1: %\documentclass{aastex}

2: 

3: %% preprint produces a one-column, single-spaced document:

4: 

5: \documentclass[preprint]{aastex}

6: 

7: %% preprint2 produces a double-column, single-spaced document:

8: 

9: % \documentclass[preprint2]{aastex}

10: 

11: %% You can insert a short comment on the title page using the command below.

12: 

13: \slugcomment{}

14: 

15: %% If you wish, you may supply running head information, although

16: %% this information may be modified by the editorial offices.

17: %% The left head contains a list of authors,

18: %% usually a maximum of three (otherwise use et al.).  The right

19: %% head is a modified title of up to roughly 44 characters.  Running heads

20: %% will not print in the manuscript style.

21: 

22: \shortauthors{Matheson et al.}

23: \shorttitle{Supernova 1993J}

24: 

25: 

26: %% This is the end of the preamble.  Indicate the beginning of the

27: %% paper itself with \begin{document}.

28: 

29: \begin{document}

30: 

31: %% LaTeX will automatically break titles if they run longer than

32: %% one line. However, you may use \\ to force a line break if

33: %% you desire.

34: 

35: \title{Detailed Analysis of Early to Late-Time Spectra of

36: Supernova 1993J}

37: 

38: 

39: 

40: 

41: \author{Thomas Matheson\altaffilmark{1}, Alexei

42: V. Filippenko\altaffilmark{1}, Luis C. Ho\altaffilmark{2}, Aaron

43: J. Barth\altaffilmark{3}, and Douglas C. Leonard\altaffilmark{1}}

44: %\affil{Department of Astronomy, University of California, Berkeley, CA

45: %94720-3411}

46: \altaffiltext{1}{Department of Astronomy, University of California,

47:     Berkeley, CA 94720-3411; matheson, alex, leonard@astron.berkeley.edu}

48: \altaffiltext{2}{The Observatories of the Carnegie Institution of

49:     Washington, 813 Santa Barbara Street, Pasadena, CA 91101-1292;

50:     lho@ociw.edu}

51: \altaffiltext{3}{Harvard-Smithsonian Center for Astrophysics,

52:     60 Garden Street, Cambridge, MA 02138; abarth@cfa.harvard.edu}

53: 

54: 

55: %\email{matheson, alex, leonard@astron.berkeley.edu,

56: %barth@cfa.harvard.edu, lho@ociw.edu}

57: 

58: 

59: %% Mark off your abstract in the ``abstract'' environment. In the manuscript

60: %% style, abstract will output a Received/Accepted line after the

61: %% title and affiliation information. No date will appear since the author

62: %% does not have this information. The dates will be filled in by the

63: %% editorial office after submission.

64: 

65: \begin{abstract}

66: 

67: We present a detailed study of line structure in early to late-time

68: spectra of Supernova (SN) 1993J.  Spectra during the nebular phase,

69: but within the first two years after explosion, exhibit small-scale

70: structure in the emission lines of some species, notably oxygen and

71: magnesium, showing that the ejecta of SN 1993J are clumpy.  On the

72: other hand, a lack of structure in emission lines of calcium implies

73: that the source of calcium emission is uniformly distributed

74: throughout the ejecta.  These results are interpreted as evidence that

75: oxygen emission originates in clumpy, newly synthesized material,

76: while calcium emission arises from material pre-existing in the

77: atmosphere of the progenitor.  Spectra spanning the range $433-2454$

78: days after the explosion show box-like profiles for the emission

79: lines, clearly indicating circumstellar interaction in a roughly

80: spherical shell.  This is interpreted within the Chevalier \& Fransson

81: (1994) model for SNe interacting with mass lost during prior stellar

82: winds.  At very late times, the emission lines have a two-horned

83: profile, implying the formation of a somewhat flattened or disk-like

84: structure that is a significant source of emission.  The very high

85: signal-to-noise ratio spectra are used to demonstrate the potential

86: significance of misinterpretation of telluric absorption lines in the

87: spectra of bright SNe.

88: \end{abstract}

89: 

90: %% Keywords should appear after the \end{abstract} command. The uncommented

91: %% example has been keyed in ApJ style. See the instructions to authors

92: %% for the journal to which you are submitting your paper to determine

93: %% what keyword punctuation is appropriate.

94: 

95: \keywords{circumstellar matter---stars: mass-loss---supernovae:

96: general---supernovae: individual (SN 1993J)---techniques:

97: spectroscopic}

98: 

99: 

100: \section{Introduction}

101: 

102: Supernova 1993J presented a great opportunity in the study of

103: supernovae (SNe).  It occurred in the nearby galaxy M81 (NGC 3031; $d$

104: = 3.6 Mpc; Freedman et al. 1994) and reached a maximum brightness of

105: $m_V = 10.8$ mag (e.g., Richmond et al. 1996).  This allowed extremely

106: detailed observations to be taken over a long time interval.  The fact

107: that SN 1993J underwent a transformation from a Type II to a Type IIb,

108: strengthening the link between SNe II and SNe Ib as core-collapse

109: events, only enhanced the importance of this object (see, e.g.,

110: Wheeler \& Filippenko 1996; Matheson et al. 2000, hereinafter Paper I,

111: for detailed discussions of the nature of SN 1993J).

112: 

113: We have previously presented analyses of the spectra of SN 1993J

114: obtained during its first year (Filippenko, Matheson, \& Ho 1993,

115: hereinafter FMH93; Filippenko, Matheson, \& Barth 1994, hereinafter

116: FMB94), while Paper I shows our complete low-resolution spectroscopic

117: data set.  The early spectra and the transition from SN II to SN IIb

118: are covered by FMH93, while the onset of the nebular phase and the

119: beginning of signs of circumstellar interaction are discussed by

120: FMB94.  In this paper, we take a closer look at the details of some of

121: the earlier spectra and we present the analysis of late-time spectra

122: of SN 1993J.  The clumpy nature of SN 1993J, as revealed in the

123: details of spectra from early to nebular times, is presented in \S 2.

124: The late-time spectra and their evidence of circumstellar interaction

125: are discussed in \S 3; we summarize our conclusions in \S 4.  In the

126: Appendix, we discuss the potential impact of telluric absorption on

127: the interpretation of the spectra of bright SNe and the implications

128: this has for future SN studies.

129: 

130: 

131: All of the low-dispersion spectra discussed herein were obtained at

132: the Lick and Keck Observatories.  In addition, a single

133: high-dispersion spectrum was obtained using the Keck~I 10-m telescope.

134: The details of the observations, including instrumental configuration,

135: dates of observation, and observational parameters, are presented in

136: Paper I.  We follow Lewis et al. (1994) in adopting 1993 March 27.5 UT

137: (JD 2,449,074) as the date of explosion.

138: 

139: 

140: \section{Clumps}

141: 

142: 

143: \subsection{Clumps in Other Supernovae}

144: 

145: The mottled appearance of supernova remnants (SNRs) has often been

146: interpreted as evidence for clumps in the ejecta of SNe.  For example,

147: fast-moving features have been identified in Cas A for quite some time

148: (e.g., Baade \& Minkowski 1954), and modern studies continue to

149: interpret the structure of Cas A as the result of clumpy SN ejecta

150: (e.g., Anderson et al. 1994).  The discovery of metal-enriched

151: fragments outside the boundary of the Vela SNR (Aschenbach et

152: al. 1995), a high-velocity ``bullet'' in Vela (Strom et al. 1995), and

153: fast-moving oxygen filaments in Puppis A (and other remnants; Winkler

154: \& Kirshner 1985, and references therein) all support the contention

155: that SN ejecta are clumped.  There is also a theoretical underpinning

156: for the expectation of clumps in the ejecta of core-collapse SNe.

157: This includes the necessity of mixing in SNe Ib/c (e.g., Wheeler et

158: al. 1987; Shigeyama et al. 1990; Hachisu et al. 1991) and the presence

159: of instabilities, such as convective instabilities in neutrino-driven

160: explosion mechanisms and Rayleigh-Taylor instabilities during the

161: expansion of the ejecta (e.g., Kifonidis et al. 2000, and references

162: therein).  In fact, with a small core mass ($3-4 M_{\sun}$), SN 1993J

163: should have relatively large Rayleigh-Taylor instabilities that mix

164: the core (Iwamoto et al. 1997).

165: 

166: The observation of small-scale structure in SN line emission requires

167: a spectrum with a high signal-to-noise (S/N) ratio and the spectral

168: resolution to distinguish narrow features.  This in turn necessitates

169: a bright source and an efficient detector with at least moderate

170: dispersion ($\sim 1-10$ \AA/pixel).  These two requirements for SNe

171: were first fulfilled with SN 1985F.  Analysis of spectra of the Type

172: Ib SN 1985F revealed several distinct clumps in the [\ion{O}{1}]

173: $\lambda\lambda$6300, 6364 lines as well (Filippenko \& Sargent 1989).

174: These clumps had widths $\sim 100-250$ km~s$^{-1}$ with amplitudes of

175: $2-10$\%.  The small-scale features changed slightly over time,

176: becoming narrower and relatively stronger.  The \ion{Mg}{1}]

177: $\lambda$4571 also had features, but they did not match those of the

178: oxygen lines.

179: 

180: The bright, nearby SN 1987A provided another opportunity to study SN

181: spectra in detail.  While examining the [\ion{O}{1}]

182: $\lambda\lambda$6300, 6364 doublet of SN 1987A, Stathakis et

183: al. (1991) discovered what appeared to be a sawtooth profile on the

184: tops of the lines: there were $\sim$ 50 small-scale features with a

185: typical size of 80 km~s$^{-1}$ (full width at half maximum [FWHM]) and

186: amplitudes of $3-10$\% of the global line profile.  The structures

187: persisted over time and were consistent between the two doublet lines.

188: Clumps were also observed in the H$\alpha$ line (Hanuschik et

189: al. 1993).  These had widths of $400-500$ km~s$^{-1}$ with amplitudes

190: of $7-10$\% of the overall line strength.  The H$\alpha$ clumps were

191: present over several hundred days, but individual clumps were not

192: persistent.  Hanuschik et al. (1993) interpreted this as changes in

193: the level of dust extinction along the lines of sight to the

194: individual clumps.  Spyromilio \& Pinto (1991) found that the oxygen

195: density derived from the line intensity ratios of [\ion{O}{1}]

196: $\lambda$6300 and [\ion{O}{1}] $\lambda$6364 gave too large an oxygen

197: mass if the oxygen filled the entire volume suggested by the expansion

198: velocity; a low filling factor for oxygen was consistent with the

199: calculated density, volume at that epoch, and a reasonable oxygen

200: mass.  Such a low filling factor can be interpreted as clumpiness.

201: Further evidence for clumps in the ejecta of SN 1987A is presented by

202: Spyromilio, Stathakis, \& Meurer (1993), who found a common feature in

203: emission lines of H$\alpha$, [\ion{Fe}{2}] $\lambda$7155, and

204: [\ion{Ca}{2}] $\lambda\lambda$7291, 7324.  Clumps were also seen at

205: later times in infrared (IR) emission lines of iron in SN 1987A

206: (Spyromilio, Meikle, \& Allen 1990; Haas et al. 1990).  The pattern

207: seen in the oxygen lines was not replicated in these other species.

208: 

209: The Type II SN 1988A may also have

210: had clumps.  They were not apparent in the spectra, but following the

211: analysis of the [\ion{O}{1}] $\lambda\lambda$6300, 6364 doublet in SN

212: 1987A of Spyromilio \& Pinto (1991), Spyromilio (1991) derived a low

213: filling factor for oxygen in SN 1988A of 0.05, and thus deduced a

214: clumpy nature for the source of the oxygen emission.  In addition,

215: Fassia et al. (1998) invoked clumps of helium to explain the strength

216: of \ion{He}{1} $\lambda$10830 line in the Type II SN 1995V given their

217: model of $^{56}$Ni mixing, although the clumps were not evident from

218: the spectra.

219: 

220: SN 1993J, as noted before, was extremely bright, so any clumps in its

221: spectrum should be readily detectable.  Li et al. (1994) reported

222: clumps in their spectra of SN 1993J.  Wang \& Hu (1994) noted the

223: presence of narrow components in the forbidden lines of oxygen, and

224: they used the clumpy nature of the SN to explain apparent blueshifts

225: in the emission lines of [\ion{O}{1}] $\lambda$5577 and [\ion{O}{1}]

226: $\lambda$6300.  Spyromilio (1994, hereinafter S94) presented a much

227: more detailed analysis of the clumps visible in the spectra of SN

228: 1993J and also concluded that there is large-scale anisotropy in the

229: SN based on the blueshifts of the lines (oxygen and magnesium) as well

230: as the differences in the distribution of clumps both between and

231: within the observed atomic species.  This interpretation of the

232: blueshifts was questioned by Houck \& Fransson (1995), however.  They

233: concluded that the blueshifts are only apparent, and did not represent

234: physical asymmetries in the lines.  The [\ion{O}{1}]

235: $\lambda\lambda$6300, 6364 doublet can be dramatically affected by

236: scattering of H$\alpha$ if oxygen and hydrogen are distributed in

237: shells, and this would mimic the observed asymmetry.  Blending affects

238: the other lines, with [\ion{O}{1}] $\lambda$5577 contaminated by

239: [\ion{Fe}{2}] $\lambda$5536 and [\ion{Co}{2}] $\lambda$5526.  The

240: \ion{Mg}{1}] $\lambda$4571 line is affected by \ion{Fe}{2} multiplets

241: near 4600 \AA.

242: 

243: \subsection{Analysis of the Clumps}

244: 

245: As the brightest supernova in the northern hemisphere since SN 1954A,

246: SN 1993J provided an object for which high S/N ratio spectra could be

247: obtained over an extended period of time.  Any intrinsic small-scale

248: structure in the emission lines of SNe is normally lost in the

249: noise of the spectra.  For SN 1993J, the noise over most of our range

250: ($3500-9000$ \AA) does not obscure the small-scale features in the

251: spectra until several hundred days after the explosion.  This allows

252: us to make an extensive study of line substructure as the supernova

253: evolves.  Telluric absorption could introduce spurious small-scale

254: structure, but we attempt to remove its effects (see Appendix).

255: 

256: Careful inspection of the spectra of SN 1993J reveals small-scale

257: structure in many emission lines, especially [\ion{O}{1}]

258: $\lambda\lambda$6300, 6364, [\ion{O}{1}] $\lambda$5577, \ion{O}{1}

259: $\lambda$7774, \ion{Mg}{1}] $\lambda$4571, and [\ion{Ca}{2}]

260: $\lambda\lambda$7291, 7324.  The [\ion{O}{1}] $\lambda\lambda$6300,

261: 6364 lines in particular show a distinctive sawtooth profile similar

262: to those described above for SN 1987A and SN 1985F (Figure 1).  Note

263: that all spectra have had the systemic velocity of SN 1993J removed

264: ($-140$ km~s$^{-1}$, determined from narrow lines in early-time

265: spectra; see, e.g., FMH93).  To isolate these fluctuations, we removed

266: the global shape of the underlying line by smoothing the profile, and

267: then subtracting the smoothed version.  The spectra were smoothed with

268: a running boxcar with a width of 60 \AA---wide enough to remove the

269: small-scale fluctuations from the line profile, but narrower than the

270: line itself, so that the subtraction would leave only the small

271: features.  This is the same method employed by Filippenko \& Sargent

272: (1989) to isolate similar substructure in the [\ion{O}{1}]

273: $\lambda\lambda$6300, 6364 lines of SN 1985F, although they used a

274: smaller boxcar width for smoothing that was better matched to the

275: velocity width of the fluctuations in SN 1985F.

