EDP Sciences
Free Access
Issue
A&A
Volume 552, April 2013
Article Number A97
Number of page(s) 9
Section Stellar structure and evolution
DOI https://doi.org/10.1051/0004-6361/201220303
Published online 05 April 2013

© ESO, 2013

1. Introduction

From the photometric point of view, Type Ia supernovae (SNIa) are characterized by a sudden rise and decay in their luminosity, followed by a slowly fading tail. From the spectroscopic point of view, they are characterized by the lack of H-lines and the presence of Si II-lines in their spectra during the maximum light and by the presence of Fe emission features during the nebular phase. A noticeable property is the spectrophotometric homogeneity of the different outbursts. Furthermore, in contrast to the other supernova types, they appear in all kinds of galaxies. These properties point to an exploding object that is compact and free of hydrogen, which can be activated on short and long time scales and is able to synthesize a minimum of 0.3 M of radioactive 56Ni to power the light curve. These constraints immediately led to the proposal that SNIa were the outcome of the thermonuclear explosion of a mass-accreting C/O white dwarf (WD) near the Chandrasekhar limit (Hoyle & Fowler 1960) in a close binary system.

Despite this homogeneity, some differences appear when SNIa are observed in detail. Now it is known that there is a group of SNIa with light curves showing very bright and broad peaks, the SN1991T class, which represents 9% of all the events. There is another group with a much dimmer and narrower peak and that lacks of the characteristic secondary peak in the infrared, the SN1991bg class, which represents 15% of all the events. To these categories a new one has been recently added that contains very peculiar supernovae: the SN2002cx class, representing  ~5% of the total. These supernovae are characterized by high-ionization spectral features in the pre-maximum, such as the SN1991T class, a very low luminosity, and the lack of a secondary maximum in the infrared, as in the SN1991bg class. The remaining ones, which amount to  ~70%, present normal behavior and are known as Branch-normal (Li et al. 2011c). However, even the normal ones are not completely homogeneous, and they show different luminosities at maximum and light curves with different decline rates (Li et al. 2011b). This variety has recently increased with the discovery of SN2001ay, which is characterized by a fast rise and a very slow decline (Baron et al. 2012). This diversity strongly suggests that different scenarios and burning mechanisms could be operating in the explosion.

From the point of view of the explosion mechanism as seen in one-dimensional models, it is possible to distinguish four cases (Hoeflich & Khokhlov 1996; Hillebrandt & Niemeyer 2000): the pure detonation model (DET), the pure deflagration model (DEF), the delayed detonation model (DDT), and the pulsating detonation model (PDD). The equivalent models in three dimensions also exist, but with a wider variety of possibilities. An additional class are the so-called sub-Chandrasekhar (SCh) models in which a detonation triggered by the ignition of He near the base of a freshly accreted helium layer completely burns the white dwarf. At present, there is no basic argument for rejecting any of the models, except the DET ones that are incompatible with the properties of the spectrum of SNIa at maximum light. Existing observations also pose severe constraints to the total amount of 56Ni that can be produced by the He-layer in SCh models (Hoeflich & Khokhlov 1996; Nugent et al. 1997; Woosley & Kasen 2011).

According to the nature of the companion, either non-degenerate or degenerate, progenitors can be classified as single degenerate systems – SD (Whelan & Iben 1973) – or double degenerate systems – DD (Webbink 1984; Iben & Tutukov 1985). The distinctions are important in order to interpret the observations since, depending on the case, the white dwarf can ignite below, near, or above the Chandrasekhar mass, and consequently the total mass ejected and the mass of 56Ni synthesized can be different. At present, it is not known whether both scenarios can coexist or just one is enough to account for the supernova variety. Observations of the stellar content in the interior of known SNIa remnants point towards one possible SD candidate (Tycho SNR, see Ruiz-Lapuente et al. 2004; González Hernández et al. 2009; Kerzendorf et al. 2009) and two almost certain DD candidates (SNR0509-67.5 and SNR 0519-69.0 Schaefer & Pagnotta 2012; Edwards et al. 2012).

The detection of γ-rays from supernovae can provide important insight into the nature of the progenitor and especially into the explosion mechanism, since the amount and distribution of the radioactive material produced in the explosion strongly depend on how the ignition starts and how the nuclear flame propagates (see Gómez-Gomar et al. 1998; Isern et al. 2008, for a detailed discussion of how these differences are reflected in the spectra). The advantages of using γ-rays for diagnostic purposes relies on their penetrative capabilities, on their ability to distinguish among different isotopes, and on the relative simplicity of their transport modeling as compared with other regions of the electromagnetic spectrum. Unfortunately, such observations have not been achieved so far because of the poor sensitivity of the instruments. For this reason it has only been possible up to now to place upper limits on the SN1991T (Lichti et al. 1994) and SN1998bu (Georgii et al. 2002) events.

