the distribution of peak lags was examined. Second, the Timmer & König (1995) method was used to create 10 000 pairs of red‑noise light curves with a power‑spectral index Γ≈2.1, matching the observed power‑law noise. The resulting CCF ensemble defined a 95 % confidence envelope. In both cases the real CCFs from the HB and transition windows lay well outside the 95 % bounds, demonstrating that the asymmetries and lags are intrinsic to the source.

Spectral analysis during lag intervals
The authors extracted EPIC‑pn spectra (0.8–10 keV) for the start and end of each transition window (first and last 500 s). Two spectral models were fitted using XSPEC v12.12.1:

1. Model 1: Tbabs × (bbodyrad + powerlaw + Gaussian).
2. Model 2: Tbabs × (bbodyrad + diskbb + Gaussian).

The hydrogen column density was fixed at N_H = 0.8 × 10²² cm⁻² (consistent with previous work). The Gaussian component models the Fe Kα line (fixed at ~6.7 keV). Markov Chain Monte Carlo (Goodman‑Weare algorithm, 200 k steps, 50 k burn‑in) was used to explore parameter uncertainties.

Key findings from the fits:

• The blackbody temperature (kT_bbr) varies between ≈1.1 keV and ≈1.4 keV across the transition intervals, while its normalization (N_bbr), which scales with the emitting area (∝ R²), changes by up to a factor of two. This suggests that the physical size of the BL (or the region where the neutron‑star surface emission dominates) expands and contracts on timescales comparable to the observed lags.
• The power‑law photon index (Γ) shows modest but statistically significant variations (e.g., from 1.70 ± 0.03 to 1.60 ± 0.03) in sections A and C, indicating changes in the Comptonizing corona or in the relative contribution of the BL versus the corona.
• The diskbb component in Model 2 (when required) yields inner‑disk temperatures around 1.1–1.5 keV and normalizations implying that the inner accretion disk extends close to the neutron‑star surface, consistent with previous reports of kHz QPOs in GX 349+2.

Physical interpretation of the lags
The authors argue that the hard‑band lags observed on the HB and during branch transitions are not due to simple light‑travel delays (which would be ≤ ms) but rather reflect dynamical readjustments of the BL/corona system. Two scenarios are discussed:

1. Viscous readjustment: The inner accretion flow (disk + BL) experiences a change in mass‑accretion rate; the viscous propagation time from the truncation radius to the BL can be of order 10² s for a low‑viscosity, geometrically thin flow. This matches the measured lags.

2. Boundary‑layer depletion: If the BL has a low effective viscosity (α ≈ 10⁻⁷), the depletion timescale t_depl = (Ω − Ω_NS) α / (Ω_k² − Ω²) can also be hundreds of seconds, where Ω is the local angular frequency, Ω_NS the neutron‑star spin, and Ω_k the Keplerian frequency at the BL radius. Using the measured changes in N_bbr to infer the BL radius (≈10–40 km), the authors obtain depletion times consistent with the observed lags.

These interpretations echo earlier RXTE results (Ding et al. 2016) that reported similar hard lags on the HB and attributed them to viscous timescales in the inner flow. The present XMM‑Newton analysis extends the energy coverage down to 0.8 keV, allowing a more robust separation of soft thermal emission (disk/BL) from the harder Comptonized component.

Connection to jet activity
The paper draws a parallel with recent work on Sco X‑1 (Gouse et al. 2025a,b), where asymmetric CCFs with long lags were linked to ballistic jet ejections, while symmetric CCFs coincided with ultra‑relativistic flow (URF) events. The authors suggest that the asymmetric CCFs observed on the HB of GX 349+2 may similarly indicate a disrupted inner flow associated with jet launching, whereas the symmetric CCFs on the NB/FB reflect a more stable accretion configuration. Although no simultaneous radio data are available for this observation, the analogy supports a broader picture in which the BL/corona geometry, jet production, and timing properties are tightly coupled in Z‑sources.

Conclusions

• Asymmetric CCFs with hard lags of a few hundred seconds are detected exclusively on the HB and during rapid HID transitions, with > 95 % statistical significance.
• Symmetric, zero‑lag CCFs dominate on the NB and FB, indicating simultaneous soft and hard variability.
• Spectral fits reveal concurrent changes in the BL temperature, emitting area, and coronal photon index, consistent with a dynamic BL whose size varies on the same timescales as the lags.
• The measured lags can be interpreted as viscous readjustment times or low‑viscosity depletion times of the BL, yielding an inferred BL size of ≈10–40 km.
• The findings align with previous RXTE studies and with recent jet‑disk coupling results in other Z‑sources, suggesting that the HB may correspond to a state where the inner accretion flow is perturbed, possibly by jet launching, while the NB/FB represent a more stable configuration.

Overall, this work provides compelling evidence that cross‑correlation asymmetries are a powerful diagnostic of the inner accretion geometry in neutron‑star LMXBs, and it offers quantitative constraints on the physical dimensions and viscous properties of the boundary layer.