Nova LMC 2009a as observed with XMM–Newton, compared with other novae

MNRAS 000,1–23 (2019) Preprint 13 May 2021 Compiled using MNRAS L A T EX style file v3.0 Nova LMC 2009a as observed with XMM-Newton, compared with other novae Marina Orio,1,2?Andrej Dobrotka,3Ciro Pinto,4Martin Henze,5Jan-Uwe Ness,6 Nataly Ospina,7,8Songpeng Pei,9Ehud Behar,10 Michael F. Bode,11 Sou Her,1Margarita Hernanz,12,13 and Gloria Sala13,14 1Department of Astronomy, University of Wisconsin, 475 N. Charter Str., Madison WI 53704 2INAF–Osservatorio di Padova, vicolo dell’ Osservatorio 5, I-35122 Padova, Italy 3Advanced Technologies Research Institute, Faculty of Materials Science and Technology in Trnava, Slovak University of Technology in Bratislava, Bottova 25, 917 24 Trnava, Slovakia 4INAF-IASF Palermo, via Ugo la Malfa, 153, 90146 Palermo, Italy 5Department of Astronomy, San Diego State University, San Diego, CA 92182, USA 6European Space Astronomy Agency (ESA), European Space Astronomy Center (ESAC), Camino Bajo del Castillo s/n, 28692 Villanueva de la Ca˜nada, Madrid, Spain 7Department of Physics and Astronomy, Padova University, via Marzolo, 3, 35131 Padova 8INFN Sezione di Padova, Via Marzolo, 8, 35131 Padova, Italy 9Department of Physics and Astronomy, Padova University, vicolo Osservatorio, 3, 35122 Padova, Italy 10Department of Physics, Technion, Haifa, Israel 11Astrophysics Research Institute, Liverpool John Moores University, IC2, Brownlow Hill, Liverpool, L3 5RF, UK 12Institut de Ciencies de l’ Espai (ICE-CSIC). Campus UAB. c/ Can Magrans s/n, 08193, Bellaterra, Spain 13 Institut d’ Estudis Espacials de Catalunya, c/Gran Capita 2-4, Ed. Nexus-201, 08034, Barcelona, Spain 14 Departament de F`ısica, EEBE, Universitat Politecnica de Catalunya. BarcelonaTech., Av. d’ Eduard Maristany 10-14, 08019, Barcelona, Spain Accepted XXX. Received YYY; in original form ZZZ ABSTRACT We examine four high resolution reflection grating spectrometers (RGS) spectra of the February 2009 outburst of the luminous recurrent nova LMC 2009a. They were very complex and rich in intricate absorption and emission features. The continuum was consistent with a dominant component originating in the atmosphere of a shell burning white dwarf (WD) with peak effective temperature between 810,000 K and a million K, and mass in the 1.2-1.4 M range. A moderate blue shift of the absorption features of a few hundred km s−1can be explained with a residual nova wind depleting the WD surface at a rate of about 10−8Myr−1. The emission spectrum seems to be due to both photoionization and shock ionization in the ejecta. The supersoft X-ray flux was irregularly variable on time scales of hours, with decreasing amplitude of the variability. We find that both the period and the amplitude of another, already known 33.3 s modulation, varied within timescales of hours. We compared N LMC 2009a with other Magellanic Clouds novae, including 4 serendipitously discovered as supersoft X-ray sources (SSS) among 13 observed within 16 years after the eruption. The new detected targets were much less luminous than expected: we suggest that they were partially obscured by the accretion disk. Lack of SSS detections in the Magellanic Clouds novae more than 5.5 years after the eruption constrains the average duration of the nuclear burning phase. Key words: X-rays: stars, stars: cataclysmic variables, novae: N LMC 2009a, galaxies: individual: LMC 1 INTRODUCTION Classical and recurrent novae (CNe, RNe) are now routinely discovered in other galaxies of the Local Group, offering useful terms of comparison of the nova phenomenon in ambients with different metallicity and star formation history. Known novae in the Magellanic Clouds are not numerous, due to the small mass of the two galaxies, but they occur in an envi- ?E-mail: [email protected] ronment of much lower metallicity than the Galaxy, and at relatively close distance, only about 5 times as high as the farthest luminous Galactic novae well studied in recent years (for instance, the RN U Sco is at 12±2 kpc distance, see Schaefer et al. 2010). 1.1 Nova LMC 2009A Nova LMC 2009a (also N LMC 2009-02 in the notation including the outburst month) was discovered on 2009 February ©2019 The Authors arXiv:2105.05346v1 [astro-ph.HE] 11 May 2021 2M. Orio et al. 05.067 UT by Liller (2009) and spectroscopically confirmed by Bond et al. (2009). It was later identified as a RN, coinciding with N LMC 1971b (Bode et al. 2016). RN are the ones recurring on human time scales, although the models indicate that all outburst recur (i.e. Prialnik 1986). An optical spectrum obtained in outburst by Orio et al. (2009) showed prominent emission lines of H, He and N, so this nova is a He/N one in the classification scheme of Williams (1992), like the other RNe we know. Orio et al. (2009) reported a large expansion velocity with the full width at half maximum of the Hαline slightly above 4200 km s−1. Additional spectra obtained by Bode et al. (2016) showed expansion velocities derived from different lines and at different phases between 1000 and 4000 km s−1. Coronal line emission before day 9 indicated shocks in the ejecta. The initial decay was fast, and the time t3for a decay by 3 magnitudes lasted from 10.4 days in the V band to 22.7 days in the infrared K. The time t2for a decay by 2 magnitudes ranged from 5 days (V filter) to 12.8 days (K filter). These parameters are pertinent to a classification as a “very fast” nova (Payne-Gaposchkin 1964). By comparison with the grid of nova models by Yaron et al. (2005), the characteristic parameters of the outburst (recurrence time of 38 years, velocity reaching ≃4000 km s−1, t3=11.4 days in the V band, amplitude of about 9 mag in V), place the nova the highest WD mass range (1.4 M), with a rather young and hot WD at the start of accretion, and mass accretion rate ˙mof a few 10−8M. However, the models include only a constant mass accretion rate ˙m, which may not be the case in some novae, and probably not for RN, whose recurrence time has been observed to vary (while the envelope mass accreted to trigger the burning outburst is expected to remain the same). Bode et al. (2016) identified the progenitor system; the optical and infrared magnitude in different filters and the colour indexes are best interpreted with the presence of a subgiant feeding a luminous accretion disk. Modulations with a period P=1.2 days, most probably orbital in nature, were evident in the UV and optical flux since day 43 (Bode et al. 2016). Two other RNe with orbital periods of the order of a day and sub-giant evolved secondaries are the Galactic novae U Sco and V394 CrA. There is also evidence that also a third RN, V2487 Oph, hosts a sub-giant, although its orbital period has not been measured yet (Strope et al. 2010). Other nova systems with suspected subgiant secondaries and day-long orbital periods are: KT Eri (also a candidate RN, but so far without previous known outbursts, Bode et al. 2016), HV Cet (Beardmore et al. 2012) and V1324 Sco (Finzell et al. 2015). Luminous CNe and RNe are monitored regularly with Swift in UV and X-rays. It is known that all nova shells emit X-rays in outburst (e.g. Orio 2012), although they are not usually luminous enough to be detected at LMC distance. However, when the ejecta become optically thin to soft X-rays, the photosphere of the WD contracts and shrinks to close to preoutburst dimension (Starrfield et al. 2012;Wolf et al. 2013) while CNO burning still occurs close to the surface, with only a thin atmosphere on top, for a period of time ranging from days to years (Orio et al. 2001;Schwarz et al. 2011;Page et al. 2020). Because the WD effective temperature Teff is in the 150,000 K to a million K range, the WD atmosphere peaks in the X-ray range or very close to it, and the WD detected as a luminous supersoft X-ray source (SSS), observable at the distance of the Clouds, also thanks to the low column density. Strengthening of the He II 4686˚ A line in the N LMC 2009a spectrum preceded the emergence of the central WD as a supersoft X-ray source (hereafter, SSS) observed in X-rays with the Swift X-Ray Telescope (XRT) since day 63. The SSS initially was at lower luminosity, but became much more luminous around day 140. The following X-ray observations indicated an approximate constant average luminosity (albeit with large fluctuations from day to day), until around day 240 of the outburst. The SSS was always variable, periodically and aperiodically. and clear oscillations with the 1.2 days period were observed (Bode et al. 2016), with a delay of 0.28P with respect to the optical modulations. Not all novae are sufficiently X-ray luminous to be studied with the gratings in detail, especially if they are as far as the Magellanic Clouds, and we did not want to miss the occasion of the X-ray luminous Nova LMC 2009, so in addition to Swift X-Ray Telescope (XRT) (Bode et al. 2016), XMM-Newton was used for longer exposures and high spectral resolution. 2 THE OBSERVATIONS The rise observed with the Swift XRT prompted two observations in the Director Discretionary Time (DDT), requested by W. Pietsch, 90 and 165 days after the optical maximum, observed on 2009 February 6 (Liller 2009). In July of 2009, the nova was sufficiently X-ray bright to trigger also two preapproved target of Opportunity (TOO) observations awarded to PI M. Orio. These exposures were done respectively days 197 and 229 after the optical discovery at the observed optical maximum. All four XMM-Newton observations are listed in Table 1. While partial results were presented in Orio et al. (2013a,2017), this paper contains the first comprehensive analysis of all the data. The XMM-Newton observatory consists of five different instruments behind three X-ray mirrors, plus an optical monitor (OM), and all observe simultaneously. For this paper, we used the spectra from the Reflection Grating Spectrometers (RGS; den Herder et al. 2001), and the light curves of the EPIC pn and MOS cameras. The calibrated energy range of the EPIC cameras is 0.15-12 keV for the pn, and 0.3-12 keV for the MOS. The RGS wavelength range is 6-38 ˚ A, corresponding to the 0.33-2.1 keV energy range. Table 1 shows that in the first two observations the EPIC cameras were used in imaging mode, the pn detector was “small window mode” to mitigate pile up and the medium filter was used, while the MOS was used with the “small” frame and the medium filter. The set up of the third and fourth observation was the same, with larger MOS2 window size and with the thin (instead of the medium) filter for both MOS. The EPIC pn and MOS detectors were used to extract light curves with the XMMSAS (XMM Science Analysis System) task XMMSELECT after applying barycentric corrections to the event files, choosing only singlephoton events (PATTERN=0) in the events’ files. A reference for this and the other XMMSAS tasks mentioned below is https://xmm-tools.cosmos.esa.int/external/xmm_ user_support/documentation/sas_usg/USG.pdf. The resulting source light curves were corrected for background variations using the XMMSAS task epiclccorr. Because there was essentially no emission above 0.8 keV, we used only the RGS with their high spectral resolution for the spectral anal- MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 15 20 25 30 35 Flux (10-11 erg/cm2/s/Angstrom) Wavelength (Angstrom) Figure 1. The RGS spectra of Nova LMC 2009 in units of measured flux versus wavelength, observed on days 90 (2009 May, in red), 165 (2009 July, in blue), 197 (2009 August, grey) and 229 (2009 September, purple). Table 1. XMM-Newton observations of Nova LMC 2009. ObsID Exp. timeaDatebMJDbDaycpn MOS1 MOS2 (ks) (UT) (d) 0610000301 37.7 2009-05-06.43 2454957.43 90.4 thin/small medium/small medium/small 0610000501 58.1 (40.0) 2009-07-20.04 2455032.04 164.9 thin/small medium/small medium/small 0604590301 31.9 2009-08-20.59 2455063.59 196.17 thin/small thin/small thin/full 0604590401 51.1 2009-09-23.02 2455097.02 228.93 thin/small thin/small thin/full Notes: a: Exposure time (cleaned of high flaring background intervals and dead time corrected) of the observation; b: Start date of the observation; c: Time in days after the discovery of Nova LMC 2009 in the optical on 2009 February 05.067 UT (MJD 54867.067, see Liller 2009). ysis. We extracted the RGS spectra with the XMMSAS task rgsproc. Periods of high background were rejected. 3 IRREGULAR VARIABILITY AND SPECTRAL VARIABILITY The averaged, fluxed RGS spectra for each of the four exposures are shown in the same plot in Fig. 1, as fluxed spectra and in units of erg cm−2s−1˚ A−1. All spectra show a luminous continuum and emission lines. While in last three observations the count rate level was comparable and the spectrum did not change very significantly, the first RGS spectrum obtained in May of 2009 (on day 90 of the outburst) still had a much lower average continuum, consistently with the rise phase observed with Swift (Bode et al. 2016). In Fig. 2 we present the EPIC-pn light curves during the four exposures, because the pn is the instrument with the largest count rate and the best time resolution. In all four exposures there was large aperiodic variability: the count rate varied by an order of magnitude on day 90, by a little more of a factor of 3 on days 165, and by about 60% on days 197 and 229. The MOS and RGS light curves of each exposure are modulated exactly like the pn one. Despite a moderate amount of pile-up in the pn spectrum, the irregular variability of the source appeared the same in all instruments. We corrected for pile-up by excluding an inner region and leaving only the PSF wing, and found that pile-up does not affect the proportional amplitude of the irregular variations and light curve trend. Irregular variability over time scales of hours has been observed at different epochs in the X-ray light curve in several other SSS-novae, most notably in N SMC 2016 (Orio et al. 2018), V1494 Aql (Drake et al. 2003), and in one of MNRAS 000,1–23 (2019) 4M. Orio et al. Figure 2. The EPIC-pn light curve, from top to bottom, as measured on days 90, 165, 197, 229. The red horizontal line in the top panel shows the pn cutoff for the averaged “low count rate” and “high count rate” spectra, shown in the inset in red and black, respectively (note that the RGS count rate varied proportionally to the pn count rate variation). the early RS Oph exposures (Nelson et al. 2008). However, RS Oph was observed many times for hours, and we know that the supersoft X-ray flux stabilized in the later exposures. Before proceeding with a spectral analysis, we examined the spectral variability to assess whether it is connected with flux variability. In the top panel of Fig. 2 a red line indicates the count rate limit of 6 counts s−1measured with the pn, chosen for the intervals in which we extracted high and low count rate high resolution RGS spectra on day 90. This is the only exposure for which we found strong variability in the strength of the spectral features, as shown in the inset in the first panel of the figure. There is especially a striking difference in the 25-28 ˚ A range, where some emission lines were almost only present when the flux increased significantly for short periods. In the other observations (days 165, 197, and 229) the count rate variability was not matched by a change in the spectral shape of the continuum, to which most of the flux variation is due. Also the strength of most emission lines varied less, and the variations mostly followed with the change in the continuum. Fig. 3 shows “time maps”, that is variations of each line during the exposures, as well as spectra extracted during two very short intervals of high and low count rate for each observation. We found a correlation of the count rate with the strength of the N VI and N VII lines. The absorption edge of N VII, that abruptly cuts the flux below 18.587 ˚ A, did not vary with the flux level, and for this reason we rule out that the variability was caused by changes in intrinsic absorption. A likely interpretation of the variability of the continuum flux is that along the line of sight a partially covering, “opaque” absorber appeared and disappeared. In this scenario, the WD surface is for some time partially obscured by the accretion disk that was not disrupted in the outburst, or by large, optically thick and asymmetric regions of the ejecta. Because of the irregular variability over short time scales, in Nova LMC 2009 we favor the second hypothesis, and elaborate this idea in the Discussion and Conclusions sections. We rule out intrinsic variations of the WD flux, because the nova models indicate that burning occurs at a constant rate, thus the thin atmosphere above the burning layer can hardly change temperature. 4 IDENTIFYING THE SPECTRAL FEATURES Identifying the spectral features for this nova appears much more challenging than it has been for other novae observed in the last 15 years (e.g. Rauch et al. 2010;Ness et al. 2011; Orio et al. 2018,2020). We started by examining the strongest features. In the spectrum of day 90, we clearly identify two strong emission features: a redshifted Lyman-αH-like line of N VII (rest wavelength 24.78 ˚ A, possibly partially blended with a weaker line of N VI Heβat 24.889 ˚ A), and the N VI MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 5 Figure 3. Visualization of the spectral evolution within the observations. The four panels show the fluxed spectra as function of wavelength (top left), colour-coded intensity map as function of time and wavelength (bottom left), rotated light curve with time on vertical and count rate on horizontal axes (bottom right). In the top right panel, a vertical bar along the flux axis indicates the colours in the bottom left panel. In the top left panel, the red (highest) and blue (lowest) spectra have been extracted during the very short sub-intervals marked in the light curve (bottom right) with shaded areas and bordered by dashed horizontal lines in the bottom left panel. The average spectrum is shown in black, and the dark blue light curve is a blackbody fit obtained assuming depleted oxygen abundance in the intervening medium (see text). He-like resonance line at 28.78 ˚ A. We fitted the two stronger lines with Gaussian functions and obtained a redshift velocity of 1983±190 km s−1for the N VII line, and, consistently, a velocity of 1926+220 −250 km s−1for N VI. The integrated flux in these two lines is 1.23±0.18 ×10−13 erg cm−2s−1and 2.37±0.18 ×10−13 erg cm−2s−1, respectively. The spectra of the three following dates, when the SSS was at plateau luminosity, show a forest of features in absorption, and also some in emission. The continua of these spectra appear remarkably similar, but many features varied from one exposure to the next, unlike in other novae observed at the peak of the SSS emission, where the absorption features were not found to change significantly in different exposures done within weeks (e.g. RS Oph, V4374 Sgr, V2491 Cyg, see Nelson et al. 2008;Rauch et al. 2010;Ness et al. 2011). We focused only on the absorption lines that are clearly common to all the last three spectra. In addition to the interstellar absorption lines of O I at 23.51 ˚ A and N I at 31.3 ˚ A, we found only five strong common features, whose profiles we show in Fig. 3. The y-axis shows the averaged RGS1 and RGS2 count rate. The line of N VII at 24.78 ˚ A appears to have a P-Cyg profile in at least two exposures. Rather than being a “true” P-Cyg, it may be due to the superposition of absorption and emission lines produced in different regions (e.g. absorption in the WD atmosphere, emission farther out in the shell, like in U Sco, see Orio et al. 2013b). The other lines in Fig.3 are only in absorption. The first three were already observed at almost the same wavelength in other novae (Ness et al. 2011), and were marked as yet “unidentified”. We suggest identification of two of these lines as argon: Ar XIII at rest wavelength 29.365 ˚ A, and Ar XIV at 27.631-27.636 ˚ A. We also propose a tentative identification of the Ca XI (30.448 ˚ A) and S XIII (32.239 ˚ A). In Table 2 we report blueshift velocity, broadening velocity and optical depth obtained with these identifications. Fits for one of the spectra are shown in Fig. 3 in the panels on the left. To calculate the blueshift velocities, we followed the method described by Ness et al. (2010) to determine the line shifts, widths, and optical depths at the line center of the absorption lines for the spectra for the four exposures. Following Ness et al. (2011), we did not include the absorption correction, because it is not important in determining velocity and optical depth. The narrow spectral region around each line was fitted with a function C(λ)×e−τ(λ) where τ(λ) is a result of the fit and is the opacity for each line. We assumed that C(λ) is a linear function for each line in modelling the continuum. We also fitted the N VII emission component with a Gaussian function. The blueshift velocity is modest compared to other novae (RS Oph, see Nelson et al. (2008), V4743 Sgr, see Rauch et al. (2010), V2491 Cyg see Ness et al. (2011), N SMC 2016, described in Orio et al. (2018)), but this nova was observed MNRAS 000,1–23 (2019) 6M. Orio et al. 24.60 24.65 24.70 24.75 24.80 24.85 24.90 24.95 25.00 Wavelength ( ) 0.000 0.002 0.004 0.006 0.008 0.010 0.012 Flux ( ph cm 2 s 1 1) N VII 1000 0 1000 2000 3000 ( km s 1) 0.0050 0.0025 0.0000 0.0025 0.0050 0.0075 0.0100 Flux ( ph cm 2 s 1 1) N VII best fit 27.50 27.55 27.60 27.65 27.70 27.75 Wavelength ( ) 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 Flux ( ph cm 2 s 1 1) Ar XIV 1500 1000 500 0 500 1000 ( km s 1) 0.001 0.000 0.001 0.002 0.003 0.004 0.005 Flux ( ph cm 2 s 1 1) Ar XIV best fit 29.1 29.2 29.3 29.4 29.5 29.6 Wavelength ( ) 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 Flux ( ph cm 2 s 1 1) Ar XIII 3000 2000 1000 0 1000 ( km s 1) 0.002 0.001 0.000 0.001 0.002 0.003 0.004 0.005 Flux ( ph cm 2 s 1 1) Ar XIII best fit 30.20 30.25 30.30 30.35 30.40 30.45 30.50 30.55 30.60 Wavelength ( ) 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 Flux ( ph cm 2 s 1 1) Ca XI 3000 2000 1000 0 1000 2000 ( km s 1) 0.002 0.001 0.000 0.001 0.002 0.003 0.004 0.005 Flux ( ph cm 2 s 1 1) Ca XI best fit 32.00 32.05 32.10 32.15 32.20 32.25 32.30 32.35 32.40 Wavelength ( ) 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 Flux ( ph cm 2 s 1 1) S XIII 2000 1500 1000 500 0 500 1000 1500 ( km s 1) 0.002 0.001 0.000 0.001 0.002 0.003 0.004 0.005 Flux ( ph cm 2 s 1 1) S XIII best fit Figure 4. On the left, the profiles of common features observed with the RGS (averaged RGS1 and RGS2) at days 165 (red), 197 (black), and 229 (blue). On the right, in velocity space the same line and the fit for day 229 (N VII) and for day 197 (other plots). MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 7 as a SSS at a much later post-outburst time. The spread of blueshift velocity was also evident in V2491 Cyg (see Ness et al. 2011). 5 SPECTRAL EVOLUTION AND THE VIABLE SPECTRAL MODELS With the gratings, we measure the integrated flux without having to rely on spectral fits like with broad band data. The flux in the 0.33-1.24 keV (10-38 ˚ A) range increased by only 10% between days 165 and 197, as shown in Table 2. On day 197, the source was also somewhat “harder” and hotter. A decrease by a factor of 1.6 was registered on day 229, when the sources had “softened” again. This evolution is consistent with the maximum temperature estimated with the Swift XRT around day 180 and the following plateau, with final decline of the SSS occurring between days 257 and 279 (Bode et al. 2016). We tried global fits of the spectra, with several steps. Although we did not obtain a complete, statistically significant and comprehensive fit, we were able to test several models and reached a few conclusions in the physical mechanisms from which the spectra originated. We focused mostly on the maximum (day 197), for which we show the result with all all models. The first steps were done by using XSPEC (Dorman & Arnaud 2001) to fit the models. •Step 1. The first panel of Fig. 5 shows the fit with a blackbody for day 197. We first fitted the spectrum with the TBABS model in XSPEC (Wilms et al. 2000), however a better fit, albeit not statistically significant yet, was with obtained with lower blackbody temperature 47.3 eV (almost 550,000 K), with the TBVARABS formulation by the same authors for the intervening absorbing column N(H). This prescription allows to vary the abundances of the absorbing medium, and in the best fit we could obtain we varied the oxygen abundance, allowing it to decrease to almost zero, because an absorption edge of O I at 22.8 ˚ A in the ISM makes a significant difference when fitting a smooth continuum like a blackbody. However, it can be seen in Fig. 5 even the best blackbody fit is not very satisfactory in the hard portion of the spectrum. •Step 2. Assuming, as the models predict (e.g. Yaron et al. 2005;Starrfield et al. 2012;Wolf et al. 2013), and several observations have confirmed (e.g. Ness et al. 2011;Orio et al. 2018), that most of the X-ray flux of the SSS originates in the atmosphere of the post-nova WD, we experimented by fitting atmospheric models for hot WDs burning in shell. We used the grid of TMAP models in Non Local Thermodynamical Equilibrium (NLTE) by Rauch et al. (2010), available in the web site http://astro.uni-tuebingen.de/#rauch/ TMAP/TMAP.html. We wanted to evaluate whether the significant X-ray brightening between May and July (day 90 to day 165) was due to the WD shrinking and becoming hotter in the constant bolometric luminosity phase predicted by the nova models (e.g. Starrfield et al. 2012), or to decreasing column density (N(H)) of absorbing material in the shell. In several novae, and most notably in V2491 Cyg (Ness et al. 2011) and N SMC 2016 (Orio et al. 2018), despite blueshifted absorption lines that indicate a residual, fast wind from the photosphere, a static model gives a good first order fit and predicts most of the absorption features. We moved towards the blue the center of all absorption lines by a given amount for all lines, leaving the blueshift as a free parameter in the 100-1500 km s−1range, compatible with the values found in Table 2. Table 3 (for all four exposures) and Fig. 5 (for day 197) show the fit with two different grids of models available in the web site http://astro.uni-tuebingen.de/#rauch/ TMAP/TMAP.html. In the attempt to optimize the fit, we calculated both the CSTAT parameter (Cash 1979;Kaastra 2017) and the reduced χ2, but we cannot fit the spectrum very well with only the atmospheric model (thus we do not show the statistical errors in Table 3), since an additional component appears to be superimposed on the WD spectrum and many fine details are due to it. Bode et al. (2016) attributed an unabsorbed luminosity 3-8 ×1034 erg s−1to the nova ejecta. Another shortcoming is that the TMAP model does not include argon. Argon L-shell ions are important between 20 and 40 ˚ A and argon may be even enhanced in some novae (the oxygen-neon ones, see Jos´e et al. 2006). Above, we have identified indeed two of these argon features, that appear strong and quite constant in the different epochs. Model 1 (M1) is model S3 of the “metal rich” grid (used by Bode et al. (2016) for the broad band Swift spectra), including all elements up to nickel, and Model 2 (M2) is from the metal poor “halo” composition grid, studied for non-nova SSS sources in the halo or Magellanic Clouds, which are assumed to have accreted metal poor material and may have only undergone very little mixing above the burning layer. The nitrogen and oxygen mass abundance in model S3 are as follows: nitrogen is 64 times the solar value, oxygen is 34 times the solar value. In contrast, carbon is depleted because of the CNO cycle, and is only 3% the solar value. Novae in the Magellanic Clouds may not be similar to the non-ejecting SSS. They may be metal-rich and not have retained the composition of the accreted material, since convective mixing is fundamental in causing the explosion, heating the envelope and bringing towards the surface β+decaying nuclei (Prialnik 1986). The first interesting fact is that adopting the TBVARABS formulation like for a blackbody does not make a significant difference in obtaining the best fit. In fact, the atmospheric models have strong absorption edges and features that “peg” the model at a certain temperature and are much more important in the fit than any variation in the N(H) formulation, especially with low absorbing column like we have towards the LMC. In the following context, we show only models obtained with TBABS (fixed solar abundances). Both the metal poor and metal enriched models imply a very compact and massive WD, with logarithm of the effective gravity log(g)=9. In fact, the resulting effective temperature is too high for a less compact configuration, the SSS would have largely super-Eddington luminosity. Table 4 does not include the blue-shift velocity, which was a variable parameter but was fixed for all lines: it is in the 300-600 km s−1for each best fit, consistently with the measurements in Table 2. In Fig. 5 we show the atmospheric fits for day 197. The metal rich model, with higher effective temperature, appears more suitable. We note that in the last three spectra, whose best fit parameters are shown in Table 3, the value of the column density N(H) for this model was limited to a minimum value of 3 ×1020 cm−2, consistently with LMC membership. The fact that the metal rich model fits the spectrum better indicates that, even in the low metallicity environment of MNRAS 000,1–23 (2019) 8M. Orio et al. 20 25 30 35 0 2×10−34×10−36×10−38×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 197, N(H)=1.9 x 10^(21) cm^(−2), T(bb)=47.3 eV 20 25 30 35 0 2×10−34×10−36×10−38×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 197, N(H)=5.61 x 10^(20) cm^(−2), T(eff)=707,000 K 20 25 30 35 0 2×10−34×10−36×10−38×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 197, T(eff)=983,000 K, T(apec)=146 eV, N(H)=3 x 10^(20) cm^(−2) 20 25 30 35 0 2×10−34×10−36×10−38×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 197, N(H)=3x10^(20) cm^(−2), T(eff)=992,000 K Figure 5. Clockwise, from the top left, comparison of the RGS fluxed spectrum on day 197 with different models, respectively: a blackbody with oxygen depleted absorbing interstellar medium, a metal poor model TMAP atmosphere, a metal rich TMAP atmosphere, and the metal rich atmosphere with a superimposed thermal plasma component (single temperature) in collisional ionization equilibrium. See Table 3 and text for the details. the LMC, the material in the nova outer atmospheric layers during a mass-ejecting outburst mixes up with ashes of the burning and with WD core material. The WD atmosphere is thus expected to be metal rich and especially rich in nitrogen. However, the metal rich model overpredicts an absorption edge of N VII at 18.587 ˚ A, which is instead underpredicted by the “halo”model. This is a likely indication that the abundances may be a little lower than in Galactic novae, for which the “enhanced” TMAP better fits the absorption edges (e.g. Rauch et al. 2010). The metal-rich model indicates that the continuum is consistent with a WD at effective temperature around 740,000±50,000 K on day 90, and hotter in the following exposures, reaching almost a million K on day 197. Although the fit indicates a decrease in intrinsic absorption after the first observation, the increase in apparent luminosity is mostly due to the increase in effective temperature and we conclude that the WD radius was contracting until at least day 197. The “enriched” model atmosphere seems to be the most suitable in order to trace the continuum, except for the excess flux above 19 ˚ A. Because the continuum shape of the atmospheric model does not change with the temperature smoothly or “incrementally” like a blackbody, and is extremely dependent on the absorption edges, we cannot obtain a better fit by assuming an absorbing medium depleted of oxygen (or other element). This is the same also for the “halo” model. Therefore, the fits we show were obtained all only with TBABS, assuming solar abundances in the intervening column density between us and the source. •Step 3. In these spectra, we do not detect He-like triplet lines sufficiently well in order to use line ratios as diagnostics (see, for instance, Orio et al. 2020, and references therein). In the last panel in Fig. 5, we show the fit with TMAP and a component of thermal plasma in collisional ionization equilibrium (BVAPEC in XSPEC,see Smith et al. 2001). The fit improvement is an indication that many emission lines from the nova outflow are superimposed on the SSS emission, but clearly a single thermal component is not sufficient to fit the whole spectrum. Nova shells may have luminous emission lines in the supersoft range after a few months from the peak of the outburst (V382 Vel, V1494 Aql Ness et al. 2005;Rohrbach et al. 2009). Such emission lines in the high MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 9 20 25 30 35 010 −32×10−33×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 90, T(eff)=737,000 K, plasma at 50 eV, 114 eV, 267 eV 20 25 30 35 0 2×10−34×10−36×10−38×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 165, T(eff)=915,000 K, apec plasma at 175 eV, 134 eV, 55 eV 20 25 30 35 0 2×10−34×10−36×10−38×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 197, T(eff)=978,000 K, plasma at 60 eV, 152 eV, 169 eV, 154 eV, 60 eV 20 25 30 35 0 2×10−34×10−36×10−38×10−3 normalized counts s−1 Å−1 cm−2 Wavelength (Å) day 229, T(eff)=906,000 K, apec plasma at 186 eV, 127 eV, 55 eV Figure 6. Fit to the RGS spectra of days 90 (of the “high” period as shown in Fig.2), 165, 197 and 229 with the composite model with parameters in Table 4. resolution X-ray spectra of novae are still measured when the SSS is eclipsed or obscured (Ness et al. 2013), indicating an origin far from the central source. Shocked ejected plasma in collisional ionization equilibrium has been found to originate the X-ray emission of several novae, producing emission features in different ranges, from the “hard” spectrum of V959 Mon (Peretz et al. 2016) to the much “softer” spectrum of T Pyx (Tofflemire et al. 2013). Adding only one such additional component with plasma temperature keV improves the fit by modeling oxygen emission lines, but it does not explain all the spectrum in a rigorous way. The fit improved by adding one more BVAPEC component, and improved incrementally, when a third one was added. The fits shown in Fig. 5 yield a reduced χ2pa- rameter still of about 3, due to many features that are are still unexplained. Also, the N VII K-edge remains too strong and this is not due to an overestimate of N(H), since the soft portion of the flux is well modeled. In Fig. 6 and in Table 4 we show fits with three BVAPEC components of shocked plasma in collisional ionization equilibrium. To limit the number of free parameters, the results we present in Table 4 were obtained with variable nitrogen abundances as free parameter, that turned out to reach even values around ≃500 for one of the components. We obtain a better fit if the nitrogen abundance is different in the the three components, but different “combinations” of temperature and nitrogen abundance give the same goodness of the fit. We also tried fits leaving free also the oxygen and carbon abundances, leaving the other abundances at solar values: although the fit always converged with al least one of these elements enhanced in at least one plasma component, there was no clear further improvement. The best composite fit with “free” nitrogen is shown in Fig. 6 for all four spectra. Some emission lines are still unexplained, indicating a complex origin, probably in many more regions of different temperatures and densities. The fits are not statistically acceptable yet, with a reduced value of χ2of 1.7 for the first spectrum, about 3 for the second and third and 4.7 for the fourth (we note that adopting cash statistics did not result in very different or clearly improved fits), however, the figure shows that the continuum is well modeled and many of the emission lines are also explained. •Step 4. Since we could not model all the emission lines with the collisional ionization code, the next step was to explore a photoionization code. We used the PION model MNRAS 000,1–23 (2019) 16 M. Orio et al. Figure 10. Comparison the spectrum of N LMC 2009 on day 165 with the spectrum of KT Eri on day 158, with the photon flux of N LMC 2009 divided by a factor of 80. In the panel on the left we indicate the lines position with a blueshift by 1410 km s−1, which is a good fit for several absorption lines of KT Eri predicted by the atmospheric models (except for O I and N I, which are local ISM lines, observed at rest wavelength). to“dwarf nova oscillations”(DNO) in an extreme“low-inertia magnetic accretor”. 7.1 Periodograms We compared periodograms of different instruments and exposures. Because the low count rate of the RGS resulted in low quality periodograms or absent signal, we focused on the EPIC cameras, despite some pile-up, which effects especially the pn. In Fig. 12 we show and compare all the calculated periodograms. Even though the reading times were uniform, we used the Lomb-Scargle (Scargle 1982) method, because it allows to better resolve the period than the Fourier transform, due to oversampling. We show a larger time interval in the left insets, while the most important features are shown in the main panels. Table 6summarizes the most significant periodicities with errors estimated from the half width at half maximum of the corresponding peaks. The periodograms for the light curves measured on days 90 and 229 do not show an obviously dominant signal in the other observations (although, as Table 3 shows, we retrieved the period in the pn light curve on day 229, with lower significance), so we did not investigate these data further. It is very likely that the modulation started only in the plateau phase of the SSS and ceased around the time the final cooling started. For the data of days 165 and 197 we found dominant peaks, especially strong on day 197. The dominant signal is absent in the MOS light curve of day 165, but was retrieved instead in the RGS data of the same date. Both periodograms suggest a single signal, with a double structure in the pn, that however may be an artifact. Moreover, the periodicity measured with the RGS data on day 165 is the average of the two values measured with the pn on the same day. This may indicate that the amplitude in the pn light curve is not stable, generating a false beat like in V4743 Sgr (Dobrotka & Ness 2017). The beat causes us to measure a splitting of the peak, while the real signal is in between (so, it may be P4detected in the RGS data). The pattern is more complex on day 197. The periodicity P3detected in July is retrieved in all the August light curves, suggesting that this feature is real and stable. Moreover, the non pile-up corrected and the pile-up corrected pn periodograms, and the MOS-1 periodograms, for day 197 show the same patterns, implying that pile-up in the pn data does not strongly effect the timing results. We were interested in exploring whether the non pile-up corrected light curves, which have higher S/N, also give reliable results. For the day 197 observation we compared the results using the pile-up corrected pn light curve, to those obtained for the MOS-1, and found that the derived periodicities (P1and P3) agree within the errors. Worth noting is the dominant peak P2in the non pile-up corrected pn data, measured also in the pileup corrected light curve, but with rather low significance. We note that also many of the low peaks and faint features are MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 17 Figure 11. Comparison of the spectrum of N LMC 2009 on day 165 with the spectrum of N SMC 2016 on day 88, with the photon flux of N LMC 2009 multiplied by a factor of 18 for the comparison. Table 6. Detected periodicities, in seconds, corresponding to the strongest signals shown in Fig. 12. We include a possible periodicity in the September data of day 229, although the corresponding peak is not by any means as dominant as on days 165 and 197. In each row we report periods that are consistent with each other within the statistical error. label pn RGS pn pn no-pile-up MOS1 pn Day 164.9 Day 164.9 Day 196.17 Day 196.17 Day 196.17 Day 228.93 P1– – 32.89 ±0.02 32.90 ±0.02 32.89 ±0.02 32.89 ±0.01 P2– – – 33.13 ±0.02 – – P333.35 ±0.02 – 33.31 ±0.02 33.32 ±0.02 33.31 ±0.02 – P4– 33.38 ±0.01 – – – – P533.41 ±0.01 – – – – – present in both periodograms, but with different power. This confirms that pile-up does not affect the signal detection, although we cannot rule out that it affects its significance. 7.2 Is the period variable? The complex pattern discovered in the light curve of day 197 suggests that the period may have varied during the exposure. We already mentioned the variable amplitude on day 165, and reminded that variable amplitude was discovered in the case of V4743 Sgr (Dobrotka & Ness 2017). In Appendix 1, we show how we used simulations to investigate the possibility of variable periodicity and amplitude. The conclusion that can be derived is that models with variable periodicity match the data better than those with a constant period. The lingering question is whether different periodicities occur simultaneously, or whether instead the period of the modulation changes on short time scales, in the course of each exposure. We reasoned that if the period varied during the exposures, this must become evident by splitting the original exposure in shorter intervals. Therefore, we experimented by dividing the day 197 light curve into two and three equally long segments. The corresponding periodograms are depicted in Fig. 13. Dividing the light curve into halves, shows that the first half comprises both dominant signals P1and P3, while the second half is dominated only by P3. We tried a further subdivision in three parts, and found that the first interval is dominated by P1, the second by P2and P3(with a slight offset), and in the third portion, none of the P1,P2and P3periods can be retrieved. This indicates that the period MNRAS 000,1–23 (2019) 18 M. Orio et al. 0 2 4 6 8 10 12 14 16 18 30 31 32 33 34 35 36 August (day 197) L-S normalised power period [s] PN PN no central 0 2 4 6 8 10 12 14 16 18 August (day 197) L-S normalised power PN MOS1 0 2 4 6 8 10 12 14 16 18 July (day 165) L-S normalised power PN RGS 0 2 4 6 8 10 12 14 16 18 May (day 90), Sept. (day 229) L-S normalised power PN May PN Sep 0 5 10 15 15 20 25 30 35 40 45 50 0 5 10 15 15 20 25 30 35 40 45 50 0 5 10 15 15 20 25 30 35 40 45 50 0 5 10 15 15 20 25 30 35 40 45 0 5 10 33.2 33.3 33.4 33.5 0 5 10 15 32.7 33.0 33.3 33.6 0 5 10 15 32.7 33.0 33.3 33.6 Figure 12. Comparison of periodograms of different instruments and observations. The pile-up corrected pn data (pn no central) were extracted after subtracting the central, most piled-up region of the source. The insets on the left show larger period intervals, and the shaded areas are the intervals plotted in the main panels. The insets on the right show instead a narrower period interval than the main panels, with the August 2009 periodogram plotted in red. of the modulation was most likely variable during the exposure. This is probably the reason for which different periods of similar length were measured in the same periodogram by analyzing the light curve of the whole exposure time. We suggest that these periods did not coexist, but instead a single modulation occurred, whose slightly changed during time intervals of minutes to hours. While Ness et al. (2015) found that the short-term periodicity may be a recurrent and transient phenomenon, here instead our interpretation is that the variations in amplitude and length of the period sometimes make it undetectable, but most likely it is always present. 0 5 10 15 20 32 32.5 33 33.5 34 L-S norm. power period [s] 2. third 3. third 0 5 10 15 20 L-S norm. power 1. third 2. third 0 5 10 15 20 P1P2P3 L-S norm. power 1. half 2. half Figure 13. Periodograms for day 197, divided into halves and thirds. The vertical dashed lines indicate the periods P1,P2and P3ob- tained by analyzing the whole exposure at once. 8 DISCUSSION: N LMC 2009A Nova LMC 2009 is only one of three novae in the Clouds that could be observed with high resolution X-ray spectroscopy. The others were N LMC 2012 (Schwarz et al. 2015) and N SMC 2016 (Orio et al. 2018). All three novae are luminous and reached a range of effective temperature that can only be explained with the presence of a massive WD, close to the Chandrasekhar limit if we compare the data with models by Yaron et al. (2005). However, a word of warning should be given concerning the fact the models do not explain all the RN characteristics in a very consistent way. In fact, the ejection velocity inferred from the optical spectra exceeds the value predicted by the models for a RN with a recurrence period of 38 years or shorter: this difficulty of modeling RN is a known problem. A possible explanation is that ˙mhas been variable over the secular evolution of these novae: if the recurrence time was longer, and ˙mwas lower in previous epochs, the closely spaced nova outbursts have only started very recently and the material in the burning layer may still be colder and more degenerate than it would be after many outburst with a very short recurrence time, thus causing a larger ejection velocity than in the models, which assume that the accreted mass is accumulated on a hotter surface. N LMC 2009a was not as X-ray luminous as N LMC 2012 and N SMC 2016. N SMC 2016 also remained a much more luminous SSS for many months. Like these and other novae observed with the gratings, LMC 2009a showed a hot continuum, compatible with a peak effective temperature of almost a million K, predicted by the models for a WD mass m(WD)⩾1.3 M. However, the absolute X-ray luminosity, estimated by fitting the spectrum in a phase when it constitutes over 98% of the bolometric luminosity, in N LMC 2009 is only a portion of the predicted Eddington luminosity. Because the X-ray flux of this nova was irregularly variable during all observations over time scales of hours, our interpretation is that the filling factor in the outflow of the ejecta varied, and was subject to instabilities even quite close to the WD surface, never becoming completely optically thin MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 19 to X-rays during the SSS phase. We suggest that the WD was observed through a large “hole” (or several “holes”) of optically thin material that changed in size as the ejecta expanded clumped due to instabilities in the outflow, became shocked, and evolved. This is likely to have happened if there if the outflow, even at late epoch after maximum, was not a continuous and smooth phenomenon. We note that Aydi et al. (2020) explain the optical spectra of novae as due to distinct outflow episodes. Also several emission and absorption features of this nova were not stable in the different exposures, varying significantly over timescales of hours, but mostly without a clear correlation with the continuum level. It is remarkable that the nova does not show all the characteristic deep and broad absorption features of oxygen, nitrogen and carbon observed in other novae and attributed to the WD atmosphere. Some of these features in this nova may be in emission and redshifted, but an attempt to fit the spectra with a “wind-atmosphere” model by van Rossum (2012) did not yield a result. Perhaps the emission cores originate in the ejecta and are not to be linked with the WD, as is the case in other novae. Although we identified and measured several absorption and emission features, we came to the conclusions that there are overlapping line systems produced in different regions and with different velocity, most of them originating in the ejected shell through which we observed only a portion of the WD luminous surface. Finally, the short period modulation observed in N LMC 2009 is intriguing, because it varies in amplitude and in period length, over time scales of few hours. The value obtained for the length of the period is not compatible with against a non-radial g-mode oscillation due to the “” mechanism during nuclear burning, which is expected to have shorter periods (Wolf et al. 2018). The non-stability of the period seems to rule out that it is due to the rotation of a WD that has been spun-up by accretion. Yet the very similar short term modulations observed in three other novae in the SSS phase and in CAL 83, a non-nova SSS, (see Ness et al. 2015), suggests that the root cause has to do with the basic physics of nuclear burning WDs. 9 ARCHIVAL X-RAY EXPOSURES OF OTHER NOVAE IN THE MAGELLANIC CLOUDS Table 7 shows details of the pointed and serendipitous X- ray observations of the MC novae in outburst in the last 20 years. We observe only about two novae a year in the LMC and one every two years in the SMC (see, among others, Mr´oz et al. 2016), but thanks to the known distance, the low column density along the line of sight and the proximity the Magellanic Clouds provide very useful constraints to study the SSS phase of novae. It is expected that in some cases even the emission of the ejecta, before the onset of the SSS, can be detected, albeit with low S/N. Since the advent of Chandra and XMM-Newton, three novae, including N LMC 2009a (2009-02) described here, have been observed with the X-ray gratings in high spectral resolution. However, the X- ray luminosity of novae in the Magellanic Clouds appears to vary greatly, as Table 7 shows. Recent Swift observations of N LMC 2017-11a carried on for 11 months did not yield any X-ray flux detection (Aydi et al. 2019). N SMC 2019 was monitored for a few weeks and only a weak X-ray source emerged after 4 months. The recurrent nova in outburst in the LMC in 2016 and in 2020 (previously known as 1968-12a and 1990b) was well followed with the XRT, but it was not X-ray luminous enough for high resolution spectroscopy (Kuin et al. 2020;Page et al. 2020). Two other X-ray-detected novae were not luminous enough for the gratings: N LMC 2000 (Orio et al. 2003a) and N LMC 1995 (Orio et al. 2003b). The latter had a long SSS phase between 5 and 8 years. In fact, it was observed again by us in 2008 with XMM-Newton, and no longer detected. XMM-Newton and Chandra typically offer much long exposures (several hours versus less than few tens of minutes for Swift XRT). We explored the archival serendipitous X-ray observations for MC novae of the last 20 years, to search for other possible detections. This work was part of Sou Her senior thesis project in Wisconsin in 2016. Table 7 shows that we found 4 previously unknown serendipitous detections in XMM-Newton deep exposures, among 13 novae that were observed with these satellites (only two were serendipitously observed with Chandra, and six more with with Swift with shallow upper limits). The upper limits on the luminosity were at least about 1035 erg s−1for the XMM-Newton exposures in Table 7. Observations of novae more than 10 years after the outburst were available in three cases: N LMC 2001- 08 (11 and 16 years), N SMC 2002-10 (10 and 15 years), N SMC 2005-08 (11 years) and yielded no detections. The most advanced epoch for which we retrieved an X-ray detection was only 5 years and 3 months after maximum. The measured count rates of the serendipitous targets are given in Table 7, and the broad band spectra with spectral fits with TMAP, yielding a reduced χ2value of 1.3 for N LMC 2008 and of about 1 for the other three novae, are shown in Table 7 and Fig. 14. The SSS flux of the newly detected novae is much lower than predicted by the models and observed in the novae measured with the X-ray gratings. Even if the absolute luminosity value may vary significantly within the 90% probability range and is poorly constrained, we can rule out that these novae were near Eddington luminosity. Rather than having observed only the final decline, which is predicted to last for only a few weeks (Prialnik, private communication) and was indeed observed to take few weeks in RS Oph (Nelson et al. 2008), we suggest that we detected novae in which the WD was not fully visible and only a small region of the surface was observed, either because of obscuration by the ejecta, or by the accretion disk. Given the elapsed postoutburst time, longer than a year, the latter is much more likely. The value of the column density N(H) inferred in the spectral fits and shown in Table 7 is in fact consisting with no significant intrinsic absorption. With the higher column density in the Galaxy, such partially visible SSS may not be sufficiently luminous for detection: this is an important factor to take into account when constraining nova parameters on the basis of the SSS detection, duration and behavior. 10 CONCLUSIONS Our exploration of the N LMC 2009a X-ray high resolution spectra has highlighted their complexity and the likely superposition of different regions of emission. We also searched the X-ray archives and discovered four more supersoft X-ray MNRAS 000,1–23 (2019) 20 M. Orio et al. Table 7. Pointed and serendipitous observations of MC novae in outburst, satellite used, time after outburst in days (d), months (m) or years (yr), number of exposures, X-ray detection (yes or no), whether the nova was a RN, and publications. The novae are identified by year and month of outburst, and the names of the (only) serendipitously observed ones are in boldface. The whole row is in boldface if there was a detection. Name Satellite pointed? when how many X-ray on? RN? reference times LMC 2020-07 Swift yes 3d-6m ≃80 yes 1968, 1990 Page et al. (2020) 2016,2020 2016 outb. Swift yes 7d-11m ≃80 yes Kuin et al. (2020) SMC 2019-7 Swift yes 10d-4m 5 yes(@4m) LMC 2018-5 eRosita survey 21m several yes Ducci et al. (2020) Swift yes 22m-25m 16 yes LMC 2018-02 Swift yes 9d-5m 50 yes yes, 1996 Page et al. (2018) LMC 2017-11 Swift yes 14d-1yr 46 no Bahramian et al. (2018) SMC 2016-10 Swift yes 6d-1yr 111 yes perhaps Orio et al. (2018) XMM-Newton yes 1 75d ” Chandra yes 2 39d,88d ” LMC 2016-04 Swift yes 14d-48d 29 yes probably LMC 2015-03 Swift yes few days 3 no LMC 2012-11 Swift yes 1d 1 no LMC 2012-10 Swift yes ≃7d 1 no SMC 2012-09 Swift no ⩾13m several no Chandra no 2,14m 2 SMC 2012-03 Swift yes 4m-14m 85 yes Schwarz et al. (2012) Page et al. (2013a,b) LMC 2012-03 Swift yes 1d-21 m 72 yes Schwarz et al. (2015) Chandra yes yes ” SMC 2011-11 Maxi no 0d? 1 yes Li et al. (2012); Morii et al. (2013) Swift yes yes LMC 2011-08 Swift no 1yr 1 no LMC 2009-05 Swift yes 3yr 4 no LMC 2009-02 Swift yes 9d-1yr 82 yes perhaps Bode et al. (2016) XMM-Newton yes 90-229d 4 yes this paper SMC 2008-10 XMM-Newton no 1yr 1 yes Swift no 4-10yr several no SMC 2006-08 XMM-Newton no >3yr 3 no LMC 2005-11 Swift yes 1m, 3y, 4y 7 no XMM-Newton no 5yr,7yr 7 no Swift no ⩾7yr several no LMC 2005-09 XMM-Newton no 21m 1 yes Swift, XMM no 3-13yr several no SMC 2005-08 XMM-Newton y 6m 1 no Swift no 1yr-11yr 7 no XMM-Newton no 4yr-11yr 8 no LMC 2004-10 XMM-Newton no 2.1yr, 3.9yr 1 yes SMC 2004-06 XMM-Newton no 5.25yr 1 yes LMC 2003-06 XMM-Newton no >10yr 4 no SMC 2002-10 XMM-Newton no 10yr, 15yr 2 no SMC 2001-10 XMM-Newton no 4.5yr, 6yr, 8yr 3 no LMC 2001-08 Swift no 5-10yr several no XMM-Newton no 11yr,16yr 2 no sources, albeit at lower luminosity than expected. These are our main conclusions. •The continuum of the spectra of N LMC 2009 indicates the supersoft emission of the atmosphere of a WD with effective temperature peaking above 810,000 K (the PION blackbody source) and likely to be around a million K (the TMAP model result), corresponding to a WD in the 1.2-1.4 M range. The continuum and its absorption edges can be well reproduced only assuming low oxygen abundance in the intervening interstellar medium and, above all, enhanced nitrogen by a factor around 100 times solar in the nova atmosphere and/or residual wind near the surface, where we assume that the absorption features originate. •The blueshift of the absorption features is explained by a mass outflow rate of the order of 10−8Myr−1. Nova models usually assume that mass loss has ceased when the SSS emerges, but like in other novae, this does not appear to be true: some residual mass loss is still occurring. •The absorption features in this nova were never as deep (or even saturated) as in other novae bserved in the Galaxy MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 21 0.5 0 0.02 0.04 0.06 normalized counts s−1 keV−1 Energy (keV) Nova SMC 2008 10.2 0.5 0.2 0.4 0.6 normalized counts s−1 keV−1 Energy (keV) N LMC 2005−09 10.2 0.5 0 0.02 0.04 0.06 0.08 0.1 normalized counts s−1 keV−1 Energy (keV) N LMC 2004−10 10.2 0.5 0 0.01 0.02 0.03 0.04 0.05 normalized counts s−1 keV−1 Energy (keV) Nova SMC 2004 Figure 14. TMAP atmospheric fit to the spectra of Novae SMC 2008-10, LMC 2005-11, LMC 2004-10, SMC 20046, at 1, 1.8, 3.9 and 5.25 years post-outburst, respectively. Table 8. Count rates and spectral fits with TMAP to the XMM-Newton observations of MC novae. For the best fit, we assumed a minimum N(H) of 3.5 ×1020 cm−2 . Nova Date counts s−1N(H) ×1020 cm−2Teff (K) Lx×1035 erg s−1 LMC 2004-10 2006-12-2 0.0145±0.0033 3.5+13 −3.5476,000±148,000 0.35+1.14 −0.34 2008-09-03 0.0218±0.0028 3.5+16 −3.5671,000±112,000 0.31+0.56 −0.23 SMC 2004-06 2009-09-13 0.0073±0.0007 8.8+8.9 −5.6404,000+49,000 −88,000 4.70+5.30 −4.69 LMC 2005-09 2007-06-19 0.1645±0.0066 4.2+2.2 −1.8715,000+24,000 −15,000 2.40+1.72 −1.18 SMC 2008-10 2009-09-27 0.0081±0.0009 3.5+1.1 −3.5520,000+165,000 −50,000 1.97+310.00 −1.87 and in the Clouds, as the comparison with KT Eri and with N SMC 2016 shows clearly. While the peak temperature in N LMC 2009 may have been 100,000-200,000 K hotter, implying a higher ionization parameter, the difference is so large that it is likely to have been due also to different abundances and density in the medium in which the absorption features originated. •The ejecta of this nova emitted additional thermal X-ray flux that produced a complex emission line spectrum, most likely arising from multiple regions at different temperatures, with a contribution of both photoionization and shock ionization. •Nova KT Eri seems to have a high mass WD and a partially evolved companion in common with N LMC 2009. However, in N LMC 2009, the turn-on time was longer, and the duration of the SSS was comparable, so the ejecta initially absorbing the SSS emission may have had higher mass, but the left-over envelope was of the same order of magnitude. MNRAS 000,1–23 (2019) 22 M. Orio et al. Since N LMC 2009a is a RN with an inter-outburst period of only 38 years, while no outburst of KT Eri have been found in tens of years before the eruption, N LMC 2009 did not have the time to accrete comparable or higher mass before the outburst unless it had high ˙m, which in turn would have resulted in ejection after accretion of a small mass, unless the WD before the outburst was still relatively cool (see Yaron et al. 2005). Most likely, this nova is at only the beginning of its RN cycle with short inter-outburst periods and the WD surface has not been heated yet by a repeated outburst cycle (see Yaron et al. 2005). •The 33s pulsation varied in amplitude and in length of the period, over time scales of hours. The search for the physical root cause of this intriguing phenomenon, already observed also in other novae, must take this fact into account. •N LMC 2009 was an order of magnitude less intrinsically luminous than the level compatible with emission arising from the whole surface. The predicted flux level was not observed in all SSS-novae, but it was indeed measured in N SMC 2016 as well as in other novae (e.g. RS Oph, Nelson et al. 2008). On the other hand, the portion of expected WD flux that is observed is much larger than in U Sco, in which only Thomson scattered radiation was detected. We suggested that in N LMC 2009a we observed the WD directly along the line of sight, but either dense remaining clumps in the ejecta, or a non-disrupted accretion disk, blocked the whole of the whole WD atmosphere. The clumps are a more likely explanation, because of the variability in the continuum level indicating varying visibility. The scenario we suggest is one in which part of the material, in an outflow in which instabilities occurred, was still opaque to X-rays. •We discovered that four other MC novae were serendipitously observed as SSS with low column density. There was no indication of flux variation and the absolute SSS luminosity was only about a factor of a thousand less than predicted (and indeed observed in the most luminous SSS-novae). We hypothesize that partial obscuration of the SSS was due to an accretion disk at high inclination, since these novae were detected long after the end of mass loss and it is unlikely that the ejecta still obscured the WD. •The statistics of the MC novae of the last 20 years indicates that novae that are still SSS after 6 or more years from the outburst are rare. We also found that half of all novae were detected as SSS for at least several months between few days and 5.5 years after the outburst. Since the sampling was done mostly at sparse post-outburst times, and it never lasted for more than a few months with Swift, or covered only one or two random epochs with XMM-Newton, the percentage of novae detectable as SSS within 6 years from the outburst would actually be quite higher if an uninterrupted survey was possible. DATA AVAILABILITY The data analyzed in this article are all available in the HEASARC archive of NASA at the following URL: https:// heasarc.gsfc.nasa.gov/db-perl/W3Browse/w3browse.pl ACKNOWLEDGEMENTS M. Orio was supported by a NASA grant for XMM-Newton data analysis. A.D. was supported by the Slovak grant VEGA 1/0408/20, and by the Operational Programme Research and Innovation for the project: “Scientific and Research Centre of Excellence SlovakION for Material and Interdisciplinary Research”, project code ITMS2014+:313011W085, co-financed by the European Regional Development Fund. REFERENCES Aydi E., et al., 2018a, MNRAS,474, 2679 Aydi E., et al., 2018b, MNRAS,480, 572 Aydi E., et al., 2019, arXiv e-prints, p. arXiv:1903.09232 Aydi E., et al., 2020, ApJ,905, 62 Bahramian A., Chomiuk L., Strader J., Kuin P., Darnley M. J., Page K., 2018, The Astronomer’s Telegram, 11301, 1 Beardmore A. P., et al., 2010, The Astronomer’s Telegram, 2423, 1 Beardmore A. P., et al., 2012, A&A,545, A116 Bode M. F., et al., 2016, ApJ,818, 145 Bond H. E., Walter F. M., Cosgrove E., Espinoza J., Liller W., 2009, IAU Circ., 9019 Cash W., 1979, ApJ,228, 939 Dobrotka A., Ness J.-U., 2017, MNRAS,467, 4865 Dorman B., Arnaud K. A., 2001, in Harnden F. R. J., Primini F. A., Payne H. E., eds, Astronomical Society of the Pacific Conference Series Vol. 238, Astronomical Data Analysis Software and Systems X. p. 415 Drake J. J., et al., 2003, ApJ,584, 448 Ducci L., et al., 2020, The Astronomer’s Telegram, 13545, 1 Finzell T., Chomiuk L., Munari U., Walter F. M., 2015, ApJ,809, 160 Heise J., van Teeseling A., Kahabka P., 1994, A&A, 288, L45 Jos´e J., Hernanz M., Iliadis C., 2006, Nuclear Phys. A,777, 550 Kaastra J. S., 2017, A&A,605, A51 Kaastra J. S., Mewe R., Nieuwenhuijzen H., 1996, in UV and X- ray Spectroscopy of Astrophysical and Laboratory Plasmas. pp 411–414 Kuin N. P. M., et al., 2020, MNRAS,491, 655 Li K. L., et al., 2012, Apj,761, 99 Liller W., 2009, IAU Circ., 9019 Mehdipour M., Kaastra J. S., Kallman T., 2016, A&A,596, A65 Morii M., et al., 2013, Apj,779, 118 Mr´oz P., et al., 2016, ApJS,222, 9 Nelson T., Orio M., Cassinelli J. P., Still M., Leibowitz E., Mucciarelli P., 2008, ApJ,673, 1067 Ness J. U., Starrfield S., Jordan C., Krautter J., Schmitt J. H. M. M., 2005, MNRAS,364, 1015 Ness J. U., Drake J. J., Starrfield S., Bode M., Page K., Beardmore A., Osborne J. P., Schwarz G., 2010, The Astronomer’s Telegram, 2418, 1 Ness J. U., et al., 2011, Apj,733, 70 Ness J.-U., et al., 2012, ApJ,745, 43 Ness J. U., et al., 2013, A&A,559, A50 Ness J. U., et al., 2015, A&A,578, A39 Odendaal A., Meintjes P. J., 2017, MNRAS,467, 2797 Odendaal A., Meintjes P. J., Charles P. A., Rajoelimanana A. F., 2014, MNRAS,437, 2948 Orio M., 2012, Bulletin of the Astronomical Society of India, 40, 333 Orio M., Covington J., ¨ Ogelman H., 2001, A&A,373, 542 Orio M., Hartmann W., Still M., Greiner J., 2003a, The Astrophysical Journal,594, 435 Orio M., Hartmann W., Still M., Greiner J., 2003b, The Astrophysical Journal,594, 435 MNRAS 000,1–23 (2019) Nova LMC 2009a with XMM-Newton 23 Orio M., Mason E., Gallagher J., Abbott T., 2009, The Astronomer’s Telegram, 1930 Orio M., Behar E., Gallagher J., Bianchini A., Chiosi E., Luna J., Nelson T., Rauch T., 2013a, in Di Stefano R., Orio M., Moe M., eds, Vol. 281, Binary Paths to Type Ia Supernovae Explosions. pp 181–185, doi:10.1017/S1743921312014950 Orio M., et al., 2013b, MNRAS,429, 1342 Orio M., Henze M., Ness J., 2017, in Ness J.-U., Migliari S., eds, The X-ray Universe 2017. p. 165 Orio M., et al., 2018, ApJ,862, 164 Orio M., et al., 2020, ApJ,895, 80 Page K. L., Walter F. M., Schwarz G. J., Osborne J. P., 2013a, The Astronomer’s Telegram, 4853, 1 Page K. L., Osborne J. P., Beardmore A. P., Schwarz G. J., 2013b, The Astronomer’s Telegram, 4920, 1 Page K. L., Kuin N. P. M., Henze M., 2018, The Astronomer’s Telegram, 11410, 1 Page K. L., Beardmore A. P., Osborne J. P., 2020, Advances in Space Research,66, 1169 Pagnotta A., et al., 2015, ApJ,811, 32 Payne-Gaposchkin C. H., 1964, The galactic novae Peretz U., Orio M., Behar E., Bianchini A., Gallagher J., Rauch T., Tofflemire B., Zemko P., 2016, ApJ,829, 2 Pietrzy´nski G., et al., 2019, Nature,567, 200 Pinto C., Ness J. U., Verbunt F., Kaastra J. S., Costantini E., Detmers R. G., 2012, A&A,543, A134 Prialnik D., 1986, Apj,310, 222 Rauch T., Orio M., Gonzales-Riestra R., Nelson T., Still M., Werner K., Wilms J., 2010, ApJ,717, 363 Rohrbach J. G., Ness J. U., Starrfield S., 2009, AJ,137, 4627 Scargle J. D., 1982, ApJ,263, 835 Schaefer B. E., et al., 2010, AJ,140, 925 Schwarz G. J., et al., 2011, ApJS,197, 31 Schwarz G. J., Osborne J. P., Page K., Walter F. M., Starrfield S., 2012, The Astronomer’s Telegram, 4501, 1 Schwarz G. J., et al., 2015, Apj,149, 95 Smith R. K., Brickhouse N. S., Liedahl D. A., Raymond J. C., 2001, ApJ,556, L91 Starrfield S., Timmes F. X., Iliadis C., Hix W. R., Arnett W. D., Meakin C., Sparks W. M., 2012, Baltic Astronomy,21, 76 Strope R. J., Schaefer B. E., Henden A. A., 2010, AJ,140, 34 Tofflemire B. M., Orio M., Page K. L., Osborne J. P., Ciroi S., Cracco V., Di Mille F., Maxwell M., 2013, ApJ,779, 22 Trudolyubov S. P., Priedhorsky W. C., 2008, ApJ,676, 1218 Williams R. E., 1992, AJ,104, 725 Wilms J., Allen A., McCray R., 2000, ApJ,542, 914 Wolf W. M., Bildsten L., Brooks J., Paxton B., 2013, ApJ,777, 136 Wolf W. M., Townsend R. H. D., Bildsten L., 2018, ApJ,855, 127 Yaron O., Prialnik D., Shara M. M., Kovetz A., 2005, AJ,623, 398 den Herder J. W., et al., 2001, A&A,365, L7 van Rossum D. R., 2012, Apj,756, 43 This paper has been typeset from a T EX/L A T EX file prepared by the author. APPENDIX A: SIMULATIONS OF THE AUGUST 2009 LIGHT CURVE We based this simulation on the August pn data, because they yielded the highest count rate and the most complex features. We first fitted the pn data with a polynomial (25th order, P25), in order to model the long-term trend. and on P25 we added a sinusoidal function and Gaussian noise G. The latter represents the residual scatter of flux after subtracting P25 from the observed Table A1. Sine parameters of the simulated signal, amplitude (a) and periodicity (p). daand dprepresent the rate of variability of the corresponding parameter (see text for details). No daor no dpvalues means (constant) Pa=aor Pp=pin equation (A2), respectively. model pdp a dabest χ2 red (s) (s) (cts/s) (cts/s) A 32.9 0.5 1.0 2.5 1.87 B 33.1 0.5 1.0 2.5 1.11 C 33.3 0.5 1.0 2.5 1.99 D 32.9 0.5 1.0 – 2.02 E 33.1 0.5 1.0 – 1.74 F 33.3 0.5 1.0 – 1.52 G 32.9 – 0.0 2.5 1.86 H 33.1 – 0.0 2.5 1.46 I 33.3 – 0.0 2.5 1.93 data. 2The flux ψthus can be expressed with this equation: ψ= P25 +asin(2 π t/p) + G (A1) We performed simulations with constant and with variable periodicity pand amplitude a. Table A1 summarizes the parameters p and a. For the variable periodicity simulations we needed a smooth function describing the variability of amplitude or/and periodicity, so we generated random points over the time interval of the exposure3, with a Gaussian distribution in which pand aare the mean values, and dpand daare the corresponding variances. We fitted these points with a polynomial (Paor Pp), the required smooth function used as input for the sine function; ψ= P25 + Pasin(2 π t/Pp) + G (A2) If the amplitude drops at or below zero, the modulation is considered to be absent (a= 0). As input values for the mean periodicity pwe used the most significant periodicities P1,P2and P3from Table 6, and we varied the amplitude until we obtained the best match. For every model, we run 10000 simulations and calculated the periodogram using a Fast Fourier transform (it allows a much faster calculation than the Lomb-Scargle method). We selected a best case, using the sum of the residual squares Σ(o−s)2calculated over a given frequency interval, where oand sare the observed and simulated powers at a given periodicity in the periodogram. The gray shaded area in the main panels of Figs. A1 and A2 shows the frequency interval we used. The sums of residual squares are relative numbers, and the minimum indicates the best value. However, it is important to evaluate also the real goodness of the model; χ2 red =1 NΣ(o−s)2 σ2,(A3) where Nis the number of degrees of freedom (number of periodogram points over which the χ2 red is calculated). In order to estimate σ, we assumed that the periodogram outside the main pattern (marked as darker shaded area in Figs. A1 and A2) contains random features attributed to noise. We assumed that the noise power is exponentially distributed, and as σwe chose the power at the 90% level in the cumulative histogram (5.27), excluding one relatively significant peak at 31.2 s, which is of unknown origin. Fig. A1 shows the comparison of the observed periodogram with the best simulated one with variable amplitude and period. In the upper panel we show a case in which both parameters are 2It yields a very similar flux distribution to the simple Poisson noise. 3Also a short time before and after the exposure, to avoid boundary fitting effects, and using a step of 200 s. MNRAS 000,1–23 (2019) 24 M. Orio et al. 0 10 20 30 40 50 30 31 32 33 34 35 36 Power (cts/s2) period [s] PN model F, χ2 red = 1.52 0 10 20 30 40 50 Power (cts/s2) PN model B, χ2 red = 1.11 33.0 33.5 0 10 20 30 period [s] time [ks] 33.0 33.5 0 10 20 30 period time [ks] 0.00 1.00 2.00 ctr Figure A1. Comparison of the observed data in August and the best of 10000 simulated periodograms for models B and F, as in Table A1. The insets show the evolution of the simulated amplitude (in count rate) and periodicity. The light shaded area is the frequency interval over which the χ2 red is calculated. The darker area represents the interval over which we calculated the power uncertainty σ(equal to the vertical extent of the shaded area). variable, and in the lower panel a case with constant amplitude and variable period. The best case is model B, in which both parameters are variable. The simulated periodogram describes almost all the observed features, supporting the interpretation that the signal is variable and the periodicity oscillates around P2. The period variability is depicted in the lower inset. Model F, with constant amplitude and variable period, also reproduces all the observed features, but the amplitudes are significantly different and the periodicity oscillates around P3, matching the dominant peaks in all the observed light curves, except the non pile-up-corrected August pn light curve. We remind that Dobrotka & Ness (2017) explained the double peak periodogram feature in V4743 Sgr as a false beat in the observed data. Two close periodicities result in a beat with low frequency, but if a single signal changes its amplitude or disappears for some time, this effect mimics a beat, so and the numerical method ”interprets” it as due to two close frequencies. The same principle can explain the non-single peak in V2491 Cyg (Ness et al. 2011). Fig. A2 shows the best case with constant periodicity but variable amplitude. However, awas set to zero, yielding the amplitude oscillate around zero, and disappearing more frequently, mimicking a false beat. Larger values of ayield only a single dominant peak, which do not describe the observed feature at all. The best case reproduces all the observed features with χ2 red even better than model F. Models G and I yield single peak solutions and do not describe the complex observed pattern. Model H is acceptable, and it supports a constant periodicity P2with variable amplitude. However, model B describes the observed periodogram much better, and indicates a variable signal. With model H we were not able to retrieve the observed power configuration, i.e. P2having lower power than P1and P34. Any periodicity introduced in the model produces the highest, or a very significant peak at the given period, while the models with variable periodicity are more flexible, and 4We repeated the 10000 simulation process several times. 0 10 20 30 40 50 30 31 32 33 34 35 36 Power (cts/s2) period [s] PN model H, χ2 red = 1.46 0 1 2 0 5 10 15 20 25 30 count rate [cts/s] time [ks] Figure A2. The same as Fig. A1 but for model H. appear to be more realistic. We note here that interesting periodicity evolution is shown in the upper inset of Fig. A1. Although the figure shows only a solution out of many possible ones, and an exact match is not expected, the significant decrease of the period length towards the end of the exposure is in agreement with the periodogram of the third time interval obtained when we split the exposure in three portions (see text). MNRAS 000,1–23 (2019)