1: \documentclass[12pt,aasms4]{aastex}

2: 

3: \begin{document}

4: \title{\bf Phase-Resolved Infrared {\it H}- and {\it K}-band Spectroscopy of 

5: EF Eridani\footnote{Partially based on observations obtained at La Silla (ESO proposal

6: 69.D-0142(A))}}

7: 

8: \author{Thomas E. Harrison\footnote{Visiting Astronomer, Kitt Peak National 

9: Observatory, National

10: Optical Astronomy Observatory, which is operated by the Association of 

11: Universities for Research in Astronomy, Inc., under cooperative agreement

12: with the National Science Foundation.}}

13: \affil{Astronomy Department, New Mexico State University, Las Cruces, 

14: New Mexico, 88003}

15: \authoremail{tharriso@nmsu.edu}

16: 

17: \author{Steve B. Howell\footnote{Visiting Astronomer, NTT Telescope at La Silla

18: Operated by ESO}}

19: \affil{WIYN Observatory and National Optical Astronomy Observatory, 950

20: North Cherry Avenue, Tucson, AZ 85719}

21: \authoremail{howell@noao.edu}

22: 

23: \author{Paula Szkody}

24: \affil{Department of Astronomy, University of Washington, Box 351580, Seattle,

25: WA 98195}

26: \authoremail{szkody@alicar.astro.washington.edu}

27: 

28: \author{Derek Homeier}

29: \affil{Department of Physics and Astronomy, University of Georgia, Athens,

30: GA 30602-2451}

31: \authoremail{derek@physast.uga.edu}

32: 

33: \author{Joni J. Johnson, Heather L. Osborne}

34: \affil{Astronomy Department, New Mexico State University, Las Cruces, 

35: New Mexico, 88003}

36: \authoremail{jojohnso@nmsu.edu, hosborne@nmsu.edu}

37: 

38: \begin{abstract}

39: We present new phase-resolved {\it H} and {\it K}-band spectroscopy of the 

40: ultra-short period magnetic cataclysmic variable EF Eri in its current,

41: prolonged ``low'' state obtained using 

42: NIRI on Gemini-North, and NIRSPEC on Keck II. These new data show that the 

43: {\it H}-band spectrum of EF Eri appears to be dominated by cyclotron emission 

44: during the entire orbital cycle. The {\it K}-band spectrum of EF Eri is 

45: likewise dominated by cyclotron emission during most of an orbital period, but 

46: near binary phase

47: 0.0, the secondary star spectrum may be visible. The lack of

48: strong CO or CH$_{\rm 4}$ absoprtion features, and the weakness of the

49: water vapor features in this spectrum, however, suggests the possibility of 

50: peculiar abundances for Carbon and/or Oxygen. We have used the \texttt{PHOENIX} 

51: stellar atmosphere code to produce

52: model brown dwarf spectra with non-solar abundances of Carbon, Nitrogen

53: and Oxygen and achieved limited success in fitting the observed spectra.

54: We conclude that strong, and highly variable cyclotron emission is responsible for the

55: photometric variation previously reported for EF Eri.

56: The nature of this cyclotron emission is complex: the {\it H}-band spectra

57: show that the dominant cyclotron harmonic at phase 0.5 peaks

58: at 1.65 $\mu$m, but at phase 0.0, the harmonic peaks near 1.72 $\mu$m.

59: At phase 0.5, there is another cyclotron feature present that peaks in between

60: the {\it H} and {\it K} bands (near 1.93 $\mu$m), but at phase 0.0, no 

61: such feature is present. These data suggest that cyclotron emission from both

62: poles is occurring. In the high state, the cyclotron emission

63: has been modeled as coming from the pole that is oriented towards the

64: secondary star. One interpretation for the phase 0.5 cyclotron emission is

65: that it originates from the opposite pole. In its current ultra-low

66: state, EF Eri reveals no outward signs of accretion (such as H I emission),

67: but continues to have a few, strong cyclotron features. Thus, EF Eri

68: joins the small group of magnetic cataclysmic variables whose accretion rate 

69: is so low that they are in the ``bombardment scenario'' regime. 

70: \end{abstract}

71: 

72: \keywords{binaries: close --- stars: individual (EF Eridani) -- stars:

73: low-mass, brown dwarfs --- stars: magnetic fields --- stars: variables: general}

74: 

75: \begin{flushleft}

76: \end{flushleft}

77: 

78: 

79: \section{Introduction}

80: Polars are magnetic cataclysmic variables that have white dwarf primaries with 

81: strong magnetic fields that dramatically effect the accretion flow 

82: from their cool Roche lobe filling companions. Instead of forming a disk, the 

83: accreted matter flows in a stream from near the L$_{\rm 1}$ point to 

84: one, or both, of the magnetic poles of the white dwarf primary 

85: (see Wickramasinghe \& Ferrario 2000 for a complete review). For reasons not 

86: yet identified, the normal ``high state'' mass transfer rate ($\approx$ 10$^{-11}$ M$_{\sun}$ yr$^{-1}$) ceases, and the system transitions to a ``low state". 

87: For the most extreme examples, accretion rates as 

88: low as 10$^{-13}$ M$_{\sun}$ yr$^{ -1}$, or lower, have been inferred (Reimers 

89: et al. 1999, Schwope et al. 1999, Szkody et al. 2003). When in these low 

90: states, polars should present an excellent opportunity to examine the very low

91: mass secondary stars of the shortest period cataclysmic variables (CVs). 

92: 

93: EF Eri is an ultra-short period (P$_{\rm orb}$ = 81 min) polar that entered a 

94: low state in 1997. Harrison et al. (2003) presented simultaneous 

95: multi-wavelength optical and infrared photometry for EF Eri in its low state in 

96: which they believed they had detected the irradiated brown dwarf-like

97: secondary star of this system. Their data showed that in the {\it BVRIJ} 

98: bandpasses, the light curves were dominated by emission 

99: from the white dwarf and some residual cyclotron emission, while in the 

100: {\it H}- and {\it K}-bands, the light curves exhibited large amplitude 

101: ($\Delta$K = 0.8 mag) variations that repeated once per orbital cycle. 

102: Harrison et al.  

103: were able to model these multi-wavelength data with 

104: an extremely cool (T$_{\rm eff} \leq$ 1,000 K) secondary star that is 

105: being irradiated by the primary. In this model, the irradiated hemisphere

106: of the secondary star is heated to T$_{\rm eff} \approx$ 1,500 K. If this 

107: model were correct, then the spectroscopic appearance of the secondary star 

108: should show dramatic changes over an orbital cycle, oscillating between a 

109: state where CO absorption is strong, as in the L-dwarfs, to one where methane 

110: might dominate, as seen in T-dwarfs. We have obtained new phase-resolved 

111: {\it H-} and {\it K}-band spectra of EF Eri using NIRSPEC on KECK II, and 

112: {\it K}-band spectra using NIRI on Gemini-North, to test the irradiated

113: brown dwarf model.

