arXiv:astro-ph/9610020v1 3 Oct 1996 o.Nt .Ato.Soc. Astron. R. Not. Mon. id .Smith, J. Linda icmtla H Circumstellar eevd19 pi .Acpe 96July. 1996 Accepted 3. April 1996 Received [email protected]) [email protected], (e-mail 3 [email protected]) (e-mail 2 [email protected]) [email protected], (e-mail 1 oetJ Cumming, J. Robert models progenitor of test observational an supernovae: yeI uenv.Tems iey(rnhe l 1995) al. et (Branch likely most The . Ia Type a a Ia, SNe o of evolution production chemical . systems, iron the efficient understanding binary the to in stumbling-block of in evolution virtue gap by stellar a and, furthermore, of is, understanding decelera- progenitors ignoranc our supernova the Our Ia and 1995). Type 1996) al. of et al. (e.g., Perlmutter et constant (e.g., lu- Sandage parameter Hubble their tion 1995; the and al. one. measure distances et important to Hamuy cosmological an used at is are issue visible minosities The are compan- progenitor. Ia the the of SNe nature of the about star know known not ion sys- do is still binary much we Ia) a how events, in (SN these given dwarf supernova though, white Ia accreting Surprisingly Type an tem. of a explosion that the accepted is generally is It INTRODUCTION 1 oa rewc bevtr,MdnlyRa,CmrdeCB Cambridge Road, Madingley Observatory, Greenwich Lond Royal College University Astronomy, & Physics of Department Saltsj¨obaden, Sweden 36 S-133 Observatory, Stockholm ubro ye fsse a,i rnil,produce principle, in can, system of types of number A 2 000 a Pettini Max – 19)Pitd2 eray2018 February 21 Printed (1996) 1–7 , ABSTRACT uenve ehv hrfr okdfrnro H s narrow progenitor for possible looked of therefore types have different We supernovae. the between discriminate to fe xlso.W eiea pe ii f2.0 of limit maximu upper before an days derive 10 We taken explosion. 1994D supernova after Ia Type normal the ymtyw n nuprlmtof limit upper an find we calculations symmetry photoionization time-dependent made have we density, 0k s km 10 msinln ttelclH local the at line emission pia,rdoadXrylmt ncntann asls rmTp apr words: use Ia Key most o Type hydrogen. the spectroscopy circumstellar from optical identifying provide loss actually high-resolution probably of mass while can loss, constraining observations mass in X-ray r progenitor the that limits assess find X-ray and We and asymmetry of radio effect optical, the discuss We progenitors. 94)–sas symbiotic stars: – 1994D) 1 α ee Lundqvist, Peter erhn o h rsneo icmtla aeili urnl t currently is material circumstellar of presence the for Searching rmS 94 n uueTp Ia Type future and 1994D SN from − 1 hslmtcnecueol h ihs-asls-aesmitcsys symbiotic highest-mass-loss-rate the only exclude can limit This . 3 n ai .King L. David and uenve–cruselrmte tr:sproa:individua supernovae: stars: – matter circumstellar – supernovae e f n odnW1 B,UK 6BT, WC1E London on, 1 E,UK 0EZ, 3 ii einvlct.T siaetelmtti uso wind on puts this limit the estimate To velocity. region hog oh oeoeflw r lopossible. also subgiant are a overflow, from lobe accreted Roche is ‘hydrogen through and H wind, where giant variables’, the red a cataclysmic where from the H systems, are accretes dwarf dwarfs Symbiotic white white candidates. C-O and promising of systems, most pairs progenitor coalescing different that the conclude of frequencies sation (Branch explosions Ia 1995). argue SN al. to most et seem for models accounting current their though against accreted, is helium of hnrska-asepoin a loocrif occur also can explosions Chandrasekhar-mass ht wr ece h hnrska as . M 1.4 progenitor mass, the Chandrasekhar when each the occur reaches In to lobe dwarf coalesce. believed white Roche dwarfs is from through explosion white He the or C-O case, or two wind H where its accretes or from overflow, dwarf either white companion, C-O a a where those are h atrtocssdffrfo h rti hthydro- that in first the from differ cases two latter The reali- likely the estimated have (1995) al. et Branch M ˙ 3 ∼ 1 . 5 × × 10 10 − α − 16 nahg-eouinsetu of spectrum high-resolution a in 5 r s erg M ⊙ − yr 1 − cm 1 ltv eiso early of merits elative esorol chance only our ffers − suigspherical Assuming . o idsedof speed wind a for n ny65days 6.5 only and m sesfrTp Ia Type for ystems 2 eol ietway direct only he o nunresolved an for u iiso the on limits ful ogenitors. ∼ esas tems . M 0.2 ⊙ Sub- . (SN l ⊙ 2 R. J. Cumming et al. gen is expected to be present in the supernova’s circum- stellar environment. A tempting way to discriminate be- tween different types of progenitor system is therefore to look for signatures of circumstellar matter. In fact, wherever the white dwarf accretes primarily hydrogen from its com- panion, the possibility exists of detecting narrow H i in emis- sion or absorption, either at early times (Wheeler 1992a,b) or after the ejecta have become optically thin (Chugai 1986; Livne, Tuchman & Wheeler 1992). If instead helium is ac- creted, narrow He i or He ii lines could be visible (Branch et al. 1995). Radio (Boffi & Branch 1995) and X-ray (Schlegel & Petre 1993; Schlegel 1995) observations might also be able to detect the interaction of the supernova with its circum- stellar medium (CSM). If the progenitor system is a super- soft X-ray source, narrow emission lines might be detectable from a pre-existing ionized (Rappaport et al. 1994; Branch et al. 1995). Finally, Wang et al. (1996) have sug- Figure 1. Spectrum of SN 1994D around Hα. The broken line −1 gested that polarisation of the supernova light could reveal marks the expected velocity of the supernova, +830 km s . the presence of a dusty asymmetric CSM. Only a few SNe Ia have shown any evidence at all of a CSM, and in all cases serious doubt remains. Branch et al. mum until 10±1 days later (Richmond et al. 1995; Meikle (1983; also Wheeler 1992a) identified unresolved Hα emis- et al. 1996). The spectrum is thus one of the earliest Type sion in a spectrum of SN 1981B taken 6 days after B max- Ia spectra ever taken. Later high-resolution spectra of SN imum (+6 days). The line is, however, blueshifted by 2000 1994D have been presented by Ho & Filippenko (1995), and −1 km s from the local interstellar Ca ii absorption, making by Patat et al. (1996). the identification hard to believe. Graham et al. (1983) and SN 1994D is located 9.0 arcsec west and 7.8 arcsec north Graham & Meikle (1986) reported the detection of an IR (Richmond et al. 1995) of the nucleus of NGC 4526, an S03 excess from SN 1982E. This supernova was never classified, in the cluster. We observed with a long slit of but exploded in an S0 galaxy, suggesting, at the time, that width 1.2 arcsec passing through both the supernova and it was a Type Ia. Type Ib and II events have since been seen the galaxy nucleus. The seeing, measured from the spatial in S0 galaxies. Polcaro & Viotti (1991) reported narrow Hα profile of the supernova, was 1.4 arcsec. Two integrations of absorption from SN 1990M at −4d, though an earlier spec- 1000 s and 750 s were added to produce the final spectrum. trum indicates that the absorption came from the galaxy CCD reduction was carried out in the usual way, wavelength- rather than the supernova (Della Valle, Benetti & Panagia calibrating with a CuAr arc lamp, and flux-calibrating by 1996). In the case of SN 1991bg, no narrow lines were seen comparison with the flux standard Feige 34 (Stone 1977). at +1d (Leibundgut et al. 1993), but narrow Hα and Na i The absolute flux-calibration is accurate to around 10 per- were tentatively identified at +197d by Ruiz-Lapuente et al. cent. (1993). Further spectra, presented by Turatto et al. (1996) The two-dimensional spectra show that faint H ii-region show that all the Fe and Co lines characteristic of SN Ia emission in Hα and [N ii] λλ6548, 6583 can be traced from at this epoch are similarly narrow. The feature identified as the nucleus of NGC 4526 out to the supernova position, but Na i is blended with [Co iii], but there seems to be no con- not beyond it. The line emission shows, however, no abrupt vincing alternative identification for the Hα feature (Turatto change at the position of SN 1994D. At 0.8±0.3 arcsec from et al. 1996). the supernova, the furthest out from the nucleus we could In section 2, we report observational limits on early cir- measure it, the H ii-region Hα emission has a heliocentric −1 cumstellar Hα from the fairly normal Type Ia SN 1994D. In velocity vH II = +830 ± 50 km s . This is our best estimate section 3 we investigate the implications of Hα limits for the of the radial velocity of the supernova itself. The velocities of progenitor systems of SN 1994D, and Type Ia supernovae in local interstellar Na i absorption, +708 km s−1, and the cen- general. In particular, we calculate the time-dependent emis- tre of the galaxy, +625 km s−1 (King et al. 1995), are both sion of the Hα in terms of a circumstellar interaction model. lower. Figure 1 shows the high-resolution spectrum around Finally, we assess what can be learnt from early Hα, radio, Hα. The spectrum has not been dereddened: measurements and X-ray limits. of the Na i interstellar lines by King et al. (1995) and Ho +0.026 & Filippenko (1995) give EB−V =0.046 and 0.026−0.013 , re- spectively, corresponding to corrections in flux of only 5−10 percent, rather less than other errors in our results. 2 OBSERVATIONS AND RESULTS The spectrum shows no sign of any narrow Hα from the On 1994 March 10.14, we used the ISIS spectrograph with supernova, in emission or absorption. (The wavelength range an EEV CCD detector on the William Herschel Telescope includes no interesting helium lines.) We have estimated 3-σ to obtain a high-resolution (R =35 km s−1) spectrum of SN detection limits for Hα absorption and emission lines of var- 1994D, centred around Hα. (The spectrum was first reported ious widths and velocities. We did this by adding Gaussian in Cumming, Meikle & Geballe [1994] and was briefly pre- emission and absorption lines of various strengths, widths sented in King et al. [1995].) The supernova was then only and velocities to the observed spectrum, and measuring the 6.5±1 days old (H¨oflich 1995), and did not reach B maxi- signal-to-noise ratio in the resulting line. Circumstellar Hα from Type Ia supernovae 3

Since detectable circumstellar material is likely to be will have overrun the densest parts of the CSM. We reject slow-moving, any CSM Hα in our spectra would proba- this ionzation source as well. bly be unresolved. For an unresolved line at the local H ii- region velocity, we found that the 3-σ limit for emission flux is 2.0×10−16 erg s−1cm−2. The corresponding equiv- 3.1.3 Preionization alent width limit for an absorption line is 0.028 A.˚ If we instead allow the true supernova velocity to differ by up to To check whether preionization by the progenitor white −1 ii dwarf is important, we assume that the CSM is spherically 200 km s from the local H region velocity, these limits in- −2 −16 −1 −2 symmetric around the progenitor, with density ρ ∝ r , crease to 2.7×10 erg s cm and 0.037 A.˚ The limits for wind where r is measured relative to the white dwarf. This is resolved lines are correspondingly higher. For vFWHM = 95 −1 −16 −1 −2 a reasonable approximation at radii a few times the sep- km s the limits are instead 3.2×10 erg s cm and ˚ × −16 −1 −2 ˚ aration, a, between the two stars. For symbiotic systems, 0.062 A(vSN=vH II), or 5.4 10 erg s cm and 0.079 A 14 −1 −1 a ∼10 cm (e.g., Nussbaumer & Vogel 1987), and we take (−200 km s < (vSN − vH II) < +200 km s ). The data this as the minimum radius for the r−2 law. We can set the easily exclude an absorption line of the strength claimed density to zero within r = a; this simplification only has by Polcaro & Viotti (1991) from SN 1990M. Their line had −1 a small effect on the results. We then compare the rate of vFWHM = 760 km s and an equivalent width of 1.3±0.2 A;˚ ionizing photons from the white dwarf to the total rate of our 3-σ limit for a line of this width from SN 1994D is 0.52 A.˚ recombinations in the wind. For a pure H wind with tem- 4 perature (T4 × 10 ) K, we find that the the wind is preion- ized (at least in the direction away from the red giant) if 3 DISCUSSION < −8 0.48 0.5 0.5 M/u˙ 10 ∼ 5.8 × 10 T4 N˙ 45 a14 . Here u10 is the wind −1 In this section we explore how the Hα flux limit for SN speed in km s , N˙ 45 the rate of ionizing photons emit- 45 −1 1994D, or indeed any , can be used to ted by the white dwarf in units of 10 s , and a14 the 14 constrain models of its progenitor system. As circumstel- separation between the components in units of 10 cm. lar Hα emission requires some source of excitation, we first Approximating the ionizing source as a black body of ef- 5 assess the different potential sources of ionizing radiation. fective temperature Teff =1.5×10 K with (photospheric) ra- 8 For the most promising of these, we use a photoionization dius Rp=9×10 cm implies that the wind is preionized if ˙ < −7 0.48 0.5 model to investigate numerically the relationship between M/u10 ∼ 1.2×10 T4 a14 . The values of Rp and Teff are the progenitor system’s mass loss rate and the luminosity those used for the hot component of a symbiotic system in of circumstellar Hα. The model results are used to place a the model of Nussbaumer & Vogel (1987). Rp is probably limit on the mass loss from SN 1994D’s progenitor system. somewhat higher than we would expect for likely Type Ia Finally, we compare our results with early radio and X-ray progenitors, and overestimates the maximum M/u˙ for which limits on other supernovae from the literature, and assess the preionization is important. relative merits of the different techniques for future events. To estimate the wind temperature in the preionized case we used a steady-state version of the photoionization code described in Lundqvist & Fransson (1988, 1996) and 3.1 Ionization sources Lundqvist et al. (1996, in preparation). We adopted a=1014 × 8 × 5 We consider four possible ionization sources: ionization by cm, Rp=9 10 cm and Teff =1.5 10 K, and assumed solar the radiation accompanying the supernova shock breakout, abundances for the elements we included (H, He, C, N, O, −8 56 ˙ × γ-rays from the decay of Ni, the possibility that the CSM Ne, Na, Mg, Al, Si, S, Ar, Ca and Fe). For M/u10=5 10 , 15 is ionized by the progenitor white dwarf before it explodes the wind is indeed preionized. At R = 10 cm, the ap- (‘preionization’), and radiation from the interaction of the proximate maximum radius of the ejecta at the time of our v ejecta with a circumstellar medium. SN 1994D observation, the wind is ionized to, e.g., C , N v-vi,O v-vi and the temperature is ∼2.5×104 K. Preion- < −7 ization is only important for M/u˙ 10 ∼ 1.9×10 , unless a is 3.1.1 Radiation from shock breakout larger than about 1014 cm, or the white dwarf hotter than T =1.5×105 K. The number of ionizing photons produced at shock break- eff out is probably quite small. For a black body with a ra- 6 dius of 0.013 R⊙ at a temperature of about 10 K, and the 3.1.4 Radiation from the circumstellar interaction region duration of the burst of 1 s (T. Shigeyama, private com- 48 munication), we expect only ∼1.5×10 photons with ener- The fourth possibility follows the suggestion of Chevalier −9 gies above 13.6 eV, enough only to ionize ∼1.3×10 M⊙ of (1984) that Type I SNe whose progenitor white dwarfs hydrogen. We can therefore exclude this possibility as the have red-giant or AGB-star companions may interact with source of excitation. their surroundings in a similar manner to core-collapse supernovae (e.g., SNe 1979C and 1980K: Fransson 1982; Lundqvist & Fransson 1988; SN 1993J: Fransson, Lundqvist 3.1.2 Radiation due to 56Ni-decay & Chevalier 1996). The supernova ejecta collide with the A day or so after the explosion, γ-rays from 56Ni-decay in CSM, setting up a forward- and reverse-shock structure. Be- the supernova ejecta begin to emerge, but these are also hind each shock the gas is heated to high temperatures and unlikely to ionize the CSM. Early on, the radiation is too produces thermal X-ray emission. Being denser, the shocked hard, and the UV tail only shows itself after a few tens of ejecta radiate more strongly than the shocked CSM. days (P. Pinto, private communication). By then, the ejecta To check whether the CSM can be ionized by the reverse 4 R. J. Cumming et al. shock, we have modelled the circumstellar interaction in a similar way to Chevalier (1982). We assume free expansion out to 1014 cm, and that the interaction can be described by Chevalier’s similarity solutions beyond that. The maximum 37 4 −1 velocity of the ejecta was taken to be vej=2.0×10 km s at 1014 cm, meaning that this radius is reached in ∼ 0.6 days. The density slopes of the outer ejecta and the wind 36 −n −s are described by power-laws, ρej ∝ r and ρwind ∝ r , respectively. We choose n = 7 (model W7 of Nomoto, Thielemann & Yokoi [1984]) and s = 2. For solar abun- dances in the ejecta, the temperature of the reverse shock 35 8 Trev ∼ 2.2 × 10 K. In fact the outer supernova ejecta are composed of heavy elements like C and O, and the higher mean molecular weight drives the temperature up 34 to ∼3.5×108 K. −8 For M/u˙ 10=5×10 , and assuming that the shocked ejecta consist mainly of oxygen (as in W7; Nomoto et al. 1984), the ionic density of the shocked ejecta is 33 7 −3 nrev,O ≃1.79×10 cm when the contact discontinuity be- tween the shocks is at r = 1014 cm. The reverse shock is adiabatic, in contrast to supernovae with larger progenitor M/u˙ . Behind the reverse shock, the electron-ion equipar- 32 tition timescale (e.g., Spitzer 1978) is longer (by a factor of ∼ 4) than the time since explosion. To simulate this we put electron temperature, Trev,e, a factor of 2 lower than 31 the equipartition temperature. Assuming the density is the -7 -6.5 -6 -5.5 -5 -4.5 same throughout the shocked ejecta, and neglecting the frequency-dependence of the free-free Gaunt factor gff (we Figure 2. Luminosity of circumstellar Hα emission from a Type set gff =1.2 at all frequencies; Rybicki & Lightman 1979), Ia supernova at 6.5 days after explosion, as a function of mass the outgoing luminosity from the reverse shock is Lrev(ǫ) ∼ loss from the companion. Filled squares show model calculations 31 −1 −1 5.2×10 exp(−ǫ/kTrev,e) erg s eV at t0 ≃ 0.6 days. For where the wind is ionized by the region of circumstellar interac- −0.6 ǫ ≪ kTrev,e, Lrev(ǫ) decreases with time as (t/t0) , and tion. The solid line shows the corresponding luminosities for a scales with wind density as (M/u˙ )2. fully-ionized wind at 2×104 K. The calculations assume a spheri- −1 The ability of the radiation from the reverse shock cally symmetric wind with a velocity of 10 km s , and maximum × 4 −1/5 −1 to ionize the wind depends on the ratio of the ionization SN ejecta velocity vej=2 10 (t/0.6 d) km s . The vertical dotted line shows the region where preionization may be impor- timescale, tion, to the dynamical timescale of the shock, tant. The horizontal dashed lines mark the sensitivity of our SN t = Rs/vs. Rs and vs are the radius and velocity of dyn 1994D observation for a typical supernova in the the forward shock. For our estimated Lrev(ǫ), we find that (“16.2 Mpc”), the Local Group (“3 Mpc”) and the LMC (“50 1.2 4 2.2 tion/tdyn ∼ 0.50(t/t0) and tion ∼ 2.3 × 10 (t/t0) s for kpc”). the circumstellar hydrogen close to the forward shock. For −8 M/u˙ 10=5×10 , the wind does not have enough time to be- < come ionized before being overtaken by the shock. At this day 6.5, the ionized fraction of hydrogen is ∼ 0.01 even −5 wind density, therefore, narrow Hα emission can only be close to the shock, while for M/u˙ 10=1.5×10 hydrogen is caused by preionization. fully ionized as far out as ∼11Rs. The temperature close For higher wind densities, the situation is more promis- to the shock in these two cases differs by a factor of ∼10 −2 4 5 ing. Since tion/tdyn ∝ (M/u˙ ) , the wind will be ionized (∼2.0×10 and ∼2.3×10 K, respectively). The luminosity- long enough to allow substantial Hα emission. To investi- weighted temperature of the Hα-emitting gas varies less dra- gate this numerically, we used a time-dependent version of matically with M/u˙ (∼2.0×104 and ∼3.9×104 K), since the −8 our computer code with M/u˙ 10 ranging from 5×10 to Hα emission is dominated by collisional excitation in regions 1.5 × 10−5. This corresponds to most of the range of plausi- where the temperature is only a few ×104 K. ble values for M/u˙ in symbiotic systems (see section 3.2). As The results from our model calculations are shown before, we assume that the maximum velocity of the ejecta graphically in Figures 2 and 3. The solid line in Figure 2 at ∼0.6 days is 2×104 km s−1 and keep the approximations shows the Hα luminosity at 6.5 days as the result of recom- > −7 4 described above. For M/u˙ 10 ∼ 2 × 10 , Trev,e actually ap- bination in a fully-ionized wind at 2×10 K. This is close to proaches the equipartition temperature, so above this value the temperature which we found for a preionized wind, and we set Trev,e = Trev. The reverse shock only begins to be- should apply when M/u˙ is low and the reverse shock ineffi- −5 come radiative at the highest M/u˙ 10 we consider, 1.5×10 . cient at ionizing the wind. The region where preionization is At higher M/u˙ we would need to include absorption of the important is to the left of the vertical dotted line. The filled ionizing photons by cool shocked ejecta (as in SN 1993J; squares mark the modelled Hα luminosity assuming that the Fransson et al.1996). wind is ionized solely by radiation from the reverse shock. As −7 Both the ionized fraction and temperature of the wind expected, this turns on at around M/u˙ 10 ∼ 2 × 10 . For −7 > −7 increase with M/u˙ . For example, for M/u˙ 10=1.5×10 on M/u˙ 10 ∼ 4 × 10 , the reverse shock excitation produces Circumstellar Hα from Type Ia supernovae 5 more Hα than pure recombination does, despite the wind not being completely ionized. This is due to efficient colli- sional excitation and large optical depths in the Lyman lines. 37 In particular, Ly β is converted into Ly α and Hα. Figure 3 shows how the Hα luminosity drops with time for a range of different M/u˙ . 36

3.2 Limits on M/u˙ for SN 1994D and future nearby supernovae 35 We have used the results of our calculations together with our flux limit on Hα from SN 1994D to put a limit on the progenitor system’s M/u˙ . From Figure 2 (compare the 34 dashed line marked ‘16.2 Mpc’ with our numerical results), the limit on an unresolved emission line at heliocentric ve- −1 < −5 locity 830 km s (section 2) gives M/u˙ 10 ∼ 1.5 × 10 (or −5 −1 −1 1.6×10 if −200 km s < vSN − vH II < 200 km s ). For 33 a supernova in the local group, at 3 Mpc, the same flux < −6 −6 limit corresponds to M/u˙ 10 ∼ 2.5 × 10 (or 3.3×10 ). As the model predicts a low temperature of the CSM < 4 32 ( ∼ 4.0 × 10 K), we expect the Hα line to be unresolved. Our limit on M/u˙ 10 for SN 1994D’s progenitor system only excludes those symbiotic systems with the very highest ˙ mass loss rates. M/u10 has been measured for about a hun- 31 dred red giants in symbiotic systems (e.g., Seaquist & Taylor 2 4 6 8 10 1990; M¨urset et al. 1991; Seaquist, Krogulec & Taylor 1992). The values lie in the range 10−8 to 2×10−5; around two- thirds of the observed systems have M/u˙ 10 within a factor Figure 3. Evolution of Hα luminosity for the model in Figure 2. ˙ × −7 of 4 of 1.0×10−7 (Seaquist et al. 1992). Figures 2 and 3 sug- Light curves are shown for M/u10 between 1.5 10 (bottom) × −5 gest that early Hα observations of closer Type Ia events at and 1.5 10 (top). The vertical dotted line marks 6.5 days after explosion, the epoch of our SN 1994D observation. As in Figure 2, comparable sensitivity can detect, or exclude, a fair fraction the horizontal dashed lines mark the sensitivity of our SN 1994D of symbiotic systems. Most, though, seem likely to remain observation for events at different distances. undetectable. Even a 10 000-s integration at the WHT on a supernova at 3 Mpc caught just 3 days after explosion can > −6 only detect Hα for M/u˙ 10 10 . ∼ (2−3)×10−6 as late as ∼16 days after maximum, assuming These limits depend, of course, on our assumption of a that any X-rays would arise from circumstellar interaction. smooth, spherically symmetric wind, and on our choice of Limits on M/u˙ from Hα emission scale roughly linearly with shock velocity and power-law index for the ejecta density. A distance. Our limit for SN 1994D corresponds therefore to full investigation of the parameters is beyond the scope of < −6 M/u˙ 10 3.3 × 10 at the distance of SN 1986G (cf. Figure this paper; we report some further calculations in Lundqvist ∼ 2). The radio, X-ray and Hα emission all decrease roughly & Cumming (1997). Our limit on M/u˙ is probably accurate as t−1 after the explosion (see Section 3.1.4 and Figure 3). to within a factor of a few, at least for a spherical wind. The This, together with the late epoch of Schlegel & Petre’s limit effects of asymmetry are harder to assess. suggests that it is X-ray observations which are the most In principle, our model calculations can also be used to promising way of detecting the CSM of a Type Ia super- interpret our limits on Hα absorption towards SN 1994D. nova. At all wavebands, observation as soon as possible after The resulting limit on M/u˙ would, however, depend more discovery is crucial. sensitively on the geometry of the CSM than the emission Upper limits on M/u˙ based, like ours, on the standard limit does. Chevalier model (Chevalier 1982) are probably robust to re- laxing the model’s assumption of spherical symmetry. The 3.3 Comparison with radio and X-ray limits presence of its white dwarf companion means that mass loss from the red giant in a symbiotic system is unlikely to be Limits on M/u˙ (sometimes quoted as limits on M˙ ) have spherically symmetric. Observations of those few systems been obtained for a few SN Ia by placing limits on radio which have circumstellar nebulae have generally shown bipo- and X-ray emission. The radio limit reported by Sramek lar structures on both large and small scales (Solf & Ulrich & Weiler (1990) for SN 1981B at ∼18 days after explo- 1985; Corradi & Schwarz 1993; Munari & Patat 1993). As- sion suggested that M/u˙ 10 was probably less than about pherical mass loss necessarily affects the interpretation of 10−6 (Boffi & Branch 1995). Eck et al. (1995) reported limits on radio, X-ray and optical line emission. For exam- that their radio non-detection of SN 1986G (distance ∼4 ple, taking optically thin emission, we estimate that if mass Mpc) at −7 days was in ‘clear conflict’ with the range loss were equatorially concentrated into a solid angle 4πf, −7 < < −6 10 ∼ M/u˙ 10 ∼ 3 × 10 . Observing in X-rays, Schlegel & where f<1, then assuming spherical symmetry in the Cheva- Petre (1993) established a limit for SN 1992A (∼17 Mpc) of lier model will always overestimate the value of M/u˙ by a 6 R. J. Cumming et al. factor of around f −1/2, for at least radio and X-ray obser- 5 ACKNOWLEDGEMENTS vations. The observational data presented here are available as A problem with radio emission in the Chevalier model part of the La Palma archive of reduced supernova data is that the radio flux depends on the efficiency of converting (Martin et al. 1994; Meikle et al. 1995). We thank Eddie kinetic energy of the supernova blast wave to magnetic field Baron, Francesca Boffi, David Branch, Claes Fransson, Pe- energy and energy of relativistic electrons. This conversion is ter Meikle, Ken Nomoto, Miguel P´erez-Torres, Phil Pinto uncertain by a factor of at least 10, corresponding to factor and Toshikazu Shigeyama for helpful discussions during the of > 3 in M/u˙ . If synchrotron self-absorption is important ∼ early stages of this work, and Jim Lewis for help with ac- for radio emission from SN Ia (as it seems to have been for cessing the raw data. PL acknowledges the financial support the early emission from SN 1987A; Chevalier & Dwarkadas of the Swedish Natural Science Research Council. LJS ac- 1995) then the uncertainty may be even greater. knowledges financial support from PPARC. Limits on early X-ray flux seem therefore likely to give The William Herschel Telescope is operated on the is- the most reliable limits on M/u˙ from the system, since the land of La Palma by the Royal Greenwich Observatory in X-rays probe the circumstellar interaction directly. M/u˙ de- the Spanish Observatorio del Roque de los Muchachos of the termined from radio and optical lines will be uncertain to Instituto de Astrof´ısica de Canarias. within a factor of at least a few. However, if detected, optical lines could provide much more powerful constraints on the composition, radiative acceleration, structure and geometry of the wind than either X-ray or radio observations. REFERENCES Boffi F., Branch D. 1995, PASP, 107, 347 Branch D., Lacy C. H., McCall M. L., Sutherland P. G., Uomoto 3.4 Limits on a supersoft X-ray source progenitor A., Wheeler J. C., Wills B. J. 1983, ApJ, 270, 123 system Branch D., Livio M., Yungelson L. R., Boffi F. R., Baron E. 1995, PASP, 107, 1019 Finally, our spectrum can in principle also test for another Chevalier R. A. 1982, ApJ, 259, 302 possible class of progenitor, the supersoft X-ray sources Chevalier R. A. 1984, ApJ, 285, L63 (SSXS; e.g., van den Heuvel et al. 1992). Some SSXS have Chevalier R. A., Dwarkadas V. V. 1995, ApJ, 452, L45 surrounding photoionized nebulae, which may still be de- Chugai N. N. 1986, Sov. Astron., 30, 563 tectable after explosion (Branch et al. 1995), assuming the Corradi R. L. M., Schwarz H. E. 1993, A&A, 268, 714 recombination timescale is long enough. We compared our Cumming R. J., Meikle W. P. S., Geballe T. R. 1994, IAU Circ., limits at Hα and [N ii] λ6583 with predictions from Rap- 5951 (correction in Cumming R. J., Meikle W. P. S., Geballe T. R. 1994, IAU Circ., 5953) paport et al. (1994) and found that the maximum expected Della Valle M., Benetti S., Panagia N. 1996, ApJ, 459, L23 line flux in λ6583, the stronger of the two lines, is a factor of Eck C., Cowan J., Branch D., Boffi F. 1995, ApJ, 451, L53 about three below our detection limit. The Hα detected from Fransson C. 1982, A&A, 111, 140 the SSXS CAL 83 in the LMC (Remillard et al. 1995) would Fransson C., Lundqvist P., Chevalier R. A. 1996, ApJ, 461, 993 have been a factor of 30 below our limit. Future searches for Graham J. R. et al. 1983, Nat, 304, 709 such emission would be better carried out around [O iii] Graham J. R., Meikle W. P. S. 1986, MNRAS, 221, 789 λ5007, much the strongest optical line expected from SSXS Hamuy M., et al. 1995, AJ, 109, 1669 nebulae. Ho L. C., Filippenko A. V. 1995, ApJ, 444, 165 H¨oflich P. A. 1995, ApJ, 443, 89 King D. L., Vladilo G., Lipman K., de Boer K. S., Centuri´on M., Moritz P., Walton N. A. 1995, A&A, 300, 881 4 SUMMARY Leibundgut B. et al. 1993, AJ, 105, 301 Livne E., Tuchman Y., Wheeler J. C. 1992, ApJ, 399, 665 Very early high-resolution optical spectroscopy of SN 1994D Lundqvist P., Fransson C. 1988, A&A, 192, 221 shows no sign of circumstellar Hα. We have used a numeri- Lundqvist P., Fransson C. 1996, ApJ, 464, 924 cal photoionization model to investigate the Hα emission as- Lundqvist P., Cumming R. J. 1997, in Rood R. T., ed, Stellar Ecology, in press suming that radiation from circumstellar interaction ionizes Martin R., Lewis J. R., Cumming R. J., Meikle W. P. S. 1994, the stellar wind from a binary companion to the progenitor Spectrum, 2, 8 white dwarf. Comparing our modelling with the observations Meikle W. P. S. et al. 1996, MNRAS, 281, 263 we have derived a limit to the progenitor mass loss which Meikle W. P. S., Catchpole R. M., Lewis J. R., Martin R., Cum- excludes a symbiotic system with a very high mass loss rate. ming R. J. 1995, Spectrum, 4, 7 Future observations of closer Type Ia supernovae can at best Munari U., Patat F. 1993, A&A, 277, 195 −6 −1 detect systems with mass loss rates around 10 M⊙ yr . Munari U., Renzini A. 1992, ApJ, 397, L87 Radio, X-ray and optical searches for circumstellar matter M¨urset U., Nussbaumer H., Schmid H. M., Vogel M. 1991, A&A, round SN Ia can complement each other: while X-ray obser- 248, 458 vations can probably set the most stringent limits on pro- Nomoto K., Thielemann F.-K., Yokoi K. 1984, ApJ, 286, 644 genitor mass loss, high-resolution optical spectroscopy offers Nussbaumer H., Vogel M. 1987, A&A, 182, 51 Patat F., Benetti S., Cappellaro E., Danziger I. J., Della Valle the exciting possibility of unambiguously detecting circum- M., Mazzali P. A., Turatto M. 1996, MNRAS, 278, 111 stellar hydrogen, if it is present. Perlmutter S. et al. 1995, ApJ, 440, L41 Polcaro V. F., Viotti R. 1991, A&A, 242, L9 Rappaport S., Chiang E., Kallman T., Malina R. 1994, ApJ, 431, 237 Circumstellar Hα from Type Ia supernovae 7

Remillard R. A., Rappaport S., Macri L. M. 1995, ApJ, 439, 646 Richmond M. W. et al. 1995, AJ, 109, 2121 Ruiz-Lapuente P. et al. 1993, Nat, 365, 728 Rybicki G. B., Lightman A. P. 1979, Radiative Processes in As- trophysics, Wiley, New York Sandage A., Saha A., Tammann G. A., Labhardt L., Panagia N., Macchetto F. D. 1996, ApJ, 460, L15 Schlegel E. M. 1995, Rep. Prog. Phys., 58, 1375 Schlegel E. M., Petre R. 1993, ApJ, 418, L53 Seaquist E. R., Taylor A. R. 1990, ApJ, 349, 313 Seaquist E. R., Krogulec M., Taylor A. R. 1993, ApJ, 410, 260 Solf J., Ulrich H. 1985, A&A, 148, 274 Spitzer L., 1978, Physical Processes in the Interstellar Medium, Wiley, New York Sramek R. A., Weiler K. W. 1990, in Petschek A. G., ed, Super- novae, Springer, Berlin Stone R. P. S. 1977, ApJ, 218, 767 Turatto M., Benetti S., Cappellaro E., Danziger I. J., Della Valle M., Mazzali P. A., Patat F. 1996, MNRAS, in press van den Heuvel E. P. J., Bhattacharya D., Nomoto K., Rappaport S. A. 1992, A&A 262, 97 Wang L., Wheeler J. C., Li Z., Clocchiatti A. 1996, ApJ (in press) Wheeler J. C. 1992a, in Kondo Y. et al., eds, Proc. IAU Symp. 151, Evolutionary Processes in Interacting Binary Stars, Rei- del, Dordrecht, p. 225 Wheeler J. C. 1992b, in Lambert D. L., ed, Frontiers of Stellar Evolution, ASP, Provo, p. 483 Yungelson L., Livio M., Tutukov A. V., Kenyon S. J. 1995, ApJ, 447, 656