276: 

277: Such a technique is similar to the process of unsharp masking in Fourier

278: space.  We experimented with this and other methods in the Fourier

279: domain to isolate high-frequency components of spectra.  The results

280: were not superior to the smoothing technique.  Moreover, the Fourier

281: techniques became less successful as the relative noise level increased

282: in later spectra.  The subtraction of a smoothed line from itself

283: does, however, introduce an artifact in the resulting residuals due to

284: poor removal of the edges of the global line profile.  These artifacts

285: have a different character than the legitimate fluctuations; they are

286: broader and generally show a gradual decrease from the subtracted

287: continuum.  Once an actual feature is reached, it clearly stands out

288: from the edge effect.  This is illustrated in Figure 2 with an

289: artificial profile from which the smoothing technique isolates the six

290: introduced components.  The edge effects are prominent, but distinctly

291: different in velocity width and character from the actual fluctuations.  To

292: eliminate the edge effects would require a model of the underlying

293: line profile.  This model would vary from line to line, and would also

294: change over time as different amounts of nearby lines contaminate

295: the feature under consideration.  For example, as illustrated in

296: Figure 4 of FMB94, the relative levels of [\ion{O}{1}]

297: $\lambda\lambda$6300, 6364 and H$\alpha$ $\lambda$6563 can change

298: dramatically over a time scale of months.  Subtracting a different

299: line profile for different spectra at different times could introduce

300: spurious components or temporal differences.  In the interest of

301: consistency, we chose to use the smoothing technique to identify

302: substructure in the lines.  In no case did it fail to find components

303: visible in the original spectra, and the only new features introduced are

304: the edge effects that are easily distinguishable from genuine components.

305: 

306: Figure 3 shows a typical example of the clumps in the [\ion{O}{1}]

307: $\lambda\lambda$6300, 6364 lines on day 209.  There are five

308: distinctive maxima marked in the figure.  S94 found six clumps, but

309: his clump \emph{f} (at $\sim$ 2900 km~s$^{-1}$) is difficult to

310: discern in this doublet; his clumps \emph{a, b, c, d,} and \emph{e}

311: correspond to our clumps 5, 4, 3, 2, and 1.  For the [\ion{O}{1}]

312: $\lambda$6300 line, these clumps are at velocities of (starting from

313: the blueshifted edge) $-3220$, $-2340$, $-1510$, $-310$, and 410

314: km~s$^{-1}$ (where 6300 \AA\ defines zero velocity).  The marked

315: minima (identified as 6, 7, and 8) have velocities of $-2750$,

316: $-1930$, and $-740$ km~s$^{-1}$, respectively.  The uncertainty in the

317: values for the velocities of clumps is $\sim$ 50 km~s$^{-1}$.  The

318: corresponding locations of these components for the [\ion{O}{1}]

319: $\lambda$6364 line are also indicated, but the features themselves are

320: difficult to distinguish.  There is also an emission feature at $\sim$

321: 2000 km~s$^{-1}$, but it is unclear if this is from [\ion{O}{1}]

322: $\lambda$6300, [\ion{O}{1}] $\lambda$6364, both, or H$\alpha$.  In

323: fact, the large velocity width of H$\alpha$ at these times (FWHM

324: $\approx$ 17,000 km~s$^{-1}$, full width at zero intensity [FWZI]

325: $\approx$ 23,000 km~s$^{-1}$, FMB94; see also below) significantly

326: contaminates the spectrum to a blue wavelength of at least 6375 \AA,

327: possibly as far as 6310 \AA.  The presence of H$\alpha$ throughout the

328: region of the spectrum where one would find [\ion{O}{1}] $\lambda$6364

329: makes the identification of the [\ion{O}{1}] $\lambda$6364 clumps

330: problematic at best.

331: 

332: The [\ion{O}{1}] $\lambda\lambda$6300, 6364 clumps described above

333: first become obvious in our day 93 spectrum.  They are not apparent on

334: day 45.  As can be seen in Figure 4, once the features appeared, they

335: remain consistent until day 433.  Beyond that day, the clumps may

336: still be present, but the supernova had faded to the point that it was

337: difficult to obtain spectra with a S/N ratio high enough to

338: distinguish the clumps.  The clumps at $-2340$ and 410 km~s$^{-1}$

339: are, in fact, not obvious even in the day 433 spectrum.  There are

340: some slight changes in the individual clumps.  The clump at $-1510$ km

341: s$^{-1}$ appears to broaden slightly, from 370 km~s$^{-1}$ on day 93

342: (all widths are FWHM) to 470 km~s$^{-1}$ on day 167, while increasing

343: in amplitude in comparison with the other clumps.  The amplitude

344: increase continues past day 167 to the time when it became difficult

345: to distinguish the clumps.  The clump at $-310$ km~s$^{-1}$ grows

346: considerably in relative strength, by almost a factor of two from day

347: 93 to day 433.  The clump at 410 km~s$^{-1}$ narrows from 540 km

348: s$^{-1}$ on day 93 to 390 km~s$^{-1}$ on day 167, after which it

349: remains constant.

350: 

351: The features seen in the [\ion{O}{1}] $\lambda\lambda$6300, 6364 lines

352: are also found in other oxygen emission lines.  Figure 5 compares

353: [\ion{O}{1}] $\lambda\lambda$6300, 6364 with [\ion{O}{1}]

354: $\lambda$5577 and \ion{O}{1} $\lambda$7774.  The clumps line up quite

355: well between [\ion{O}{1}] $\lambda$6300 and [\ion{O}{1}]

356: $\lambda$5577. Li et al. (1994) found a similar correspondence for a

357: single epoch.  The only significant differences are slight changes in

358: relative intensity of some of the clumps (compare the clumps at $-310$

359: and 410 km~s$^{-1}$ in [\ion{O}{1}] $\lambda$6300 and [\ion{O}{1}]

360: $\lambda$5577).  These may suggest intrinsic differences in physical

361: conditions in the two clumps, but are more likely the result of

362: contamination of the lines by other species (see above and Houck \&

363: Fransson [1996] for a discussion of potential line contaminants in SN

364: 1993J).  S94's clump \emph{f} at $\sim$ 2900 km~s$^{-1}$ is apparent

365: in [\ion{O}{1}] $\lambda$5577 and \ion{O}{1} $\lambda$7774.  The

366: \ion{O}{1} $\lambda$7774 line is missing the $-2340$ km~s$^{-1}$

367: component completely, and the $-3220$ km~s$^{-1}$ clump is actually at

368: $-3110$ km~s$^{-1}$, slightly redshifted in comparison with the other

369: oxygen lines.  The shift of a clump is most likely the result of

370: contamination, but the absence of one clump from \ion{O}{1}

371: $\lambda$7774 that is clearly evident in [\ion{O}{1}] $\lambda$6300

372: and [\ion{O}{1}] $\lambda$5577 is puzzling.  The \ion{O}{1}

373: $\lambda$7774 line is contaminated by telluric absorption (see

374: Appendix), but this is not likely to result in the apparent loss of a

375: single component, especially not consistently over many epochs.  The

376: difficulty in determining actual flux values for the clumps

377: effectively precludes any evaluation of the differing physical

378: parameters for the various clumps, but the changing physical

379: conditions from clump to clump are most likely the cause of the slight

380: differences seen when comparing the various oxygen lines.  In

381: addition, the contamination of all the lines makes even global

382: evaluations highly uncertain.

383: 

384: There is evidence for clumps in lines of other species.  Figure 6

385: shows the small-scale structure of \ion{Mg}{1}] $\lambda$4571 from day

386: 139 to day 298, while Figure 7 shows them for [\ion{Ca}{2}]

387: $\lambda\lambda$7291, 7324 from day 123 to day 298 (7291 \AA\ defines

388: the zero velocity for calcium).  (The fact that

389: small-scale fluctuations in magnesium and calcium lines are visible

390: for only a subset of the number of days that they are present in

391: oxygen is more indicative of the vagaries of observation than a

392: difference in the physical conditions producing the lines; earlier

393: spectra are contaminated by other lines, while later spectra are

394: noisier.)  Again, the features are fairly consistent over time, but,

395: as Figure 8 shows in a detailed comparison of [\ion{O}{1}]

396: $\lambda\lambda$6300, 6364, \ion{Mg}{1}] $\lambda$4571, [\ion{Ca}{2}]

397: $\lambda\lambda$7291, 7324, and H$\alpha$ on day 209, the pattern in

398: the clumps is very different between the species, although Li et

399: al. (1994) found that some features did correspond between H$\alpha$

400: and [\ion{O}{1}] $\lambda$6300.  Figure 9 shows [\ion{O}{1}]

401: $\lambda\lambda$6300, 6364, H$\alpha$, and \ion{Mg}{1}] $\lambda$4571

402: from day 433, and the differences between the lines are even more

403: apparent.  The H$\alpha$ line will be discussed in more detail in \S

404: 3.  The \ion{Mg}{1}] $\lambda$4571 line does have one clump (410

405: km~s$^{-1}$) that lines up with oxygen, but the rest of the structure

406: is very different, with relatively fewer fluctuations, as in SN 1985F

407: (Filippenko \& Sargent 1989).  (The minimum at $-2740$ km s$^{-1}$ in

408: [\ion{O}{1}] $\lambda\lambda$6300, 6364 marginally lines up with a

409: minimum in \ion{Mg}{1}] $\lambda$4571, but it is probably

410: coincidental.)  This may indicate that the magnesium and oxygen

411: emission arise from substantially different clumps.

412: 

413: The [\ion{Ca}{2}] $\lambda\lambda$7291, 7324 lines have some

414: substructure, but they are relatively smooth.  Aside from the peaks

415: of the two lines and the edge effects described above, there is very

416: little small-scale structure that is consistent.  There is a small

417: peak at $\sim$ $-1700$ km~s$^{-1}$, but no other significant evidence

418: for persistent clumps.  In addition, these lines coincide with a

419: telluric water absorption feature (cf. Figure 16), so incomplete

420: removal of the water band profile may contaminate this region.

421: 

422: The scale of the clumps is not hidden by the resolution of the

423: observations.  The low-dispersion spectra used here have typical

424: resolutions of $6-7$ \AA\ (FWHM).  At 6300 \AA, that corresponds to

425: $\sim$ 300 km~s$^{-1}$.  Comparison with the high-resolution HIRES

426: spectrum illustrates that there is little, if any, substructure with

427: scales smaller than this (see Figure 9 of Paper I).  Virtually all the

428: clumps that are apparent in the HIRES spectrum appear in the

429: low-dispersion spectrum taken four days later.  The clumps that appear

430: with lower contrast in the low-dispersion spectrum, but are easily

431: found using our smoothing technique, do stand out more clearly in the

432: HIRES spectrum.  The HIRES spectrum shows that we are not missing any

433: of the significant details of the clumps in the low-resolution

434: spectra.

435: 

436: \subsection{Discussion of the Clumps}

437: 

438: The contrast between the presence of clumps in the emission lines of

439: oxygen (and other species) and the relative smoothness of the calcium

440: emission lines implies that the sources of emission for the various

441: species are distributed differently in the supernova ejecta.  One

442: possibility is that the oxygen emission arises in distinct clumps of

443: material while the source of calcium emission is distributed uniformly

444: throughout the ejecta.  While attempting to model the [\ion{O}{1}]

445: $\lambda\lambda$6300, 6364 lines and the \ion{Ca}{2} lines

446: ([\ion{Ca}{2}] $\lambda\lambda$7291, 7324 and the near-infrared

447: [near-IR] triplet), Li \& McCray (1992, 1993) developed a similar

448: concept to explain the structure seen in emission lines of SN 1987A

449: that parallel those described above for SN 1993J.

450: 

451: Li \& McCray (1992) initially focused on the forbidden oxygen lines.

452: They developed a model based upon the relative line strengths at a

453: given epoch for the [\ion{O}{1}] $\lambda\lambda$6300, 6364 doublet

454: that predicted a value for the total oxygen mass to filling-factor

455: ratio.  By choosing a reasonable value for the total oxygen mass (from

456: theoretical models), they concluded that the oxygen emission must come

457: from clumps distributed uniformly throughout the ejecta with a small

458: filling factor.  The clumps would develop from material synthesized in

459: earlier evolution or during the explosion itself.  As already

460: mentioned, fast-moving oxygen components are seen in SNRs (Winkler \&

461: Kirshner 1985).  When Li \& McCray (1993) subsequently analyzed the

462: calcium lines, they found that the observed lines were too weak given

463: the probable amount of calcium produced during prior evolution and

464: from explosive nucleosynthesis.  Li \& McCray hypothesized that clumps

465: of calcium do form but do not contribute to the emission.  These

466: clumps of calcium intercept $\gamma$-ray radiation in proportion to

467: their mass fraction.  As this fraction is small, they receive only a

468: small amount of radioactive luminosity and thus they do not achieve

469: temperatures necessary for emission, although oxygen lines can be

470: excited at these temperatures.  The pre-existing envelope of the star

471: is heated generally as hydrogen and helium do not radiate efficiently,

472: allowing calcium that was thoroughly mixed into the atmosphere, most

473: likely present from the time of the progenitor's formation, to reach

474: the temperatures required to produce \ion{Ca}{2} emission.

475: 

476: This scenario of emission from smoothly distributed, pre-existing

477: calcium while oxygen is clumped with a small filling factor explains

478: the detailed line structure seen in SN 1993J.  Other attempts to model

479: the spectra of SN 1987A, however, concluded that the source of the

480: oxygen emission consisted chiefly of the envelope, and so was also the

481: result of pre-existing constituents (Swartz, Harkness, \& Wheeler

482: 1989; Swartz 1991), although later analysis by the same authors

483: acknowledged that some oxygen from the core may contribute to the

484: emission and that it might be clumped (Wheeler, Swartz, \& Harkness

485: 1993) .  Smoothly distributed oxygen is difficult to reconcile with

486: the clumps exhibited by the oxygen-line profiles.  The underlying

487: smooth line may be the result of the pre-existing oxygen, but clumps

488: are clearly present in SN 1993J and SN 1987A, implying that there is a

489: distinct difference in the site of oxygen and calcium emission.

490: 

491: The sawtooth appearance of the [\ion{O}{1}] $\lambda\lambda$6300, 6364

492: lines indicates a rough scale for the size of the clumps in the

493: ejecta.  Chugai (1994) used similar data for SN 1987A to develop a

494: technique for determining the oxygen mass.  Given the minimum cloud

495: size (on a velocity scale), the total velocity width of the lines, and

496: the relative level of fluctuation in the clumps (the scale of the

497: clumps compared to the overall line strength), one can then estimate

498: the filling factor of the clumps ($f \approx 0.1$ in the case of SN

499: 1987A).  Several groups used the relative ratio of the [\ion{O}{1}]

500: $\lambda\lambda$6300, 6364 lines in SN 1987A to derive an oxygen mass

501: to filling-factor ratio ($M_O / f$) of 11 $M_{\sun}$ (Li \& McCray

502: 1992) to 15 $M_{\sun}$ (Spyromilio \& Pinto 1991; Chugai 1992;

503: Andronova 1992).  This implies a total oxygen mass of $1.1-1.5$

504: $M_{\sun}$ (Chugai 1994).  As the line widths for SN 1993J are much

505: larger (FWHM $\approx$ 5400 km~s$^{-1}$ for SN 1993J on day 167

506: vs. FWHM $\approx$ 2700 km~s$^{-1}$ for SN 1987A on day 173 [Li \&

507: McCray 1992]), the two doublet lines are blended, as well as severely

508: affected by the broad H$\alpha$ line, and this effectively precludes a

509: new evaluation for $M_O / f$.  If one makes the admittedly dangerous

510: assumption that this ratio is relatively constant among core-collapse

511: SNe, then Chugai's method can be applied.  For SN 1993J, the sizes of

512: the clumps imply a typical cloud radius (in velocity space) of 150

513: km~s$^{-1}$ with a relative fluctuation compared to the total line

514: flux of $\delta\rm{F}/\rm{F} \approx 0.1$.  To estimate the effective

515: radius (in velocity space) for Chugai's calculation, we use Li \&

516: McCray's (1992) definition that emission at $|v| \lesssim v_{eff}$

517: constitutes 90\% of the flux in the line (Li \& McCray refer to

518: $v_{eff}$ as the ``expansion velocity'').  For SN 1993J, the FWHM is

519: $\sim$ 5400 km~s$^{-1}$, and, assuming a Gaussian profile for the

520: line, this implies that 90\% of the emission of the line comes from

521: within $v \approx \pm 3800$ km s$^{-1}$.  With $v_{eff} \approx 3800$

522: km~s$^{-1}$, we find a filling factor of 0.06 by Chugai's method.

523: Using the oxygen mass to filling-factor ratios derived for SN 1987A,

524: these clump parameters yield an oxygen mass of $0.7-0.9 M_{\sun}$ for

525: SN 1993J.

526: 

527: The preferred models of Nomoto et al. (1993) and Shigeyama et

528: al. (1990) for SN 1993J predict an oxygen mass of $\sim$ 0.4

529: $M_{\sun}$.  This is similar to the value of $0.47 M_{\sun}$ for

530: oxygen in the model of Swartz et al. (1993).  The models of Woosley et

531: al. (1994) and Woosley, Langer, \& Weaver (1995) yield $0.5-0.7

532: M_{\sun}$ for SN 1993J.  Utrobin (1994) finds a smaller value for the

533: oxygen mass of $\sim$ 0.2 $M_{\sun}$, but also predicts a slightly

534: narrower velocity profile for the oxygen lines ($v \lesssim 4600$

535: km~s$^{-1}$); the day 167 spectrum has a blue edge for the

536: [\ion{O}{1}] $\lambda$6300 line that indicates an expansion velocity

537: of $\sim$ 4900 km~s$^{-1}$.  For SN 1987K, another Type IIb SN,

538: Schlegel \& Kirshner (1989) found an oxygen mass of $\sim$ 0.3

539: $M_{\sun}$.  The value we derive for SN 1993J is probably a bit high,

540: but the blanket assumption that $M_O / f$ is constant among

541: core-collapse SNe definitely increases the uncertainty of the result.

542: A slightly smaller $M_O / f$ ratio would make our derived oxygen mass

543: entirely consistent with the model predictions.

544: 

545: 

546: \section{Late-Time Spectra}

547: 

548: \subsection{Late-Time Studies of Supernovae}

549: 

550: 

551: 

552: Few SNe have had relatively frequent spectroscopic observations for

553: more than three years.  Several have have been recovered many years

554: (even decades) after explosion.  These include SN 1957D (Long, Blair,

555: \& Krzeminski 1989; Cappellaro, Danziger, \& Turatto 1995), SN 1970G

556: (Fesen 1993), SN 1978K (Ryder et al. 1993; Chugai, Danziger, \& Della

557: Valle 1995; Chu et al. 1999), SN 1979C (Fesen \& Matonick 1993; Fesen

558: et al. 1999, hereinafter F99), SN 1980K (Fesen \& Becker 1990;

559: Leibundgut et al. 1991; Uomoto 1991; Fesen \& Matonick 1994; F99), SN

560: 1985L (Fesen 1998), SN 1986E (Cappellaro et al. 1995), SN 1986J

561: (Leibundgut et al. 1991; Uomoto 1991), and SN 1994aj (Benetti et

562: al. 1998).  SN 1988Z showed evidence for circumstellar interaction

563: from its early-time spectra, and continues to be observed

564: spectroscopically with $\sim$ 10 years of coverage (Filippenko 1991a,

565: b; Stathakis \& Sadler 1991; Turatto et al. 1993; Aretxaga et

566: al. 1999).  SN 1987A represents a special case; its proximity allows

567: continued monitoring that would most likely not be possible for this

568: SN if it were in a more distant galaxy, and so we will exclude it from

569: this discussion\footnote{As of early 2000, though, signs of more

570: circumstellar interaction in SN 1987A are beginning to appear (e.g.,

571: Bouchet et al. 2000)}.

572: 

573: SNe such as SN 1988Z that show the spectroscopic characteristics of

574: circumstellar interaction from an early time do so through very

575: strong, narrow emission, especially in the Balmer lines.  They are

576: often referred to as a subclass of SNe II, either as ``Seyfert 1''

577: types or the more commonly used Type IIn (Filippenko 1989, 1991a;

578: Schlegel 1990).  If we ignore the SNe IIn, and consider only the SNe

579: II observed at very late times that appeared normal initially, then

580: all of the SNe recovered after many years have fairly similar spectra.

581: The pre-condition of normalcy at maximum brightness is not necessarily

582: applicable to all the SNe listed above, as some (SN 1957D, SN 1978K,

583: and SN 1986J) were not actually observed until much later in their

584: evolution.  All but SN 1978K have fairly broad emission lines (FWZI

585: $\approx$ 10,000 km~s$^{-1}$) of H$\alpha$, [\ion{O}{3}]

586: $\lambda\lambda$4959, 5007, and [\ion{O}{1}] $\lambda\lambda$6300,

587: 6364 at late times.  Most of these lines show a largely box-like

588: profile, as expected from an expanding, roughly spherical, shell.

589: 

590: \subsection{Prior Analysis of our Nebular to Late-Time Spectra}

591: 

592: Spectra of SN 1993J up to day 433 are discussed by FMB94.  They

593: concluded that evidence for circumstellar interaction is visible as

594: early as day 235 when the box-like profile of H$\alpha$ became

595: distinct.  This shape is predicted by the circumstellar interaction

596: models of Chevalier \& Fransson (1994; hereinafter CF94).  Indeed,

597: FMB94 found an H$\alpha$ luminosity of $4.4 \times 10^{38}$

598: ergs~s$^{-1}$ on day 433, in approximate agreement with the models of

599: CF94 (and assuming the X-ray luminosity and density profile as

600: described in Fransson, Lundqvist, \& Chevalier [1996]).  There is a

601: considerable amount of evidence for circumstellar interaction in SN

602: 1993J from observations at radio and X-ray wavelengths (see Fransson

603: et al. 1996, and references therein).  Houck \& Fransson (1996) and

604: Patat, Chugai, \& Mazzali (1995) also conclude that circumstellar

605: interaction is present based on late-time optical spectra.  The

606: spectra described here are dominated by the effects of circumstellar

607: interaction, and thus are discussed within that context.  These

608: late-time spectra are shown in their entirety in Figures 7 and 10 of

609: Paper I.

610: 

611: For our present purposes, we restrict our definition of late-time

612: spectra to those obtained from day 433 onward.  Nebular features begin

613: to appear by day 93 (cf. Figures 3, 5, and 7 of Paper I), and were

614: noted earlier ($\sim$ day 62) by others with better temporal coverage

615: during this period (Barbon et al. 1995).  From day 209, the nebular

616: features completely dominate the spectra, although an unusual

617: box-shaped H$\alpha$ line may also be apparent at that point.  This

618: box-shaped line is clearly visible by day 355.  From that day, the

619: emission characteristics of the SN begin to be dominated by evidence

620: for circumstellar interaction.  Nebular features in the ejecta

621: continue to play a significant role in the spectra, but are less

622: important from day 473 onward.  This section will focus on the

623: late-time spectra during this interaction phase, from day 473 to day

624: 2454.  Figure 10 displays three high-quality Keck spectra of SN 1993J

625: from days 976, 1766, and 2454 to illustrate the discussion of the

626: late-time spectra.

627: 

628: 

629: 

630: \subsection{Extinction}

631: 

632: Analysis of the late-time spectra depends upon the amount of reddening

633: that affects them.  As absolute values for the line fluxes are

634: difficult to obtain, line intensity ratios must suffice, and here the

635: effects of reddening can be extreme.  Various values for the

636: extinction were found in photometric and spectroscopic studies of

637: SN 1993J.  Wheeler et al. (1993) used reddened blackbody models in

638: comparison with early-time spectra to conclude that $A_V = 0.47 \pm

639: 0.06$ mag (all reported values of $A_V$ assume $R = A_V / E(\bv)

640: = 3.1$).  With similar techniques, Lewis et al. (1994) calculated $A_V

641: = 0.58 \pm 0.05$ mag, while Clocchiatti et al. (1995) found $A_V

642: \approx 0.7$ mag.  Richmond et al. (1994) summarized several

643: measurements of the extinction toward M81 in general and SN 1993J in

644: particular.  The average value is $A_V \approx 0.6$ mag, but ranges

645: from $A_V = 0.2$ mag to $A_V = 1.0$ mag.  Barbon et al. (1995)

646: estimated $A_V = 0.9$ from \ion{Na}{1}~D line equivalent widths.  The

647: Schlegel, Finkbeiner, \& Davis (1998) dust maps show that the Galactic

648: component is $A_V = 0.25$ mag, not an insignificant fraction.  For the

649: purposes of this discussion, we will take the extinction toward SN

650: 1993J to be $A_V = 0.6$ mag.  All fluxes and line ratios will be

651: reported for the spectra as measured directly as well as corrected for

652: this amount of reddening using the extinction corrections of Cardelli,

653: Clayton, \& Mathis (1989), including the O'Donnell (1994)

654: modifications at blue wavelengths.

655: 

656: \subsection{Line Measurements}

657: 

658: The emission-line intensity ratios of the late-time spectra of SN

659: 1993J are fairly difficult to measure.  The large velocity width of

660: the lines (FWHM $\gtrsim 15000 $ km~s$^{-1}$; see, e.g., Figures 7, 8,

661: 10, 11, 12, and 13 of Paper I and below) and consequent overlapping

662: make the deblending of the lines almost impossible.  Without a clean,

663: isolated line to provide a well-defined profile, all measurements will

664: be affected by a relatively large uncertainty.  Despite this problem,

665: we were able to extract some information from the spectra.

666: 

667: The first step in the line measurement was the removal of the

668: underlying ``continuum.''  This is not an actual continuum from a

669: photosphere, but rather a mixture of blended lines that combine to

670: produce a relatively smooth pseudo-continuum, extending from the

671: ultraviolet (UV) edge of our spectra to $\sim$ 5500 \AA.  (The

672: pseudo-continuum was removed over the entire range of our spectra;

673: redward of 5500 \AA\ it is much weaker.)  The circumstellar

674: interaction itself can also produce a weak blue continuum (Fransson

675: 1984).  The stronger region of the pseudo-continuum is probably iron

676: and iron-group element emission.  Beyond 5500 \AA, these species also

677: contribute, but there may be other sources responsible for the red

678: light.  As this is not a well-defined continuum, its removal is an

679: additional factor in the uncertainty of any measurements.  The

680: subtraction of the pseudo-continuum was accomplished with a spline fit

681: by hand to the spectra, but in a consistent manner for all of the

682: spectra to be analyzed.

683: 

684: Line fluxes were calculated by summing the contributions from each

685: pixel over a range determined by the best approximation of the edges

686: of the line.  In the cases of blended lines, total fluxes were

687: measured.  The edges of the lines provided the values for the maximum

688: blue and red velocities for the given line(s).  Table 1 lists the

689: fluxes of the lines available in our spectra relative to H$\alpha$,

690: with the second number indicating the line ratios when dereddened by

691: $A_V = 0.6$ mag.  The relative spectrophotometry of the spectra is

692: excellent (see Paper I for discussion of the observing and calibration

693: details), so the major sources of uncertainty in the fluxes are line

694: blending, choice of line boundaries, and the subtraction of the

695: continuum.  The relative flux values probably have uncertainties of

696: $\sim$ 10\%; individual measurements with larger uncertainties are

697: denoted in Table 1.  Table 2 reports the velocities of the lines

698: determined from the same measurements; similar caveats about the

699: uncertainties of the fluxes apply to the velocities, although the

700: values for FWHM are more secure than either the blue velocity at zero

701: intensity (BVZI) or the red velocity at zero intensity (RVZI).

702: 

703: For the day 976 spectrum (see Figure 10), we felt that the line shapes

704: were sufficiently simple and well-defined compared to the other epochs

705: to attempt a decomposition of the \ion{He}{1} $\lambda$5876 +

706: \ion{Na}{1}~D blend.  We modeled the line as a simple box shape with

707: an appropriate velocity width (15,000 km~s$^{-1}$).  Using the SPECFIT

708: task in the STSDAS package of IRAF, we found a good fit to the blend

709: with approximately equal contributions from both \ion{He}{1}

710: $\lambda$5876 and \ion{Na}{1}~D (Figure 11).  Later spectra were

711: either too noisy or did not show as clean a box shape for the profile

712: as did the day 976 spectrum.  Visual inspection of the later spectra,

713: however, indicates that the relative ratio of these two lines does not

714: appear to change significantly.  This can be seen in Figure 10 of Paper

715: I.  The blue edge of \ion{He}{1} $\lambda$5876 and the red edge of

716: \ion{Na}{1}~D are not as sharp in later spectra as they are on day

717: 976.  For days 2028 to 2176, the lower S/N ratio is at least partly

718: responsible for degrading the line profile.  By day 2454, though, the

719: edges still appear, but the shoulders of the lines are more rounded.

720: This may indicate that the spherical structure that produces the

721: box-like profiles is beginning to lose its coherence.  When comparing

722: the relative heights of the blue and red shoulders of the \ion{He}{1}

723: $\lambda$5876 + \ion{Na}{1}~D blend, one must consider that the

724: pseudo-continuum affects the blue half more than the red.  Taking this

725: into account, the relative strengths of the two lines appear to remain

726: effectively constant.  There may be a slight increase of the strength

727: of \ion{He}{1} $\lambda$5876 in comparison with \ion{Na}{1}~D, but it

728: is not a significant change.  Therefore, the line ratio for the

729: \ion{He}{1} $\lambda$5876 + \ion{Na}{1}~D blend reported in Table 1

730: can probably be interpreted as approximately half \ion{He}{1}

731: $\lambda$5876 and half \ion{Na}{1}~D.

732: 

733: We also employed a decomposition technique that used one of the lines

734: in the spectra to model the others.  The individual species were

735: different enough that we chose to use the line of a given element to

736: isolate only lines from the same species, although different

737: ionization states were allowed.  Given the lines available, this

738: restricted us to the use of H$\alpha$ to model H$\beta$, and

739: [\ion{O}{2}] $\lambda\lambda$7319, 7330 to model [\ion{O}{1}]

740: $\lambda\lambda$6300, 6364 and [\ion{O}{3}] $\lambda\lambda$4959,

741: 5007.  The line used as the model was isolated from the spectrum,

742: shifted in wavelength space to match the other line, and then scaled

743: until the subtraction of the model gave the most effective removal of

744: the other line.  The choice of scaling was subjective, as

745: contamination by other species and noisy spectra precluded a more

746: formal comparison.  (The [\ion{O}{1}] $\lambda\lambda$6300, 6364

747: doublet was removed from the blue edge of H$\alpha$ before it was

748: used.)  The fact that [\ion{O}{2}] $\lambda\lambda$7319, 7330 matched

749: the other oxygen lines so well suggests that there is very little, if

750: any, contamination by [\ion{Ca}{2}] $\lambda\lambda$7291, 7324 in this

751: wavelength region at late times.  Calcium is probably the more

752: significant line at earlier times, as discussed in \S 2.  The relative

753: line fluxes determined with this technique are reported in Table 3.

754: The line-subtraction technique was only effective in the spectra

755: from day 976 onward.  In addition, the results were more accurate for

756: the high S/N ratio spectra from Keck (days 976, 1766, and 2454) than

757: for the more noisy spectra from Lick (days 2028, 2069, 2115, and

758: 2176).

759: 

760: \subsection{Discussion}

761: 

762: Most of our interpretation of the late-time spectra of SN 1993J falls

763: within a comparison with the circumstellar interaction model of CF94.

764: The CF94 model envisions cool, freely expanding SN ejecta colliding

765: with circumstellar material from a pre-explosion stellar wind.  A

766: forward shock propagates into the wind, while a reverse shock moves

767: back into the ejecta.  The SN ejecta have a fairly steep density

768: gradient, leading to a slow reverse shock with emission at far-UV

769: wavelengths (possibly in X-rays, with a different gradient).  This

770: produces emission from highly ionized species.  Absorption by a shell

771: formed at the shock boundary can yield low-ionization lines, although

772: these can also originate in the ejecta themselves.  CF94 treat two

773: different models for the structure of the wind.  One is a power law,

774: most applicable to a relatively compact progenitor, while the other

775: uses the density structure of a red supergiant (RSG) from stellar

776: evolution models.  They make specific predictions for their model,

777: including line intensity ratios and line profiles.

778: 

779: \subsubsection{Velocities}

780: 

781: The uncertainty of the edges of the lines at zero intensity, as well

782: as the large number of potentially overlapping lines, makes

783: interpretation of the measured velocities problematic.  One of the

784: very specific predictions of CF94's interaction model is that the line

785: widths decrease with time.  A pulsar-powered model would have velocity

786: widths that \emph{increase} slowly with time, and would have $v

787: \lesssim 1000$ km~s$^{-1}$ (Chevalier \& Fransson 1992).  The

788: velocities listed in Table 2 appear to suggest a general decrease over

789: time.  The H$\alpha$ BVZI is severely contaminated by [\ion{O}{1}]

790: $\lambda\lambda$6300, 6364 at early times, and this contamination

791: never completely vanishes.  In addition, H$\alpha$ may be affected on

792: the red side by \ion{He}{1} $\lambda$6678.  Given that \ion{He}{1}

793: $\lambda$5876 is approximately one-third the strength of H$\alpha$

794: (cf. Table 1 and \S 3.4), \ion{He}{1} $\lambda$6678 probably only has

795: a small effect on H$\alpha$, unless unusually high densities are

796: altering the typical \ion{He}{1} line intensity ratios (e.g., Almog \&

797: Netzer 1989).  The RVZI for [\ion{O}{3}] $\lambda\lambda$4959, 5007

798: is difficult to isolate, and thus is highly uncertain.  The BVZI for

799: H$\beta$ appears to grow fairly dramatically at late times, but this

800: is more likely the result of contamination by another species than any

801: change in H$\beta$ itself.  A comparison of H$\alpha$ and H$\beta$ in

802: velocity space on day 1766 and day 2454 (Figure 12), though, shows

803: that the blue edge of H$\beta$ does not look like H$\alpha$ at late

804: times and is probably some other species.  The CF94 model does predict

805: a strong \ion{Mg}{1}] $\lambda$4571 line, but this feature near

806: H$\beta$ is at the wrong wavelength to be magnesium.

807: 

808: The BVZI of \ion{He}{1} $\lambda$5876 and the RVZI of \ion{Na}{1}~D

809: are more likely to reflect the actual limits of velocity space

810: occupied by their respective species.  The most accurate velocity

811: information comes from the FWHM of H$\alpha$.  This is much less

812: affected by errors in the subtraction of the pseudo-continuum, and

813: there is less impact by weaker lines.  There is still some

814: contamination by [\ion{O}{1}] $\lambda\lambda$6300, 6364, but the

815: line-subtraction technique described above indicates that it widened

816: H$\alpha$ minimally.  Perhaps the best evidence that the FWHM of

817: H$\alpha$ is fairly accurate is how remarkably constant the values

818: are.  Probably from day 881, and definitely from day 1766, the FWHM is

819: $\sim$ 15,000 km~s$^{-1}$, within our measurement uncertainty.  The

820: decrease found by day 2454 appears to be real.  If this is the case,

821: then the line widths are decreasing in accordance with the CF94

822: model.  Marcaide et al. (1997) also found an apparent deceleration in

823: the ejecta using very long baseline interferometry.

824: 

825: The CF94 model has other predictions relevant to the observed

826: velocities.  One is that highly ionized lines should have different

827: velocity widths than low-ionization lines.  Unfortunately, we cannot

828: test this predication, as the difference is only $\sim$ 830

829: km~s$^{-1}$ in the power-law model for the wind structure.  Another is

830: the scale of the line-profile widths.  In the power-law model, the

831: half width at zero intensity (HWZI) is $\sim$ 5000 km~s$^{-1}$, while

832: for the RSG model, the HWZI is $\sim$ 4200 km~s$^{-1}$.  For the lines

833: of SN 1993J, the HWZI is $\sim$ 7500 km~s$^{-1}$.  This is not

834: surprising as the structure of the progenitor of SN 1993J would not

835: have been that of a typical RSG.  It may indicate that the particular

836: power law used is not applicable to SN 1993J either.  The four SNe

837: discussed by F99 (1986E, 1980K, 1979C, and 1970G) have expansion

838: velocity widths (HWZI) on the order of 5500 km~s$^{-1}$, implying that

839: they have higher-mass envelopes than SN 1993J.  This is

840: understandable, as most models for SN 1993J postulate that it had a

841: low-mass ($0.2-0.5 M_{\sun}$) hydrogen envelope, so a smaller mass

842: implies a larger velocity for a given amount of input energy (see

843: Paper I and references therein).  In all five cases, though, the

844: velocities are larger than predicted by any of CF94's models.  In

845: addition, the mass-loss rate of SN 1993J as determined from radio

846: observations (Van Dyk et al. 1994) is $(2-15) \times 10^{-5}

847: M_{\sun}$~yr$^{-1}$, and this is comparable to the values for the

848: other four SNe [($2-19) \times 10^{-5} M_{\sun}$~yr$^{-1}$; F99, and

849: references therein].

850: 

851: \subsubsection{Emission Lines}

852: 

853: Table 4 lists the predicted line fluxes from the CF94 model, both for

854: the power-law wind structure and the RSG model.  The measured fluxes

855: from our data range from 1.8 to 6.7 years, thus providing a comparison

856: with several epochs of the model.  Table 4 only contains the lines

857: with which we can make any comparisons.  The limitations of the

858: spectra in wavelength coverage and quality at red wavelengths restrict

859: the set of usable lines.  The blending of the lines also complicates

860: some direct comparisons.

861: 

862: The spectra from days 653, 670, 881, and 976 are closest to the CF94

863: year-2 predictions.  For these days, assuming \ion{Na}{1}~D is half

864: the blend of \ion{He}{1} $\lambda$5876 + \ion{Na}{1}~D as described

865: above, the sodium line is nearly half the predicted strength.  The

866: [\ion{O}{1}] $\lambda\lambda$6300, 6364 doublet is only a quarter of

867: the predicted value and [\ion{O}{2}] $\lambda\lambda$7319, 7330 is

868: three times stronger than in the model, but the [\ion{O}{3}]

869: $\lambda\lambda$4959, 5007 strength may be comparable.  If the line

870: subtraction technique is correct, and H$\beta$ is $0.2-0.3$ of the

871: H$\beta$ + [\ion{O}{3}] blend, then the observed oxygen-line flux is

872: similar to the predicted value.  Unfortunately, the [\ion{Ne}{3}]

873: $\lambda$3968 line was not available in all our spectra.  Its value on

874: day 653 is three times larger than predicted, but this region has the

875: most uncertainty due to potential line contamination and the

876: decreasing quality of the spectra toward blue wavelengths.

877: 

878: The later spectra are closer in time to the year-5 model of CF94, but

879: we will also consider the year-10 models.  The \ion{Na}{1}~D strength

880: is still weaker than predicted, but closer to calculated values, and

881: would actually fit well with the prediction for year 2.  The relative

882: strength of the sodium line does increase, as is the general trend in

883: the models.  The RSG model actually predicts a quite low value for

884: \ion{Na}{1}~D, in contrast to the power-law model.  While continuing

885: to gain in relative strength, [\ion{O}{1}] $\lambda\lambda$6300, 6364

886: is still far below the predicted values for the power-law model.  In

887: this case, the oxygen doublet is in agreement with the RSG value at

888: year 10.  The biggest difference between the observations and the

889: models may be the strength of [\ion{O}{2}] $\lambda\lambda$7319, 7330.

890: In the models, this feature barely contributes, while the observations

891: show it to be a significant line, with $70-80\%$ of the strength of

892: H$\alpha$.  Unfortunately, the CF94 RSG model does not report the

893: predicted strength of [\ion{O}{2}] $\lambda\lambda$7319, 7330.  The

894: RSG model predicts an even higher strength for [\ion{O}{3}]

895: $\lambda\lambda$4959, 5007 relative to H$\alpha$ than the power-law

896: model, and this is not what is observed, so we will concentrate on the

897: predictions of the power-law model.

898: 

899: As [\ion{O}{2}] $\lambda\lambda$7319, 7330 is expected to originate in

900: the ionized ejecta along with [\ion{O}{3}] $\lambda\lambda$4959, 5007,

901: it is unclear why the singly ionized oxygen lines should be so strong,

902: when the doubly ionized lines are slightly weaker than the models

903: indicate unless the ionization structure is fairly different from the

904: models assumptions.  The density may explain the ratios, as will be

905: discussed below.  Chevalier \& Fransson (1989) show that [\ion{O}{2}]

906: $\lambda\lambda$7319, 7330 is stronger in models of typical late-time

907: spectra of core-collapse SNe without circumstellar interaction.

908: Unfortunately, [\ion{Ca}{2}] $\lambda\lambda$7291, 7324 is predicted

909: to be an order of magnitude stronger than the [\ion{O}{2}] lines, so

910: the circumstellar model is more appropriate, although the depletion of

911: calcium onto dust grains formed as the ejecta cool may affect the

912: [\ion{Ca}{2}] $\lambda\lambda$7291, 7324 line strength (e.g., Kingdon,

913: Ferland, \& Feibelman 1995).

914: 

915: The [\ion{Ne}{3}] $\lambda$3968 line is again uncertain in these very

916: late-time spectra, but it appears to be significantly stronger than

917: the models predict.  This makes the issue of ionization structure even

918: more complicated, as this high-ionization line also originates in the

919: ejecta.  If the relative strengths of [\ion{O}{2}] and [\ion{O}{3}] are

920: the result of a lower ionization of the ejecta, then the strength of

921: [\ion{Ne}{3}] is difficult to explain.  This may imply that there are

922: also abundance issues that affect the relative line strengths.  As SN

923: 1993J was assumed to have lost most of its hydrogen envelope (see

924: Paper I and references therein), perhaps the relative amount

925: of hydrogen has been depleted in the ejecta compared to the abundances

926: used in CF94's models, which were computed for normal SNe II.  This

927: could explain the apparent excess of emission from the lines that

928: originate in the ejecta ([\ion{O}{2}], [\ion{Ne}{3}]), while the lines

929: that originate in the shell (\ion{Na}{1}~D, [\ion{O}{1}]) seem too

930: weak, as all the line ratios are relative to H$\alpha$.  The strength

931: of [\ion{O}{3}] actually fits with the predictions, and thus does not

932: necessarily support this scenario.

933: 

934: Some of the lines predicted by CF94 are notable by their absence from

935: our spectra.  These include [\ion{C}{1}] $\lambda$8729, [\ion{N}{2}]

936: $\lambda\lambda$6548, 6583, [\ion{O}{2}] $\lambda\lambda$3726, 3729,

937: \ion{Mg}{1}] $\lambda$4571, and [\ion{S}{3}] $\lambda\lambda$9069,

938: 9532.  The carbon line would not be particularly strong, but the

939: near-IR sulfur lines would be comparable in strength to H$\alpha$.

940: Neither of these features is apparent, although hints of the sulfur

941: line may exist in the day 2176 spectrum (Figure 10 of Paper I), but

942: clearly at a lower strength.  The [\ion{O}{2}] $\lambda\lambda$3726,

943: 3729 line could be contributing to the [\ion{Ne}{3}] $\lambda$3968

944: line, but the distinctive double-horned oxygen line profiles described

945: below do not appear and the wavelengths do not match well.  As

946: discussed above, the \ion{Mg}{1}] $\lambda$4571 line might be

947: appearing at the blue edge of H$\beta$, but the velocities are not

948: correct.  In addition, the magnesium line should be two-thirds of the

949: strength of H$\alpha$, and that is obviously not the case.  The

950: nitrogen lines could be hidden within the H$\alpha$ structure, but

951: there should then be a shoulder on the blue edge of H$\alpha$ caused

952: by the $\lambda$6548 component.  The predicted total [\ion{N}{2}]

953: $\lambda\lambda$6548, 6583 flux is about one-third that of H$\alpha$,

954: and if the line ratios are not affected by density considerations,

955: then the $\lambda$6548 line may be too weak to show up on the

956: H$\alpha$ profile.  As we show below, there is contamination on the

957: blue side, but from [\ion{O}{1}] $\lambda\lambda$6300, 6364.

958: 

959: Most of these apparently missing lines have one attribute in

960: common---a low critical density.  For [\ion{O}{2}]

961: $\lambda\lambda$3726, 3729, [\ion{N}{2}] $\lambda\lambda$6548, 6583,

962: and [\ion{S}{3}] $\lambda\lambda$9069, 9532, the critical densities

963: are all less than $10^6$ cm$^{-3}$ for $T \approx 10,000$ K. (These,

964: and subsequent values for the critical density and other line

965: diagnostics, were calculated with the IRAF/STSDAS NEBULAR suite of

966: tasks.)  The [\ion{C}{1}] $\lambda$8729 line has a high critical

967: density ($> 10^7$ cm$^{-3}$), so its absence, along with the lack of

968: \ion{Mg}{1}] $\lambda$4571, may be explained by an origin in the

969: shell, whose lines appear to be weaker than those from the ejecta.

970: The critical density for [\ion{O}{2}] $\lambda\lambda$7319, 7330 at $T

971: \approx 10,000$ K is $5.7 \times 10^6$ cm$^{-3}$, while it is only

972: $6.2 \times 10^5$ cm$^{-3}$ for [\ion{O}{3}] $\lambda\lambda$4959,

973: 5007.  This may explain the relative weakness of [\ion{O}{3}], if the

974: absence of the other lines does imply a density $\gtrsim 10^6$

975: cm$^{-3}$.

976: 

977: The [\ion{O}{3}] lines may be able to further constrain the density.

978: The [\ion{O}{3}] $\lambda$4363 and $\lambda\lambda$4959, 5007 lines

979: are contaminated by H$\gamma$ and H$\beta$, respectively.  This

980: contamination is potentially significant from days 553 to 1766, but

981: probably less than 20\% thereafter (cf. Tables 1 and 3).  The relative

982: contamination of H$\beta$ would be greater, and this would tend to

983: drive the [\ion{O}{3}] ($\lambda\lambda$4959, 5007/$\lambda$4363)

984: ratio lower as it dilutes [\ion{O}{3}] $\lambda\lambda$4959, 5007 if

985: removed properly.  Residual H$\beta$ contamination could drive the

986: [\ion{O}{3}] ($\lambda\lambda$4959, 5007/$\lambda$4363) ratio higher.

987: Without considering the effects of the hydrogen lines, though, the

988: [\ion{O}{3}] ($\lambda\lambda$4959, 5007/$\lambda$4363) ratio is

989: already $\sim$ 1 at day 670, implying that, to within the

990: uncertainties of our measurements, we can probably ignore the hydrogen

991: without affecting our interpretation significantly.  To get

992: [\ion{O}{3}] ($\lambda\lambda$4959, 5007/$\lambda$4363) as low as 1,

993: the temperature must be at least 10,000 K, even at the large densities

994: considered here.  The electron density under these conditions would be

995: $n_e \gtrsim 10^8$ cm$^{-3}$.  By day 2454, the ratio is $\sim$ 3,

996: implying a density of $n_e \approx 10^7$ cm$^{-3}$.  For higher

997: temperatures, the required density drops.  If $T \approx$ 30,000 K,

998: for example, $n_e \approx 10^7$ cm$^{-3}$ on day 670 and $n_e \approx

999: 2 \times 10^6$ cm$^{-3}$ by day 2454.  No matter what the temperature

1000: is, the increase of the [\ion{O}{3}] $\lambda\lambda$4959, 5007 to

1001: $\lambda$4363 ratio with time implies a decreasing density, as one

1002: would expect in the expanding ejecta.  Given the discussion of

1003: critical densities above, a density $\gtrsim 10^6$ cm$^{-3}$ with a

1004: temperature of $T \approx 10,000-30,000$ K appears to be consistent

1005: with the line emission.

1006: 

1007: The relative ratios observed are, for the most part, consistent with

1008: values determined by F99 from other late-time spectra of core-collapse

1009: SNe.  Both SN 1980K and SN 1979C have very similar line ratios when

1010: compared with SN 1993J.  This includes the large amount of emission in

1011: [\ion{O}{2}] $\lambda\lambda$7319, 7330.  SN 1986E has a much weaker

1012: [\ion{O}{2}] feature (see also Cappellaro et al. 1995; they identify

1013: the line at $\sim$ 7300 \AA\ as the [\ion{Ca}{2}]

1014: $\lambda\lambda$7291, 7324 doublet).  SN 1979C is the only one of

1015: these four to have a measured [\ion{Ne}{3}] $\lambda$3968 line, and it

1016: is not as strong as in SN 1993J.  The lines predicted by CF94 that are

1017: missing in SN 1993J are also missing in these SNe, suggesting that

1018: they, too, have a density higher than the critical density of these

1019: lines.

1020: 

1021: A few of the late-time spectra were observed on photometric nights.

1022: For these, we calculated the luminosity of the H$\alpha$ line using a

1023: distance to M81 of $d = 3.6$ Mpc (Freedman et al. 1994).  On days 553,

1024: 670, and 881, the H$\alpha$ luminosity was (in units of $10^{38}$

1025: ergs~s$^{-1}$) 2.5, 3.0, and 4.8, respectively.  By day 1766, it was

1026: 0.86, and 0.37 on day 2454.  (To correct for an extinction of $A_V =

1027: 0.6$ mag, the luminosities would be multiplied by 1.5.)  Both the day

1028: 553 and day 670 spectra were observed through relatively narrow slits

1029: on nights with variable seeing, implying a larger uncertainty for the

1030: luminosities calculated for those observations.  The day 881 spectrum

1031: was taken with a 4$\arcsec$ slit width; thus, the value of 4.8 $\times

1032: 10^{38}$ ergs~s$^{-1}$ is probably a better indication of the

1033: H$\alpha$ luminosity during these days.  Note that FMB94 found the

1034: luminosity of the H$\alpha$ line to be $4.4 \times 10^{38}$

1035: ergs~s$^{-1}$ on day 433---a comparable value within our probable

1036: uncertainties.  The day 1766 and day 2454 spectra were also taken

1037: through narrow slits, with the seeing comparable to the slit width.

1038: Therefore, the luminosities at these later times could be larger (or

1039: smaller) than measured, but most likely by not more than a factor of

1040: two, implying a real decrease in H$\alpha$ luminosity by such late

1041: times.  CF94 predicted an H$\alpha$ luminosity of 0.45, 0.19, and 0.09

1042: $\times 10^{38}$ ergs~s$^{-1}$ at 2, 5, and 10 years, respectively,

1043: for the power-law model.  The RSG model gave an H$\alpha$ luminosity

1044: of 0.48 $\times 10^{38}$ ergs~s$^{-1}$ at 10 years.  Patat et

1045: al. (1995) found an H$\alpha$ luminosity of $1.3 \times 10^{38}$

1046: ergs~s$^{-1}$ (corrected for extinction of $A_V = 0.1$ mag) on day 368

1047: (their day 367) and they interpreted this level of emission as a

1048: requirement for circumstellar interaction; radioactive decay alone was

1049: not enough.  Our values are consistent with the Patat et al. (1995)

1050: model that requires circumstellar interaction to power the emission.

1051: The luminosities we observe at late times for SN 1993J agree with the

1052: values reported for other SNe observed at similar times (see F99 for a

1053: table of values).

1054: 

1055: \subsubsection{Overall Comparison with the Circumstellar Interaction

1056: Model}

1057: 

1058: The CF94 model makes many predictions for the spectrum resulting from

1059: circumstellar interaction.  The comparison of these predictions with

1060: the late-time spectra of SN 1993J and the prior study of F99 indicates

1061: that the model is somewhat successful, but needs refinement.  The

1062: velocity widths in SN 1993J and the four SNe of F99 are larger than

1063: the model predicts, especially larger than those of the RSG model, but

1064: they do follow the expected trend of a general decrease with time.

1065: The line intensity ratios of the RSG model seem fairly inconsistent

1066: with the observed values.  The power-law model is slightly better, as

1067: the trends of the relative ratios over time are consistent.  One

1068: exception to this is the prediction of a strengthening of H$\beta$

1069: relative to H$\alpha$ as the interaction continues; we do not see this

1070: in SN 1993J (cf. Tables 3 and 4), although the density of the emitting

1071: material may affect this result.  The trend of a decreasing luminosity

1072: for the H$\alpha$ line at late times is also consistent between the

1073: observations and the models, although the absolute values differ.

1074: 

1075: The power-law model predictions may not be so different from the

1076: observed results if density effects are considered.  Many of the

1077: emission lines predicted by the model that have low critical densities

1078: are absent from the spectra of SN 1993J and the four SNe discussed by

1079: F99.  SN 1993J may have a relatively lower hydrogen abundance, as

1080: shown by its transformation from Type II to Type IIb.  This will

1081: increase the relative intensity of other lines.  The fact that similar

1082: values are seen in the four SNe of F99, though, implies that the

1083: low-mass envelope of SN 1993J is not altering the circumstellar

1084: interaction significantly.  A larger density for the emitting region

1085: associated with the circumstellar interaction may be the main

1086: explanation for the deviation of the line intensity ratios from the

1087: predictions of the CF94 models.  The relative weakness of lines

1088: predicted to arise in the ejecta, as opposed to the shell, argues that

1089: a depleted level of hydrogen may have some impact on the

1090: line-intensity ratios.

1091: 

1092: \subsubsection{H$\alpha$ Line Structure}

1093: 

1094: One other aspect of the late-time H$\alpha$ line is that it shows a

1095: persistent, narrow (unresolved, FWHM $\lesssim$ 250 km~s$^{-1}$ on day

1096: 1766) feature at zero velocity (see Figure 9 for a detailed view on

1097: day 433).  Careful inspection of Figure 10 shows that the line is

1098: present in all three of the high S/N ratio, late-time spectra.  It is

1099: not clear if this is emission from flash-ionized circumstellar

1100: material around SN 1993J (as noted in early spectra, see Paper I and

1101: references therein), or merely an \ion{H}{2} region along the line of

1102: sight.  Finn et al. (1995) noted the presence of this feature and

1103: found it to be slightly variable in strength.  They also show (their

1104: Figure 7) an image of the environment of SN 1993J taken through a

1105: narrow-band filter centered on H$\alpha$.  There is faint \ion{H}{2}

1106: region emission evident around the SN.  Aldering, Humphreys, \&

1107: Richmond (1994), while studying earlier images of M81 to analyze the

1108: progenitor of SN 1993J, found that the colors indicated more than one

1109: star, and perhaps a faint OB association, that could be the source of

1110: the narrow hydrogen emission.  Finn et al. (1995) concluded that the

1111: feature was probably not directly connected with SN 1993J.  If the

1112: emission is not associated with SN 1993J itself, then it cannot

1113: originate very far from the line of sight to SN 1993J.  Data from the

1114: Keck telescope on day 1766 were taken through a 1$\arcsec$ slit with

1115: $\sim$ 1$\arcsec$ seeing (the extraction width was 4.$\arcsec$2).  The

1116: narrow feature is still present, and thus is likely directly along the

1117: line of sight.  The H$\beta$ line shows no evidence for this feature,

1118: but the S/N ratio of the spectra in this region would make

1119: identification difficult.

1120: 

1121: Another possible source for the narrow H$\alpha$ line is material

1122: ablated from a companion star.  As discussed in Paper I, many models

1123: for SN 1993J postulate a companion to facilitate the loss of the

1124: hydrogen envelope from the progenitor.  Chugai (1986), Livne, Tuchman,

1125: \& Wheeler (1992), and Marietta, Burrows, \& Fryxell (2000) have

1126: considered the effects of the interaction of a SN explosion with a

1127: companion.  Their calculations indicated that material from a

1128: companion to an exploding white dwarf could be swept up in the ejecta

1129: with emission becoming visible several hundred days after the

1130: explosion.  The predicted widths, though, range from a few hundred to

1131: over a thousand kilometers per second.  Given the low velocity width

1132: of the observed late-time narrow H$\alpha$ line in SN 1993J, it is

1133: unlikely that ablated material from the companion star is the source.

1134: 

1135: As can be seen in Figure 10, there are small-scale features covering

1136: the flat top of the H$\alpha$ profile similar to the clumps described

1137: in \S 2 for the oxygen lines.  In this case, though, the structures

1138: change over time.  As noted by FMB94, they probably represent

1139: Rayleigh-Taylor instabilities in the hydrogen-emitting region behind

1140: the shock (Chevalier, Blondin, \& Emmering 1992).  The instabilities

1141: are enhanced if the shock is radiative (Chevalier \& Blondin 1995), as

1142: Fransson et al. (1996) suggest for SN 1993J.

1143: 

1144: \subsubsection{Late-Time Oxygen Lines}

1145: 

1146: The relative strengths of the various oxygen lines are shown by the

1147: ratios determined using the line-subtraction technique described in \S

1148: 3.4 and presented in Table 3.  The [\ion{O}{1}] $\lambda\lambda$6300,

1149: 6364 to [\ion{O}{2}] $\lambda\lambda$7319, 7330 ratio is approximately

1150: constant.  The values for the ratio of [\ion{O}{3}]

1151: $\lambda\lambda$4959, 5007 to [\ion{O}{2}] $\lambda\lambda$7319, 7330

1152: are difficult to interpret given their apparently large and random

1153: range.  The values from the Keck spectra (days 976, 1766, and 2454)

1154: are the most reliable, and, considering only these values, the

1155: relative strength of [\ion{O}{3}] $\lambda\lambda$4959, 5007 appears

1156: to weaken slightly compared with [\ion{O}{2}] $\lambda\lambda$7319,

1157: 7330 over the course of the observations.  This may indicate a

1158: reduction in the strength of the interaction and thus a decrease in

1159: the ionization level within the ejecta.  Given the uncertainty in

1160: these values, this is highly speculative.

1161: 

1162: A striking change occurs in the spectra of SN 1993J between days 976

1163: and 1766.  The oxygen lines ([\ion{O}{1}] $\lambda\lambda$6300, 6364,

1164: [\ion{O}{2}] $\lambda\lambda$7319, 7330, and [\ion{O}{3}]

1165: $\lambda$4363 and $\lambda\lambda$4959, 5007) all develop narrower

1166: profiles on top of the underlying box-like shape (cf. Figure 10).

1167: These features are not contamination by other lines, as they appear

1168: with the same relative velocities for all of the oxygen lines

1169: accessible in our spectra (Figure 13).  The peaks of the narrower

1170: profiles are separated by $\sim$ 7100 km~s$^{-1}$, so these are not

1171: the two components of the oxygen doublets (the maximum velocity

1172: separation for the oxygen doublets is $\sim$ 3000 km~s$^{-1}$ for

1173: [\ion{O}{1}] $\lambda\lambda$6300, 6364).  We believe that these

1174: profiles represent the blue and red peaks of a double-horned profile

1175: due to the ejecta colliding with a disk-like, or at least a somewhat

1176: flattened, region.  (For a derivation of the two-horned profile from

1177: an expanding disk-like structure, see, e.g., Leonard et al. 2000;

1178: Gerardy et al. 2000.)  The disk-like structure may be due to the

1179: interaction of circumstellar material with a binary companion, thought

1180: to be present according to models for mass loss from the progenitor of

1181: SN 1993J (see Paper I and references therein).  The red horn is weaker

1182: than the blue horn due to differential extinction; we see the former

1183: through a much longer path in the ejecta.  The large gap in our

1184: coverage between days 976 and 1766 means that the onset of the disk

1185: features was not recorded.  There are vague hints of the incipient

1186: two-horned structure on day 976 (Figure 14), but it is certainly not

1187: as significant as in the later spectra.  This type of structure is not

1188: evident in any of the four SNe whose late-time spectra are discussed

1189: by F99.

1190: 

1191: Starting with the day 1766 spectrum, a notch develops in the middle of

1192: the H$\alpha$ line.  This loss of flux at zero velocity may indicate a

1193: slight flattening of the spherical distribution that had been the

1194: source of the hydrogen emission.  Blondin, Lundqvist, \& Chevalier

1195: (1996) present a model for an axisymmetric circumstellar interaction

1196: in which the SN is interacting with a dense equatorial structure.

1197: Such a scenario would result in little emission along the axis of

1198: symmetry because of a lower abundance of circumstellar material, while

1199: the equatorial structure would emit as in a typical interaction.  This

1200: could create a disk-like or flattened source of circumstellar emission

1201: as is seen here.  As the ejecta expand, they can interact with the

1202: disk-like or flattened circumstellar material producing the two-horned

1203: profiles.  The emission-line profile from an expanding disk is double

1204: peaked and resembles that of a bipolar ejection.  In fact, some new

1205: models for the core-collapse mechanism may require a bipolar outflow

1206: (Khokhlov et al. 1999), although they are fairly speculative.  In

1207: addition, the optical spectropolarimetric studies of SN 1993J

1208: indicated asymmetries at early times (Trammell, Hines, \& Wheeler

1209: 1993; Tran et al. 1997).  Polarimetry of other SNe (e.g., Leonard et

1210: al. 2000; Wang et al. 2000) also suggests that asymmetries might be

1211: common, although the late appearance of a non-spherical emission

1212: source may indicate that the circumstellar matter is the cause of the

1213: asymmetry.

1214: 

1215: 

1216: The appearance of this disk-like structure may complicate the

1217: interpretation of the line-intensity ratios discussed above.  There is

1218: still the box-like profile from a roughly spherical source, in

1219: addition to the flattened region.  The discussion of line ratios

1220: assumed that the conditions were the same for the entire emitting

1221: region.  It may be that the shell and the disk have distinctly

1222: different physical conditions, but such differences would be very

1223: difficult to separate in our spectra.  The wings of the disk-like

1224: profile are obvious, but the precise decomposition of the line into

1225: disk and shell components would be model dependent.  This is made even

1226: more complicated by the fact that the box-like profile shows no

1227: significant asymmetry, while the two-horned profile is clearly

1228: stronger in the blue component.  This implies that the flattened

1229: region suffers from differential extinction, while the roughly

1230: spherical shell does not.

1231: 

1232: 

1233: \section{Conclusions}

1234: 

1235: 

1236: 

1237: The detailed substructure of the ejecta revealed by the spectra from

1238: early to nebular times indicates that SN 1993J is a clumpy supernova.

1239: (For a discussion of substructure that is the result of telluric

1240: absorption, see the Appendix.)  Both oxygen and magnesium lines show

1241: clumps.  The oxygen clumps appear in both permitted and forbidden

1242: lines, with some slight variations.  The calcium lines do not seem to

1243: have clumps.  This is consistent with a scenario in which the observed

1244: oxygen emission originates in clumps of the newly synthesized material

1245: (either through prior evolution or explosive nucleosynthesis), while

1246: the calcium emission arises mostly from pre-existing material

1247: distributed uniformly throughout the envelope, following the models of

1248: Li \& McCray (1992, 1993).  Using the technique of Chugai (1994), we

1249: find a total oxygen mass based on the observed clumps of $M_O \approx

1250: 0.7-0.9 M_{\sun}$.  This value is very uncertain and it is larger than

1251: the values predicted by models of SN 1993J.

1252: 

1253: We also present late-time spectra of SN 1993J.  The lines clearly

1254: exhibit circumstellar interaction through their box-like emission

1255: profiles.  They indicate a gradual decrease in velocity width from day

1256: 433 to day 2454, with the FWHM of H$\alpha$ remaining relatively

1257: constant from day 976 to day 2176 at $\sim$ 15,000 km~s$^{-1}$,

1258: decreasing slightly by day 2454.  The line flux for H$\alpha$ is

1259: approximately constant from day 433 to day 881.  It decreases by about

1260: a factor of five by days 1766 and 2454.  Other strong lines in the

1261: spectrum include [\ion{O}{2}] $\lambda\lambda$7319, 7330, [\ion{O}{3}]

1262: $\lambda$4363, [\ion{O}{3}] $\lambda\lambda$4959, 5007, [\ion{Ne}{3}]

1263: $\lambda$3968, \ion{He}{1} $\lambda$5876 + \ion{Na}{1}~D (probably

1264: each of equal strength), H$\beta$, and [\ion{O}{1}]

1265: $\lambda\lambda$6300, 6364.  These lines, and their relative ratios,

1266: agree fairly well with other studies of interacting SNe at late times

1267: (e.g., F99).  While the circumstellar interaction model of CF94

1268: predicts some lines that do not appear, and that [\ion{O}{2}]

1269: $\lambda\lambda$7319, 7330 should be weak, the model is somewhat

1270: consistent with the general line-intensity ratios, and the predicted

1271: evolution of the line ratios agrees with the observations.  Our

1272: spectra, though, indicate that the density is probably higher than

1273: that used by CF94.  This may be because SN 1993J is unusual, but it is

1274: inconsistent with both the power-law model that assumes a compact

1275: progenitor and the RSG model that has a more extended progenitor.

1276: Considering SN 1993J and the four SNe of F99, the power-law model

1277: appears to be a better match in its predictions, although a higher

1278: density may be required to reproduce accurately the observed

1279: line-intensity ratios.

1280: 

1281: The late-time oxygen lines imply a relatively high density ($n_e

1282: \approx 10^7-10^8$ cm$^{-3}$) on day 670, decreasing to $n_e \approx

1283: 10^6-10^7$ cm$^{-3}$ by day 2454, assuming a temperature of 10,000 K.

1284: The low ratio [\ion{O}{3}] ($\lambda\lambda$4959, 5007/$\lambda$4363)

1285: $\approx$ 1 at early times excludes temperatures much below 10,000 K.

1286: The increase of this ratio to $\sim$ 3 on day 2454 shows the general

1287: expansion of the oxygen-emitting region.  The appearance of the

1288: two-horned profile on the oxygen lines by day 1766 exhibits the

1289: development of emission from a somewhat flattened, disk-like structure

1290: in the interaction region.  This may indicate a pre-existing

1291: equatorial belt of circumstellar material around SN 1993J, possibly

1292: similar to that observed in SN 1987A (e.g., Panagia et al. 1996, and

1293: references therein), although the emission in SN 1993J is likely the

1294: result of the ejecta colliding with the equatorial belt to produce the

1295: emission.

1296: 

1297: %Finally, we have shown that the effects of telluric absorption are

1298: %apparent in bright SNe, even when observed at low airmass.  The fact

1299: %that a small water feature can create the appearance of an absorption

1300: %line at 6515 \AA\ is relevant to the detailed study of SNe, especially

1301: %when looking for H$\alpha$ absorption in spectra of SNe Ia.  The

1302: %presence of hydrogen absorption in the spectrum of a SN Ia would

1303: %greatly affect the theory of SNe Ia progenitors, potentially

1304: %eliminating the double-degenerate scenario.  Observers must make sure

1305: %that only real lines are identified as such.

1306: 

1307: \acknowledgments This research was supported by NSF grants AST-9115174

1308: and AST-9417213 to A.V.F as well as by NASA through grants GO-7434,

1309: GO-7821 and GO-8243 from the Space Telescope Science Institute, which

1310: is operated by AURA, Inc., under NASA contract NAS 5-26555.  We are

1311: grateful to the staffs of the Lick and Keck Observatories for their

1312: assistance with the observations.  The W. M. Keck Observatory is

1313: operated as a scientific partnership among the California Institute of

1314: Technology, the University of California, and NASA. The Observatory

1315: was made possible by the generous financial support of the W. M. Keck

1316: Foundation.

1317: 

1318: \appendix

1319: 

1320: \section{Telluric Absorption}

1321: 

1322: One aspect of spectroscopy that is often overlooked in the study of

1323: SNe is the impact of weak telluric absorption, which can sometimes

1324: mimic intrinsic absorption features in SN spectra.  Virtually all

1325: observers are familiar with the prominent A-band ($\sim$ 7600 \AA) and

1326: B-band ($\sim$ 6850 \AA) absorptions produced by molecular oxygen, and

1327: some make an attempt to remove them.  There are also several strong

1328: water absorption bands in the near-IR region.  Weaker water bands in

1329: the range $5000-6700$ \AA, however, are much less well known.  Typical

1330: SN spectra are often so noisy that the weaker absorptions are not

1331: obvious.  This is especially true at blue wavelengths where detectors

1332: are usually less efficient; for example, there are ozone absorption

1333: features in the region $3200-3450$ \AA\ (the Huggins bands; e.g.,

1334: Schachter, Filippenko, \& Kahn 1989; Schachter 1991, and references

1335: therein).  An extremely weak absorption, the Chappuis ozone band,

1336: extends from 5000 \AA\ to 7000 \AA, but it is unlikely to affect SN

1337: spectra significantly (see, e.g., Kondratyev 1969).  Even in spectra

1338: with a moderate S/N ratio, some of the weaker bands can appear at high

1339: airmass or high humidity.  The fortuitous appearance of a few

1340: relatively bright SNe in the past decade, however, shows that these

1341: features will be present even at low airmass in the spectra of all

1342: objects.

1343: 

1344: Figure 15 displays spectra of three bright SNe (SN 1993J, SN 1994D, and SN

1345: 1999em) over the range $6360-6700$ \AA.  Each appears to contain an

1346: absorption line at $\sim$ 6515 \AA, but the spectra are shown with

1347: \emph{observed} wavelengths, and these three SNe have differing

1348: redshifts ($-140$ km~s$^{-1}$, 850 km s$^{-1}$, and 720 km~s$^{-1}$,

1349: respectively), clearly implying a local source for the absorption.

1350: Figure 16 is a plot of the telluric absorption spectrum as determined

1351: from solar observations obtained with the Fourier-transform

1352: spectrometer at the McMath-Pierce telescope at Kitt Peak National

1353: Observatory\footnote{NSO/Kitt Peak FTS data used here were produced by

1354: NSF/NOAO.} (Wallace, Hinkle, \& Livingston 1993, 1998).  This telluric

1355: absorption spectrum has had the amount of water absorption increased

1356: by a factor of 1.7 to match better the conditions at Lick Observatory

1357: (G. Marcy, 1999, personal communication), and it has been smoothed to

1358: correspond to a spectral dispersion of $\sim$ 2 \AA/pixel.  The inset of

1359: Figure 16 shows the telluric absorption in the region $6250-6670$ \AA.

1360: Note that at this resolution a strong, narrow absorption appears at

1361: 6515 \AA\ within a broader water band.

1362: 

1363: The presence of the absorption at 6515 \AA\ and other, weaker features

1364: near H$\alpha$ indicates that great caution must be taken when

1365: interpreting small-scale fluctuations in spectra near this wavelength

1366: range.  This telluric line was noted and discussed as a possibly

1367: intrinsic feature in SN 1993J by Wheeler \& Filippenko (1996) and Finn

1368: et al. (1995), although neither group attributed any significance to

1369: it.  Since even experienced spectroscopists can be fooled by this

1370: line, it is important to be reminded of the potential effects of weak

1371: telluric absorption.  This is especially true in efforts to search for

1372: the presence of H$\alpha$ in spectra of SNe Ia.  A bright SN Ia would

1373: be an obvious choice for such an analysis, but a brighter object is

1374: more likely to show the effects of telluric absorption and lend itself

1375: to misinterpretation.  The presence or absence of hydrogen in emission

1376: or absorption in the spectra of SNe Ia could have tremendous

1377: significance in the evaluation of their proposed progenitor systems

1378: (see, e.g., Branch et al. 1995).  One must therefore be very careful

1379: with SN data to avoid the pitfalls of telluric absorption.  With the

1380: development of ever larger telescopes, this problem will extend to

1381: even fainter SNe.  In addition, the (relatively) strong O$_2$

1382: absorption near $6260-6340$ \AA\ can appear to be an intrinsic SN

1383: feature in spectra with a low S/N ratio.

1384: %mention Ohta et al. (1994) call this Halpha??

1385: 

1386: It is possible to remove most of the effects of telluric absorption,

1387: especially at high spectral resolution.  The division by the spectrum

1388: of a comparison star that is intrinsically featureless in the

1389: wavelength regions affected by absorption can correct object spectra

1390: fairly well (Wade \& Horne 1988; Bessell 1999).  It can sometimes be

1391: difficult to find a star that is sufficiently featureless at the

1392: wavelengths of interest.  This technique requires a well-exposed star

1393: to minimize the addition of any noise to the object spectrum.  It is

1394: preferable to have a comparison star that is observed at an airmass

1395: similar to the object, but it is possible to scale the comparison star

1396: to match.  We use our own procedures to remove telluric absorption

1397: following the technique of Wade \& Horne (1988)\footnote{We note that

1398: the IRAF task TELLURIC divides the object by the comparison star

1399: properly, but does not scale the spectrum of the comparison star

1400: according to the method of Wade \& Horne (1988).  With both TELLURIC

1401: and Wade \& Horne, the comparison star is exponentiated by a function

1402: of the ratio of the airmasses of the object and the comparison star.

1403: In TELLURIC, however, this function is the ratio of airmasses

1404: multiplied by a scale factor, while with Wade \& Horne it is the ratio

1405: of the airmasses raised to a power.  Wade \& Horne choose a power as

1406: the strongest telluric lines are saturated and the equivalent widths

1407: will grow with approximately the square root of the airmass.  The

1408: TELLURIC scaling of the ratio of the airmasses effectively assumes an

1409: optically thin atmospheric absorption.  When the airmasses of object

1410: and comparison star are similar, this different treatment has little

1411: impact.  With the Wade \& Horne method, variation of the power

1412: depending on the source and level of saturation of the absorption may

1413: result in better removal.  We are conducting tests in this area.}.

1414: 

1415: In the normal course of data reduction, we remove the stronger

1416: telluric absorptions, but we only include the regions of weaker bands

1417: when observing at high airmass (or sometimes when the humidity is

1418: high).  This is to avoid adding undue noise in these regions when the

1419: weak bands are already hidden in the noise of the spectrum.  High S/N

1420: ratio objects, though, will show the weak bands.  When we make an

1421: extra effort to remove them, as shown in Figure 15 for the 6515 \AA\

1422: line in SN 1999em, we achieve some success.  There are still residuals

1423: at 6515 \AA; it appears unlikely that any telluric absorption removal

1424: scheme can be 100\% effective, but it can be done to a fairly good

1425: level, as long as interpretation of the spectra includes consideration

1426: of its effects.  Observers must make sure that only real lines are

1427: identified as such.

1428: 

1429: 

1430: \begin{thebibliography}{}

1431: \bibitem{} Aldering, G., Humphreys, R. M., \& Richmond, M. 1994, \aj,

1432: 107, 662

1433: \bibitem{} Almog, Y., \& Netzer, H. 1989, \mnras, 238, 57

1434: \bibitem{} Anderson, M. C., Jones, T. W., Rudnick, L., Tregillis,

1435: I. L., \& Kang, H. 1994, \apj, 421, L31

1436: \bibitem{} Andronova, A. A. 1992, Sov. Astron. Lett., 18, 360

1437: \bibitem{} Aretxaga, I., Benetti, S., Terlevich, R. J.,

1438:  Fabian, A. C., Cappellaro, E., Turatto, M., \&

1439:  Della Valle, M. 1999, \mnras, 309, 343

1440: \bibitem{} Aschenbach, B., Egger, R., \& Tr\"umper, J. 1995, \nat, 373,

1441: 587

1442: \bibitem{} Baade, W., \& Minkowski, R. 1954, \apj, 119, 206

1443: \bibitem{} Barbon, R., Benetti, S., Cappellaro, E., Patat, F.,

1444: Turatto, M., \& Iijima, T. 1995, \aaps, 110, 513

1445: \bibitem{} Benetti, S., Cappellaro, E., Danziger, I. J., Turatto, M.,

1446: Patat, F., \& Della Valle, M. 1998, \mnras, 294, 448

1447: \bibitem{} Bessell, M. S. 1999, \pasp, 111, 1426

1448: \bibitem{} Blondin, J. M., Lundqvist, P., \& Chevalier, R. A. 1996,

1449: \apj, 472, 257

1450: \bibitem{} Bouchet, P., Lawrence, S., Crotts, A., Sugerman, B.,

1451: Uglesich, R., \& Heathcote, S. 2000, \iaucirc\ 7354

1452: \bibitem{} Branch, D., Livio, M., Yungelson, L. R., Boffi, F. R., \&

1453: Baron, E. 1995, \pasp, 107, 1019

1454: \bibitem{} Cappellaro, E., Danziger, I. J., \& Turatto, M. 1995,

1455: \mnras, 277, 106

1456: \bibitem{} Cardelli, J. A., Clayton, G. C., \& Mathis, J. S. 1989,

1457: \apj, 345, 245

1458: \bibitem{} Chevalier, R. A., \& Blondin, J. M. 1995, \apj, 444, 312

1459: \bibitem{} Chevalier, R. A., Blondin, J. M., \& Emmering, R. T. 1992,

1460: \apj, 392, 118

1461: \bibitem{} Chevalier, R. A., \& Fransson, C. 1989, \apj, 343, 323

1462: \bibitem{} Chevalier, R. A., \& Fransson, C. 1992, \apj, 395, 540

1463: \bibitem{} Chevalier, R. A., \& Fransson, C. 1994, \apj, 420, 268 (CF94)

1464: \bibitem{} Chu, Y.-H., Caulet, A., Montes, M., Panagia, N., Van Dyk,

1465: S. D., \& Weiler, K. W. 1999, \apj, 512, L51

1466: \bibitem{} Chugai, N. N. 1992, Sov. Astron. Lett., 18, 239

1467: \bibitem{} Chugai, N. N. 1994, \apj, 428, L17

1468: \bibitem{} Chugai, N. N. 1986, Sov. Astron., 30, 563

1469: \bibitem{} Chugai, N. N., Danziger, I. J., \& Della Valle, M. 1995,

1470: \mnras, 276, 530

1471: \bibitem{} Clocchiatti, A., Wheeler, J. C., Barker, E. S., Filippenko,

1472: A. V., Matheson, T., \& Liebert, J. W. 1995, \apj, 446, 167

1473: \bibitem{} Fassia, A., Meikle, W. P. S., Geballe, T. R., Walton,

1474: N. A., Pollacco, D. L., Rutten, R. G. M., \& Tinney, C. 1998, \mnras,

1475: 299, 150

1476: \bibitem{} Fesen, R. A. 1993, \apj, 413, L109

1477: \bibitem{} Fesen, R. A. 1998, \aj, 115, 1107

1478: \bibitem{} Fesen, R. A., \& Becker, R. H. 1990, \apj, 351, 437

1479: \bibitem{} Fesen, R. A., et al. 1999, \aj, 117, 725 (F99)

1480: \bibitem{} Fesen, R. A., \& Matonick, D. M. 1993, \apj, 407, 110

1481: \bibitem{} Fesen, R. A., \& Matonick, D. M. 1994, \apj, 428, 157

1482: \bibitem{} Filippenko, A. V. 1982, \pasp, 94, 715

1483: \bibitem{} Filippenko, A. V. 1989, \aj, 97, 726

1484: \bibitem{} Filippenko, A. V. 1991a, in Supernovae, ed. S. E. Woosley

1485: (New York: Springer-Verlag), 467

1486: \bibitem{} Filippenko, A. V. 1991b, in SN 1987A and Other Supernovae,

1487: eds. I. J. Danziger \& K. Kj\"ar (Garching: ESO), 343

1488: \bibitem{} Filippenko, A. V., \& Sargent, W. L. W. 1989, \apj, 345, L43

1489: \bibitem{} Filippenko, A. V., Matheson, T., \& Barth, A. J. 1994, \aj,

1490: 108, 2220 (FMB94)

1491: \bibitem{} Filippenko, A. V., Matheson, T., \& Ho, L. C. 1993, \apj,

1492: 415, L103 (FMH93)

1493: \bibitem{} Finn, R. A., Fesen, R. A., Darling, G. W., Thorstensen,

1494: J. R., \& Worthey, G. S. 1995, \aj, 110, 300

1495: \bibitem{} Fransson, C. 1984, \aap, 133, 264

1496: \bibitem{} Fransson, C., Lundqvist, P., \& Chevalier, R. A. 1996,

1497: \apj, 461, 993

1498: \bibitem{} Freedman, W. L., et al. 1994, \apj, 427, 628

1499: \bibitem{} Gerardy, C. L., Fesen, R. A., H\"oflich, P., \& Wheeler,

1500: J. C. 2000, \apj, submitted (astro-ph/9912433)

1501: \bibitem{} Haas, M. R., Colgan, S. W. J., Erickson, E. F., Lord,

1502: S. D., Burton, M. G., \& Hollenbach, D. J. 1990, \apj, 360, 257

1503: \bibitem{} Hachisu, I., Matsuda, T., Nomoto, K., \& Shigeyama,

1504: T. 1991, \apj, 368, L27

1505: \bibitem{} Hanuschik, R. W., Spyromilio, J., Stathakis, R.,

1506: Kimeswenger, S., Gochermann, J., Seidensticker, K. J., \& Meurer,

1507: G. 1993, 261, 909

1508: \bibitem{} Houck, J. C., \& Fransson, C. 1996, \apj, 456, 811

1509: \bibitem{} Iwamoto, K., Young, T. R., Nakasato, N., Shigeyama, T.,

1510: Nomoto, K., Hachisu, I., \& Saio, H. 1997, \apj, 477, 865

1511: \bibitem{} Kingdon, J., Ferland, G. J., \& Feibelman, W. A. 1995,

1512: \apj, 439, 793

1513: \bibitem{} Khokhlov, A. M., H\"oflich, P. A., Oran, E. S., Wheeler,

1514: J. C., Wang, L., \& Chtchelkanova, A. Yu. 1999, \apj, 524, L107

1515: \bibitem{} Kifonidis, K., Plewa, T., Janka, H.-Th., \& M\"uller,

1516: E. 2000, \apj, submitted (astro-ph/9911183)

1517: \bibitem{} Kondratyev, K. Ya. 1969, Radiation in the Atmosphere (New

1518: York: Academic Press)

1519: \bibitem{} Leibundgut, B., Kirshner, R. P., Pinto, P. A., Rupen,

1520: M. P., Smith, R. C., Gunn, J. E., \& Schneider, D. P. 1991, \apj, 372,

1521: 531

1522: \bibitem{} Leonard, D. C., Filippenko, A. V., Barth, A. J., \&

1523: Matheson, T. 2000, \apj, in press (astro-ph/9908040)

1524: \bibitem{} Lewis, J. R., et al. 1994, \mnras, 266, L27

1525: \bibitem{} Li, A., Hu, J., Wang, L., Jiang, X., \& Li, H. 1994, \apss,

1526: 211, 323

1527: \bibitem{} Li, H., \& McCray, R. 1992, \apj, 387, 309

1528: \bibitem{} Li, H., \& McCray, R. 1993, \apj, 405, 730

1529: \bibitem{} Livne, E., Tuchman, Y., \& Wheeler, J. C. 1992, \apj, 399,

1530: 665

1531: %\bibitem{} Livio, M. 1999, preprint (astro-ph/9903264)

1532: \bibitem{} Long, K. S., Blair, W. P., \& Krzeminski, W. 1989, \apj,

1533: 340, L25

1534: \bibitem{} Marcaide, J. M., et al. 1997, \apj, 486, L31

1535: \bibitem{} Marietta, E., Burrows, A., \& Fryxell, B. 2000, \apj,

1536: submitted (astro-ph/9908116)

1537: \bibitem{} Matheson, T., et al. 2000, \aj, in press (Paper I)

1538: %\bibitem{} Montes, M. J., Van Dyk, S. D., Weiler, K. W., Sramek,

1539: %R. A., \& Panagia, N. 1997, \apj, 482, L61

1540: \bibitem{} Nomoto, K., Suzuki, T., Shigeyama, T., Kumagai, S.,

1541: Yamaoka, H., \& Saio, H. 1993, \nat, 364, 507

1542: \bibitem{} O'Donnell, J. E. 1994, \apj, 422, 158

1543: \bibitem{} Panagia, N., Scuderi, S., Gilmozzi, R., Challis, P. M.,

1544: Garnavich, P. M., \& Kirshner, R. P. 1996, \apj, 459, L17

1545: \bibitem{} Patat, F., Chugai, H., \& Mazzali, P. A. 1995, \aap, 299,

1546: 715

1547: \bibitem{} Richmond, M. W., Treffers, R. R., Filippenko, A. V., Paik,

1548: Y., Leibundgut, B., Schulman, E., \& Cox, C. V. 1994, \aj, 107, 1022

1549: \bibitem{} Richmond, M. W., Treffers, R. R., Filippenko, A. V., \&

1550: Paik, Y. 1996, \aj, 112, 732

1551: \bibitem{} Ryder, S., Staveley-Smith, L., Dopita, M., Petre, R.,

1552: Colbert, E., Malin, D., \& Schlegel, E. 1993, \apj, 416, 167

1553: \bibitem{} Schachter, J. 1991, \pasp, 103, 457

1554: \bibitem{} Schachter, J., Filippenko, A. V., \& Kahn, S. M. 1989,

1555: \apj, 340, 1049

1556: \bibitem{} Schlegel, D. J., Finkbeiner, D. P., \& Davis, M. 1998,

1557: \apj, 500, 525

1558: \bibitem{} Schlegel, E. M. 1990, \mnras, 244, 269

1559: \bibitem{} Schlegel, E. M., \& Kirshner, R. P. 1989, \aj, 98, 577

1560: \bibitem{} Shigeyama, T., Nomoto, K., Tsujimoto, T., \& Hashimoto,

1561: M. 1990, \apj, 361, L23

1562: \bibitem{} Spyromilio, J. 1991, \mnras, 253, 25P

1563: \bibitem{} Spyromilio, J. 1994, \mnras, 266, L61 (S94)

1564: \bibitem{} Spyromilio, J., Meikle, W. P. S., \& Allen, D. A. 1990,

1565: \mnras, 242, 669

1566: \bibitem{} Spyromilio, J., \& Pinto, P. A. 1991, in SN 1987A and Other

1567: Supernovae, eds. I. J. Danziger \& K. Kj\"ar (Garching: ESO), 423

1568: \bibitem{} Spyromilio, J., Stathakis, R. A., \& Meurer, G. R. 1993,

1569: \mnras, 263, 530

1570: \bibitem{} Stathakis, R. A., Dopita, M. A., Cannon, R. D., \& Sadler,

1571: E. M. 1991, in Supernovae, ed. S. E. Woosley (New York: Springer), 95

1572: \bibitem{} Stathakis, R. A., \& Sadler, E. M. 1991, \mnras, 250, 786

1573: \bibitem{} Strom, R., Johnston, H. M., Verbunt, F., \& Aschenbach,

1574: B. 1995, \nat, 373, 590

1575: \bibitem{} Swartz, D. A. 1991, in Supernovae, ed. S. E. Woosley (New

1576: York: Springer), 434

1577: \bibitem{} Swartz, D. A., Clocchiatti, A., Benjamin, R., Lester,

1578: D. F., \& Wheeler, J. C. 1993, \nat, 365, 232

1579: \bibitem{} Swartz, D. A., Harkness, R. P., \& Wheeler, J. C. 1989,

1580: \nat, 337, 439

1581: \bibitem{} Trammell, S. R., Hines, D. C., \& Wheeler, J. C. 1993,

1582: \apj, 414, L21

1583: \bibitem{} Tran, H. D., Filippenko, A. V., Schmidt, G. D., Bjorkman,

1584: K. S., Jannuzi, B. T., \& Smith, P. S. 1997, \pasp, 109, 489

1585: \bibitem{} Turatto, M., Cappellaro, E., Danziger, I. J., Benetti, S.,

1586: Gouiffes, C., \& Della Valle, M. 1993, \mnras, 262, 128 

1587: \bibitem{} Uomoto, A. 1991, \aj, 101, 1275

1588: \bibitem{} Utrobin, V. 1994, \aap, 281, L89

1589: \bibitem{} Van Dyk, S. D., Weiler, K. W., Sramek, R. A., Rupen, M. P.,

1590: \& Panagia, N. 1994, \apj, 324, L115

1591: \bibitem{} Wade, R. A., \& Horne, K. D. 1988, \apj, 324, 411

1592: \bibitem{} Wallace, L., Hinkle, K., \& Livingston, W. 1993, An Atlas of

1593: the Photospheric Spectrum from 8900 to 13,600 cm$^{-1}$ (7350 to 11,230

1594: \AA) (NSO: Tech. Rep. 93-001)

1595: \bibitem{} Wallace, L., Hinkle, K., \& Livingston, W. 1998, An Atlas of the 

1596: Spectrum of the Solar Photosphere from 13,500 to 28,000 cm$^{-1}$ (3570 to 

1597: 7405 \AA) (NSO: Tech. Rep. 98-001)

1598: \bibitem{} Wang, L., Howell, D. A., H\"oflich, P., \& Wheeler,

1599: J. C. 2000, \apj, submitted (astro-ph/9912033)

1600: \bibitem{} Wang, L., \& Hu, J. 1994, \nat, 369, 380

1601: \bibitem{} Wheeler, J. C., et al. 1993, \apj, 417, L71

1602: \bibitem{} Wheeler, J. C., \& Filippenko, A. V. 1996, in Supernovae and

1603: Supernova Remnants, ed. R. A. McCray \& Z. Wang (Cambridge: Cambridge

1604: University Press), 241

1605: \bibitem{} Wheeler, J. C., Harkness, R. P., Barker, E. S., Cochran,

1606: A. L., \& Wills, D. 1987, \apj, 313, L69

1607: \bibitem{} Wheeler, J. C., Swartz, D. A., \& Harkness, R. P. 1993,

1608: \physrep, 227, 113

1609: \bibitem{} Winkler, P. F., \& Kirshner, R. P. 1985, \apj, 299, 981

1610: \bibitem{} Woosley, S. E., Eastman, R. G., Weaver, T. A., \& Pinto,

1611: P. A. 1994, \apj, 429, 300

1612: \bibitem{} Woosley, S. E., Langer, N., \&  Weaver, T. A. 1995, \apj,

1613: 448, 315

1614: 

1615: \end{thebibliography}

1616: 

1617: 

1618: %% Generally speaking, only the figure captions, and not the figures

1619: %% themselves, are included in electronic manuscript submissions.

1620: %% Use \figcaption to format your figure captions. They should begin on a

1621: %% new page.

1622: 

1623: \clearpage

1624: 

1625: %% No more than seven \figcaption commands are allowed per page,

1626: %% so if you have more than seven captions, insert a \clearpage

1627: %% after every seventh one.

1628: 

1629: %% There must be a \figcaption command for each legend. Key the text of the

1630: %% legend and the optional \label in curly braces. If you wish, you may

1631: %% include the name of the corresponding figure file in square brackets.

1632: %% The label is for identification purposes only. It will not insert the

1633: %% figures themselves into the document.

1634: %% If you want to include your art in the paper, use \plotone.

1635: %% Refer to the on-line documentation for details.

1636: 

1637: 

1638: \figcaption{Spectrum of the [\protect\ion{O}{1}] $\lambda\lambda$6300,

1639: 6364 blend on day 209.  Note the sawtooth structure superposed on the

1640: global line profile.\label{fig1}}

1641: 

1642: \figcaption{Artificial spectral line with six clumps added.  The lower

1643: spectrum indicates the result of the smoothing and subtraction

1644: technique described in the text.  The six clumps are readily apparent,

1645: along with the edges of the line, although the edges look distinctly

1646: different from the clumps.  For this, and all subsequent figures of

1647: smoothed and subtracted line profiles, the flux-density scaling is

1648: linear but arbitrary; the profiles are offset by arbitrary, additive

1649: amounts.\label{fig2}}

1650: 

1651: \figcaption{Smoothed and subtracted [\protect\ion{O}{1}]

1652: $\lambda\lambda$6300, 6364 line on day 209.  Pairs of maxima and

1653: minima with the same relative velocity for the [\protect\ion{O}{1}]

1654: $\lambda$6300 and [\protect\ion{O}{1}] $\lambda$6364 lines are

1655: indicated.  The zero point for velocity is 6300 \AA.  See discussion

1656: in the text.\label{fig3}}

1657: 

1658: \figcaption{Time evolution of the [\protect\ion{O}{1}]

1659: $\lambda\lambda$6300, 6364 line from day 93 to day 433.  The zero

1660: point for velocity is 6300 \AA.  Note the remarkable consistency of

1661: the substructure in the line.\label{fig4}}

1662: 

1663: \figcaption{Comparison of the clumps in the [\protect\ion{O}{1}]

1664: $\lambda$5577, [\protect\ion{O}{1}] $\lambda\lambda$6300, 6364, and

1665: \protect\ion{O}{1} $\lambda$7774 emission lines on day 209.  The zero

1666: point for velocity for [\protect\ion{O}{1}] $\lambda\lambda$6300, 6364

1667: is 6300 \AA.  All three have the same basic structure, although the

1668: one permitted line ($\lambda$7774) is missing the component at $-2340$

1669: km~s$^{-1}$; see text for details.\label{fig5}}

1670: 

1671: \figcaption{Time evolution of the \protect\ion{Mg}{1}] $\lambda$4571

1672: line from day 139 to day 298.  The major clumps persist over time, but

1673: there are changes in the details of the substructure.\label{fig6}}

1674: 

1675: \figcaption{Time evolution of the [\protect\ion{Ca}{2}]

1676: $\lambda\lambda$7291, 7324 line from day 123 to day 298.  The zero

1677: point for velocity is 7291 \AA.  The two components of the doublet

1678: stand out over the entire series, with relatively little other

1679: structure.\label{fig7}}

1680: 

1681: \figcaption{Comparison of the clumps in [\protect\ion{Ca}{2}]

1682: $\lambda\lambda$7291, 7324, H$\alpha$, [\protect\ion{O}{1}]

1683: $\lambda\lambda$6300, 6364, and \protect\ion{Mg}{1}] $\lambda$4571 on

1684: day 209. The zero point for velocity for [\protect\ion{O}{1}]

1685: $\lambda\lambda$6300, 6364 is 6300 \AA, while that for

1686: [\protect\ion{Ca}{2}] $\lambda\lambda$7291, 7324 is 7291 \AA.  The

1687: individual clumps in each species do not appear correlated.  The

1688: magnesium line may show some clumps that line up with oxygen.  Calcium

1689: and hydrogen are distinctly different.\label{fig8}}

1690: 

1691: \figcaption{Comparison of the clumps in [\protect\ion{O}{1}]

1692: $\lambda\lambda$6300, 6364, H$\alpha$, and \protect\ion{Mg}{1}]

1693: $\lambda$4571 on day 433.  The zero point for velocity for

1694: [\protect\ion{O}{1}] $\lambda\lambda$6300, 6364 is 6300 \AA.  By late

1695: times, there is no correspondence between the lines for the

1696: substructure.  Note the narrow component of H$\alpha$ at zero

1697: velocity.\label{fig9}}

1698: 

1699: 

1700: \figcaption{Three examples of the late-time spectra of SN 1993J on

1701: days 976, 1766, and 2454.  The day 976 spectrum was obtained under

1702: non-photometric conditions, and so has been arbitrarily scaled to

1703: match the other two.  A constant has also been added to the day 976

1704: and day 1766 spectra to offset the spectra from each other.  The day

1705: 2454 spectrum suffers from minor second-order contamination beyond

1706: $\sim$ 7600 \AA.  Note the relative increase over time of the oxygen

1707: lines in comparison with H$\alpha$.  In addition, the substructure of

1708: the H$\alpha$ line varies over time, although a small feature at zero

1709: velocity persists.\label{fig10}}

1710: 

1711: \figcaption{Decomposition of the \protect\ion{He}{1} $\lambda$5876 +

1712: \protect\ion{Na}{1}~D blend on day 976 using simple box-like profiles

1713: to model the individual lines.  The dotted lines indicate the two

1714: separate components (\protect\ion{He}{1} $\lambda$5876 and the

1715: \protect\ion{Na}{1}~D blend), while the dashed line is the sum that

1716: produces the overall blend.  This best fit indicates that the two

1717: lines contribute approximately equally to the blend.\label{fig11}}

1718: 

1719: \figcaption{Comparison of the H$\alpha$ and H$\beta$ lines on days

1720: 1766 and 2454.  The red half of H$\beta$ is heavily contaminated by

1721: [\protect\ion{O}{3}] $\lambda\lambda$4959, 5007, so it has been

1722: removed for clarity.  The relatively strong, narrow H$\beta$ feature

1723: at $\sim$ $-8,000$ km~s$^{-1}$ does not appear to correspond with the

1724: structure of H$\alpha$, which is contaminated by [\protect\ion{O}{1}]

1725: $\lambda\lambda$6300, 6364 on the blue side (cf. Figure

1726: 13).\label{fig12}}

1727: 

1728: \figcaption{Four oxygen lines on day 1766.  The two-horned structure

1729: is clearly evident for [\protect\ion{O}{2}] $\lambda\lambda$7319, 7330

1730: (zero point: 7325 \AA) and [\protect\ion{O}{3}] $\lambda\lambda$4959,

1731: 5007 (zero point: 5007 \AA).  The weakness of the lines, contamination

1732: from other species, and the noisy nature of their regions of the

1733: spectra obscure the features in [\protect\ion{O}{3}] $\lambda$4363 and

1734: [\protect\ion{O}{1}] $\lambda\lambda$6300, 6364 (zero point: 6300

1735: \AA), but they are still apparent.  The red peak of

1736: [\protect\ion{O}{1}] $\lambda\lambda$6300, 6364 is visible as a

1737: distinct shoulder on the blue side of H$\alpha$.  The consistency of

1738: the two-horned structure over several lines, as well as the large

1739: velocity separation of the two horns ($\sim$ 7100 km~s$^{-1}$),

1740: indicates that these profiles do not represent the two separate

1741: components of any of the oxygen doublets, but rather show the presence

1742: of a flattened, disk-like structure.\label{fig13}}

1743: 

1744: \figcaption{Line profiles of [\protect\ion{O}{2}]

1745: $\lambda\lambda$7319, 7330 on 976, 1766, and 2454.  The zero point for

1746: the velocity is 7325 \AA.  The two-horned structure is obvious in the

1747: last two spectra, but the beginnings are apparent even at day 976,

1748: especially for the blue peak.\label{fig14}}

1749: 

1750: \figcaption{Spectra of three bright SNe illustrating the potential

1751: impact of weak telluric absorption on high S/N ratio spectra.

1752: Spectrum (a) is SN 1993J from day 19 (1993 April 15), (b) is SN 1994D

1753: from 1994 April 18, (c) is SN 1999em from 1999 November 5, and (d) is

1754: the same SN 1999em spectrum after an attempt to remove some of the

1755: effects of telluric absorption.  All three SNe were observed with the

1756: same instrument, but have different redshifts, indicating that the

1757: feature at 6515 \AA\ is not intrinsic to the SNe.  When telluric

1758: effects are considered, the feature can be removed reasonably well.

1759: Note the presence of even weaker features (such as around 6470 \AA)

1760: which we did not attempt to remove.\label{fig15}}

1761: 

1762: 

1763: \figcaption{Spectrum of telluric absorption as determined from solar

1764: Fourier-transform spectra from the National Solar Observatory database

1765: at an airmass of 1.0.  The spectrum has been smoothed to 2 \AA/pixel,

1766: and the relative strength of the water features is enhanced (see

1767: text).  The inset shows the region around the narrow feature at 6515

1768: \AA.\label{fig16}}

1769: 

1770: 

1771: \clearpage

1772: 

1773: \begin{deluxetable}{lccccccc}

1774: 

1775: 

1776: \tablenum{1}

1777: \tabletypesize{\small}

1778: \tablewidth{0pt}

1779: \tablecaption{RELATIVE LINE FLUXES\label{tbl-1}}

1780: \tablehead{\colhead{Day\tablenotemark{a}} &

1781: \colhead{H$\alpha$} &

1782: \colhead{\protect\ion{He}{1} + \protect\ion{Na}{1}}  &

1783: \colhead{[\protect\ion{O}{1}]} &

1784: \colhead{[\protect\ion{O}{2}]} &

1785: \colhead{[\protect\ion{O}{3}] + H$\beta$} &

1786: \colhead{[\protect\ion{O}{3}]} &

1787: \colhead{[\protect\ion{Ne}{3}]}  \\

1788: \colhead{} &

1789: \colhead{$\lambda$6563} &

1790: \colhead{$\lambda$5876} &

1791: \colhead{$\lambda\lambda$6300,} &

1792: \colhead{$\lambda\lambda$7319,} &

1793: \colhead{$\lambda\lambda$4959,} &

1794: \colhead{$\lambda$4363} &

1795: \colhead{$\lambda$3967} \\

1796: \colhead{} &

1797: \colhead{} &

1798: \colhead{$\lambda\lambda$5890,} &

1799: \colhead{6364} &

1800: \colhead{7330} &

1801: \colhead{5007} &

1802: \colhead{} &

1803: \colhead{} \\

1804: \colhead{} &

1805: \colhead{} &

1806: \colhead{5896} &

1807: \colhead{} &

1808: \colhead{} &

1809: \colhead{$\lambda$4861} &

1810: \colhead{} &

1811: \colhead{}}

1812: \startdata

1813: 

1814: 

1815: 553     & 100, 100& 30, 32&   8.8, 9.0&   23, 21  &  33, 39 &   34, 44&   45, 62   \\ 

1816: 670     & 100, 100& 29, 30&   4.9, 5.0&   23, 21  &  29, 35 &   27, 35&   \nodata \\ 

1817: 881     & 100, 100& 40, 43& 4.6, 4.6:  &   \nodata&  37, 44: & \nodata& \nodata   \\ 

1818: 976     & 100, 100& 37, 39&   3.2, 3.3&   25, 23  &  33, 40 &   27, 35&   \nodata \\ 

1819: 1766    & 100, 100& 54, 57&   6.8, 6.9&   46, 44  &  40, 48 &   19, 25&   \nodata \\ 

1820: 2028    & 100, 100& 82, 86:&   18, 18:  &   85, 80:  &  80, 94: &   33, 44:&   62, 85:   \\

1821: 2069    & 100, 100& 90, 95&  12, 12:   &   52, 50  &  87, 104&   27, 36:&   73, 100:  \\

1822: 2115    & 100, 100& 66, 70&  12, 12   &   68, 65  &  79, 94 &   28, 36:&   72, 100:  \\

1823: 2176    & 100, 100& 70, 74&  16, 16:   &   71, 67  &  67, 80 &  30, 39 &  96, 132:   \\ 

1824: 2454    & 100, 100& 74, 78&  13, 13   &   89, 84  &  76, 91 &  24, 31 & \nodata   \\

1825: \enddata

1826: \tablecomments{Listed relative to H$\alpha \equiv 100$.  First value is relative flux as 

1827: directly measured; second value is relative flux corrected for a reddening of $A_V = 0.6$ mag using the extinction corrections of Cardelli et

1828: al. (1989), including the O'Donnell (1994) modifications.  In all cases a

1829: continuum was removed from beneath the individual lines.  When 

1830: two (or more) lines are listed, the flux indicated is the sum for all.  Values are uncertain by $\sim$ 10\% (those

1831: indicated with a colon are uncertain by $\gtrsim$ 20\%).}

1832: \tablenotetext{a}{Day since explosion, 1993 March 27.5 UT.}

1833: \end{deluxetable}

1834: \clearpage

1835: \begin{deluxetable}{lccccccc}

1836: 

1837: 

1838: 

1839: \tablenum{2}

1840: \tabletypesize{\small}

1841: \tablewidth{0pt}

1842: \tablecaption{MEASURED VELOCITIES\label{tbl-2}}

1843: \tablehead{\colhead{Day\tablenotemark{a}} &

1844: \colhead{H$\alpha$ $\lambda$6563} &

1845: \colhead{H$\alpha$ $\lambda$6563} &

1846: \colhead{H$\alpha$ $\lambda$6563} &

1847: \colhead{H$\beta\ \lambda$4861} &

1848: \colhead{[\protect\ion{O}{3}] $\lambda$5007} &

1849: \colhead{\protect\ion{He}{1} $\lambda$5876}  &

1850: \colhead{\protect\ion{Na}{1} $\lambda$5890} \\

1851: \colhead{} &

1852: \colhead{BVZI\tablenotemark{b}} & 

1853: \colhead{RVZI\tablenotemark{b}} & 

1854: \colhead{FWHM\tablenotemark{b}} & 

1855: \colhead{BVZI\tablenotemark{b}} & 

1856: \colhead{RVZI\tablenotemark{b}} & 

1857: \colhead{BVZI\tablenotemark{b}} & 

1858: \colhead{RVZI\tablenotemark{b}} } 

1859: \startdata

1860: 

1861: 

1862: 433     &  $-$16600 &  11700 &   23100& \nodata & \nodata  & \nodata & \nodata \\

1863: 473     &  $-$16100 &  10900 &   22000& \nodata & \nodata  & \nodata & \nodata \\

1864: 523     &  $-$10400 &  13100:&   17400& \nodata & \nodata  & \nodata & \nodata \\

1865: 553     &  $-$10000 &  9800  &   17200&  $-$10500 &   9100   &  $-$16200 &   8400  \\

1866: 670     &  $-$10500 &  10100 &   16900&   $-$9700 &    9600  &  $-$16000 &    9200 \\

1867: 881     &  $-$10300 &  9500  &   16400&   $-$9500 &  10000   &  $-$16200 &   8700  \\

1868: 976     &  $-$10600 &   9800 &   16000&   $-$9500 &    8000  &  $-$16000 &    8900 \\

1869: 1766    &  $-$9900  &   9700 &   15000&   $-$9700 &    7600: &  $-$15600 &    8500   \\

1870: 2028    &  $-$8900  &   9100 &   15000&  $-$10400 &    9800  &  $-$21100:&    8400  \\

1871: 2069    &  $-$9400  &   9100 &   15000&  $-$12000 &    7500  &  $-$17200 &    8700  \\

1872: 2115    &  $-$9900  &   9300 &   14700&  $-$14900:&    7200: &  $-$15400 &    8300   \\

1873: 2176    & $-$9600   &   9000 &   14900&  $-$11800:&    6700: &  $-$15900 &    7700   \\

1874: 2454    & $-$9300   &   8700 &   13800&  $-$13400:&    8400: &  $-$15200 &    7500   \\

1875: 

1876: \enddata

1877: \tablecomments{All velocities are km~s$^{-1}$.  Values are uncertain by $\sim$ 10\% (those indicated by a colon are

1878: uncertain by $\sim 20-30$\%).}.

1879: \tablenotetext{a}{Day since explosion, 1993 March 27.5 UT.}

1880: \tablenotetext{b}{BVZI = blue velocity at zero intensity; RVZI = red velocity at zero intensity; FWHM = full width at

1881: half maximum.}

1882: 

1883: 

1884: 

1885: \end{deluxetable}

1886: 

1887: \clearpage

1888: 

1889: \begin{deluxetable}{lccc}

1890: 

1891: \tablenum{3}

1892: 

1893: \tablewidth{300pt}

1894: \tablecaption{LINE DECOMPOSITION FLUXES\label{tbl-3}}

1895: \tablehead{\colhead{Day\tablenotemark{a}} &

1896: \colhead{H$\beta$\tablenotemark{b}} &

1897: \colhead{[\protect\ion{O}{1}]\tablenotemark{c}} &

1898: \colhead{[\protect\ion{O}{3}]\tablenotemark{c}}  \\

1899: \colhead{} &

1900: \colhead{$\lambda$6563} &

1901: \colhead{$\lambda$6300} &

1902: \colhead{$\lambda$5007} }

1903: 

1904: \startdata

1905: 976  &  0.2, 0.3  &  0.3, 0.3  &  1.5, 1.9  \\

1906: 1766 &  0.2, 0.2  &  0.3, 0.3  &  1.3, 1.6  \\

1907: 2028 &  0.3, 0.4  &  0.3, 0.4  &  1.3, 1.8  \\

1908: 2069 &  0.3, 0.4  &  0.4, 0.4  &  1.9, 2.3  \\

1909: 2115 &  0.2, 0.3  &  0.3, 0.3  &  1.5, 1.8  \\

1910: 2176 &  0.2, 0.3  &  0.3, 0.4  &  1.4, 1.7  \\

1911: 2454 &  0.1, 0.1  &  0.3, 0.3  &  1.2, 1.4  \\

1912: \enddata

1913: \tablecomments{Line fluxes as a fraction of another line in the spectrum.  Details of decomposition are presented in the

1914: text.  The first number is the fraction as measured; the second is the fraction after the spectrum has been dereddened

1915: by $A_V = 0.6$ mag using extinction corrections of Cardelli et

1916: al. (1989), including the O'Donnell (1994) modifications.  The values for days 2028, 2069, 2115, and 2176 are much less

1917: certain than those for days 976, 1766, and 2454.}

1918: \tablenotetext{a}{Day since explosion, 1993 March 27.5 UT.}

1919: \tablenotetext{b}{Fraction relative to the strength of H$\alpha$.}

1920: \tablenotetext{c}{Fraction relative to the strength of \protect[\ion{O}{2}] $\lambda\lambda$7319, 7323.}

1921: 

1922: 

1923: \end{deluxetable}

1924: \clearpage

1925: 

1926: \begin{deluxetable}{lccccccc}

1927: 

1928: \tablenum{4}

1929: \tabletypesize{\small}

1930: \tablewidth{0pt}

1931: \tablecaption{PREDICTED LINE FLUXES\label{tbl-4}}

1932: \tablehead{\colhead{Age} &

1933: \colhead{H$\alpha$} &

1934: \colhead{H$\beta$} &

1935: \colhead{[\protect\ion{O}{1}]} &

1936: \colhead{[\protect\ion{O}{2}]} &

1937: \colhead{[\protect\ion{O}{3}]} &

1938: \colhead{[\protect\ion{Ne}{3}]} &

1939: \colhead{\protect\ion{Na}{1}~D} \\

1940: \colhead{(yr)} &

1941: \colhead{$\lambda$6563} &

1942: \colhead{$\lambda$4861} &

1943: \colhead{$\lambda\lambda$6300,} &

1944: \colhead{$\lambda\lambda$7319,} &

1945: \colhead{$\lambda\lambda$4959,} &

1946: \colhead{$\lambda\lambda$3868,} &

1947: \colhead{$\lambda\lambda$5890,} \\

1948: \colhead{} &

1949: \colhead{} &

1950: \colhead{} &

1951: \colhead{6364} &

1952: \colhead{7330} &

1953: \colhead{5007} &

1954: \colhead{3969} &

1955: \colhead{5896} }

1956: \startdata

1957: PL\tablenotemark{a} &  &     &    &    &     &    &     \\

1958:  \ \  1     & 100 & 14 & 26 &  3 &   7 & 11 &  27 \\

1959:  \ \  2     & 100 & 23 & 26 &  6 &  27 & 18 &  38 \\

1960:   \ \ 5     & 100 & 30 & 35 & 14 & 100 & 30 &  61 \\

1961:   \ \ 10    & 100 & 32 & 52 & 15 & 220 & 43 & 100 \\

1962:   \ \ 17.5  & 100 & 34 & 72 & 14 & 470 & 55 & 140 \\

1963:   \ \ 30    & 100 & 34 & 74 & 12 & 640 & 56 & 140 \\

1964:  RSG\tablenotemark{b} &  &     &    &    &     &    &     \\

1965:   \ \ 10    & 100 & 33 & 14 &\nodata & 340 & 30 & 17 \\

1966: \enddata

1967: \tablecomments{Line fluxes from the circumstellar interaction model of Chevalier \& Fransson (1994), listed relative to

1968: H$\alpha \equiv 100$.  Only the lines relevant to our observations are shown.}

1969: \tablenotetext{a}{Power-law model.}

1970: 

1971: \tablenotetext{b}{Red-supergiant model.}

1972: 

1973: \end{deluxetable}

1974: \clearpage

1975: 

1976: %% The following command ends your manuscript. LaTeX will ignore any text

1977: %% that appears after it.

1978: 

1979: \end{document}

1980: