Evolution of the recent high-accretion state of the recurrent nova T CrB: HST, Swift, NuSTAR, and XMM-Newton observations

TL;DR

This study presents a comprehensive, multiwavelength view of the recurrent nova T CrB as it enters a super-active state (SAS) and transitions to a faint state. By integrating HST, Swift, NuSTAR, XMM-Newton, and AAVSO optical data, it demonstrates that the SAS is driven by an increased mass accretion rate, causing the boundary layer to become optically thick and producing a soft X-ray (blackbody-like) component, while the optically thin cooling-flow emission rises as the system approaches eruption. The analysis finds no evidence for a ~6000 s periodic signal after accounting for red noise, argues against dust obscuration for the dip, and highlights intrinsic accretion-related variability as the dominant driver of the observed changes. A self-consistent framework links the SAS evolution to accretion-disk structure changes and boundary-layer dynamics, though precise eruption timing remains uncertain due to large scatter in recurrence intervals. The work provides a solid baseline for predicting pre-eruption behavior and informs models of mass accumulation in quiescent phases of recurrent novae.

Abstract

As the recurrent nova T Coronae Borealis (T CrB) approaches its next predicted thermonuclear eruption, it is currently exhibiting a "super-active state" (SAS) characterized by enhanced multiwavelength emission similar to the behavior recorded prior to the 1946 outburst. We present a multiwavelength analysis of the SAS and the subsequent "faint state" using observations from HST, Swift, NuSTAR, and XMM-Newton. Our results indicate that the SAS was driven by an increase in the mass accretion rate, which caused the accretion disk's boundary layer to become optically thick. A weighted least squares regression analysis quantifies the evolution of the accretion components, displaying a highly significant (4.5$σ$) increase in the luminosity of the optically thin cooling flow (L$_{cf}$) and a marginal (2.58$σ$) decrease in the optically thick boundary layer luminosity (L$_{bb}$) as the system transitioned into the faint state. We find that this dimming is consistent with an intrinsic change in the accretion flow rather than dust obscuration, supported by the lack of infrared excess and the stability of the 2175 Å feature. Additionally, a time-series analysis using autoregressive modeling to account for correlated red noise revealed no significant periodicities, thereby disputing the previously reported $\sim$6000 s signal. These findings suggest that the pre-outburst evolution of T CrB is characterized by significant changes in the accretion disk structure and boundary layer, providing a self-consistent physical framework for the system's behavior as it approaches eruption.

Figures (20)

• Figure 1: Long-term, multiwavelength light curves of T CrB since March, 2012 until August 2025. From top to bottom: ($i$) Swift/XRT light curve in the 0.3-10 keV energy range; ($ii$) Swift/BAT 14–50 keV light curve with 100-day bins. ($iii$) AAVSO B-band light curve (Johnson magnitudes). Downward arrows mark the dates of the XMM-Newton (blue), HST (orange), and NuSTAR (red) observations. Note: Because some XMM-Newton and HST observations were quasi-simultaneous, their marks overlap in this panel and we only mark the dates of those HST observations flagged as Good in Table \ref{['tab:hst']}. The purple-shadowed region comprises the period known as the SAS, while the pink-shaded region marks the period when the optical/UV brightness dropped to pre-SAS levels, known as the dip, and the green-shadowed region shows the period that started around late 2023 when the brightness in optical/UV has started to rise again in the recovery phase; ($iv$) Swift/UVOT light curve in the $uvm2, uvw2, uvw1, u$ filters; ($v$) AAVSO B-V color, where red indicates a redder, cooler emission, while blue shows the opposite.
• Figure 2: First and second columns:XMM-Newton/pn and NuSTAR (in red) spectra of T CrB taken during the SAS. The model parameters, listed in Table \ref{['tab:1']}, were obtained by simultaneously fitting the pn, MOS 1 and MOS 2 spectra during each pointing. Here we show only the $pn$ spectra for clarity. Third column:NuSTAR/FPMA, XMM-Newton, and Swift spectra of T CrB taken during the faint state after the SAS. The Swift spectra were grouped at one count per bin and modeled using Cash statistic.
• Figure 3: Evolution of the luminosity from the black-body (red points) and hard X-ray emitting, cooling flow (black points) spectral components over time. The shaded areas mark the SAS, Dip and Recovery phase using the same color code as in Fig. \ref{['fig:xrtlc']}. A WLS fit to the luminosity of the optically thick (BB) component (dotted red line) indicates that its slope is different from zero at the 2.58$\sigma$ level. We include Nu2 and Nu3 observations taken in 2024 and Obs7 taken in 2025, whose spectra arise from the optically thin portion of the boundary layer, i.e., the cooling flow. Dotted black lines mark the WLS fit to the $L_{CF}$ evolution whose slope is different than zero at the 4.49$\sigma$ level. The increase in its luminosity suggests that the accretion rate in the disk decreased over time, with the boundary layer becoming optically thinner and hotter.
• Figure 4: Mg II doublet. IUE spectra from before the SAS are plotted together with the HST STIS spectra from the end of the SAS period. Two small black lines are added to guide the eye to the absorptions that match the two Mg II profiles: blueshifted outflow in the line of sight for the 2022-09-09 (MJD59469, phase 0.56) spectrum, but not in the 2022-09-09 (MJD 59831, phase 0.36) spectrum. Note: the IUE Mg II spectra seem to be missing the blue wing. The background during SAS is higher than in the IUE spectra. The flux scale is listed above the left corner of the image.
• Figure 5: Time-series analysis of the Obs1 XMM-Newton/$pn$+MOS1+MOS2 light curve of T CrB. (a) Full, original light curve, showing significant aperiodic variability (red noise) across the observation. (b) ACF of the original mean-subtracted data. The slow decay confirms the presence of strong, correlated red noise. (c) PACF of the original data. The significant power concentrated in the first few lags suggests that an autoregressive (AR) process is a suitable model for the red noise. (d) ACF of the residuals after subtracting the best-fit AR(p) model. The lack of significant correlation spikes indicates that the model has successfully "whitened" the data, removing the dominant red noise component. (e) PACF of the residuals. Similarly to the residual ACF, the absence of significant spikes confirms the goodness-of-fit of the AR(p) model. (f) log-log plot of the LS power spectrum of the original data (blue), which shows a classic red-noise profile with power increasing toward lower frequencies. The theoretical power spectrum of the fitted AR(p) model (red) are overplotted, demonstrating an excellent match to the stochastic properties of the light curve. (g) LS power spectrum of the "whitened" residuals. This plot should reveal periodic signals that were previously buried under the red noise. The horizontal dashed line marks the 99.99% FAP level. The most prominent peak does not rise significantly above this threshold, indicating that its periodicity is not statistically significant. The arrow marks the frequency corresponding to the 6000 s period reported by zhekov19.