114: 

115: \section{Observations}

116: 

117: EF Eri was observed continuously from 7:21 until 9:00 UT on 24 December

118: 2002 using NIRI\footnote{http://www.us-gemini.noao.edu/sciops/instruments/niri/NIRIIndex.html}

119:  on Gemini-North in queue scheduled mode. For these observations

120: the F/6 camera, a four pixel slit and the {\it K-grism} were used to produce 

121: {\it K-}band spectra with a resolution of R $\approx$ 780. Individual spectra, 

122: with exposure times of 120 s, were obtained at five different positions along 

123: the slit. After each exposure, the telescope was offset by 6" along the slit,

124: and a new exposure started. In this way forty two individual spectra were

125: obtained in 100 minutes, covering 123\% of an orbital cycle for EF Eri.

126: In addition to this set of continuous spectra, four additional spectra

127: were obtained in the period between 6:10 and 6:19 UT.

128: 

129: To remove telluric features in the spectra of EF Eri, the G2V star HD16358 was 

130: observed using an identical procedure, but with exposure times of 5 s. At the 

131: time of the observation, HD16358 had an airmass of 1.44, which is very close 

132: to the mean airmass attained by EF Eri during the 100 minute observation period.

133: The resulting dataset was reduced using a procedure common for reducing

134: infrared spectral data in which the images are grouped into sets of five 

135: individual exposures. A median of four of these exposures is subtracted 

136: from the spectrum of interest to remove the sky contamination. A spectrum is 

137: then extracted from this sky-subtracted image using the normal longslit 

138: spectroscopic reduction packages within IRAF, and the process is repeated. 

139: Each individual spectrum was wavelength calibrated using the spectrum of an 

140: argon lamp obtained shortly before our observations began.

141: The wavelength calibrated spectra of EF Eri were then divided by the 

142: median of the five wavelength calibrated spectra of HD16358 to remove 

143: telluric absorption features. The latter procedure introduces spurious 

144: features into the spectra of interest due to weak atomic and molecular

145: absorption lines in the G2V spectrum. To remove these weak features, we have 

146: employed the procedure described by Maiolino, Rieke, \& Rieke (1996).

147: 

148: With only a single standard to work with, the telluric 

149: correction for the entire data set is not as good as can be produced 

150: (see Harrison et al.  2004) when a large number of such spectra are available.

151: This is evidenced by dips at 2.0 $\mu$m

152: and 2.05 $\mu$m in the NIRI spectra presented below.

153: The flux calibration of these data was fairly good, and our spectra revealed

154: the same large amplitude ($\Delta${\it K} = 0.8 mag) flux variations

155: as reported by Harrison et al. (2003). We have assumed that EF Eri has

156: not changed its photometric behavior between the time of the last photometric 

157: data set 

158: (see below) and the epoch of these observations, allowing us to use the 

159: {\it K}-band light curve to produce flux-calibrated spectra.

160: 

161: EF Eri was observed using NIRSPEC\footnote{http://www2.keck.hawaii.edu/inst/nirspec/nirspec.html} on Keck II on 6 September 2003. We used NIRSPEC in the

162: low-resolution mode with the 0.38'' slit. We employed the four-nod script

163: to obtain spectra at four different positions along the slit.

164:  For the {\it H}-band spectra the

165: grating tilt was set to cover the wavelength range 1.54 to 1.83 $\mu$m with

166: a dispersion of 2.91 \AA/pix. The {\it K}-band setup covered the 

167: wavelength range 2.00 to 2.42 $\mu$m, with a dispersion of 4.27 \AA/pix.

168: Due to the strong curvature of the spectra produced by NIRSPEC,

169: each spectrum has a slightly different wavelength coverage

170: that depends on the position of the object along the slit.

171: This effect will be apparent in some of the NIRSPEC spectral sequences presented

172: below.  The {\it H}-band data were obtained from 12:29 to 13:57 UT, covering 

173: 109\% of an orbital cycle. The {\it K}-band spectra were obtained in the 

174: interval 14:24 to 15:25 UT, and covered 88\% of an orbital period. HD16358

175: was used as the telluric standard for both bandpasses. While the

176: conditions were photometric, we found that the nodding routine did not

177: reliably place our targets exactly on the slit during the four-position

178: nod, and thus we were unable to extract photometric information from these data.

179: We have decided to use the infrared light curves to produce 

180: flux-calibrated spectra.

181: 

182: Three months before the Gemini observations, on 2002 September 24, we were 

183: able to obtain additional, simultaneous {\it JHK} photometry of EF Eri using

184: SQIID on the KPNO 2.1 m. These data were reduced in the same fashion as our

185: earlier observations

186: (see Harrison et al. 2003 for full details). The conditions during this

187: run were poorer than those of our 2001 run, and

188: the {\it J}-band light curve (see Fig. 1) has a lower S/N than the earlier

189: effort. The {\it H}- and {\it K}-band data, however, reveal identical light 

190: curves to the 2001 data set, showing no evidence for changes in either the 

191: absolute flux levels, or in the amplitude of the variations.

192: 

193: Harrison et al. (2003) found that the phasing of EF Eri by Bailey et al. 

194: (1982), where phase 0 corresponded to the appearance of a sharp

195: dip in the {\it J}-band light curve, appeared to be identical with binary 

196: orbital phase. We have used the same set of ephemerides to phase the data here.

197: Because of EF Eri's extremely short period, the spectral smearing caused

198: by the orbital motion is significant. We have used the evolution

199: code predictions by Howell, Nelson, \& Rappaport (2001) to estimate the

200: masses of the two components in the EF Eri system as M$_{\rm 1}$ = 0.6

201: M$_{\sun}$ and M$_{\rm 2}$ = 0.06 M$_{\sun}$. Using an orbital inclination

202: of 45$^{\circ}$ (see Harrison et al. 2003, and references therein), we

203: estimate the radial velocity semi-amplitude of the secondary in EF Eri

204: to be K$_{\rm 2}$ = 340 km s$^{\rm -1}$. The full orbital variation corresponds 

205: to only about two resolution elements in the NIRI spectra, but six 

206: resolution elements for the Keck spectra. Thus,

207: to allow us to coadd the individual spectra to search for features

208: from the secondary star, we have Doppler corrected our data using this 

209: estimated value for K$_{\rm 2}$.

