CTCV J2056-3014: An X-ray-faint Intermediate Polar Harboring An Extremely Fast-spinning White Dwarf

J2056 was observed for about 18 ks on 2019 October 24 by XMM-Newton (ObsID 0842570101; PI: R. Lopes de Oliveira). The snapshot was focused on X-ray spectrophotometry with the European Photon Imaging Camera (EPIC), namely MOS1, MOS2, and pn cameras. The Reflection Grating Spectrometers RGS1 and RGS2 did not collect enough photons to allow us to carry out high-resolution X-ray spectroscopy. The UV observations with the optical monitor in timing mode barely covered the source position, rendering its data unusable for timing analysis.

The EPIC observations were partially contaminated by solar particles, with only a low background level during about 15.2 ks for the MOS cameras, and 8.7 ks for the pn camera. No pile-up or technical issues were identified in these data. The observations were reduced and data products were extracted following standard procedures using the Science Analysis System (SAS) v18.0.0. In particular, they were reprocessed using the epproc (for the pn data) and emproc (for the MOS1-2 data) tasks. We used calibration files downloaded on 2020 January 2. Spectral analysis was accomplished using the xspec software version 12.9.1m.

We revisited the optical photometric observations of J2056 presented by Bruch (2018) with the sole purpose of checking for high-frequency periodicities. This effort was motivated by the detection of pulsation in X-rays (Section 3.2). The optical observations were carried out on four nights in 2015 (June 9-12) and on two nights in 2016 (September 7-8) with the 0.6m Zeiss telescope of Observatório do Pico dos Dias – Laboratório Nacional de Astrofísica, Brazil. Light curves spanning from 55 to 340 minutes were obtained with a time resolution of 5 s.  To maximize the count rates within the short exposures, no filter was used. The throughput of the instrumentation corresponded roughly to V magnitude (Bruch, 2018). Basic data reduction (bias removal, flat-fielding) and aperture photometry were performed using default procedures with IRAF (Tody, 1986) and with the MIRA (Bruch, 1993, “MIRA: A Reference Guide”, Univ. Münster) software system, respectively.

The net count rates of J2056 at 0.3–12 keV were 0.194$\pm$0.004 counts s^-1, 0.202$\pm$0.004 counts s^-1, and 0.739$\pm$0.010 counts s^-1 for the MOS1, MOS2, and pn cameras, respectively. The spectra were binned such that each bin had at least 25 counts and thus the $\chi^{2}$ method was applied to both fit and test statistics in their modeling with xspec.

Even with relatively short exposures, the observations resulted in good-quality EPIC spectra (Fig. 1). The X-ray energy distribution of J2056 extends over the whole energy range covered by the EPIC cameras. An excess emission due to ionized lines of the Fe K$\alpha$ complex at 6.6–7 keV is seen in the pn data and supports the interpretation, as usual in accreting WDs, of the predominantly thermal nature for the X-ray emission. There is no evidence of an optically thick, blackbody-like component from the WD surface. Thus, we applied the following xspec models that account for the emission from collisionally ionized diffuse gas due to accretion as the primary energy source: apec, describing a single thermal plasma, and mkcflow, a multi-temperature model representing a cooling-flow (Mushotzky & Szymkowiak, 1988). The mkcflow model was interpolated by using the AtomDB (Foster et al., 2012) data so that it is equivalent in assumptions to those used in the apec model. We adopted the abundance table of Asplund et al. (2009). The phabs model was used to account for the photoelectric absorption effects on X-rays.

A single temperature component (phabs$*$apec) does not match the spectra well, failing to explain the continuum below 2 keV and the Fe K$\alpha$ complex ($\chi^{2}_{\nu}$ = 2.18). The inclusion of a second thermal component (phabs$*$(apec+apec)) improves the fit but fails especially in the description of the 0.9-1.5 keV region and slightly overpredicts the Fe K$\alpha$ lines ($\chi^{2}_{\nu}$ = 1.19). Despite similar statistics ($\chi^{2}_{\nu}$ = 1.18), the cooling flow model (phabs$*$mkcflow) visually improves the description of both continuum and iron lines. Finally, an acceptable description ($\chi^{2}_{\nu}$ = 1.09) is found by adding a single thermal component to the cooling flow model, providing a good fit of the 1 keV region, which is expected to be rich in unresolved emission lines. As for the mkcflow, we fixed the unconstrained $k$T_low parameter to its minimum value of 0.0808 keV and the required redshift parameter to 6.1$\times$10^-8 from the Gaia distance and standard cosmological values of xspec. Table 1 lists the best-fit spectral parameters of the models described above. Figure 1 shows the EPIC spectra and the final model phabs$*$(apec+mkcflow); henceforth, this is the model discussed in this letter.

