V723 CASSIOPEIA STILL ON IN X-RAYS: A BRIGHT SUPER SOFT SOURCE 12 YEARS AFTER OUTBURST

Classical novae (CNe) are the observable events caused by thermonuclear runaways on the surfaces of white dwarfs (WDs), fueled by material accreted from a companion star. Material dredged up from below the WD surface is mixed with the accreted material and violently ejected. The outburst continues until either nuclear burning has converted all the remaining hydrogen on the WD surface to helium or it has been ejected from off the WD. The bolometric luminosity at the peak of the explosion is near to or exceeds the Eddington luminosity, ∼1 × 10³⁸ erg s⁻¹ for a 1 M_☉ WD (see, for example, Schwarz et al. 2001). Initially, the radiative energy output of the nova peaks in the optical, since the binary is surrounded by expanding ejecta that are not transparent to the high-energy radiation produced on the surface of the WD by nuclear burning. As the ejecta expand and thin (and the remaining material on the WD burns hydrostatically in a shell), the photosphere recedes to inner and hotter layers. The peak of the spectral energy distribution then shifts to higher energies (Gallagher & Starrfield 1978).

The X-ray bright phase of the evolution has only been observed sporadically, with the first systematic observations obtained with ROSAT (e.g., Krautter 2002). The X-ray spectrum during this phase was found to sometimes resemble that of the class of super soft X-ray binary sources (SSS; van den Heuvel et al. 1992) such as Cal 83 in the LMC (e.g., Lanz et al. 2005). This phase is now called the SSS phase. V1974 Cyg was the first nova to be followed in X-rays from before the SSS phase until decline. Krautter et al. (1996) reported a slow rise and fast decline of the X-ray brightness after the peak had been reached. The first spectral modeling of V1974 Cyg was carried out by Krautter et al. (1996) using blackbody fits, but Balman et al. (1998) reanalyzed the spectra using LTE atmosphere models, because blackbodies did not fit the spectra. It is further known that luminosities and temperatures derived from blackbody fits are inconsistent with realistic physical conditions, with a tendency to overestimate the luminosity and thus underestimate the temperature (Kahabka & van den Heuvel 1997). The LTE models presented by Balman et al. (1998) were an attempt to account for complex absorption patterns that change the shape of the spectrum. Although only 24 spectral bins were available, an attempt was made to determine a large number of parameters (temperature, surface gravity, bolometric luminosity, neutral hydrogen column density, and the abundances of carbon, oxygen, and neon). Further, LTE and a static WD atmosphere were assumed. The combination of a small ratio of independent data points to adjustable parameters and the limitations of current atmosphere models call for caution when drawing conclusions from these models. The effects of the expansion, especially during the early phases of the evolution, have been investigated with the PHOENIX code by Petz et al. (2005).

In addition to the SSS spectrum Krautter et al. (1996) found a hard component that was fit with a MEKAL model by Balman et al. (1998) and thus was assumed to be an optically thin plasma that is radiatively cooling. This phenomenon has been observed in a number of novae (e.g., Ness et al. 2007a), the most recent example being RS Oph (e.g., Ness et al. 2007c).

A systematic search of all ROSAT observations of novae including the ROSAT All Sky Survey and the archives of pointed ROSAT observations was carried out by Orio et al. (2001). Of a sample of 108 CNe and recurrent novae (RNe) that exploded within the last ∼100 years, very few SSS spectra were found, implying that the SSS phase must be short. A recent survey of all novae observed with Swift also resulted in few detections of CNe in the SSS phase (e.g., Ness et al. 2007a). Few Galactic (e.g., Orio et al. 2001; Ness et al. 2007a and references within) and extra-galactic (Pietsch et al. 2005, 2007) novae have had the duration of this phase determined. In the extra-galactic surveys candidates for novae in the SSS phase can be confused with supernova remnants, foreground neutron stars, or black hole transients (Orio 2006).

SSSs are believed to be WDs with accretion rates high enough to sustain steady nuclear burning (van den Heuvel et al. 1992; Kahabka & van den Heuvel 1997). Stable shell burning can cause the WD to grow in mass via accretion with a buildup of hydrogen deficient material. Thus, the SSSs are strong candidates for single-degenerate supernovae Ia (SN Ia) progenitors (Starrfield et al. 2004). In general, CNe are not considered to be SN Ia progenitors because the outburst is believed to eject more material than is accreted.

While theoretical models predict novae to be active for tens to hundreds of years (e.g., Starrfield 1989; Sala & Hernanz 2005), all novae observed so far in X-rays have turned off after much less than a decade (Orio et al. 2001; Ness et al. 2007a), implying that mass must be lost during the outburst by some mechanism (e.g., MacDonald et al. 1985). The typical observed outburst duration is of order 1 year with the longest observed Galactic nova in outburst being GQ Mus at ∼9 years (Ögelman et al. 1993; Shanley et al. 1995). The outburst duration depends on the nuclear burning rate, the amount of hydrogen left on the WD after the explosion, and the rate of any ongoing mass loss. Unfortunately, all these properties including the WD mass, the most critical driver, are poorly known.

A different approach was proposed by Greiner et al. (2003), who found from the available data that systems with shorter orbital periods display long durations of supersoft X-ray phases, while long-period systems show very short or no SSS phase at all. They speculate that shorter periods may be related to a higher mass transfer rate (e.g., by increased irradiation) and thus to the amount of material accreted before the explosion.

In this paper we present Swift observations in X-rays of the slow nova V723 Cas (Section 2). In Sections 3 and 4 we describe the observations and our analysis, and address the possible future X-ray evolution in Section 5.

V723 Cas was discovered on 1995 August 24 (Hirosawa et al. 1995), and reached visual maximum on 1995 December 17 at 7.1 mag (Munari et al. 1996). Its pre-2001 evolution has been extensively observed from ultraviolet (Gonzalez-Riestra et al. 1996) to radio (Heywood et al. 2005) wavelengths. A detailed description of the optical photometric evolution since 2001 is given by Goranskij et al. (2007 and references therein). The orbital period is ∼16.6 h with a sinusoidal-like shape of the V-band variations (Goranskij et al. 2007).

We have monitored the optical spectral evolution of V723 Cas since the outburst. Beginning in 2001, V723 Cas evolved to an extreme "coronal" phase with the emergence of the high-ionization line of [Fe X]λ6376 Å (Figure 1; see also Iijima 2006). This iron line was also strong in the late-time spectra of GQ Mus (Krautter & Williams 1989; Ögelman et al. 1993) which was observed by ROSAT to have an SSS X-ray spectrum. In Figure 1 we show the development of [Fe X] relative to [Fe VII]λ6089 Å, between 2001 January and 2006 November in V723 Cas (see Table 2; for details on the observations we refer to G. Schwarz et al. in preparation). By 2006 the [Fe X] line flux exceeded that of [Fe VII] indicating that the ejecta were being photoionized by a hot (>2 × 10⁵ K) central source, implying ongoing nuclear burning on the WD (see also Greenhouse et al. 1990). The [Fe X] and [Fe VII] lines can be used to monitor the temporal evolution of the photoionizing X-ray source and constrain the turn-off timescale. The spectral similarity of V723 Cas to GQ Mus motivated our initial Swift X-ray observation. Further analysis of the optical spectra, along with data from other wavelengths, will appear in a forthcoming paper.

Zoom In Zoom Out Reset image size

2.1. Extinction and Distance Determinations

The first observation with the Swift X-Ray Telescope (XRT; Burrows et al. 2005) of V723 Cas was taken on 2006 January 31.27 and yielded a clear detection with no counts above 0.6 keV and a peak at ∼0.35 keV, thus typical of a SSS spectrum (Ness et al. 2006, 2007a). We have obtained six more Swift observations on the dates listed in Table 2, and the mean count rate for all observations is 0.022 ± 0.01 cts s⁻¹. In this table we also list the ObsIDs, exposure times, measured count rates and spectral hardness ratios, HR = (H − S)/(H + S) with S and H denoting the count rates in the energy ranges 0.25–0.38 keV and 0.38–0.60 keV, respectively. The extraction of count rates and spectra is described in Ness et al. (2007a). Briefly, we use a circular extraction region with radius 10 pixels for the source and an annular extraction region with inner radius 10 pixels and outer radius 100 pixels for the background. The extracted count rates are independent of the size of the extraction region. In Figure 2 we show the evolution of count rate and hardness ratio with time. While the count rate varies by over a factor of two from the mean, no changes of the hardness ratio can be established at a significant level (see also Table 2).

Zoom In Zoom Out Reset image size

Due to the orbit of the spacecraft, only short continuous observations (snapshots) can be taken. Each observation consists of between 1 and 20 short snapshots separated by hours to days. We extracted the count rates for each snapshot individually and show the evolution of count rate for ObsID 00030361008 in the top panel of Figure 3. During the first part of the observation a flare with an amplitude of a factor three above the pre-flare and post-flare rates can be seen. The other observations also show considerable variability. In the bottom panel of Figure 3 we show all observed count rates as a function of orbital phase, which was computed based on the ephemerides given by Goranskij et al. (2007). While the count rates between phase ϕ = 0 and 0.5 appear to be slightly higher, there is no systematic trend with phase in this dataset.

Zoom In Zoom Out Reset image size

Next we extracted spectra for each observation and found all to be typical SSS spectra. Since the hardness ratio does not change at a significant level, no significant difference between model parameters related to the spectral shape (e.g., temperature) can be expected, and we combined all observations into a single spectrum (31.4 ks) for the spectral analysis (see Figure 4).

Zoom In Zoom Out Reset image size

V723 Cas was also detected in the XMM-Newton Slew Survey Full Source Catalog (v1.1, Read et al. 2005) in an observation taken on 2007 February 1 (22:33:51 start time; ObsID 9130900003) with a count rate of 1.2 cts s⁻¹. We converted this count rate to an equivalent Swift XRT rate using the HEASARC tool PIMMS, assuming a blackbody spectrum with kT = 30 eV, and the converted value is consistent with our contemporaneous observations. There is no detection in the hard energy band, defined as 2–12 keV.

As a confirmation that the Swift X-ray source is associated with the 1995 outburst of V723 Cas, we inspected the ROSAT archive for any observations prior to August 1995 near the coordinates of V723 Cas. We found one match, an observation on 1990 July 30 (ObsID 930702) from which we determine an upper limit of 4.4 × 10⁻⁴ cts s⁻¹ at the coordinates of V723 Cas. Converting the mean Swift count rate to ROSAT counts with PIMMS, results in an expected ROSAT count rate ≈1000 times brighter than the measured upper limit. Thus we conclude that there was no X-ray source at this position prior to the 1995 outburst and that the current Swift observations are associated with the 1995 outburst of V723 Cas.

V723 Cas was included in the survey of RXTE All-Sky Monitor Long-Term Observed Sources. Between 1996 January 4 and 2007 September 20, 75,691 observations were taken in the energy band between 2 and 12 keV with exposure times between 11 and 92 s. Out of this sample there are 30 detections at a significance level >3σ and five at >4σ, with no systematic trend in time. Meanwhile, there are 2793 observations where the background-subtracted count rate is below zero at the >3σ significance level, and five events have negative count rates at a 4σ significance level. We conclude that none of these observations delivers convincing detections of the source in hard X-rays. This assertion is consistent with the non-detection of V723 Cas in the 2–12 keV band of the XMM-Newton observation and with our soft Swift spectra yielding no signal above 0.6 keV (Figure 4).

Given the relatively modest spectral resolution of the XRT, we utilized blackbody fits, B_λ(T_eff), to determine the range of temperature consistent with the observed spectrum. We took the temperatures, T_eff, and the emitting radii, R_EM (derived from the normalization assuming a distance of d = 2.7 kpc), to determine bolometric luminosities, L_bol. We corrected for interstellar absorption by fitting a parameterized bound–free absorption model implemented in PINTofAle (Kashyap & Drake 2000). The total column density from free–bound absorption by neutral elements in the line of sight was computed from the hydrogen column density, N_H, assuming solar abundances. In an iterative process we determined the combination of the three parameters T_eff, R_EM, and N_H that yielded the best agreement with the measured spectrum, using a maximum likelihood estimator. We determined uncertainties of the parameters by application of the likelihood ratio test, i.e., we consider two models to be different with probabilities 68.3%, 84.3%, 95.45%, and 99% if the likelihood increases by 3.53, 6.25, 8.02, and 11.3, respectively, from the lowest of all likelihood values (for three free parameters). We carried out rigorous optimization of all three parameters, but found an unrealistically high value of luminosity and a value of N_H that is inconsistent with the Galex observations (Section 2.1). We therefore explore the full uncertainty range and constrain the range of solutions within these uncertainties plus physical arguments.

In Figure 4 we compare four fits with fixed N_H = (6.0,  4.8,  4.0,  1.6) × 10²¹ cm⁻² but fitted T_eff and R_EM and find the values listed in the legend. In the bottom right we show the likelihood values compared to the fit with the lowest likelihood. It can be seen that better fits are achieved with values of N_H greater than the upper bound found from the Galex observations (see Section 2.1). However, the likelihood for the fit with N_H = 4.8 × 10²¹ cm⁻² increases by less than 3.53 in value and is thus consistent within 68.3 per cent probability. In addition to statistical arguments from the spectral fitting, physical arguments must be considered. For example, the fit with N_H = 1.6 × 10²¹ cm⁻² gives a radius that is too small for a WD, while the fit with N_H = 6.0 × 10²¹ cm⁻² is too luminous. The super-Eddington luminosity phase in novae is extremely short, lasting at best days and certainly not over 12 years (Schwarz et al. 2001).

In Figure 5 we show the results from computing a grid of models as a contour plot with the confidence intervals from the likelihood ratio test shown in shades of gray. Overplotted are the contours of the upper and lower N_H values derived in Section 2.1 and contours of emitting radii. The R_EM values can be interpreted as a proxy of the WD mass. A static 1 M_☉ WD has a radius of ∼6000 km (e.g., Sirius B; Barstow et al. 2005). WDs with a lower mass will have a larger radius, or the radius can be larger as a result of surface hydrogen burning. A 1.4 M_☉ WD has a radius ∼2000 km (computed from Equation (3) in Truran & Livio 1986), and no radii below R_EM = 2000 km are physically realistic. We also have to exclude luminosities above the Eddington limit, which is ∼2 × 10³⁸ erg s⁻¹ for a 1 M_☉ WD. The most likely region in the luminosity–T_eff diagram shown in Figure 5 is, therefore, the 3-σ contour area bounded by the N_H = 4.8 × 10²¹ cm⁻² contour on the left and top. From Figure 5 it is apparent that this boundary also excludes the physically unrealistic luminosities above log(L_bol) = 38.3. Meanwhile, the lower bound of 1.6 × 10²¹ cm⁻² is less likely, according to our fits than the higher value of N_H = 4.8 × 10²¹ cm⁻². With these boundaries, we find a range in T_eff ∼ (2.8 − 3.8) × 10⁵ K and L_bol greater than 5 × 10³⁶ erg s⁻¹, but less than 2 × 10³⁸ erg s⁻¹. Within these uncertainties and those in the distance, the derived V-band magnitude from the range of blackbody fits is consistent with the current V-band photometry.

Zoom In Zoom Out Reset image size

In an attempt to test the viability of our blackbody fits, we have applied the same method to a Chandra LETGS spectrum of the canonical SSS Cal 83, which is sufficiently well exposed and has enough spectral resolution for an atmosphere analysis (Lanz et al. 2005). We found a significantly higher luminosity and lower temperature than the atmosphere results. This confirms that blackbody fits systematically overestimate the luminosity and underestimate the temperature (see also, e.g., Greiner et al. 1991) suggesting that the actual luminosity is lower and the temperature is higher.

The long lifetime of V723 Cas is a significant departure from the typical behavior of Galactic CNe. It requires that a reservoir of hydrogen-rich material be maintained over timescales much longer than usual. For normal CNe this can be done with either a low-mass WD leading to a lower burning rate, a large amount of hydrogen left on the WD after the outburst, or inefficient mass loss mechanisms (i.e. winds, common envelope frictional drag, etc) after outburst. With a conversion efficiency of 6 × 10¹⁸ erg g⁻¹ and a constant luminosity of 10³⁷ erg s⁻¹, only 3 × 10⁻⁷ M_☉ of hydrogen is required to power the nova over a 12 year period, assuming no mass loss. Lower luminosities could further extend the duration of nuclear burning. Regardless of the combination of factors involved, if V723 Cas is only burning residual accreted hydrogen, it will eventually turn off.