210: 

211: In our ongoing program of continued, occasional monitoring EF Eri, we have 

212: obtained a new optical

213: spectrum using the 

214: EMMI\footnote{http://www.ls.eso.org/lasilla/Telescopes/NEWNTT/emmi/} 

215: spectrograph in the Red Imaging Low Dispersion (RILD) mode on the {\it NTT}. 

216: This spectrum, shown in Figure 2, was obtained

217: on 10 August 2002, is a 600 s exposure, and has a resolution of 2.3 \AA. It is 

218: interesting

219: to compare this spectrum with those presented in Harrison et al. (2003),

220: and Beuermann et al. (2000): the Zeeman absorption features are now more

221: pronounced, and the emission from H I has continued to decline to the point

222: where it is no longer detectable.

223: 

224: \section{Results}

225: 

226: In the preceding section we noted that we have produced flux-calibrated

227: spectra by using the infrared light curves for EF Eri. We did this so as 

228: to produce a consistent set of data that would enable us to ignore

229: slit losses in either set of spectra. Comparison of data sets obtained on widely

230: separated epochs like those analyzed here could be compromised by intrinsic

231: variations in EF Eri. While we are confident about the photometric ``stability''

232: of EF Eri up to, and including the Gemini data, there is no assurance that

233: EF Eri was in the same state when we observed it with Keck. But to make

234: sense of the Gemini NIRI data has required us to assume that the 

235: {\it H}-band spectra obtained with NIRSPEC represents a similar state in

236: EF Eri. The fact that the NIRSPEC {\it K}-band data had similar phase-dependent

237: morphology gives us confidence in this assumption.

238: 

239: In Fig. 3 we present the {\it H}- and {\it K}-band spectra for EF Eri at

240: a number of orbital phases. Each of these spectra is the median of four 

241: ({\it H}) or five ({\it K}) individual spectra. The {\it H}-band spectra in this plot have

242: been produced as described above. The telluric correction scheme of

243: Maiolino et al. (1996) truncates the blue end of the {\it K}-band at

244: 1.94 $\mu$m. Due to the blueward rise seen in the NIRI spectra of EF Eri, we

245: have also reduced the {\it K}-band spectra by simple raw division of these 

246: data by the telluric standard, and then multiplied the divided spectra

247: by the spectrum of the appropriate blackbody. This allowed us to extend our 

248: wavelength coverage 

249: down to $\approx$ 1.92 $\mu$m. This latter method, however, introduces weak 

250: ``emission'' features from the G-star division (see Fig.1 in Harrison et al. 

251: 2004).  Given the low S/N of our data set, the only obvious such feature is 

252: the false H I Br-$\gamma$ emission line at 2.16 $\mu$m visible in the

253: spectra with orbital phases near 0.5. 

254: 

255: \subsection{The {\it H}-band Spectra}

256: 

257: During one orbital cycle, the {\it H}-band spectra undergo remarkable

258: changes. Near phase 0.0 (i.e., minimum light), the spectra show a pronounced

259: blueward rise. Note that the data have relatively high S/N in this region,

260: as it is not near the blue cutoff of the {\it H}-band. Alternatively, one could

261: suggest that there is a single, broad absorption feature located at

262: 1.574 $\mu$m. After this dip, the spectrum rises to a broad maximum whose

263: peak is near 1.72 $\mu$m. As we show below, this is not the expected behavior 

264: for the spectrum of a normal late-type/brown dwarf star.

265: 

266: As the orbital phase progresses from phase 0.0, the blueward rise rapidly

267: disappears, and the flux peak in the spectrum moves to the blue. It is

268: unclear whether this blue feature completely disappears, or is simply

269: hidden by the much higher flux levels associated with the phase 0.5 cyclotron 

270: emission.  During the transition to maximum light, a weak rise at the extreme 

271: red end of the {\it H}-band appears. 

272: While the S/N of the data in this region is quite low, the combination of the

273: {\it H} and {\it K}-band data sets suggest that this rise is real, and is due 

274: to a broad feature that peaks somewhere between the edges of the {\it H} and 

275: {\it K} bandpasses, in the middle of the region strongly affected by telluric

276: water vapor absorption. As the orbital phase continues to increase on its

277: way back to phase 0.0, the spectra retrace their pre-maximum evolution.

278: 

279: \subsection{The {\it K}-band Spectra}

280: 

281: The {\it K}-band spectra obtained with NIRI are shown

282: in Fig. 3, and the {\it K}-band spectra obtained with NIRSPEC are presented in 

283: Fig. 4. Both data sets show similar evolution over an orbital cycle,

284: suggesting that little has changed in the intervening nine months. Near phase 

285: 0.0, the spectrum shows a continuum that is 

286: consistent with that of a late-type star, with declining fluxes at both

287: blue and red ends, presumably caused by water vapor absorption in the

288: photosphere of the secondary star. The co-addition of the NIRSPEC

289: phase 0.0 spectra (Fig. 5) do not confirm this, however, with no clear sign

290: of the expected spectral features for a late-type star. {\it This is

291: true even when we correct the spectra for the predicted radial velocity

292: of the secondary star.} While

293: a feature consistent with $^{\rm 12}$CO$_{\rm (2,0)}$ (at 2.294 $\mu$m) 

294: appears to be present, other features with a similar depth do not correspond 

295: to any of the expected atomic features (e.g., Na I at 2.20 $\mu$m) for

296: a cool star. By coadding

297: the entire NIRSPEC data set, the first overtone feature of $^{\rm 12}$CO$_{\rm (2,0)}$ appears to remain marginally detectable. This would be the first 

298: detection of the secondary star in EF Eri.

299: 

300: As the orbital phase advances, the spectra show rapid changes

301:  with the development of a pronounced hump blueward of 2.10 $\mu$m, and 

302: an increasingly steeper

303: continuum beyond 2.2 $\mu$m. This blue hump is better defined in the NIRI

304: spectra due to their extended blue coverage. Note that the telluric water

305: vapor absorption below 2.0 $\mu$m is very strong, and thus the shape of the 

306: continuum here is highly sensitive to the telluric correction. While the

307: blueward rise begins well before this difficulty, the exact form of

308: the slope of the continuum below 2.0 $\mu$m is not well-defined with the

309: current data set.

310: 

311: \subsection{The Secondary Star of EF Eri}

312: 

313: As we have just noted (see Fig. 5), combining the entire NIRSPEC data

314: set has allowed us to tentatively detect the secondary star in EF Eri. While

315: the S/N of these data are not high, if the secondary star were relatively

316: normal, it should be more prominent than seen here. 

317: By modeling the optical spectrum of EF Eri, Beuermann et al. 

318: (2000) have derived distances for EF Eri in the range 80 $\leq$ d $\leq$

319: 128 pc, while Thorstensen (2003) has measured a distance of 163 $\pm$ 58 pc

320: via parallax. Thorstensen finds that an astrometric solution with a distance 

321: of 113 pc is possible by adjusting some of the input criteria. 

322: As shown in Harrison et al. (2003), the optical photometry of EF Eri is 

323: well-modeled by a 9500 K white dwarf, and this white dwarf provides no more 

324: than 10\% of the flux in the {\it K}-band. If we assume that 

325: the secondary star is responsible for the remainder of the {\it K}-band flux 

326: at phase 0.0 ({\it K} = 15.55), the absolute magnitude of EF 

327: Eri is in the range 10.0 $\leq$ M$_{\rm K} \leq$ 11.0. This absolute 

328: magnitude corresponds to a spectral type in the range M8 to L1 (ignoring any 

329: effects due to the Roche lobe filling nature of this object).

330: 

331: A secondary star with a spectral type near M9 should have strong

332: CO features, as well as a detectable Na I doublet. We have compared the 

333: combined {\it H}- and {\it K}-band spectra of EF Eri with published spectra

334: for stars and brown dwarfs in the range M7V to T8. While the overall 

335: slope of the continuum of the phase 0.0 {\it K}-band spectrum

336: is consistent with a very late-type M dwarf (see Fig. 6), the {\it H}-band

337: spectrum bears no resemblence to such an object. The departure 

338: becomes more pronounced when we compare the spectrum of EF Eri 

339: to those of L-dwarfs. While brown dwarfs around L5 show a similar spectral 

340: slope at

341: 1.55$-$1.6\,$\mu$m, there is no feature comparable to the strong flux

342: peak at 1.7\,$\mu$m. While the absence of Na\,I in the $K$-band is 

343: consistent with a spectral type later than L0 (McLean et al. 2003), the

344: water vapor absorption bands of such an object are much deeper and

345: depress the blue side of the $K$-band spectrum much further to the red

346: ($>$\,2.1\,$\mu$m) than is seen in EF Eri. Also, the first overtone CO bandhead at 2.29\,$\mu$m 

347: remains strong in brown dwarfs until CH$_4$ takes over in the late L types.

348: 

349: In addition to the difference in $H$- and $K$-band spectral features,

350: the $JHK$ photometry is almost impossible to reconcile with any known

351: dwarf. While the {\it H} to {\it K}-band flux ratios for

352: EF Eri are more-or-less consistent with those of late type objects, giving

353: rise to ($H - K$) = 0.42 ($\approx$ M8V), the observed ($J  -  K$) color 

354: ($>$ 2.5) exceeds even the reddest IR colors observed in mid-L

355: dwarfs (Legget et al. 2002). 

356: 

357: \subsubsection{Model Spectra for Carbon/Oxygen-depleted Atmospheres}

358: 

359: As shown in Harrison et al. (2004), many CV secondary stars show evidence

360: for weak CO absorption features. Since the water vapor features in those

361: objects appear to be relatively normal, Harrison et al. suspect that

362: Carbon is deficient in those systems.  This view is supported by analysis

363: of FUV observations of CVs by G$\ddot{a}$nsicke et al. (2003) where 

364: large N V/C IV ratios have been found, suggesting Nitrogen enhancements, and 

365: Carbon

366: deficits in the photospheres of the white dwarf primaries, presumably

367: arising from matter transferred from the secondary star. It would therefore

368: not be unexpected for the secondary star in EF Eri to have anomalously

369: weak CO absorption features. A lowered Carbon abundance would also reduce the

370: CH$_4$ absorption expected for late-L and T dwarfs, and thus could 

371: produce much redder ($J\!-\!K$) colors through the range of

372: possible spectral types for the secondary of EF Eri.

373: 

374: To test this hypothesis we have calculated theoretical spectra using

375: the general stellar atmosphere code \texttt{PHOENIX} (Hauschildt et al. 1999)

376: for a variety of chemical compositions that might be expected in a

377: CNO-processed stellar core.

378: Depth-dependent partial pressures for more than 650 species are

379: calculated using the EOS of Allard \& Hauschildt (1995) by solving chemical

380: equilibrium for 40 elements, as described in Allard et al. (2001), including 

381: dust formation.

382: Model structures and flux spectra are computed self-consistently,

383: treating line blanketing by the direct opacity sampling method, with

384: line strengths taken from a master database of 43 million atomic and 700 

385: million molecular lines. In addition, dust extinction profiles for 31 grain

386: species are included.

387: Major updates relative to the AMES-Cond and AMES-Dusty models

388: of Allard et al. (2001) are the addition of 30 million methane lines from

389: Homeier et al. (2004a), and the explicit treatment of the gravitational settling

390: of dust grains developed by Allard et al. (2003).

391: In addition, the Ames list of water line opacities from

392: Partridge \& Schwenke (1997) has been replaced by the earlier list from

393: Miller et al. (1994) for these models. The latter

394: list is less complete for high temperature steam, but has been found

395: to give better agreement with the IR spectra of cool brown dwarfs

396: (Homeier et al. 2004b).

397: Nevertheless we emphasize that none of the currently available

398: databases for molecular opacities accurately reproduces the observed

399: shape of all absorption bands in the IR. These models also suffer from

400: incomplete or missing data for several molecules, such as CrH

401: and FeH. For example, the $H$-band spectra of mid-L dwarfs show

402: a moderately strong absorption feature near 1.6\,$\mu$m (see Fig. 6) that

403: is not reproduced by any of our models. Several features in this

404: region have been identified by Cushing et al. (2003) with the

405: $E^4\Pi$--$A^4\Pi$ system of FeH, for which no line opacity data are

406: available at this time.

407: Possibly due to this missing opacity source, and other uncertainties

408: in input data such as H$_2$O opacities and dust extinction

409: profiles, our theoretical spectra tend to produce

410: ($H\!-\!K$) colors that are too blue when compared to observed L dwarfs

411: (Legget et al. 2001).

412: 

413: We have constructed grids of models based on revised solar abundances

414: as compiled by Lodders (2003), except for Carbon, Oxygen and

415: Nitrogen, which were altered as indicated for the individual

416: models. The basic atmospheric parameters span the range

417: 1400~K $\leq$ T$_{\rm eff}$ $\leq$ 2500~K, and

418: log\,$g$\,=\, (4.5, 5.0, 5.5). First, investigating the effects of

419: lowering the Carbon abundance to [C/Fe]\,=\,($-$0.5, $-$1.0, $-$1.5),

420: we find that a depletion to at least 0.1\,$\times$\,solar is required

421: to explain the weakness or absence of the CO overtone bands.

422:  From synthetic photometry we can also exclude models with

423: $T_{\rm eff}$ significantly higher than 2000\,K, as these disagree

424: with the observed $J$ magnitudes even assuming the upper limit of 5.5

425: for the distance modulus. Below 1600\,K, the strong methane

426: bands at 2.2\,--\,2.3\,$\mu$m become prominent even at

427: [C/Fe]\,=\,$-$1.0, but lowering the Carbon abundance further does not

428: improve the fit to the observed $K$-band spectrum  at

429: such low temperatures. We find best agreement for the red part of the

430: $K$-band spectrum for models with 1600 $\leq$ T$_{\rm eff}$ $\leq$ 1800\,K, as 

431: shown by the

432: bottom spectrum in Fig. 7. But a Carbon deficiency alone cannot explain the

433: early onset of absorption to the blue side of the $K$-band, which appears at

434: much shorter wavelengths than typically seen in both our models and

435: in observations of L dwarfs. Nor can a lack of Carbon be responsible for

436: the peculiar shape of the $H$-band spectra. The

437: main opacity source at these wavelengths is water vapor, and

438: significant changes can only be expected if the Oxygen abundance is

439: reduced as well. Such effects might be expected if the secondary CNO

440: cycle has been active for some time, though models by Marks \&

441: Sarna (1998) do not indicate significant changes in the Oxygen

442: abundance for CNO processed material. We have nonetheless produced test models 

443: for [O/Fe]\,=\,$-$0.5 to explore this case.

444: As one would expect, these models (Fig. 7) show

445: an onset of the water vapor bands both in the $H$ and $K$-band at

446: shorter wavelengths, with a $K$-band spectrum that fits our

447: observations quite well. For comparison, a model with additionally

448: enhanced Nitrogen abundance, as would be expected with the depleted

449: Carbon and Oxygen transformed into $^{14}$N, is also shown. 

450: 

451: Synthetic

452: colors calculated for these models show that the Nitrogen-enriched

453: atmosphere produces the closest match to the ($J\!-\!K$) and

454: ($H\!-\!K$) photometry. We have not attempted

455: to modify our input chemistry further to optimize these fits, since

456: such results would necessarily be of limited significance given the

457: uncertainties in the available opacity data and the limited wavelength

458: coverage of our spectra. Clearly a direct detection of the water absorption in

459: the $H$-band, which we expect to occur at 1.5\,--\,1.55\,$\mu$m, and

460: $J$-band spectra, would help in resolving these uncertainties. Such data

461: would also allow for the study of irradiation, which is expected to have

462: the strongest impact at shorter wavelengths (Barman et al. 2000).

463: In conclusion, a brown dwarf with a considerable deficiency of Carbon,

464: a lesser depletion of Oxygen, and an enhancement of Nitrogen,

465: reproduces much of the secondary's IR SED, and the shape of the

466: $K$-band at phase 0.0. The additional absorption at 1.6\,$\mu$m cannot be 

467: accounted for in the current models. If this band is indeed due to FeH, we 

468: expect it to become relatively stronger when the steam bands are reduced.

469: In that case the spectra could probably reproduce the ``base'' of

470: the $H$-band emission. The shape of the actual peak, however, does not

471: correspond to any known stellar features, and also shows changes with

472: phase that cannot be explained by irradiation effects.

473: We conclude that cyclotron emission at 1.6\,--\,1.8\,$\mu$m

474: contaminates the infrared spectrum for all orbital phases.

475: 

476: \subsection{The Cyclotron Features}

477: 

478: It is clear that the {\it H}-band spectra are inconsistent with that of a 

479: normal (or chemically altered) late-type or brown dwarf secondary star. 

480: If the phase 0.0 dip near 1.6 $\mu$m is associated with FeH, then we estimate

481: that $\approx$ 30\% of the $H$-band flux at this juncture is due to cyclotron

482: emission.  Thus, to produce the observed, phase-dependent changes in

483: the {\it H}-band spectra probably requires two separate cyclotron features 

484: in this bandpass. If we assume the phasing and geometry from 

485: Bailey et al. (1982), the phase 0.0 cyclotron features that we see in the

486: {\it H}-band are 

487: presumably associated with the pole that accretes during the high state. This 

488: suggests that the single, dominant cylcotron feature at phase 0.5 might be from 

489: the other pole.

490: 

491: While the minimum light {\it K}-band spectrum is fairly consistent with 

492: the shape of a cool star spectrum, with little evidence for cyclotron

493: emission, the maximum light spectrum is clearly contaminated

494: by at least one, possibly two cyclotron features. Obviously, there is a 

495: cyclotron harmonic that peaks below 2.0 $\mu$m. The rapid change in slope of

496: the red half of the $K$-band spectrum, and the dip near 2.07 $\mu$m, suggest

497: the possibility of an additional, weaker cyclotron feature at phase 0.5. 

498: To better examine the phase 0.5 cyclotron features, we have subtracted the 

499: phase 0.0 spectrum from the phase 0.50 spectrum, to produce the residual 

500: spectrum shown in Fig. 8. The result shows two strong cyclotron features, one

501: that peaks in the {\it H}-band near 1.58 $\mu$m, and one that (we estimate) 

502: peaks near 1.93 $\mu$m. The possibility of a second cyclotron feature in the

503: {\it K}-band seems more likely after this subtraction. If the secondary star

504: has a similar spectrum at phases 0.0 and 0.5, then any water vapor feature

505: should disappear upon subtraction. Ignoring the false H I Br-$\gamma$ emission

506: feature, there continues to be a modest dip near 2.07 $\mu$m and a small

507: rise that peaks near 2.15 $\mu$m and that falls off slowly beyond 2.2 $\mu$m. 

508: 

509: \section{Discussion}

510: Regardless of the exact interpretation of the spectra presented here, it is

511: clear that the model proposed by Harrison et al. (2003) is incorrect. The

512: photometric modulations in the {\it K}-band are certainly being driven by the 

513: changing strength of the cyclotron feature at the blue end of this bandpass

514: (and possibly aided by a broader, and weaker feature that peaks near 2.15 

515: $\mu$m).  There is very little evidence for the secondary star in our 

516: {\it K}-band spectra, with only the suggestion of a weak first overtone CO 

517: feature in the median of the {\it K}-band data set.

518: While the presence of weak CO features in a CV is not surprising (see Harrison 

519: et al.  2004), stars with the late spectral type expected for the secondary in 

520: EF Eri have {\it H}-band spectra that are completely different to what

521: we have observed. While model brown dwarf spectra with dramatic alterations in 

522: the abundances of the major opacity sources at these wavelengths do a somewhat

523: better job at fitting the data, neither the observations or the models 

524: are currently adequate enough to truly investigate the nature of the underlying

525: secondary star. 

526: 

527: The fact that cyclotron emission can still dominate the spectrum of EF Eri

528: in this low state is surprising given that EF Eri exhibits no outward signs of 

529: continued accretion. The optical and infrared spectra lack Balmer 

530: and Brackett emission from H I. The optical {\it BVRI} spectral energy 

531: distribution and spectrum are completely consistent with that of an inactive

532: magnetic white dwarf. The model white dwarf atmosphere, with T$_{\rm eff}$ =

533: 9500 K, used by Harrison et al. (2003) to compare with their 2001 March 11

534: spectrum of EF Eri, does an excellent job at fitting the 2002 August 10 optical

535: spectrum we obtained with the NTT. Clearly, the white dwarf photosphere alone 

536: is too cool to supply significant quantities of ionizing photons. It appears

537: that the white dwarfs in ultra-short period, low mass transfer rate CVs are

538: significantly cooler than those in longer period, or higher mass transfer rate

539: systems (c.f., Szkody et al. 2002).

540: 

541: Four magnetic CV systems have now been found that display a single,

542: dominant cyclotron feature in their optical spectra that have

543: been attributed to very low mass transfer rates: SDSS 

544: J155331.12+551614.5 and 

545: SDSS J132411.57+032050.5 (Szkody et al. 2003), RBS0206 (= CQ Cet; Schwope, 

546: Schwarz, \& Greiner 1999), and HS 1023+3900 (= WX LMi; Reimers, Hagen \& 

547: Hopp 1999). Along with the single very large cyclotron feature is evidence for

548: a second, weaker feature at the next lower/higher harmonic number covered by

549: their optical spectra. Of these four polars, both SDSS J155331.12+551614.5 and 

550: J132411.57+032050.5 have optical light curves that resemble the

551: infrared light curves of EF Eri (see Figs. 2 and 6 of Szkody et al. 

552: 2003). SDSS1553 shows optical variations with peak-to-peak amplitudes of

553: $\Delta$m $\approx$ 1 mag in {\it V} and {\it R}, while SDSS1324 has variations 

554: in the Sloan {\it r}-band of $\Delta r$ = 1.3 mags. Both sets of light curves

555: are roughly sinusoidal, like the {\it H} and {\it K}-band light curves of

556: EF Eri. Unlike EF Eri, however, the spectra of both SDSS1553 (Fig. 1 in

557: Szkody et al. 2003) and SDSS1324 (G. Schmidt, priv. comm.) reveal H I emission.

558: 

559: The interpretation of the spectra of SDSS1553, SDSS1324, CQ Cet and WX LMi

560: is that they have entered phases of extremely low accretion rates,

561: the so-called ``bombardment solution'' (Kuijpers \& Pringle 1982), where any 

562: heating is from particle collisions instead of shocks. This state

563: is characterized by low plasma temperatures, and mass transfer rates

564: of \.{M} $\leq$ 10$^{\rm -13}$ M$_{\sun}$ yr$^{-1}$. In each of these

565: four polars, the harmonic numbers of the dominant cyclotron features are in the

566: range 2 $\leq  n  \leq$ 4, with magnetic field strengths above 45 MG.

567: Magnetic field strength estimates for EF Eri

568: range from 10 to 21 MG. Assuming B $\sim$ 14 MG, 

569: the $n$ = 4 harmonic would peak at 1.96 $\mu$m, and the $n$ = 5 harmonic 

570: would peak at 1.58 $\mu$m. The $n$ =3 harmonic would peak near 2.6 $\mu$m,

571: while the $n$ = 6 and 7 harmonics would appear in the {\it J}-band. It

572: is worth noting that during the {\it H}- and {\it K}-band maxima, the 

573: {\it J}-band light curve shows a minimum. Given this fact, and the need

574: for very high accretion rates to produce power in these higher harmonics,

575: it is likely that the $n$ = 6 and 7 harmonics are not significant sources of 

576: flux in EF Eri. The fact that the {\it J}-band light curve has a maximum 

577: during the {\it H}- and {\it K}-band 

578: minima suggests activity from another pole with a significantly higher

579: magnetic field strength. The flux peak at 1.72 $\mu$m, and the upturn at the 

580: blue end of the phase 0.0 {\it H}-band 

581: spectra, along with the lack of significant cyclotron emission in the $K$-band 

582: indicate that the phase 0.0 cyclotron emission comes from a region with a 

583: significantly different field strength.  A fit to the cyclotron features of

584: the high state $JHK$ spectra of EF Eri by Ferrario et al. (1996) suggested 

585: emission from two regions with magnetic field strengths of 16.5 and 21 MG.

586: An accreting pole with a field strength of 21 MG, would not produce a 

587: cyclotron feature in the $K$-band, but would produce harmonics at 1.7 and 

588: 1.3 $\mu$m, consistent with our phase 0.0 spectra.

589: 

590: That the magnetic field structure of EF Eri is complex, including strong

591: evidence for emission from more than one pole, has been demonstrated by 

592: Hutchings et al. (1982), Piirola et al. (1987), Meggit \& Wickramasinghe

593: (1989), and Ferrario et al. (1996).

594: Recent ``magnetic field topology'' mapping of the magnetic field of the

595: white dwarf in EF Eri by Reinsch et al. (2003) reveals this complexity:

596: there are two main polar caps with field strengths as high 100 MG

597: at their centers, as well as evidence for multiple, zonal components.

598: The complex structure of the field in EF Eri must be partly to blame

599: for the results presented here. Modeling the cyclotron features that would

600: arise from accretion onto such a field would be helpful in unraveling our

601: spectroscopic results.

602: 

603: The lack of detectable H I emission from EF Eri, when compared to the other

604: members of its small family, is interesting. It is highly likely 

605: that the current accretion rate in EF Eri is lower than the other four 

606: objects, and is simply not sufficient to supply the heating necessary to 

607: generate the UV photons required to create H I emission. In the bombardment

608: scenario, the accretion shock disappears, and the white dwarf photosphere

609: is directly heated by particle collisions. A lower accretion rate means less 

610: heating, and therefore less ionizing radiation.

611: 

612: The secondary star of EF Eri remains elusive. Beuermann et al. (2000)

613: have presented arguments that the spectral type of the secondary star

614: in EF Eri must be later than M9, with the most probable spectral type

615: later than L3. We can estimate the maximum amount of flux from the

616: secondary by assuming that the flux level of the dip seen in the phase 0.0 

617: {\it H}-band spectrum (at 1.57 $\mu$m) represents the continuum level of

618: the underlying secondary star. This corresponds to {\it H}$_{\rm sec} \geq$ 

619: 16.5. Assuming 80 $\leq$ d $\leq$ 128 pc, the absolute {\it H} magnitude 

620: of the secondary star is in the range 11.0 $\leq$ M$_{\rm H}$ $\leq$ 12.0,

621: consistent with an early/mid-L dwarf. Thus, EF Eri joins WZ Sge (see Howell

622: et al. 2004) as the cataclysmic variables with the strongest evidence for 

623: harboring brown dwarf-like secondary stars.

624: 

625: \section{Conclusions}

626: 

627: We have obtained phase-resolved infrared spectroscopy of EF Eri and conclude

628: that the large amplitude variations seen in its infrared light curve are not 

629: due to a heated brown dwarf, but has its origin in cyclotron emission. EF Eri

630: appears to join the small family of polars where the accretion rate has 

631: dropped to a very low level. 

632:  Unfortunately, our data are inadequate to allow us unravel the 

633: complex cyclotron emission that is present in EF Eri. 

634: Phase-resolved {\it J}-band spectroscopy is required to make further progress

635: on disentangling the cyclotron spectrum of EF Eri. 

636: Given that the mean {\it J}-band magnitude of EF Eri is 17.3, multiple

637: orbits of data on 8 m-class telescopes will be needed to obtain spectra with

638: sufficient S/N. To fully interpret the spectroscopic observations,

639: phase-resolved, broad-band {\it JHK} polarimetric observations of EF Eri will 

640: be essential.  

641: It would also be extremely useful to have a radial velocity curve for the EF 

642: Eri system to allow us to properly phase, and coadd the infrared spectra to 

643: search for features from the secondary star. Such

644: data could be acquired through optical spectroscopy of the 

645: Zeeman absorption features of H I. 

646: 

647: \acknowledgements{TEH acknowledges partial support under NSF grant

648: AST 99-86823. We would like to thank T. Geballe for his help in planning

649: and executing our NIRI observations, and R. Campbell and G. Wirth for

650: assistance with NIRSPEC. DH acknowledges support from the National

651: Science Foundation under NFS grant N-Stars RR185-258, and from NASA

652: under grant NLTE RR185-236. Models presented in this work are based in part on 

653: calculations performed at the NERSC IBM SP at LBNL with support from the DoE.

654: We would also like to thank F.~Allard and P.~Hauschildt for assistance in

655: construction of the model atmospheres, T.~Barman for helpful discussions

656: on the effects of irradiation, and I.~Baraffe for providing us with results of

657: evolutionary calculations for interacting brown dwarfs. We also would like

658: to thank our referee D. Wickramasinghe, and editor J. Leibert for additional

659: insight on EF Eri, and for suggesting changes to improve the clarity of our

660: paper.  Based partly on observations obtained at the Gemini 

661: Observatory, which is operated by the Association of Universities for Research 

662: in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of 

663: the Gemini partnership: the National Science Foundation (United States), the 

664: Particle Physics and Astronomy Research Council (United Kingdom), the

665: National Research Council (Canada), CONICYT (Chile), the Australian Research 

666: Council (Australia), CNPq (Brazil) and CONICET (Argentina). Additional

667: data presented herein were obtained at the W.M. Keck Observatory, which is 

668: operated as a scientific partnership among the California Institute of 

669: Technology, the University of California and the National Aeronautics and 

670: Space Administration. The Observatory was made possible by the generous 

671: financial support of the W.M. Keck Foundation. The authors wish to recognize 

672: and acknowledge the very significant cultural role and reverence that the 

673: summit of Mauna Kea has always had within the indigenous Hawaiian community.  

674: We are most fortunate to have the opportunity to conduct observations from 

675: this mountain.

676: 

677: \begin{flushleft}

678: {\bf References}

679: 

680: Allard, F. \& Hauschildt, P.~H. 1995, ApJ, 445, 433\\

681: Allard, F., Hauschildt, P., Alexander, D., Tamanai, A., \& Schweitzer, A. 

682:   2001, \apj, 556, 357\\

683: Allard, F., Guillot, T., Ludwig, H., Hauschildt, P.~H., Schweitzer,

684: A., Alexander, D.~R., \& Ferguson, J.~W. 2003, in IAU Symposia, Vol. 211,

685: {\it Brown Dwarfs}, ed. E.~Mart{\'i}n, International Astronomical Union (San

686: Francisco: Astronomical Society of the Pacific), p. 325\\

687: Bailey, J., Hough, J. H., Axon, D. J., Gatley, I., Lee, T. J., Szkody, P.,

688: Stokes, G., \& Berriman, G. 1982, MNRAS, 199, 801\\

689: Barman, T. S., Hauschildt, P.~H., \&  Allard, F. 2002, in ASP Conf. Ser.

690:  261: The Physics of Cataclysmic Variables and Related Objects, (San

691:  Francisco: Astronomical Society of the Pacific), p49\\

692: Beuermann, K., Wheatley, P., Ramsay, G., Euchner, F., \& Gansicke, B. T.

693: 2000, A\&A, 354, L49\\

694: Cushing, M.~C., Rayner, J.~T., Davis, S.~P., \& Vacca, W.~D. 2003,

695: \apj, 582, 1066\\

696: Ferrario, L., Bailey, J., \& Wickramasinghe, D. 1996, MNRAS, 282, 218\\

697: Harrison, T. E., Howell, S. B., Huber, M. E., Osborne, H. L., Holtzman, J. A.,

698: Cash, J. L., \& Gelino, D. M. 2003, AJ, 125, 2609\\

699: Harrison, T. E., Osborne, H. L., Howell, S. B., 2004, AJ, 127, 3493\\

700: Hauschildt, P. H., Allard, F., \& Baron, E. 1999, ApJ, 512, 377\\

701: Homeier, D., Allard, F., Hauschildt, P., \& Boudon, V. 2004a, \apj, in prep.\\

702: Homeier, D., Burgasser, A., Hauschildt, P., Allard, F., Allard, N.,  

703:   McLean, I., McGovern, M., Kirkpatrick, J., \& Prato, L. 2004b, \apj, to be  

704:   submitted \\

705: Howell, S. B., Harrison, T. E., \& Szkody, P. 2004, ApJ, 602, L49\\

706: Howell, S. B., Nelson, L. A., \& Rappaport, S. 2001, ApJ, 550, 897\\

707: Hutchings, J. B., Cowley, A. P., Crampton, D., Fisher, W. A., \& Liller, M. H. 

708: 1982, ApJ, 252, 690\\

709: Kuijpers, J., \& Pringle, J. E. 1982, A\&A, 114, L4\\

710: Lancon, A., \& Rocca-Volmerange, B. 1992, A\&A Supp., 96, 593\\

711: Leggett, S.~K., Allard, F., Geballe, T.~R., Hauschildt, P.~H., \&

712:   Schweitzer, A. 2001, \apj, 548, 908\\

713: Lodders, K. 2003, ApJ, 591, 1220\\

714: Maiolino, R., Rieke, G. H., \& Reike, M . J. 1996, AJ, 111, 537\\

715: Marks, P. B., \& Sarna, M. J. 1998, MNRAS, 301, 699\\

716: McLean, I.S., McGovern, M.R., Burgasser, A.J., Kirkpatrick, J.D., Prato, L., 

717:   \& Kim, S.S. 2003, ApJ, 596, 561\\

718: Meggitt, S. M. A., \& Wickramasinghe, D. T. 1989, MNRAS, 236, 31\\

719: Miller, S., Tennyson, J., Jones, H., \& Longmore, A. 1994, in IAU  

720:   Colloquia, Vol. 146, Molecular Opacities in the Stellar Environment, ed.  

721:   P.~Thejll \& U.~G. J{\o}rgensen, International Astronomical Union (Niels Bohr 

722:   Institute and Nordita press, Copenhagen), 296\\

723: Partridge, H. \& Schwenke, D.~W. 1997, \jcp, 106, 4618\\

724: Piirola, V., Reiz, A., \& Coyne, G. V., 1987, A\&A, 186, 120\\

725: Reimers, D., Hagen, H., \& Hopp, U. 1999, A\&A, 343, 157\\

726: Reinsch, K., Euchner, F., \& Beuermann, K. 2003, astro-ph/0302056\\

727: Schwope, A. D., Schwarz, R., \& Greiner, J. 1999, A\&A, 348, 861\\

728: Szkody, P. et al. 2003, ApJ, 583, 902\\

729: Szkody, P., Gansicke, B. T., Sion, E. M., \& Howell, S. B. 2002, ApJ, 574, 950\\

730: Thorstensen, J. R. 2003, AJ, 126, 3017\\

731: Wickramasinghe, D. T., \& Ferrario, L. 2000, PASP, 112, 873\\

732: \end{flushleft}

733: 

734: \begin{figure}

735: \epsscale{0.95}

736: \plotone{f1.eps}

737: \caption{The infrared light curves of EF Eri obtained with SQIID on the

738: KPNO 2.1 m. The absolute fluxes and light curve amplitudes have not changed 

739: when compared to those presented in Harrison et al. (2003). The data

740: sets have been plotted over two orbital cycles for clarity.}

741: \end{figure}

742: 

743: \begin{figure}

744: \epsscale{0.85}

745: \plotone{f2.eps}

746: \caption{An optical spectrum of EF Eri obtained with EMMI on the {\it NTT} on 

747: 2002 August 10. The

748: accretion rate in EF Eri has declined to the point where there is no longer

749: detectable emission at H$\alpha$. The $\sigma^{\pm}$ features from the Zeeman

750: splitting of the H I lines, however, are clearly seen. The central,

751: unshifted absorption components of H I are weaker than expected, suggesting

752: the possibility of continued low-level emission.}

753: \end{figure}

754: 

755: \begin{figure}

756: \epsscale{0.90}

757: \plotone{f3.eps}

758: \caption{The Keck NIRSPEC {\it H-}band, and Gemini NIRI {\it K}-band, spectra

759: of EF Eri. Both data sets have been flux-calibrated using the infrared

760: light curves. For clarity, the {\it H}- and {\it K}-band spectra for each 

761: orbital phase have been offset in flux. The mean orbital

762: phases of the spectra are listed.}

763: \end{figure}

764: 

765: \begin{figure}

766: \epsscale{0.95}

767: \plotone{f4.eps}

768: \caption{The NIRSPEC {\it K}-band phase-resolved spectra of EF Eri. These

769: data have been flux-calibrated using the infrared light curves and then

770: offset for clarity. The same phase-dependent morphology seen in the Gemini 

771: data set is present here. Each spectrum is the median of several

772: three minute exposures.}

773: \end{figure}

774: 

775: \begin{figure}

776: \epsscale{0.95}

777: \plotone{f5.eps}

778: \caption{The median of the entire NIRSPEC {\it K}-band data set (top, offset

779: in flux),

780: and a smoothed (to FWHM = 60 \AA) median of the seven spectra closest to phase 

781: 0.0 (bottom). A possible, very weak

782: absorption feature at the position of the first overtone of 

783: $^{\rm 12}$CO$_{\rm (2,0)}$  (at 2.294 $\mu$m) is seen in both spectra,

784: and appears to be the only indication of the presence of the secondary star, 

785: besides the decline at the red end of the {\it K}-band that might be associated with water vapor 

786: absorption.}

787: \end{figure}

788: 

789: \begin{figure}

790: \epsscale{0.95}

791: \plotone{f6.eps}

792: \caption{Comparison of the phase 0.0 spectrum of EF Eri to those of

793: M9 and L5 dwarfs (from McLean et al. 2003). The sharp feature at 2.294 $\mu$m

794: in the comparison spectra is due to the first overtone feature of CO.}

795: \end{figure}

796: 

797: \begin{figure}

798: \epsscale{0.80}

799: \plotone{f7.eps}

800: \caption{Comparison of the phase 0.0 spectrum of EF Eri to brown dwarf model

801: spectra generated using \texttt{PHOENIX}. The temperature of each model is listed,

802: and each of these models has a gravity of log{\it g} = 5.0.

803: Below the temperature are the abundances of Carbon, Nitrogen, and Oxygen (in

804: dex, vs. solar). The top model has deficits of Carbon and Oxygen, and enhanced

805: levels of Nitrogen (4$\times$ solar). The bottom-most model simply has

806: a Carbon deficit (0.1$\times$ solar). The Oxygen deficient models fit the

807: observed {\it K}-band spectrum due to their shallower water vapor absorption.

808: None of the models we developed, however, could explain the shape of the

809: {\it H}-band spectrum of EF Eri.}

810: \end{figure}

811: 

812: \begin{figure}

813: \epsscale{0.95}

814: \plotone{f8.eps}

815: \caption{The residual spectrum created by subtracting the phase 0.0 spectra

816: from the phase 0.5 spectra. The result is two apparent cyclotron humps

817: that peak near 1.58 and 1.93 $\mu$m.}

818: \end{figure}

819: \end{document}

820: