PAN-CHROMATIC OBSERVATIONS OF THE RECURRENT NOVA LMC 2009a (LMC 1971b)

We obtained optical and near-IR (NIR) photometry with the SMARTS 1.3 m telescope at the Cerro Tololo Inter-American Observatory (CTIO) using the ANDICAM dual-channel imager from 2009 February 6 through 2015 April 2 (days 1 to 2247 after discovery).

ANDICAM employs a dichroic filter to enable simultaneous optical and NIR imaging. The optical detector is a Fairchild 447 2048 × 2048 CCD. The 62 × 62 image is oversampled; data are taken with 2 × 2 pixel binning for a 0369 pixel⁻¹ plate scale. Bias subtraction and flat fielding are done by the SMARTS pipeline prior to data distribution.

The NIR detector is a Rockwell 1024 × 1024 HgCdTe Array. The field is 24 × 24; the images are rebinned to 512 × 512 on the ground (0274 pixels). No other processing is done prior to data distribution.¹⁵ We obtain three images dithered with a 20'' throw.

The optical and NIR images are obtained simultaneously in pairs. On most nights we obtained full ${{BVR}}_{{\rm{C}}}{I}_{{\rm{C}}},{JHK}$ coverage. However, due to problems with the filter wheel we were only able to obtain a single pair of exposures each night between 2009 February 6 and 16. Consequently, we have poor temporal coverage in the first 12 days, when the nova is decaying fastest. We obtained two observations per night on many nights between days 40 and 60 in order to search for optical modulation on the 1.2 day period seen in the Swift observations (see below). These observations were separated by 2–4 hr, in order to mitigate against aliasing with a 1 day period.

We have obtained between 160 and 173 exposures in each band. Exposure times were increased from 2 to 300 s in the optical and from 12 to 195 s in the NIR channel as the nova faded.

2.1.1. Data Reduction

We obtained low dispersion spectra with the RC spectrograph on the SMARTS 1.5 m telescope at CTIO. All observations were made by SMARTS service observers. The spectroscopic record extends from 2009 February 7 though 2009 December 26 (days 2–323, with the vast majority of the data before day 160) by which time it had become too faint for the 1.5 m. We secured a total of 56 spectra on 56 nights. Observing conditions ranged from photometric to thick overcast.

The RC spectrograph is a long-slit instrument; the slit length subtends 5 arcmin on the sky. We used a 1'' slit width. The detector is a Loral CCD. We used five spectroscopic set-ups, as detailed in Table 1.

We generally obtained three observations with integration times between 200 and 1200 s depending on the set-up and the target brightness. The three observations are median-filtered to minimize contamination by cosmic rays.

We reduced the data using our spectroscopic data reduction pipeline. We subtract the bias and trim the overscan then flatten the image using dome flats. The spectra are extracted by fitting a Gaussian plus a linear background at each column. The extracted spectrum is the area of the Gaussian fit at each wavelength. Uncertainties are based on counting statistics, including uncertainties in the fit background level. Wavelength calibration utilizes an arc lamp spectrum obtained before each set of images.

We obtained a spectrum of a spectro-photometric standard, generally either Feige 110 or LTT 4364, to provide a counts-to-flux conversion factor each night. Because this is slit spectroscopy, slit losses preclude an absolute flux calibration, but we do recover the shape of the spectrum. Small errors in the calibration arise from the fact that the target and the standard are generally observed at different air masses and consequent slit losses may change, and the spectroscopic slit is oriented E–W, not at the parallactic angle.

We calibrated the absolute fluxes of the spectra using the spectra obtained with the low dispersion grating 13 set-up, which spans the entire optical range including the B, V, R_C, and I_C passbands. However, we only obtained four such spectra. Additionally, we were able to calibrate the absolute fluxes of the blue (grating 26 set-up) spectra in a two-step process. The low dispersion blue spectra span the full width of the Johnson B filter. We convolved these spectra with the filter response, and scaled the total flux to match the B-band flux interpolated to the time of observation. Correction factors ranged from 0.7 to 3, reflecting air mass corrections, slit losses, and changes in sky transparency between the time of these observations and those of our spectral flux calibrator.

We then interpolated the continuum flux at λ 4250 Å for the calibrated low dispersion blue spectra to the times of observation of the high dispersion (grating 47/II) blue spectra, and scaled the latter accordingly.

The Swift satellite (Gehrels et al. 2004) began observing Nova LMC 2009a on 2009 February 14, nine days after the optical discovery, with the last observation occurring almost a year later, on 2010 January 30. The resulting data were processed using the standard Swift tools within HEASoft version 6.16 together with the most up-to-date calibration files. Count rates were estimated using the XIMAGE package, while upper limits were calculated using the Bayesian method of Kraft et al. (1991). Grades 0–12 were used for the Photon Counting (PC; time resolution of 2.51 s) mode X-ray Telescope (XRT; Burrows et al. 2005) data, while grades 0–2 were used for Windowed Timing data (WT; time resolution of 18 ms), which were collected when the source count rate was above about 1 count s⁻¹. A circle with a 20 pixel (1 pixel = 236) radius centered on the source was used for the WT data, with the background determined from the same-sized region offset from the source position. The PC data were considered to be piled-up above 0.4 count s⁻¹, during which times annular extraction regions were used, excluding between two and seven pixels depending on the actual count rate. At other times, a 20 pixel radius circle was also used for the PC data, though this was decreased to 10 pixels when the source count rate was below ∼0.1 count s⁻¹.

UV/Optical Telescope (UVOT; Roming et al. 2005) source magnitudes were extracted within a 5'' circular region centered on the source (since the aspect-correction failed for this field, the region was carefully re-centered on the source for every individual snapshot) and the background value calculated within a 15'' radius offset from, but close to, the nova.

The UV source was detected throughout the Swift observations, with an X-ray source being detected from days 63 to 302. The latter in particular showed great variability and the cadence of observations was adjusted accordingly. Full details are given in Section 3.3 below.

After Liller et al. (2009) reported the discovery of the nova on 2009 February 05.067 with an unfiltered TechPan film magnitude = 10.6, he found the nova had faded to magnitude 11.8 at 2009 February 07.092 (JD 4869.592¹⁶ , t = 2 days). Our first observation, through an I-band filter only, occurred 2 minutes earlier than this latter observation, with I = 11.55. We make no attempt to convert Liller's magnitudes to BVR_CI_C.

For the first four months of observations, the nova decayed monotonically, with no evidence for any re-brightenings or dust-formation-related dimmings (see Figure 1). The V- and I-band light curves over the whole range of observations into the quiescent phase are shown in Figure 2. While we can fit our early light curve data fairly well with a single exponential decay through to day ∼60, that simple decay does not fit the first few data points, and the extrapolation of the exponential toward time t = 0 obviously becomes meaningless. It is clear, however, that there is some indication of a trend toward Liller's discovery magnitude in the earliest data points. Data from day ∼70 through day 200 can be well fit with a second exponential, and data from day 400 and 800 onward, for the optical and NIR data, respectively (assumed to be quiescence), can be fit with a constant magnitude. An example of such a fit to the V-band data can be seen in Figure 2, with the residual from this fit shown in the lower panel. A clear break in all the optical and NIR light curves is seen at day 65 ± 2, which coincides with the first detection of X-rays from the eruption (Bode et al. 2009a, see Section 3.3). From this point onward, the emission begins to show more scatter.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Table 2 we provide the t₂ and t₃ timescales (days to decay 2 or 3 mag from peak, respectively). To determine these timescales we assume a peak magnitude of 10.6 in all bands (we note, however, the caveat that there is a possibility that the exact peak had been missed and the nova might have been even brighter at peak). The measurements of t₂ and t₃ were determined using an exponential fit with a timescale of the form ${e}^{-A+{Bt}}$, which is necessary to get close to the peak magnitude. A second (empirical) pair of t₂ and t₃ values were generated from a linear interpolation between points and gave very similar results. The t₂ and t₃ timescales generally increase with wavelength, therefore the nova colors must initially get redder with time. By the time we have good color data, after day 2, the observed colors are becoming bluer. This may indicate that the peak B and V magnitudes did not reach 10.6 or that the evolution of the emission lines, generally in B and R, is important. We come back to the emission lines later.

The light curve became more complex after the system returned from behind the Sun after ∼150 days. Rather than the expected plateau while the soft X-ray source was bright (Kato et al. 2008), the brightness continued to decline until about day 200, after which it re-brightened by about two magnitudes in all bands (see Figure 2; this apparent "re-brightening" is further discussed in Section 4.4). Following the turn-off of the X-ray source (around day 250–300; see Section 3.3) the system began to fade toward quiescence. The optical fading ceased after around day 400; we take the mean brightness after this date to be the quiescent level. Quiescent magnitudes obtained post-eruption by SMARTS are presented in Table 3. Note that the system had faded below detectability in the NIR channels by day 110. We discuss the quiescent (progenitor) system in more detail in Section 3.5.

We show the early color evolution in Figure 3. The broadband colors reflect the evolution of both the continuum and the emission lines. The color evolution through the first 120 days is essentially monotonic with some variations superimposed, except in B − V, which becomes bluer until about day 40, and then becomes redder.

Zoom In Zoom Out Reset image size

We measured the fraction of the flux in the emission lines directly using our spectra (Table 4, where there are four nights where we obtained low dispersion grating 13 spectra covering the full optical range). The grating 26/Ia spectra span the entire B filter, but other spectral set-ups do not sample full broadband filter bandpasses, so we ignore them. We fitted a smooth continuum by eye between the emission lines. It is hard to define the continuum shortward of 4000 Å at the lowest resolutions, so the uncertainty in the B-band line fractions can be up to 10% in those spectra. Emission lines are most important in B, with the lines contributing over 1/3 of the flux until about day 50. The Hα line is an important contributor to the R-band flux. Correcting the optical colors for the line contribution alters the color trends and shows that the strongest contributions to the color changes are the strong line emission in the B band and the Hα emission in the R band (see Figure 3).

Table 4.  Fractional Broadband Flux in Lines through Given Filters

$t-{t}_{0}$ (days)	B	V	RC	IC
7.068	0.31	0.11	0.49	0.21
10.114	0.27	⋯	⋯	⋯
14.963	0.31	⋯	⋯	⋯
20.940	0.29	⋯	⋯	⋯
25.944	0.35	⋯	⋯	⋯
30.930	0.37	⋯	⋯	⋯
33.944	0.34	⋯	⋯	⋯
35.928	0.36	⋯	⋯	⋯
36.975	0.35	⋯	⋯	⋯
42.920	0.35	⋯	⋯	⋯
43.921	0.38	⋯	⋯	⋯
45.965	0.41	0.07	0.23	0.06
50.920	0.37	⋯	⋯	⋯
52.926	0.31	⋯	⋯	⋯
54.914	0.23	⋯	⋯	⋯
57.918	0.26	⋯	⋯	⋯
64.911	0.18	⋯	⋯	⋯
65.914	0.24	0.07	0.14	0.05
66.908	0.14	0.05	0.13	0.05
68.934	0.16	⋯	⋯	⋯
72.908	0.12	⋯	⋯	⋯
84.905	0.10	⋯	⋯	⋯
96.894	0.09	⋯	⋯	⋯

Download table as:  ASCIITypeset image

The evolution of the broadband photometric data from the eruption and subsequent decline of Nova LMC 2009a are shown in the distance and extinction-corrected spectral energy distribution (SED) presented in Figure 4. We have followed the methodology of Schaefer (2010) to enable comparison with their Galactic RN SEDs (see their Figure 71). For Nova LMC 2009a we assume a distance to the LMC of d = 48.1 ± 2.3_r ± 2.9_s kpc (Macri et al. 2006) and an estimated extinction toward the LMC of A_V = 0.6 ± 0.2 (see Section 4). In this figure, SEDs are shown for days 10, 19, 28, 35, 47, 68, 90, 239, 302, and finally an average over day 400–2247 (in red), which is assumed to be quiescence. A number of inter-eruption data points are also shown—the gray points indicate upper limits from 2MASS data and the blue points optical detections (see Section 3.5 for further details).

Zoom In Zoom Out Reset image size

The optical behavior of novae around peak has been observed to resemble blackbody emission (Gallagher & Ney 1976; Gehrz 2008) and has been modeled in terms of the development of a pseudo-photosphere (PP) in the optically thick ejecta (see Bath & Harkness 1989 and references therein). The PP radius r_p is greatest and the effective temperature T_p is at a minimum at optical peak. Thereafter, the mass loss rate from the WD surface declines, and at constant bolometric luminosity, r_p shrinks while T_p rises, shifting the peak of the emission further into the UV with time past maximum light. In CNe, with relatively high ejected masses, T_p ∼ 10,000 K at optical maximum, placing the peak emission in the optical part of the spectrum.

Figure 4 shows that if the simple PP model is correct, then the peak of the continuum emission throughout the epochs sampled is at wavelengths shorter than that of the uvw2 filter on Swift (1928 Å; i.e., T_p > 15,000 K). For this very fast nova, with first SED observations taken 10 days after peak, this is not unexpected (for a more detailed treatment of the evolving nova spectrum at outburst, see Hauschildt 2008).

In Figure 5 we present five low/medium resolution SMARTS spectra of Nova LMC 2009a, taken on day 6 (black spectrum; set-up 13/I; see Table 1), 14/15 (red; set-ups 26/Ia and 47/Ib, respectively), 25/26 (gray; set-ups 26/Ia and 47/Ib, respectively), 36/37 (blue; set-ups 26/Ia and 47/Ib, respectively), and 65/66 (green; co-added, both set-up 13/I). All spectra are dominated by broad emission lines of the Balmer series, as well as emission from He and N, confirming the classification of LMC 2009a as a He/N nova in eruption. The early spectra show a flat continuum that evolves toward a bluer continuum at later times (see, e.g., Figure 5). Figure 6 shows some of the early spectral evolution in more detail, including that of the observed P Cygni absorptions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.2.1. He ii λ4686

Because the first ionization potential of He is about 50 eV, the detection of He ii λ4686 indicates the presence of high-energy photons or particle collisions. He ii λ4686 lies on the side of the strong λλ4640–4650 Å Bowen fluorescence blend, the strength and shape of which varies with time. We see no evidence of any excess emission at λ4686 Å prior to day 29.

A narrow emission feature above the local continuum (the red edge of the Bowen blend) becomes visible on day 29. Although it is difficult to discern in this earliest spectrum due to the relative dominance of the Bowen blend, the emission line is double-peaked (see, e.g., Figure 7), with the mean velocity centered about the rest wavelength of He ii at the 278 km s⁻¹ redshift of the LMC. As shown in Figure 8, the emission flux rises rapidly until about day 55, when it begins to decay approximately equally fast. Weak emission is detectable through our last observation on day 112.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We fit the emission line spectra as two Gaussian components atop a background parameterized as a third-order polynomial. The two components do not have the same width. Early on, the weak emission, in particular the bluer component, is difficult to discern against the steep background. We note that the existence of any spectral features in this region for which we have not accounted could bias the measurements.

The emission lines were double-peaked upon discovery. The velocity splitting increased to about ±800 km s⁻¹ and then decreased. Within the limits of our signal-to-noise ratio (S/N) we cannot quantify exactly when the trend reversed, but the velocity splitting appears to have peaked before the flux peaked.

3.2.2. Hα

The Hα line exhibits a complex and evolving profile (Figure 9). Early on the emission is dominated by a broad (±∼2500 km s⁻¹) component that rapidly fades while maintaining its full velocity width (the "shoulders" in Figure 9). The two stationary high-velocity peaks are commonly seen in RNe (e.g., Walter & Battisti 2011). A narrower double-peaked central component becomes prominent after a few days and soon dominates the flux in the line (Figure 10).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We fit the central component as the sum of two Gaussians superposed on the pedestal. It is clear that the profile is more complex, but this suffices to quantify the major structures in the line. Some of these are shown in Figure 10. The emission centroids of the two components behave like those of He ii λ4686, with a velocity separation that increased to about ±700 km s⁻¹ by day 50 and decreased thereafter. The mean of the two velocities is +34 km s⁻¹ with respect to the LMC standard of rest. The mean velocity of the pedestal component, taken to be the mean of the FWHM velocities, is −52 km s⁻¹. Given the uncertainties, these are not significantly different.

3.2.3. He i

3.2.4. Other Lines

At the time of the first Swift observation at t = 9 days, a UV source was clearly detected in all three UVOT NUV filters (uvw1, uvm2 and uvw2, with central wavelengths of 2600, 2246, and 1928 Å respectively), as first announced by Bode et al. (2009a). Observations continued every few days for the next two months, with the UV magnitude fading from about 11.5 to 14, but also showing short-term variability with peak-to-peak amplitude ∼0.6 mag (see Figure 11). High cadence observations (snapshots of data obtained almost every 96 minute spacecraft orbit) from days 43.0–46.9 after eruption demonstrated that the variability included a periodic component with P ∼ 1.2 days (Bode et al. 2009a, see Section 3.4).

Zoom In Zoom Out Reset image size

On day 63.0, there was the first marginal detection (99.7% confidence) of an X-ray source, at a count rate of (3 ± 1) × 10⁻³ count s⁻¹ (Bode et al. 2009b). From day 70, the X-ray source was clearly detected, with a strongly variable count rate observed; the emission was found to be dominated by a super-soft spectral component (see below).

Daily observations were performed between days 70 and 79 and again between days 82 and 90, at which point the cadence decreased slightly. Throughout this time, the UV source showed an overall fading trend, although with the periodic oscillations superimposed.

Further high-cadence observations between days 97 and 152 after eruption demonstrated that the X-ray source was also varying periodically (again, see further discussion in Section 3.4).

Observations then continued approximately once a week for six months. The UV source faded until around day 200, after which there was an apparent re-brightening (see Section 4), followed by a final decline. The maximum X-ray count rate (∼2.5 count s⁻¹) was detected on day 173, after which a fading trend (although still with the variations superimposed) began. Some time between day 257 and 279, when no data were collected, the count rate began to fall away, until the final X-ray detection occurred 302 days after eruption. The UV source remained visible until the end of the observing campaign.

Figure 12 shows the evolution of the observed X-ray spectrum. Figure 13 then shows the results of fitting the XRT spectra with a plane-parallel, static, non-local thermal equilibrium (NLTE) atmosphere model (grid 003¹⁷ ) and an optically thin plasma model (Mekal) component (with absorption fixed at N_H = 1.13 × 10²¹ cm⁻², derived from HI maps by Kalberla et al. 2005, known as the Leiden/Argentine/Bonn (LAB) survey). Most of the spectra show at least a few counts above 1–2 keV. One spectrum was extracted per Obs ID (median duration 0.2 days), some of which span a significant variation in intensity and hardness ratio. Data taken after day 280 are not included because there are insufficient counts to usefully constrain spectral fit parameters during the decline phase.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

From the model atmosphere fits to the evolving XRT spectrum, the temperature, radius, and luminosity were determined for the SSS (see Figure 13); the normalization of atmosphere models is defined in terms of the effective emitting radius and distance to the object. A slow rise in temperature to a peak of kT = 87 eV around day 170–180, followed by an unexpectedly noisy decline is evident. The measured radius and luminosity are consistent with the high WD mass implied by the peak temperature (see also Section 4). Although a hard component is definitely present, with a count rate of a few × 10⁻⁴ cps above 1.5 keV, its low flux meant that meaningful investigation of the evolution of its flux or temperature was rendered impossible. Parameterizing the hard spectral component as optically thin emission at an assumed temperature of 5 keV, the 90% upper limit on the unabsorbed 0.3–10 keV X-ray luminosity is ∼3–8 × 10³⁴ erg s⁻¹ between days 74 and 170.

As noted in Section 3.3, periodic oscillations were evident in the UVOT data from day 43. To explore this further, several sets of high-cadence data were secured. Figure 14 shows the UVOT photometry between days 43 and 47, where four peaks separated by ∼1.2 days are evident.

Zoom In Zoom Out Reset image size

The period was derived from a Lomb–Scargle periodogram of the Swift UVOT light curve, using data out to day 200, after first detrending it by subtracting a fourth-order polynomial least-squares fit (Figure 15). At early times, data were collected using all three UVOT UV filters; these were normalized to the uvw2 filter prior to periodogram analysis.

Zoom In Zoom Out Reset image size

The time of phase zero was obtained by fitting a sinusoidal profile to the maximum seen in the folded UVOT light curve, excluding 0.3 cycles around the minimum where a secondary maximum is sometimes evident. The resulting ephemeris, giving the time of UV flux maximum, is

$$\begin{eqnarray}&&{t}_{{\rm{max}}}={\rm{HJD}}2454913.174(15)+1.1897(42){\rm{N}}\end{eqnarray} \tag{ 1 }$$

where the uncertainties are 1σ estimates (see, for example, Larsson 1996). A Lomb–Scargle periodogram of the XRT data is shown in the lower panel of Figure 15, clearly demonstrating that a modulation is also detected in the X-rays, at a period of 1.1832 ± 0.0078 day (using data between days 80 and 200).

Figure 16 shows the UVOT and XRT data phase-folded as a function of time. For the UV data, the modulation is clearly visible between days 42 and 48 and 130 and 200, while the X-rays are most coherently modulated between days 80 and 90. The X-ray-modulated flux peaks 0.28 cycles (corresponding to ∼0.3 day) later than the UV.

Zoom In Zoom Out Reset image size

We also examined the SMARTS data for an optical counterpart to the modulation seen in the UV light curve. We recovered a sinusoidal modulation in the V-band data over the first 120 days (Figure 17). The phasing is that seen in the UV within the uncertainties.

Zoom In Zoom Out Reset image size

Based on the SMARTS photometry of the 2009 eruption, we assumed that the system had returned to quiescence from 2010 March (2011 April for the NIR data); see Figure 2 and Table 3. We also undertook a search of archival data sets to attempt to recover the system before the 2009 eruption.

We computed the spatial transformation between a V-band SMARTS image of the nova in eruption, taken on 2009 December 28 (day 326) and the same field from the SERC-J_DSS1 archival survey (plate taken 1975 January 10) using 21 stars resolved and unsaturated in both images (see Figure 18). This approach, being independent of astrometric calibration, yields the most accurate results. The uncertainty in the derived transformations (0.26 pixels, equivalent to 01) dominates over the average positional error of the nova in the SMARTS image. There is a resolved object within 0.14 pixels (equivalent to 005, 0.4σ) seen in the archival data. Given the distribution of objects within both data sets, the probability of finding an object at least as close to the position of the nova by chance is <0.01% (following the methodology outlined in Bode et al. 2009c). We conclude therefore that this is the progenitor system.

Zoom In Zoom Out Reset image size

Having identified the pre-eruption progenitor system, a search through other archival images was made. While present in the SERC-J data, the object is out of the field of view for the SERC-I and SERC-V surveys. However, it was present on the ESO-R (1986 February 03 UT) and ESO-B (1991 January 17 UT) images (in addition to SERC-J on 1975 January 16). Five stars in the field around the nova were selected and the B and R magnitudes of those objects were found in the USNO-B1 catalog. Differential photometry with the GAIA¹⁸ data analysis package was performed to find the magnitudes of the quiescent nova system as B = 18.32 ± 0.37, R = 19.20 ± 0.62, and B_J = 18.17 ± 0.20 (see also Table 3).

The 2MASS online catalog yielded no source at, or near, the position of Nova LMC 2009a, and a re-analysis of the 2MASS JHK_S data did not detect any source at the position. Based on the photometry of nearby faint, just-resolved sources in the 2MASS data we derived the following 3σ upper limits for the NIR photometry of the progenitor system J > 19.3, H > 18.2, and K_S > 18.7 (see Table 3).

The 2MASS upper limits, the SERC-J photometry, and the ESO photometry are also plotted in the SED evolution plot; see Figure 4.

In Figure 19 we present the distance and extinction-corrected SED of the quiescent Nova LMC 2009a based on the SMARTS and 2MASS photometry, those of the Galactic RNe RS Oph and T CrB (RG-novae), and U Sco (SG-nova), and that of the suspected Galactic recurrent RG-nova KT Eri. Photometry for the Galactic RNe is derived from Schaefer (2010, see their Table 30), and has been extended redward of the K-band using WISE data (Evans et al. 2014), distances and extinction are from Darnley et al. (2012; see their Table 2 and references therein), additional RS Oph photometry is from Darnley et al. (2008, and the Liverpool Telescope; Steele et al. 2004), KT Eri photometry is from Jurdana-Šerpić et al. (2011), and optical and NIR absolute calibrations are from Bessell (1979) and Campins et al. (1985), respectively. We assume a distance to RS Oph of $d={1.4}_{-0.2}^{+0.6}$ kpc (Barry et al. 2008; see also Bode 1987).

Zoom In Zoom Out Reset image size

Figure 20 presents a B versus (B − R) and a J versus (J − H) color–magnitude diagram (CMD) showing stars from the Hipparcos data set (Perryman et al. 1997) merged with the NOMAD1 (Zacharias et al. 2005) and 2MASS data sets, respectively, via the VizieR database (Ochsenbein et al. 2000), with parallax errors < 10%. These stars have been translated to the distance, and estimated extinction, toward the LMC. We discuss the conclusions to be drawn from the data shown in Figures 19 and 20 in the next section.

Zoom In Zoom Out Reset image size

As the distance to the nova is reasonably well constrained (see Section 3.1) compared to the typical case for Galactic novae, we can derive a fairly robust estimate of its absolute magnitude at peak. The main uncertainty lies in the extinction. Using $E(B-V)={N}_{{\rm{H}}}/(6\pm 2)\times {10}^{21}$ cm⁻², with N_H = 1.13 × 10²¹ cm⁻² (see Section 3.3 and Kalberla et al. 2005) and ${A}_{V}=3.1E(B-V)$, A_V = 0.6 ± 0.2. The online LMC extinction calculator¹⁹ derived from Zaritsky et al. (2004) gives $\langle {A}_{V}\rangle$ = 0.31 ± 0.30 from the results for 22 stars within 1' of the position of the nova. This is at least consistent with the result from the H column. Almost all of this extinction lies within the LMC itself (Zaritsky et al. 2004, note that the foreground Galactic extinction is minimal—A_V ∼ 0.05).

The peak magnitude of the eruption is subject to at least two uncertainties. First of all, there is only a single observed magnitude on the night of discovery (SMARTS observations beginning the next night) and this is an unfiltered magnitude, compared to those obtained subsequently by us. However, if we take ${[{m}_{V}]}_{{\rm{max}}}=10.6\pm 0.1$ and the values of d and A_V above, M_V = −8.4 ± 0.8_r ± 0.7_s. As an aside, we note that this is very similar to M_V = −8.7 derived for the Galactic nova KT Eridani by Ragan et al. (2009) which is discussed further below. Using the MMRD relationship from della Valle & Livio (1995), with t₂ = 5 days for LMC 2009a (see Table 2), the implied M_V = −8.9, which is comparable to the observed peak magnitude, within the quoted errors. It should be noted that della Valle & Livio (1995) use a fit to novae observed in M31 where the intrinsic scatter due to distance uncertainties is less than for Galactic novae and the sample includes several novae with declines at least as fast as that of LMC 2009a, again unlike the linear MMRD relationships derived for Galactic novae (e.g., Downes & Duerbeck 2000; we note, however, that Kasliwal et al. 2011 have cast some doubt on the overall validity of the MMRD relations).

Although the peak absolute magnitude is consistent with those found for CNe of a similar speed class, Nova LMC 2009a would appear anomalous in terms of the amplitude of the eruption, A. We estimate A ∼ 9 mag in the V band. This compares to A ≳ 13 for a CN of this speed class from the A versus t₂ relationship for CNe given in Warner (2008). Such a low amplitude of eruption implies either (i) the peak luminosity is much lower than that typical of CNe or (ii) the luminosity of the quiescent system is much higher than is typical in such objects. As the peak absolute magnitude does not appear anomalous, either the secondary star is more luminous than is normal in a CN at quiescence, or there is a greater contribution from the accretion disk (assuming emission from the WD itself is negligible in the optical—see further discussion below), or a combination of the two. Such low amplitude eruptions are, however, typical of RNe due to their relatively high quiescent luminosity.

4.2.1. SSS Duration and WD Mass

4.2.2. Mean Mass Accretion Rate

The results of a grid of nova models by Yaron et al. (2005) illustrate the fact that for an interval between eruptions Δt_eruption ∼ 38 years, both M_WD and the mean mass accretion rate ${\dot{M}}_{{\rm{acc}}}$ need to be high (we note this as the mean rate as, e.g., Shaviv et al. 2014 show that ${\dot{M}}_{{\rm{acc}}}$ may be a highly variable parameter). We may estimate the mass of the accreted envelope, M_acc, required to give rise to a TNR from

$$\begin{eqnarray}&&{M}_{{\rm{acc}}}\simeq \displaystyle \frac{4\pi {R}_{{\rm{WD}}}^{4}}{{{GM}}_{{\rm{WD}}}}{P}_{{\rm{crit}}}\end{eqnarray} \tag{ 2 }$$

(see, e.g., Shara 1981; Starrfield 1989).

If we take M_WD = 1.2 M_⊙ then R_WD = 4.3 × 10⁸ cm (from fitting to the formula for the Chandrasekhar mass–radius relation by Eggleton and reported in Truran & Livio 1986). We note that this is of the same order as the radius of the emitting region derived from the model atmosphere fits—see Figure 13. With P_crit ≃ 10¹⁹ dyne cm⁻² (Yaron et al. 2005), then M_acc ≃ 1.4 × 10⁻⁵M_⊙. Thus with Δt_eruption ∼ 38 years, ${\dot{M}}_{{\rm{acc}}}\simeq {3.6}_{-2.5}^{+4.7}\times {10}^{-7}{M}_{\odot }$ yr⁻¹ for the deduced range of WD mass, which as noted above is extremely high (we may also note that Truran & Livio 1986 give P_crit ≃ 10²⁰ dyne cm⁻² which obviously would increase the derived mass accretion rate further still, as would a lower value of Δt_eruption if there were any eruptions missed between 1971 and 2009 as might be the case for such a fast declining nova).

Consulting the Yaron et al. (2005) grid of nova eruption models, M_WD = 1.25 M_⊙ with ${\dot{M}}_{{\rm{acc}}}={10}^{-7}\ {M}_{\odot }$ yr⁻¹ gives Δt_eruption ∼ 20 years, which is compatible with that for LMC 2009a, although the ejection velocities they derive are much lower than those we have observed in LMC 2009a (underprediction of the ejecta velocities by current theoretical models is, however, well known).

4.2.3. Origin of the Periodic Oscillations

4.2.4. Other Parameters of the Quiescent System and Long-term Behavior

As can be seen in Figures 2 and 4, the nova returned to a long lived (>1800 days) reduced luminosity level around 400 days after the eruption, which, for now, we will assume is the quiescent state. As both the optical and the NUV parts of the SEDs in both Figures 4 and 19 indicate, the high luminosity and blue color of the emission from the system at quiescence are consistent with a system whose luminosity is most likely dominated by a bright accretion disk. As the comparative plots in Figure 19 show, the luminosity of the Nova LMC 2009a disk is significantly higher than that present in U Sco, and on par with that contained in the RS Oph and KT Eri systems. Such a bright disk implies a high mass transfer rate, which is in turn consistent with the conclusion above from consideration of the accreted mass and inter-eruption timescale. For example, using

$$\begin{eqnarray}&&{L}_{{\rm{acc}}}=(1-\alpha )\displaystyle \frac{{{GM}}_{{\rm{WD}}}{\dot{M}}_{{\rm{acc}}}}{2{R}_{{\rm{WD}}}}\end{eqnarray} \tag{ 3 }$$

with α = 0.5 (Starrfield et al. 1988), M_WD = 1.2 M_⊙, R_WD = 4.3 × 10⁸ cm, and ${\dot{M}}_{{\rm{acc}}}=3.6\times {10}^{-7}{M}_{\odot }$ gives L_acc ≃ 550 L_⊙ (which is far less than L_Edd of course, so the accretion will not be disrupted).

Turning to the quiescent NIR emission, the quiescent system is not detected in the 2MASS data. This and the ∼1 day orbital period seen in Nova LMC 2009a allow us to confidently rule out the presence of a luminous red giant secondary (akin to the RS Oph or T CrB systems) within the Nova LMC 2009a system (see Figure 19).

The quiescent SED shows remarkable similarity to that of the galactic RN candidate KT Eri. Jurdana-Šerpić et al. (2011) proposed that the KT Eri system has a 737 day orbital period and contains a low luminosity (i.e., young) red giant secondary. As noted above, the 1.2 day period seen in the Nova LMC 2009a system implies that the secondary here is actually a sub-giant, akin to the secondary in the U Sco system. The quiescent NIR photometry from SMARTS is consistent with emission from a secondary of similar luminosity to that seen in the U Sco system.

We also note that Nova LMC 2009a lies in a different region of the CMD (see Figure 20) from RS Oph and all other RG-novae with the exception of KT Eri, and is grouped with those systems whose optical emission is accretion disk dominated. The location of the quiescent Nova LMC 2009a on the NIR CMD is consistent with the U Sco system and therefore again with the secondary star being a sub-giant.

From the data presented in Figure 19, the accretion disk is in fact among the most luminous seen in any nova. This implies an extremely high mass accretion rate, and/or a disk close to face-on; however, a very low system inclination is ruled out by our detection of the 1.2 day flux modulation, which we assume is orbital in origin. In comparing LMC 2009a with other novae in Figure 19, it should also be noted that U Sco is a high inclination, eclipsing source (Schaefer et al. 2011 and references therein). The implied higher accretion luminosity in LMC 2009a compared to U Sco is also consistent with its slightly lower observed ejection velocities, in line with the results of the Yaron et al. (2005) models as discussed above.

When comparing the post-2009 eruption quiescent data to the pre-2009 eruption data we first note that the ESO-R data from 1986 February (see Section 3.5) are consistent with the SMARTS post-eruption data. However, both the SERC-J (1975) and ESO-B (1991) data show significantly brighter emission (∼2 mag brighter). The SERC-J and ESO-B observations are both consistent with the luminosity seen ∼300 days after the 2009 eruption, that is ∼100 days before reaching quiescence. There are two possibilities that could explain these apparent discrepancies. (1) It is possible that the recurrence time of this system is much shorter than the observed 38 years and that eruptions in ∼1974 and ∼1990 were missed. However, it seems quite unlikely that two, effectively random, observations would sample essentially the same portion of the decline phase. (2) Alternatively, if the accretion disk in the system is significantly disrupted by an eruption, it could take some time to return to its pre-eruption luminosity, therefore, it is possible that even 2250 days post-eruption the disk has yet to fully "recover." However, the SERC-J observation took place only ∼1000 days after the 1971 eruption. The SMARTS data over nearly 2000 days at quiescence do not imply a slow re-brightening with time, nor do they show any high amplitude flickering that could explain the archival data. Neither of these scenarios seems to satisfactorily explain this discrepancy. Indeed, if we take the combination of the SERC and ESO data as the true quiescent level, then the accretion disk is even more luminous than is implied by the SMARTS data.

Figure 21 shows a plot of the X-ray hardness ratio versus XRT count rate (normalized to the peak counts) for each of the novae RS Oph, U Sco, KT Eri, and LMC 2009a observed in detail by Swift and showing SSS phases (see also Schwarz et al. 2011). There are several points to note stemming from this diagram.

Zoom In Zoom Out Reset image size

First of all, the locus of the SSS phase above a value of normalized counts of ∼0.02 cps is very similar for RS Oph, KT Eri, and LMC 2009a (that for U Sco shows more scatter). The vertical offsets in the loci for these three novae are consistent with the differences in interstellar column noted on the figure. At low source counts, we immediately see the group of points in RS Oph at relatively high HR. These reflect emission dominated by that from the shocked wind at early and late times. The relative absence of this feature in the other novae is consistent with there being a far less dense pre-existing circumstellar envelope in these sources, in turn consistent with the presence of a less evolved secondary than in RS Oph. Similarities between the locus of points for LMC 2009a and KT Eri are also evident, and we now consider the comparison of these two objects in more detail.

Figure 22 shows the evolution of the observed XRT counts (normalized to peak counts) and HR for LMC 2009a and KT Eri. It is immediately apparent that the behavior of LMC 2009a is very similar to that of KT Eri on this plot. Indeed, the evolution of the optical spectra in these two novae is also very similar (F. M. Walter et al. 2016, in preparation). Although the first detection of the SSS in KT Eri was at 55.5 days post-eruption (Bode et al. 2010), this would not have been detected at the distance of LMC 2009a and the dramatic rise in counts in KT Eri was observed on day 65.7, which compares to the detection of the SSS in LMC 2009a at t ≲ 70 days (Bode et al. 2009b). Similarly, although the later time XRT light curve of KT Eri shows some complexity, the start of the final decline of the SSS appears to be at ∼260 days, compared to 270 ± 10 days for LMC 2009a as given above. Furthermore, taking into account the different distances of KT Eri (6.5 kpc, Ragan et al. 2009) and LMC 2009a (48.1 kpc), the peak SSS count rates are within a factor of two of one another, which may be accounted for at least in part by the lower column to KT Eri (N_H ≃ 5 × 10²⁰ cm⁻²; Ragan et al. 2009; Bode et al. 2010, see also Figure 20). Finally, the amplitudes of the eruptions in KT Eri and LMC 2009a (A ∼ 8–9 mag in V) are also very similar. There are some similarities in the quiescent SEDs of the two novae as noted above, and intriguingly short period quasi-periodic oscillations have been detected in both novae with P = 33 s in LMC 2009a (with XMM; Ness et al. 2014, 2015) and 35 s in KT Eri (with Swift; Beardmore et al. 2010; Ness et al. 2015).²⁰

Zoom In Zoom Out Reset image size

The inference then is that LMC 2009a and KT Eri are very similar systems, particularly in terms of their characteristics at eruption. As well as similarities in SSS development and outburst optical spectra,²¹ we note that their optical light curves are very similar in form until ∼100 days, after which the main difference is the distinct "plateau" observed in KT Eri, lasting until around the time of the re-brightening seen in LMC 2009a; see the next section. The eruption characteristics of course depend on the WD mass and composition, composition of material undergoing TNR, and mass accretion rate. From Figure 19 it appears from the observed flux from the disk that ${\dot{M}}_{{\rm{acc}}}$ is also similar in KT Eri and in LMC 2009a. However, the similarity could be caused by the combination of different disk luminosities and inclinations (we note as an aside that for KT Eri, Ribeiro et al. 2013 find $i={58}_{-7}^{+6}$ deg from kinematical modeling of the KT Eri ejecta). It is then interesting to note that the accretion rate in both systems may be similar despite them harboring rather different secondary star types.

Finally, we turn to an unusual feature of the optical light curve. As can be clearly seen in Figure 2, in both the V- and I-band light curves and the V-band residual plot, there is a clear, apparent, re-brightening of the nova around day 250. In Figure 23 we show the residuals following a subtraction of a pair of exponential decays and a quiescent level from each of the SMARTS B, V, R, and I light curves (see Section 3.1). Here the "bump" is clearly visible, beginning after day ∼180 and peaking around day 250. What is of most interest is perhaps the achromatic nature of this feature, being remarkably similar in morphology and amplitude in all optical bands. This feature could of course be easily dismissed as an artifact in the SMARTS data processing due to the low S/N of the data at these late times, however, it is also seen in the Swift uvw2 data, again with a similar morphology and amplitude.

Zoom In Zoom Out Reset image size

Such features in light curves of explosive transients, with no apparent color dependency (from ∼0.2 to ∼1.25 μm) are unusual (the secondary maximum seen in the RN T CrB, for example, appears rather different and was explained by Webbink et al. (1987) as emission from the secondary being heated by a Chandrasekhar mass WD at the Eddington limit and it occurs ∼1 orbital period after the eruption). If seen unconnected to a transient event, one would be suspicious of a gravitational microlensing event. A standard Paczynski (1986) curve was fitted to the combined B, V, R, and I-band data between days 100 and 400 (data outside this range was excluded to avoid any effects due to poor modeling of the nova light curve behavior and the 1.2 day periodicity). This fit to the "bump" was reasonable and returned a time of maximum magnification of day 271 ± 1 (consistent with the sampling of the SMARTS data), and an Einstein radius crossing time (event timescale) of 108 ± 8 days. We note that this timescale is broadly consistent with LMC microlensing events, and that this "bump" would have passed the Paczynski parameter cuts employed by Wyrzykowski et al. (2011; see their Table 3, cuts 10 and 11). If this "bump" were indeed due to microlensing, however, we would expect to see a similar feature in the Swift X-ray data. As can be seen in the XRT data (see Figure 11), there is some interesting behavior around day 250, but also there is a gap in the XRT data between day 257 and 279. The "bump" is also coincident with the start of the X-ray count decreasing and the end of the SSS phase, and the two are most likely connected, but it is hard to reconcile the apparent achromatic nature of the "bump" with the physical processes ongoing in the nova system at that time.

Considering the purely statistical likelihood of this bump being related to a microlensing event, Bennett (2005) estimated the total LMC microlensing rate to be Γ ∼1.5 × 10⁻⁹ events source⁻¹ yr⁻¹, so clearly the probability of witnessing a microlensing event coincident with a given nova (or any given star for that matter) is negligible, no matter how long one observes that nova. However, the question we should be asking is, given all the known novae, what is the probability that we witness at least one microlensing event? Well over 1500 novae are now known, the majority of these residing in either the Milky Way or M31. If we assume that both the Galaxy and M31 have the same microlensing rate Γ ∼ 2 × 10⁻⁵ events source⁻¹ yr⁻¹ (Sumi et al. 2013), then the probability that one of these systems has been lensed is ∼3% per year of follow-up. This is low, but not prohibitively so.