DISCOVERY OF A NEW PHOTOMETRIC SUB-CLASS OF FAINT AND FAST CLASSICAL NOVAE

P60-FasTING was designed with the specific goal of probing new phase space, particularly, fast transients with peak luminosity in the gap between novae and supernovae. The sample of galaxies included the brightest and nearest galaxies (<20 Mpc, majority around 10 Mpc). The survey was undertaken in a single filter (primarily Gunn-g and some Gunn-i data just around full moon). The limiting magnitude was typically Gunn-g <21 and cadence was <1 day. The field of view of P60 is 135 × 135 and all galaxies except M31 were covered in a single pointing. For M31, five pointings were chosen to cover a larger fraction of the galaxy (Table 1).

A real-time data reduction and transient search pipeline was written and implemented in 2008 April. P60-FasTING ended in 2009 March. The search pipeline was written in python. A deep reference image was constructed by combining images from several of the best seeing dark nights. Next, wcsremap was used to align every new image with the reference and hotpants was used to compute a convolution kernel prior to image subtraction (both codes supplied by A. Becker⁷). Although the image subtraction software was quite sophisticated in its convolution of the new image to match the reference prior to subtraction, we suffered from a large number of false positives. To distill the false subtraction residuals from the bona fide astrophysical sources, a variety of automatic filters were used (e.g., the shape characteristics of the point-spread function (PSF) of the candidate, how well it resembles the PSF characteristics of other stars in the image). However, the final step in the vetting process was done by human eye on candidate thumbnails every morning. Due to the myriad tradeoffs for maximum completeness and minimum contamination, the complex issue of quantifying the completeness of the nova sample is beyond the scope of this paper.

Our survey was sensitive to classical novae only in a handful of the nearest galaxies in our sample (distance, d < 4 Mpc). Classical novae discovered by P60-FasTING are summarized in Table 2. Some novae in M31 were announced by different groups before P60-FasTING's first detection (usually due to bad weather at Palomar); these are summarized in Table 3.

Table 2. Novae Discovered by P60-FasTING

Nova	Host	Discovery Date	R.A. (J2000)	Decl. (J2000)	Offset from Host	Reference
P60-NGC 2403-090314	NGC 2403	2009 Mar 14.160	07:36:35.00	+65:40:20.8	1010W, 2520N	Kasliwal et al. (2009a)
P60-M82OT-090314	NGC 3034	2009 Mar 14.496	09:56:12.60	+69:41:32.3	1042E, 482N	...
P60-M81OT-090213	NGC 3031	2009 Feb 13.404	09:55:35.96	+69:01:51.0	15"E, 124"S	Kasliwal et al. (2009b)
P60-M31OT-081230 (2008-12b)	NGC 224	2008 Dec 30.207	00:43:05.03	+41:17:52.3	2334E,1038N	Kasliwal et al. (2009c)
P60-M81OT-081229	NGC 3031	2008 Dec 29.373	09:55:38.15	+69:01:43.6	267E, 1314S	Rau et al. (2009a)
P60-M81OT-081203	NGC 3031	2008 Dec 3.303	09:55:16.92	+69:02:17.7	872W, 974S	Kasliwal et al. (2008a)
P60-M82OT-081119	NGC 3034	2008 Nov 19.536	09:55:58.39	+69:40:56.2	295E, 104N	Kasliwal et al. (2008b)
P60-M81OT-081027	NGC 3031	2008 Oct 27.402	09:55:36.11	+69:03:22.0	158E, 331S	Kasliwal et al. (2008g)
P60-M81OT-080925	NGC 3031	2008 Sep 25.49	09:55:59.35	+69:05:57.1	235E, 203N	Kasliwal et al. (2008i)
P60-M31OT-080915 (2008-09c)	NGC 224	2008 Sep 15.36	00:42:51.42	+41:01:54.0	134E, 1424S	Kasliwal et al. (2008f)
P60-M31OT-080913 (2008-09a)	NGC 224	2008 Sep 13.18	00:41:46.72	+41:07:52.1	108W, 83S	Kasliwal et al. (2008e)
P60-NGC 891OT-080813	NGC 891	2008 Aug 13.45	02:22:32.70	+42:21:56.1	8"W, 59"N	Kasliwal et al. (2008c)
P60-M31OT-080723 (2008-07b)	NGC 224	2008 Jul 23.33	00:43:27.28	+41:10:03.3	81E, 61S	Kasliwal et al. (2008d)
P60-M82OT-080429	NGC 3034	2008 Apr 29.24	09:55:21.00	+69:39:42.0	165"W, 64"S	Kasliwal et al. (2008h)
P60-M81OT-071213	NGC 3031	2007 Dec 13.40	09:55:25.98	+69:04:34.8	40"W, 40"N	Kasliwal et al. (2007)

Download table as:  ASCIITypeset image

The robotic Palomar 60 inch has a standard data reduction pipeline (Cenko et al. 2006). This pipeline performs basic detrending (flat fielding and bias subtraction) and computes an astrometric solution. In 2008 August, we added a new functionality: computation of a photometric solution. We used the Sloan Digital Sky Survey (SDSS) catalog where available, otherwise the NOMAD catalog. Note that where NOMAD was used (e.g., M31), the transformation from Johnson UBVRI magnitudes to SDSS ugriz magnitudes was done following Jordi et al. (2006).

To compute a light curve, we first measured the magnitude of the nova on the subtracted image. The subtracted image was scaled to the same flux level as the new image. Thus, we measured the magnitude of ∼150 reference stars on the new image to compute a relative zero point with appropriate outlier rejection. Finally, we applied this relative zero point to the instrumental magnitude of the nova.

An integral part of P60-FasTING was follow-up spectroscopy to confirm and classify discovered candidate transients. Since we were looking for fast evolving phenomenon, we triggered our Target of Opportunity program on the Keck I and Palomar Hale telescopes soon after discovery. Sometimes due to bad weather or bright moon-phase⁸, neither of these was an option. We resorted to the queue-scheduled service-observed programs on the Gemini or HET telescopes. A log of spectroscopic observations can be found in Table 4.

We emphasize that spectroscopy is crucial in distinguishing between an optical transient which happened to be co-incident with a nearby galaxy and a classical nova. For instance, we took spectra of several optical transient candidates which did not turn out to be novae: a foreground M-dwarf flare in the Milky Way spatially coincident with NGC 7640; a background supernova; a luminous blue variable in NGC 925.

We present light curves of novae in M 31 (Figure 5), M 81 (Figure 6), M 82 (Figure 7), NGC 2403 (Figure 7) and NGC 891 (Figure 9). Additional light curves of novae in M 31 are presented in Figure 8.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Data were reduced in iraf using standard tasks in the NOAO package onedspec and the spectra are shown in Figure 10.

Zoom In Zoom Out Reset image size

Motivated by the fast evolution, we obtained Chandra observations (Director's Discretionary Time; PI: P. Jonker) of P60-M81OT-071213 to explore whether this could be a neutron star or black hole binary event. In a 5.17 ks observation on UT 2007 December 21.779, we did not detect a single X-ray photon. Thus, 8.4 days after discovery, we derive a 99% confidence upper limit on the 0.3–7 keV flux of <9 × 10⁻⁴ counts s⁻¹. Assuming a power-law spectrum with index of two and column density (n_H) of 6 × 10²⁰cm⁻², we derive a luminosity limit of <8.9 × 10³⁶ erg s⁻¹. Extensive X-ray monitoring of classical novae in Andromeda (Henze et al. 2010a, 2010b) suggests that this upper limit is not inconsistent with classical novae.

A multitude of methods have been used in the literature to measure extinction to novae. Darnley et al. (2006) compared a synthetic dust-free stellar r − i map of M31 to an observed r − i color map of M31 and used the difference between the maps to generate a dust map of M31. The location of the nova on this map determined how much extinction needed to be applied. This assumed that the novae were behind the galaxy and suffered extinction due to the entire column of dust. The average extinction as determined by this method is A_i = 0.8. The galactic extinction along the line of sight of M31 of A_i = 0.13 (Schlegel et al. 1998). Shafter et al. (2009) compared the observed color of the new nova to that of a well-studied nova to derive the extinction. Kogure (1961) used the Balmer decrement (H_α/H_β) and attributed the excess in the ratio over the theoretical Case B value to dust.

Our preferred method of computing extinction is by using the spectroscopic Balmer decrement where spectra dominated by line emission are available. We subtract the continuum, measure the flux ratio of the two lines, and then use

$$\begin{displaymath} A_{g} = 3.793 \,\times \, E(B-V) = 3.650 \,\times \, E(g-r) \end{displaymath}$$

$$\begin{displaymath} = 3.650 \,\times \, \log _{10}(H_{\alpha }/H_{\beta }) - 1.75 \pm 0.14. \end{displaymath}$$

The uncertainty comes from the range in expected ratios for Case B of 2.76–3.30.

Our second choice is to use the g − i color of the nova at maximum, compare against the typical g − i color, and attribute the reddening to dust.

van den Bergh & Younger (1987) compiled photometry of several Galactic novae and derived an average color at maximum of 〈B − V〉₀ = 0.23 ± 0.06 mag. Following Shafter et al. (2009), we can use the colors of an A5V star (T = 8200 K) to translate 〈B − V〉₀ to 〈g − i〉₀. Using colors of an A5V star from Kraus & Hillenbrand (2007), we get

$$\begin{displaymath} \langle g-i \rangle _{0} = 1.88-2.15 = -0.27\; {\rm mag.} \end{displaymath}$$

Now,

$$\begin{displaymath} A_{g} = 3.793 \,\times \, E(B-V) = 2.223 \,\times \, E(g-i). \end{displaymath}$$

Hence,

$$\begin{eqnarray*} A_{g} &=& 2.223 \,{\times} \, [(g\,{-}\,i)_{\rm obs}- \langle g\,{-}\,i \rangle _{0}]\nonumber\\ &=& 2.223 \,{\times }\, [(g\,{-}\,i)_{\rm obs} + 0.27]. \end{eqnarray*}$$

For the case of M31N2007-11d (Shafter et al. 2009), where data for both the above options are available, we get consistent answers: g − i = 0.7 mag at maximum suggests A_g = 0.75 mag and a Balmer decrement of 4.6 suggests A_g = 0.65 mag.

If neither a color at maximum nor a spectrum at late-time is available, we use the average line-of-sight extinction to the host galaxy using the Schlegel maps. The uncertainty in extinction calculation significantly contributes to the uncertainty in the peak magnitude.

We note here that for the case of P60-M82OT-081119, the light curve was unusually red for a nova and the extinction correction may be overestimated.

The heterogeneity in nova light curves suggests that a single parameter may not characterize the decline well. Traditionally, the time to decay from peak by one magnitude (t1), two magnitudes (t2), or three magnitudes (t3) is used. For several novae (e.g., M31N2007-10a, M31N2008-08c, M31N2008-11a), the decline is more or less linear and t1 can be approximated as half of t2. For some novae (e.g., M31N2008-10b), the light-curve behavior is more complex and this simplification is not applicable. In Table 5, we see that t2 values (where available) are sometimes larger and sometimes smaller than twice the t1 value. Recently, Strope et al. (2010) put together a catalog of very well observed Galactic novae and tried to classify the diversity based on light-curve morphology—they found that 20% of the novae in their sample have "jitters" or "oscillations" superposed on the decline.

We note here that we did not have data to measure the decline of the nova in NGC 891 and hence it is excluded from further MMRD analysis.

Given the cadence of P60-FasTING, we were able to catch several novae on the rise. We define the rate of rise as the average slope between first detection and peak detection and summarize in Table 5. We find a wide range of rise times, from >1.8 mag day⁻¹ (e.g., M31N2008-11a) to 0.2 mag day⁻¹ (e.g., M31N2008-09a). It is not clear how previous determinations of the MMRD in the literature dealt with the uncertainty in the peak magnitude due to inadequate coverage. In particular, since previous surveys likely had a relatively slower cadence, missing the peak may be a substantial source of error. Slower cadence and/or shallower depth would correspond to a weaker constraint on the rise time of the nova. Due to gaps on account of weather, some of the P60 light curves have constraint weaker than 0.1 mag day⁻¹ on the rate of rise. Hence, we do not use the light curves of P60-M81OT-080926 or P60-M82OT-080429 for subsequent analysis of the MMRD relation.

For spectroscopy the primary analysis was to classify the novae by their spectra. The taxonomy of novae were laid out by Payne-Gaposchkin (1964) and McLaughlin (1960). The most prominent feature in all classical novae is Balmer emission. Williams (1992) proposed that there is a two-component structure of the emitting gas—discrete shell and continuous wind. If the wind mass-loss rate is low, the effective photosphere is smaller, the radiation temperature higher, the level of ionization of the shell higher, resulting in a shell-dominated He/N spectrum. If the wind mass-loss rate is high, it results in a wind-dominated Fe ii spectrum.

Thus, classical novae are divided into two principal families—the "Fe" class (dominated by Fe ii lines, often low velocity and showing P-Cygni profiles) and "He/N" class (dominated by He and N lines, often high velocity and flat or jagged-topped profiles). These evolve into nebular spectra with four classes based on the prominent forbidden lines—standard (e.g., [N ii], [O ii], [O iii]), neon (e.g., [Ne v], [Ne iii]), coronal (e.g., [Fe x]), or no forbidden lines. The Fe class novae are expected to evolve into standard or neon nebular spectra. The He/N class are expected to evolve into neon, coronal, or no forbidden line spectra. Some novae are classified as "hybrid" as they start out with high velocity Fe ii features and quickly evolve into showing He/N features (e.g., V745 Sco, V3890 Sgr, M31N2008-11a).

Majority of the P60-FasTING spectra show clear permitted lines from Fe ii (42), Fe ii (37,38) and O i. The line velocities are low, and typical Gaussian FWHM are <2500 km s⁻¹ with the exception of P60-M81-080925 where H_α velocity is 3000 km s⁻¹. P60-NGC 2403-090314 shows weak Fe II(42) and weak P-Cygni profiles in Balmer lines and can tentatively also be classified as Fe ii class. P60-M81-071213 and P60-M82-080429 have very low signal-to-noise ratio (S/N) and no features other than the Balmer lines are detected; hence, we cannot classify them. Multiple spectra of P60-M81-081203 were taken; initially, the spectra showed a featureless continuum (obtained around maximum light) and later (about 12 days after maximum light), evolved to show Balmer lines, Fe ii (42), O i. We summarize spectral classifications in Table 4. For five novae in M31, other groups obtained spectra and we summarize their classification in Table 3.