Several authors have examined the γ-ray emission from SNIa (Gehrels et al. 1987; Ambwani & Sutherland 1988; Burrows & The 1990; Ruiz-Lapuente et al. 1993; Hoeflich et al. 1994; Kumagai & Nomoto 1997; Timmes & Woosley 1997; Gómez-Gomar et al. 1998; Sim & Mazzali 2008). To explore the above model variants, we used as a guide the properties obtained with the code described in Gómez-Gomar et al. (1998), which is based on the methods described by Pozdnyakov et al. (1983) and Ambwani & Sutherland (1988). To test the consistency of this model, the results of this code were successfully cross-checked with those obtained by other authors (Milne et al. 2004). This code was later generalized to three dimensions (Hirschmann 2009).

Before and around the epoch of maximum of the optical light curve, the γ-ray emission can be characterized (Fig. 2 of Gómez-Gomar et al. loc.cit.) as follows: i) a spectrum dominated by the 56Ni 158 and 812 keV lines; ii) because of the rapid expansion, the lines are blueshifted but their energy peak quickly evolves back to the red as matter becomes more and more transparent. The emergent lines are broad, typically from 3% to 5%. Because of the Doppler effect the 812 keV line blends with the quickly growing 56Co 847 keV line, forming a broad feature; iii) the intensity of the 56Ni lines rises very quickly with time, after being very weak at the beginning, even in the case of SCh models. This fact, together with the relatively short lifetime of 56Ni, makes the observational window rather short.

SN 2011fe (RA = 14:03:05.81, Dec = +54:16:25.4; J2000) was discovered by M101 on August 24, 2011 (Nugent et al. 2011a). The absence of hydrogen and helium, coupled with the presence of silicon in the spectrum, clearly indicates that it belongs to the SNIa class. Since it was not visible on August 23, this supernova must have been detected about one day after the explosion (Nugent et al. 2011a). Furthermore, since M101 is at a distance of 6.4 Mpc (Stetson et al. 1998; Shappee & Stanek 2011), SN2011fe is the brightest SNIa detected in the last 25 years. This distance is slightly less than the maximum distance at which current gamma-ray instruments should be able to detect an intrinsically luminous SNIa. The closeness of SN2011fe has made it possible to obtain the tightest constraints on the supernova and its progenitor system to date in a variety of observational windows. Red giant and helium star companions, symbiotic systems, systems at the origin of optically thick winds or containing recurrent novae are excluded for SN2011fe (Li et al. 2011a; Bloom et al. 2012; Brown et al. 2012; Chomiuk et al. 2012), leaving only either DD or a few cases of SD as possible progenitor systems of this supernova.

In this study we analyze the data obtained by INTEGRAL during the optical pre-maximum observations spanning from 4258.8733 IJD (August 29, 20:59 UT) to 4272.1197 IJD (September 12, 2:52 UT) with a total observation time of 975 419 s. This schedule was essentially determined by the constraints imposed by the Sun, according to the TVP tool of INTEGRAL, which prevented the observation just beyond the optical maximum where the 56Ni lines are expected to peak. Despite this limitation, these early observations were triggered to constrain any predicted early gamma-ray emission as may be expected from some variants of SCh models. In the next section, we present the data obtained with the instruments onboard INTEGRAL. Then, we discuss the limits they can put on present models of SNIa explosion and conclude.

2. INTEGRAL data

INTEGRAL (Winkler et al. 2003) is able to observe in gamma-rays, X-rays and visible light. It was launched in October 17, 2002 and was injected into a highly eccentric orbit with a period of about three days in such a way that it spents most of its time well outside the radiation belts of the Earth. The spacecraft contains two main instruments, SPI, a germanium spectrometer for the energy range of 18 keV to 8 MeV with a spectral resolution of 2.2 keV at 1.33 MeV (Vedrenne et al. 2003), and IBIS, an imager able to provide an imaging resolution of 12 arcmin FWHM (Ubertini et al. 2003), which has a CdTe detector, ISGRI, able to provide spectral information of the continuum and broad lines in the range of 15 keV to 1 MeV (Lebrun et al. 2003). Other onboard instruments are an X-ray monitor, JEM-X, that works in the range of 3 to 35 keV (Lund et al. 2003), and an optical camera, OMC, able to operate in the visible band of the spectrum up to a magnitude of 18 (Mas-Hesse et al. 2003).

2.1. OMC data

thumbnail Fig. 1

Light curves in the V-band obtained from DDT models that produce, from top to bottom, 0.63, 0.45, 0.27, 0.09 M of 56Ni and satisfy the brightness-decline relationship.

Open with DEXTER

The height of the maximum of the optical light curve depends on the total amount of 56Ni synthesized during the explosion, its distribution in the debris, the total kinetic energy of matter and the opacity (Arnett 1996). The brightness decline relation (Δm15) relates the absolute brightness at maximum light and the rate of the post-maximum decline over 15 days. This relationship is well understood from theory: light curves are powered by the radioactive decay of 56Ni (Colgate & McKee 1969). More 56Ni increases the luminosity and causes the envelopes to be hotter. Higher temperature means higher opacity, hence longer diffusion time scales and slower decline rates after maximum light (Hoeflich et al. 1996; Nugent et al. 1997; Umeda et al. 1999; Kasen et al. 2009). The Δm15-relation holds up for virtually all explosion scenarios as long as there is an excess amount of stored energy to be released (Hoeflich et al. 1996; Baron et al. 2012). The tightness of the relation observed for Branch-normal SNe Ia is about 0.3m (Hamuy et al. 1996; Perlmutter et al. 1999), and it is consistent with explosions of models of similar mass. Since DDT models (see Fig. 1) provide a reasonable spectral evolution, their maximum brightness is consistent with the Hubble constant and the Δm15-relationship is consistent with the observations with a factor between 1 and 1.3, we use these models to estimate the total amount of 56Ni freshly synthesized.

The properties and the photometric characterization of the OMC can be found in Mas-Hesse et al. (2003). The data obtained during the orbits 1084–1088 plus the data corresponding to orbits 1097–1111, were analyzed with the Offline Scientific Analysis Software (OSA, version 9) provided by the ISDC Data Centre for Astrophysics (Courvoisier et al. 2003). The fluxes and magnitudes were derived from a photometric aperture of 3 × 3 pixels (1 pixel = 17.504 arcsec), slightly circularized, i.e. removing one quarter pixel from each corner (standard output from OSA). The default centroiding algorithm was used, i.e. the photometric aperture was centered on the source coordinates. We checked that the photometric aperture of 3 × 3 pixels does not include any significant contribution by other objects. Because the source was bright enough, combining several shots was not required to increase the signal-to-noise ratio. To only include high-quality data, some selection criteria were applied to individual photometric points by removing those measurements with a problem flag. Shots were checked against saturation, rejecting those with long exposures (200 s) for V < 10.5. To avoid noisy measurements, the shortest exposures (10 s) were not used.

The extinction along the line of sight has two components, one due to the Milky Way and the other to M101. Since M101 has a galactic latitude of  ~60° the Milky Way contribution is expected to be small. Shappee & Stanek (2011) estimate E(B − V) = 0.009 which means an extinction of AV ~ 0.03 mag. The M101 contribution is also expected to be small since it is a face-on galaxy, and SN2011fe is placed in a region with a low concentration of interstellar dust and gas (Suzuki et al. 2007, 2009). It is possible to obtain a rough upper limit for the extinction using the observations (Shappee & Stanek 2011) of two regions containing Cepheids that overlap just at the position of SN2011fe. The average of the two extinctions, E(B − V) = 0.2 mag, would imply an absorption of AV = 0.62 mag. However, since the region containing the supernova has a low concentration of interstellar matter, the comparison of the light curve at different bands and the absence of strong sodium lines in the spectrum suggest that extinction is small. Thus the adopted absorption affecting the supernova is taken to be AV ~ 0.03.

thumbnail Fig. 2

Light curve of SN2011fe in the V-band obtained with the OMC/INTEGRAL assuming a distance of 6.4 Mpc (dots) for the orbits 1086–1088 and 1097–1101. The continuous line represents a reasonable fit obtained with a DDT model (see text).

Open with DEXTER

The V-band light curve reached the maximum, V = 9.99, at the day IJD = 4271.44, in agreement with Richmond & Smith (2012). When taking a canonical distance of 6.4 Mpc (see, however, Tammann & Reindl (2011)) and assuming no extinction, the absolute magnitude should be MV = −19.04 at its maximum, thus indicating that the SN2011fe was a slightly dim average SNIa. The V-band light curve can be well fitted with a delayed detonation model (see Fig. 2) of a Chandrasekhar mass white dwarf igniting at ρC = 2 × 109 g cm-3, making the transition deflagration/detonation at ρtr = 2.2 × 107 g cm-3. This model produces 0.51 M of 56Ni, although if extinction and distance uncertainties are taken into account, this value could easily be around ten percent larger (Höflich et al. 2002). A value that is also consistent with the estimations by Röpke et al. (2012) using an independent DDT model and a violent merger model, and it is roughly equivalent to the one obtained with model DDTe of Table 2. We note that this yield of 56Ni is consistent with the value derived by Nugent et al. (2011b). From this theoretical model, Δm15(B) = 1.2 ± 0.2, in agreement with Tammann & Reindl (2011) but slightly less than the value found by Richmond & Smith (2012). The observation of the early light curve, 4 h (Bloom et al. 2012), and 11 h (Nugent et al. 2011b) after the explosion strongly supports the scenario based on the explosion of a C/O white dwarf.

2.2. SPI data

thumbnail Fig. 3

Time variation of the event rate in the ACS of SPI during the pre-maximum observations of SN2011fe.

Open with DEXTER

thumbnail Fig. 4

Time variation of the event rate in the GeD of SPI during the pre-maximum observations of SN2011fe.

Open with DEXTER

Only data in the low-energy range (≲2 MeV) have been analyzed in this study of the SPI output. The energy calibration was performed for each orbit by fitting parameters of a four degree polynomial function with the channel positions of the instrumental background lines at 23.4 keV, 198.4 keV, 309.9 keV, 584.5 keV, 882.5 keV, and 1764.4 keV. The precision of the resulting calibration is better than  ~0.1 keV at 1 MeV. The calibrated single-detector and multiple-detector1 event data have been binned into separated spectra at 0.5 to 50 keV per bin.

The time-averaged energy spectrum of the supernova was extracted by a model fitting method. The flux of the source and the instrumental background are both fitted to the data (counts per detector per pointing at each energy bin), assuming a point source at the SN2011fe position. The instrumental background is fitted per pointing by assuming that the relative background rate in each detector of the Ge camera is fixed, and is obtained by summing the counts per detector of all the pointings of the observation. We have verified that the detector pattern did not change with time. This background modeling method is adapted to the analysis of data that show strong instrumental background variations on time scales  ≲3 days2. Figures 3 and 4 display the count rate in the anticoincidence system (ACS) and in the germanium detector (GeD) of SPI. The ACS count rate could be used to trace the instrumental background fluctuations in the germanium detectors (Jean et al. 2003). The passage through the radiation belts and the impact of the solar activity are clearly seen. Solar activity was particularly influential during this period of observations because of two solar flares occurring on  ~4266 and 4267 IJD followed by a Forbusch decrease around 4268.5 IJD. To check that these background variations do not produce artifacts, we also performed the model-fitting analysis using a filtered data set without periods with strong rate variations. The results obtained with filtered and non-filtered data are statistically equivalent. Consequently, we decided to use unfiltered data in the next steps of the analysis.

Figure 5 displays the spectrum of the whole observation. It was obtained by combining the spectra extracted by model fitting from single-detector and multiple-detector events. In this case, the model fitting was performed with data rebinned in 50 keV bins.The spectrum does not show any significant feature. A χ2 test shows that it is consistent with a Poissonian background (χ2 = 12.4 with a d.o.f. = 10). A similar conclusion is obtained when the spectra are extracted by model fitting with data rebinned in 5 keV or 2 keV bins.

thumbnail Fig. 5

Spectrum of the whole early observation of SN2011fe extracted from SPI data with 50 keV bins.

Open with DEXTER

No significant excess was found in the spectrum, even at the energies of the strongest gamma-ray lines that are expected from the decay of 56Ni (blueshifted 158 and 812 keV lines). Table 1 presents the upper-limit fluxes derived from the analysis at the energies of interest for several band widths (Isern et al. 2011).

Table 1

Upper limit of the flux in selected spectral regions for SPI (2σ), JEM-X (2σ), and IBIS/ISGRI (3σ) for the entire pre-maximum observation period.

2.3. IBIS/ISGRI data

IBIS uses two detection layers to cover the same energy range as SPI. The low-energy camera, ISGRI, uses 16 384 thin CdTe detectors operated at ambient temperature. The imaging is much better (12′ FWHM) than for SPI, but the spectral resolution is more limited, and the efficiency begins to drop above 100 keV. As a result, the IBIS/ISGRI sensitivity to the 812 keV line from 56Ni is far worse than the SPI one at the same energy even in the case of a broad line. Conversely, IBIS/ISGRI is much better at work than SPI to reveal a low-energy continuum or broad lines, such as the 158 keV 56Ni or 122 keV 57Co ones. To summarize the picture, one could say that SPI is better at detecting narrow lines in a spectrum, while IBIS/ISGRI is better at detecting point sources in broad energy-range sky images. Two data processing methods have been followed in parallel. One uses the standard OSA-9 version and the other takes advantage of the developments for the forthcoming OSA-10. The latter approach includes two new corrections: for the spectral drift along the mission and for the flat field. In either case no significant (greater than 3σ) signal was found at the position of SN2011fe. Table 1 gives the upper limits obtained for the 60–170 keV, 90–172 keV and the 150–172 keV bands. Similarly, maps using the last six days of the observing period for the same energy ranges do not show any significant emission at the position of SN2011fe.

2.4. JEM-X data

Both JEM-X units were simultaneously operating at the time of these INTEGRAL observations and they can be used to constrain the continuum emission using a source search in broad band images as in the IBIS/ISGRI case. Owing to the smaller field of view of the JEM-X monitors compared to those of SPI and IBIS (Lund et al. 2003), we selected and analyzed only those pointings where SN2011fe was within 5° of the pointing direction. The data obtained during the orbits 1084–1088 (August 29 to September 12, 2011) were analyzed with OSA-9 following the standard procedure. Images from single pointings were combined into one mosaicked image for each X-ray monitor. The two composite images from JEM-X1 and JEM-X2 were then merged to obtain the final image, providing an on-source effective exposure time of 450 ks. The imaging analysis was performed in the 3–25 keV band, and in the 3–10 keV and 10–25 keV sub-bands. SN2011fe is not detected in any of the JEM-X images. Assuming a Crab-like spectrum, we estimate 2σ upper limits on the 3–10 keV and 10–25 keV fluxes of 5 × 10-4 ph cm-2 s-1 and 4 × 10-4 ph cm-2 s-1, respectively (see Table 1). The source is also not detected when analyzing the JEM-X data separately from orbits 1086–1088 (Sept 4th to 12th, 2011), corresponding to the observations closer to the epoch of the expected 56Ni line maximum.

3. Discussion

The γ-ray spectrum of SNIa mainly depends on the total amount and distribution of 56Ni synthesized during the explosion, as well as on the chemical structure and velocity distribution of the debris (Gómez-Gomar et al. 1998). For instance, when 56Ni is present in the outer layers of some of the SCh models, the corresponding lines should appear very early in the spectrum. However, for a given flux, there are at least two more factors that determine whether the γ-signal is detectable: the change in the signal with time and the width of the lines.

Table 2

Kinetic energy (K) and mass of 56Ni produced by different models of explosion.

To check the influence of the mass and distribution of 56Ni, several models obtained under different hypotheses about the burning regime or the explosion mechanism have been considered. Most are one-dimensional, spherically symmetric models, which should be sufficient for SN2011fe in view of the small amount of global asymmetry suggested by spectropolarimetric measurements (Smith et al. 2011). We have not included models of DD explosions for two reasons. First, Nugent et al. (2011b) satisfactorily reduced the rising part of the light curve of SN2011fe with a simple analytic model involving a Chandrasekhar mass white dwarf, and found only small amounts of unburned carbon in the early spectra of this supernova. Second, in spite of recent advances on simulations of almost normal SNIa from mergers of massive white dwarfs (Pakmor et al. 2012), theoretical models of DD explosions are not as mature at present as those involving either Chandrasekhar or sub-Chandrasekhar mass white dwarfs accreting from a non-degenerate companion. The number of free parameters involved in DD explosions is therefore too large to allow for efficient constraints derived from upper limits on the gamma-ray emission of SN2011fe.

Table 2 displays the main characteristics of the models used in the present study. The DETO model (Badenes et al. 2003) corresponds to a pure detonation of a white dwarf near the Chandrasekhar limit. It is also representative of the most massive models computed by Fink et al. (2010). The W7 is the classical model of Nomoto et al. (1984). The DD202c (Hoeflich et al. 1998) and the DDTc,e (Badenes et al. 2005) models are delayed detonation models that produce different amounts of 56Ni and have different expansion energies. The HED6 model corresponds to the explosion of a 0.6 M C/O WD that has accreted 0.17 M of helium in Hoeflich & Khokhlov (1996), and SC1F and SC3F (E. Bravo, unpublished) are also SCh models equivalent to models 1 and 3 of Fink et al. (2010). Finally, the SCOP3D model is a three-dimensional SCh model that corresponds to model A of García-Senz et al. (1999).

thumbnail Fig. 6

Comparison of the mean flux obtained from the theoretical models during the early observation period with the 2σ upper-limit flux in 5 keV bins. Theoretical spectra were obtained with the code described in Gómez-Gomar et al. (1998).

Open with DEXTER

Table 3

Measured and predicted flux in spectral regions where 56Ni lines are expected to be as a function of models.

Figure 6 displays the SPI 2σ upper limit spectrum obtained with 5 keV bins as, well as the gamma ray spectra predicted by several models for this early observation, assuming that the distance of SN2011fe is 6.4 Mpc. All the models are well below the upper-limit. The predicted intensity of the 56Ni 158 keV, 270 keV, 480 keV, 750 keV, and 812 keV lines is larger for the detonation model (DETO) and the SCh models SCOP3D, SC3F, and SC1F, as expected. Since the spectral shape (width and centroid) and intensity change from one model to another, the energy band used to extract the fluxes of every line or complex of lines was chosen to provide the optimum significance for each model. Results are displayed in Table 3.

Despite this non-detection and the bounds imposed by the optical light curve, the compatibility of the different models with the zero flux hypothesis has been tested. Since the measured fluxes are compatible with zero, the χ2 value for each model was computed as the quadratic sum of the expected significance of the predicted fluxes3. Then, the probability of rejection of each particular model was derived by computing the probability that the χ2 is smaller than the measured value within χ2 statistics. The rejection probability of the selected models is given in the fifth column of Table 3. None of them can be firmly rejected by SPI data. Although the probability of rejecting the DETO model is 92.4%, this value corresponds to a significance of  ~1.4σ. To reject it at a 99% of confidence level, the intensity of the lines should be higher by a factor 1.45 times or, equivalently, the distance to SN2011fe should be less than 5.3 Mpc. Since the DETO model is the one that produces the brightest lines, SPI can only provide interesting results during this epoch if the distance to the supernova is less than this value.

The influence of the temporal behavior of the line intensity and line width on the detectability of the γ emission can be easily seen by estimating the significance of the observation. In the limit of weak signals this significance is given by (Jean 1996) (1)where Δt is the observation time, Aeff the effective area at the corresponding energies, ϕ is the flux (cm-2 s-1) in the energy band ΔE, V is the volume of the detector, and b the specific noise rate (cm-3 s-1 keV-1), where it has been assumed that it is weakly dependent on the energy and time in the interval of interest.

If the flux grows like , the significance reached by integrating the time interval (ti − Δt,ti) is (2)For αΔt ≪ 1, the significance behaves as and has a maximum at αΔt = 1.26. Furthermore, the dependence on ΔE clearly shows the convenience of taking a value that maximizes the signal to noise ratio. Unfortunately, since the value of α is not known a priori, the optimal observing time is not known in advance. Since the time dependence of the model fluxes does not strictly follow an exponential growth, we analyzed the optimum observation periods for each model. Table 4 displays the results. In general, models and energy ranges showing weak variation of the flux are best detected when the observation period is long whereas models with a strong variation in their flux during the first two weeks are best detected using data at those times.

It is well known (Gómez-Gomar et al. 1998) that the width of the line has a strong influence on their detectability. In the case of the present observations it is possible to estimate the significance of a narrow line, ΔE ≲ 2 keV, by comparing the measured 1σ flux uncertainty (see Table 3), ϕ1, with the flux ϕ and the width ΔE predicted by the model (3)In the case of the 158 keV line (Table 3), the flux predicted by the DETO model is ϕ = 8.1 × 10-5 ph s-1 cm-2, ϕ1 = 7.8 × 10-5 ph s-1 cm-2, and ΔE = 25 keV, where ΔE has been chosen to maximize the signal to noise ratio. With these data, n ~ 3.8, and in this case with this hypothesis, the 158 keV line would have been detected by SPI if it had been narrow. In contrast, in the model SC1F, ϕ = 3.2 × 10-5 ph s-1 cm-2, ϕ1 = 6.7 × 10-5 ph s-1 cm-2, ΔE = 20 keV and the result is n ~ 1.5. Here, the line would only be marginally detectable even in the narrow case.

The luminosity at the maximum of the optical light curve is proportional to the total mass of 56Ni, while the shape of the optical peak is determined by the amount of 56Ni and the kinetic energy of the remnant (Arnett 1997). As already been mentioned, the favored model by the optical light curve is the DDTe one, and it seems worthwhile to interpret the SPI observations in terms of this model.

Figure 7 (upper panel) displays the early gamma ray spectra predicted by DDTe for several instants after the explosion. The way the different lines grow and the width of them is clearly seen. The main problem with the 812 keV line is that, owing to the Doppler effect and energy degradation of photons by Compton scattering, it blends with the 847 keV line of 56Co (see Fig. 7, lower panel) and makes it more difficult to interpret the signal. During this observation, and for this model, the energy interval that maximizes the signal to noise ratio is  ~25 keV centered at 826 keV, and the corresponding flux at day 18 is ϕ = 3.4 × 10-6 ph s-1 cm-2. In contrast, the 158 keV line has a more regular behavior and it is better suited to diagnostic purposes. At day 20 after the explosion, the flux in this line is ϕ = 4.7 × 10-6 ph s-1 cm-2 for ΔE = 20 keV. In both cases, the signal is too weak to be detected with SPI, and the source should be at a distance of  ≲2 Mpc to be detectable.

Table 4

Optimal observation period for several models and energy ranges with SPI.

thumbnail Fig. 7

Early gamma ray spectra predicted by the delayed detonation model that best fits the OMC (see Fig. 2) at days 8, 11, 14, 17, 20 (green, blue, cyan, magenta, black) at a distance of 6.4 Mpc (upper figure). The lower panel displays the evolution of the profile of the 56Ni 812 kev-56Co 847 keV features.

Open with DEXTER

The 200–540 keV band contains almost all the photons produced by the annihilation of positrons produced during the 56Ni chain decay. Since the emission of 511 keV only represents the 25% of the total number of the annihilation photons, this band offers, in principle, better possibilities of detection than the 511 keV line itself. In the DDTe model, the maximum of the band occurs at day 50 and amounts of 1.05 × 10-4 ph s-1 cm-2, while the maximum of the 511 keV line occurs at day 74 with a flux of 1.2 × 10-5 ph s-1 cm-2, within a band of 30 keV. The expected emission at day 18 after the explosion, the last day of this observation window, are 4.9 × 10-5 ph s-1 cm-2 and 2.4 × 10-6 ph s-1 cm-2. The 3σ sensitivity in this band is estimated to be 8.7 × 10-4 ph s-1 cm-2 for an integration time of 106 s, well above the value of the expected emission for this model, for which reason this band is not detectable by SPI/INTEGRAL except if the explosion occurs at a distance d ≲ 1−2 Mpc (assuming an explosion similar to what is predicted by the DDTe model).

As in the case of SPI, it is possible to discuss the results from ISGRI in more depth using detailed predictions by different theoretical models. In this case we use the model to produce a daily spectrum corresponding to the period of observation and combine these spectra with the ISGRI sensitive area (ARF) and spectral response (RMF) to predict the observable ISGRI spectrum as a function of time. Using the latest ISGRI background estimate (Mattana, priv. comm.), the expected relative value of the signal-to-noise ratio can be computed as a function of energy and time. This was done for the DDTe, W7, DETO, and SC3F models. The first two of these models predict fluxes that are out of the reach of the ISGRI sensitivity by an order of magnitude even if only the last days of the observing period are used, where the expected signal is higher. Figure 8 illustrates an attempt in the case of the DDTe model. The 60–172 keV upper limit on the ISGRI rate during the last six days of the first observation period is 1.7 × 10-3 s-1 keV-1, which represents 2.2 × 10-4 cm-2 s-1.

thumbnail Fig. 8

ISGRI expected rate spectrum from a DDTe model averaged over the last 6 days of the first observation period. The 3σ ISGRI upper limit for the 60–172 keV band is displayed as a horizontal bar.

Open with DEXTER

thumbnail Fig. 9

ISGRI expected rate spectrum from a DETO model averaged over the entire observation period. The 3σ ISGRI upper limit for the 150–172 keV band is displayed as a horizontal bar.

Open with DEXTER

thumbnail Fig. 10

ISGRI expected rate spectrum from an SC3F model averaged over the entire observation period. The 3σ ISGRI upper limit for the 90–172 keV band is displayed as a horizontal bar.

Open with DEXTER

Since the DETO and SC3F models produce important amounts of 56Ni near the surface, they display an important emission feature in the  ~150–170 keV band. Figure 9 displays the spectrum that could be expected from the DETO model. This spectrum is an average over all the time that ISGRI has been pointed at the source and the optimal band corresponds to 150–172 keV. In this case the 3σ limit corresponds to 7.1 × 10-5 cm-2 s-1. If a source like DETO were been present, the significance of the signal would have been 2.8σ, but there was nothing detected by ISGRI in the field. Model SC3F exhibits similar behavior (Fig. 10), and for the same specifications, the significance in the band is 1.4σ. However, in this case, the continous emission at low energies is important and the optimal observation band is 90–172 keV, while the 3σ limit is 1.1 × 10-4 cm-2 s-1 and the significance of the signal would have been 2.1σ. If these models, which synthesize 1.16 and 0.69 M of 56Ni, respectively, are scaled to the  ~0.5 M demanded by the observations of SN2011fe in the optical, their expected gamma-ray flux would be marginally above the ISGRI sensitivity limit.

4. Conclusions

SN2011fe was observed with the four instruments SPI, ISGRI/IBIS, JEM-X, and OMC operating onboard INTEGRAL just before the maximum of the optical light curve for a period of time of 1000 ks, and the data were compared with the predictions of several theoretical models. SPI data in the bands containing the 158 keV and the 812 keV emission of 56Ni would have allowed us to reject at 98% confidence level (2σ) models that produce large amounts of 56Ni near the surface, if the supernova were closer. For instance, models with a mass fraction of 56Ni in the outer 0.15 M of 0.05 and 0.5 would have been rejected if the supernova distance was closer than 1.4 and 3.7 Mpc respectively, assuming non-detection. Furthermore, ISGRI/IBIS has proved to be very efficient at exploring the low-energy region (~100 keV) and has confirmed that there were not significant amounts of radioactive elements in the outer layers. This picture is consistent with the light curve obtained with the OMC. The observations in the optical suggest that the total amount of 56Ni produced in the event is  ~0.5 M triggered by a mild DDT explosion.


1

We used only double-detector events because including events with greater detector multiplicity does not improve the sensitivity for this analysis (Attié et al. 2003).

2

The analyses performed with background models that use background tracers show strong systematics due to these large instrumental background variations.

3

The expected significance is the predicted flux divided by the uncertainty of the flux measurement in the corresponding energy band.

Acknowledgments

This work was supported by the MINECO-FEDER grants AYA08-1839/ESP, AYA2011-24704/ESP, AYA2011-24780/ESP, AYA2009-14648-C02-01, CONSOLIDER CSC2007-00050, by the ESF EUROCORES Program EuroGENESIS (MINECO grants EUI2009-04170), by the grant 2009SGR315 of the Generalitat de Catalunya. In parts, this work was also supported by the NSF grants AST-0708855 and AST-1008962 to PAH. The INTEGRAL SPI project was completed under the responsibility and leadership of the CNES. We also acknowledge the INTEGRAL Project Scientist Chris Winkler (ESA, ESTEC) and the INTEGRAL personnel for their support of these observations.

References

All Tables

Table 1

Upper limit of the flux in selected spectral regions for SPI (2σ), JEM-X (2σ), and IBIS/ISGRI (3σ) for the entire pre-maximum observation period.

Table 2

Kinetic energy (K) and mass of 56Ni produced by different models of explosion.

Table 3

Measured and predicted flux in spectral regions where 56Ni lines are expected to be as a function of models.

Table 4

Optimal observation period for several models and energy ranges with SPI.

All Figures

thumbnail Fig. 1

Light curves in the V-band obtained from DDT models that produce, from top to bottom, 0.63, 0.45, 0.27, 0.09 M of 56Ni and satisfy the brightness-decline relationship.

Open with DEXTER
In the text
thumbnail Fig. 2

Light curve of SN2011fe in the V-band obtained with the OMC/INTEGRAL assuming a distance of 6.4 Mpc (dots) for the orbits 1086–1088 and 1097–1101. The continuous line represents a reasonable fit obtained with a DDT model (see text).

Open with DEXTER
In the text
thumbnail Fig. 3

Time variation of the event rate in the ACS of SPI during the pre-maximum observations of SN2011fe.

Open with DEXTER
In the text
thumbnail Fig. 4

Time variation of the event rate in the GeD of SPI during the pre-maximum observations of SN2011fe.

Open with DEXTER
In the text
thumbnail Fig. 5

Spectrum of the whole early observation of SN2011fe extracted from SPI data with 50 keV bins.

Open with DEXTER
In the text
thumbnail Fig. 6

Comparison of the mean flux obtained from the theoretical models during the early observation period with the 2σ upper-limit flux in 5 keV bins. Theoretical spectra were obtained with the code described in Gómez-Gomar et al. (1998).

Open with DEXTER
In the text
thumbnail Fig. 7

Early gamma ray spectra predicted by the delayed detonation model that best fits the OMC (see Fig. 2) at days 8, 11, 14, 17, 20 (green, blue, cyan, magenta, black) at a distance of 6.4 Mpc (upper figure). The lower panel displays the evolution of the profile of the 56Ni 812 kev-56Co 847 keV features.

Open with DEXTER
In the text
thumbnail Fig. 8

ISGRI expected rate spectrum from a DDTe model averaged over the last 6 days of the first observation period. The 3σ ISGRI upper limit for the 60–172 keV band is displayed as a horizontal bar.

Open with DEXTER
In the text
thumbnail Fig. 9

ISGRI expected rate spectrum from a DETO model averaged over the entire observation period. The 3σ ISGRI upper limit for the 150–172 keV band is displayed as a horizontal bar.

Open with DEXTER
In the text
thumbnail Fig. 10

ISGRI expected rate spectrum from an SC3F model averaged over the entire observation period. The 3σ ISGRI upper limit for the 90–172 keV band is displayed as a horizontal bar.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.