Our best-fit model indicates that the X-rays are absorbed by the equivalent in hydrogen column density (N_H) of 2.7${}^{+0.7}_{-0.7}\times$10²⁰ cm^-2. The X-ray emission is dominated by a moderately hard thermal component that cools down from $k$T = 14.34${}^{+1.18}_{-1.21}$ keV. A secondary contribution is well described by a plasma component having $k$T = 0.79${}^{+0.04}_{-0.04}$ keV, which accounts for about 6.2% of the total unabsorbed flux at 0.3-12 keV. A subsolar abundance of 0.81${}^{+0.16}_{-0.14}$ $Z_{\odot}$ is inferred, forced to be the same for both thermal components during the fits, but its determination strongly depends on the Fe K$\alpha$ lines. From the mkcflow component, the mass accretion rate is  5.9${}^{+0.5}_{-0.4}\times$10^-12 M_☉ yr^-1. The total luminosity of the system at 0.3-12 keV is 1.8$\times$10³¹($d$/261.6 pc)² erg s^-1. These results are discussed in Section 4.

Time flags were converted to the Barycentric Dynamical Time scale using the online tool¹¹1See http://astroutils.astronomy.ohio‑state.edu/time of Eastman et al. (2010) for the optical observations and the barycen/SAS task for the X-ray data. Background-corrected X-ray light curves from each EPIC camera were produced considering a binning of 10 s. We considered three energy ranges: 0.3–10 keV, 0.3–2 keV (“soft”), and 2–10 keV (“hard”), in order to access the energy dependence of any variable signal. The light curves of each EPIC camera were investigated separately. Optical light curves were constructed retaining the original resolution of 5 s.

The search for periodicities was carried out using the Lomb-Scargle periodogram (Lomb, 1976; Scargle, 1982) in optical and X-ray light curves. We explored X-ray light curves considering two datasets. The first one incorporates the entire observations. It includes spikes in background count rates, which can be as high as the source signal, especially in the 2–10 keV band. This condition lasted for about 2.6 ks. The second one considers only data that were collected during the last $\sim$ 7.5 ks, the longest continuous time interval with low particle background.

A high-significance peak associated with a period of 29.6 s is clearly seen in the Lomb-Scargle periodogram applied to the 0.3-10 keV and 0.3-2 keV light curves of all EPIC cameras (Fig. 2). As for the 2-10 keV band, the peak is recovered only in light curves produced considering a time interval with low background contamination, and with a lower power when compared to results of the other two energy ranges. This is mainly due to the low signal-to-noise ratio in the hard energy range.

As reported by Bruch (2018), the optical light curve of J2056 displays variations on timescales of hours superposed by flickering. A search for high-frequency periodicities in optical data was first carried out individually for each of the six nights. The 29.6 s pulsation seen in X-rays is also identified in all optical light curves, and the periods in both regions are the same at the level of the formal errors. The upper left and center panels of Fig. 2 show the periodogram of the 2015 June 10 optical light curve. The optical data taken in subsequent nights do not indicate migration of phase. We refine the periodicity in the optical by merging the light curves of each of the two observing seasons, resulting in an average period of 29.6098$\pm$0.0014 s.²²2Note that the time difference between the two seasons is too large to concatenate the two data sets without cycle count ambiguities and thus to refine the period even more. The same value was derived from the 0.3-2 keV pn light curve considering the entire observation, but with an uncertainty of $\pm$0.0213 s.

Figure 2 (right frames) shows folded optical and X-ray light curves. Before folding the optical data on the 29.6 s period, variations on longer timescales were removed from the light curve by subtraction of a filtered version generated by the Savitzky-Golay algorithm (Savitzky & Golay, 1964) which eliminates variations below a cutoff timescale, here chosen to be 1 min. We adopted a conservative approach and used the free-flare dataset to construct the phase-folded, X-ray light curves. It still covers $\sim$250 pulsation cycles and avoids features that may be background-induced.

The waveform in the optical phase diagram is characterized by a deep minimum and two maxima of approximately equal height, separated by about 0.4 in phase (see the upper right frame of Fig. 2). The semi-amplitude of these variations is $\sim$ 0.016 mag, and its shape is quite similar to that observed in soft X-rays (0.3–2 keV), except for the difference in height of the X-ray maxima. The pulsed fraction is significant in all light curves. In the soft band, the variability reaches about 25% of the mean level. A similar value is obtained for the integrated band because the total counts are dominated ($\sim$80%) by soft X-ray photons. As for “hard” (2-10 keV) X-rays, even though the pulsation is not as clearly observed in the periodograms, the folded light curve indicates that the 29.6 s modulation is present and has a significant ($\sim$ 50%) pulsed fraction (see Fig. 2).

Optical and X-ray modulations can be understood within the same scenario. The stability of the 29.6 s modulation over two years in the optical (2015 and 2016) and its presence three years later in X-rays (2019) lead us to the interpretation that it represents the spin period of the WD in J2056.