Other systems with long outburst lifetimes such as AG Peg and RR Tel (Ness et al. 2007b) are all symbiotic novae which are different from CNe because they have longer orbital periods (hundreds of days) and giant secondaries with large wind mass loss rates. The WDs of these long-lived systems are probably of lower mass than those in CNe (e.g, RR Tel; Jordan et al. 1994) which allows a larger envelope of hydrogen-rich material prior to outburst. The turnoff of the hydrogen burning shell occurs on a nuclear burning timescale rather than the faster hydrogen envelope depletion timescale which includes mass loss from the ejection of previously accreted material and wind mass loss (MacDonald et al. 1985). The temperature of the WD in RR Tel is lower than the one we find for V723 Cas or for the canonical SSS Cal 83 (Jordan et al. 1994 found T_eff = 142, 000 K), again showing that symbiotics are different from CNe and SSSs. V723 Cas is not a symbiotic nova, since the orbital period is much shorter, and the companion is not a giant, which, at a distance of 2.7 kpc and E(B − V) = 0.5 (Section 2.1) would lead to a higher observed magnitude than V ∼ 15. In addition, a recent optical spectrum of V723 Cas taken during the optical minimum revealed no signature of a late-type, giant companion (G. Schwarz et al. in preparation). In order for the V723 Cas secondary to fill its Roche lobe, it must be a subgiant. The majority of CNe have shorter orbital periods.

There is, therefore, the interesting possibility that accretion has been re-established in the V723 Cas system and is supplying the WD with additional fuel. No signs of V723 Cas being an intermediate polar (IP) are detected. IPs typically have strong hard X-ray components as in AM Her's and at least two periods in their optical light curves, the orbital and the spin period of the WD. In Figure 5 of Ness et al. (2007a) the X-ray spectra of V723 Cas and the IP V4743 Sgr are compared. Although V4743 Sgr is further away, the X-ray count rate at high energies is much higher, while there are no counts at high energies that can be attributed to V723 Cas. Further, a strong period of 22 min was found in the X-ray light curve of V4743 Sgr (Ness et al. 2003). The only optical period for V723 Cas seems to be the orbital period of ∼16.6 h (Goranskij et al. 2000). Also, Goranskij et al. (2007) report of no other period that would indicate an IP.

Although other CNe have been observed to be active for several years, V723 Cas is now confirmed to be the longest CN observed in outburst and as of December 2007 shows no indication it is turning off. While the observed count rate was variable, the spectral hardness remained unchanged. From blackbody fits we find a temperature range from 2.8 × 10⁵ to 3.8 × 10⁵ K and the luminosity greater than 5 × 10³⁶ erg s⁻¹. We caution that these numbers have limited physical meaning, as it is well known that blackbody fits lead to overestimated luminosities and thus underestimated temperatures (Kahabka & van den Heuvel 1997). Nevertheless, the high temperatures (especially if the numbers are underestimated) indicate that nuclear burning is still continuing. V723 Cas may be burning the remaining hydrogen left on the WD surface after the initial explosion. If so, the X-ray emission will decline once the fuel is exhausted. It is unknown when this event will happen but we plan on continuing our Swift monitoring program to detect a possible X-ray turn off and to further investigate the changes in the X-ray count rate.

There is also the intriguing possibility that the nova is now fed via renewed accretion similar to that observed in the prototype SSS, Cal 83. If so, V723 Cas may have evolved into a permanent SSS. Evidence that points to this possibility is the exceptional length of time in outburst and the long orbital period similar to that in Cal 83 and Cal 87. Note also that Greiner et al. (2003) found that long-period systems show shorter SSS phases than short-period systems. V723 Cas does not fit into this picture, suggesting that this is a different situation. Establishing the WD mass and determining whether accretion has re-established will assist in understanding the fate of V723 Cas. G. Schwarz et al. (in preparation) are presently obtaining optical spectra during the different phases of the orbit to determine whether the V723 Cas system is dominated by an illuminated subgiant secondary, a re-established accretion disk, or some mixture of both. Any orbital variations in the lines may also be used to provide mass estimates of the system components.

J.-U.N. gratefully acknowledges support provided by NASA through Chandra Postdoctoral Fellowship grant PF5-60039 awarded by the Chandra X-ray Center, which is operated by the Smithsonian Astrophysical Observatory for NASA under contract NAS8-03060. S.S. received partial support from NSF and NASA grants to ASU. J.P.O., K.L.P, and A.P.B. acknowledge the support of PPARC. We thank N. Gehrels for granting the observations. We also thank the anonymous referee for efficient handling and useful comments.