1: \documentclass[11pt,preprint]{aastex}

2: 

3: \def\ltsima{$\; \buildrel < \over \sim \;$}

4: \def\lsim{\lower.5ex\hbox{\ltsima}}

5: \def\gtsima{$\; \buildrel > \over \sim \;$}

6: \def\gsim{\lower.5ex\hbox{\gtsima}}

7: 

8: \slugcomment{}

9: 

10: \shorttitle{Limits on eclipses of KH~15D} 

11: \shortauthors{Winn, Garnavich, Stanek, \& Sasselov}

12: 

13: \begin{document}

14: 

15: \title{ Limits on eclipses of the pre-main-sequence star KH~15D\\ in

16: the first half of the 20th century }

17: 

18: \author{

19: Joshua N.\ Winn\altaffilmark{1,2},

20: Peter M.\ Garnavich\altaffilmark{3},

21: K.~Z.\ Stanek\altaffilmark{1},

22: Dimitar D.\ Sasselov\altaffilmark{1}

23: }

24: 

25: \email{jwinn@cfa.harvard.edu; pgarnavi@miranda.phys.nd.edu;

26: kstanek@cfa.harvard.edu; sasselov@cfa.harvard.edu}

27: 

28: \altaffiltext{1}{Harvard-Smithsonian Center for Astrophysics, 60

29: Garden st., Cambridge, MA 02138}

30: 

31: \altaffiltext{2}{National Science Foundation Astronomy \& Astrophysics

32: Postdoctoral Fellow}

33: 

34: \altaffiltext{3}{University of Notre Dame, Notre Dame, IN 46556}

35: 

36: \begin{abstract}

37: Over the last decade, the pre-main sequence star KH~15D has exhibited

38: periodic eclipses that are surprisingly deep ($\sim$3~mag) and

39: long-lasting ($\sim$40\% of the 48.4-day period). The cause of the

40: eclipses is unknown, but it could be a feature in a nearly edge-on

41: protoplanetary disk.  Here we report on an analysis of archival

42: photographs of KH~15D from the Harvard College Observatory plate

43: collection, most of which were taken during the years

44: 1913-1951. During this time range, the data are consistent with no

45: eclipses; the duty cycle of 1~mag eclipses was less than 20\%. The

46: decadal timescale of this change in eclipse behavior is compatible

47: with the expected timescale of protoplanet/disk interactions. Archival

48: images from more recent epochs should reveal the onset of the

49: eclipses.

50: \end{abstract}

51: 

52: \keywords{ stars: pre-main sequence --- stars: individual (KH~15D) ---

53: open clusters and associations: individual (NGC~2264) --- planetary

54: systems: protoplanetary disks }

55: 

56: \section{Introduction}

57: \label{sec:intro}

58: 

59: The unique eclipses of the pre-main sequence star KH~15D have recently

60: gained attention because of the possibility that they are related to

61: processes of planet formation.  The eclipses were discovered by

62: \citet{kh98}\footnote{The name KH~15D was derived from their initials

63: and the star's catalog number in field D of \citet{keh+97}.} during a

64: variability study of the young cluster NGC~2264. The optical spectrum

65: of KH~15D is that of a T Tauri star \citep{hhs+01,hhm+03}, a category

66: believed to represent a late stage in the accretion and dispersal of

67: circumstellar material and the condensation of planets \citep{b89}.

68: \citet{hhs+01} estimated the age of the star to be 2--10~Myr, from its

69: position on a color-magnitude diagram.

70: 

71: During the eclipses, which recur every 48.36 days and last $\approx$20

72: days, the star plummets in brightness by over 3 magnitudes in a manner

73: consistent with a knife-edge crossing the face of the star (see Herbst

74: et al.\ 2002 for a striking presentation and discussion of the light

75: curves). The obscuring matter must be composed of large dust grains or

76: macroscopic objects, because small dust grains would cause reddening

77: of the starlight that is not observed. If the obscuring material is

78: orbiting the star, it must be spread out over a large fraction of the

79: orbit in order to explain the long duration of the

80: eclipses. Strangely, the star re-brightens near mideclipse, as if the

81: obscuring material is distributed symmetrically about a central

82: opening.

83: 

84: The eclipses are sufficiently dramatic that it appeared possible to

85: investigate their history in photographic plate archives, despite the

86: poor sensitivity and resolution of old photographs compared to modern

87: detectors. In \S~\ref{sec:data}, we describe our findings from the

88: Harvard College Observatory plate collection. In \S~\ref{sec:analysis}

89: we use the data to place limits on the eclipse properties of KH~15D in

90: the first half of the 20th century. Finally, in

91: \S~\ref{sec:discussion}, we place our results in the context of

92: previous measurements and previously offered theories for the origin

93: of the eclipses.

94: 

95: \section{Data}

96: \label{sec:data}

97: 

98: We searched the collection of the Harvard College Observatory, one of

99: the world's largest archives \citep{h94}, for plates of NGC~2264. The

100: most useful part of the archive was the ``MC series,'' one of the

101: highest-quality plate series in the collection.  It was obtained with

102: the 16-inch $f/5.2$ Metcalf Doublet refracting telescope, in Cambridge

103: and Oak Ridge, Massachusetts, between 1909 and 1988.  We found 66

104: suitable plates in the MC series.  In each case we digitized a

105: $30\arcmin\times 30\arcmin$ region surrounding the expected position

106: of KH~15D, using an optical scanner with a resolution of 3000

107: pixels~inch$^{-1}$ and 14 bits~pixel$^{-1}$.

108: 

109: We also found one suitable plate in the ``A series,'' taken with the

110: 24-inch $f/5.6$ Bruce Doublet in Bloemfontein, South Africa, on 1936

111: January 24. This plate could not be digitized due to its large

112: physical size ($14\times 17$ inches). However, the A plate was of much

113: higher quality than any of the MC plates, allowing the analysis

114: described below to be performed visually.

115: 

116: To determine the threshold of detectability of each plate, we chose 10

117: stars with a range in brightness of roughly 2 magnitudes bracketing

118: the magnitude of KH~15D in its uneclipsed state, and searched for

119: those 10 stars in each image. These reference stars are identified by

120: KH number in Fig.~1a, which is an image of the field from the Palomar

121: Observatory Sky Survey (POSS)\footnote{The National Geographic

122: Society---Palomar Observatory Sky Atlas (POSS-I) was made by the

123: California Institute of Technology with grants from the National

124: Geographic Society.}.

125: 

126: Next, we looked for KH~15D. This was not a trivial task because its

127: uneclipsed magnitude was usually close to the limiting magnitude. In

128: addition, a nearby bright star (HD~47887) was often overexposed to

129: such a degree that its ``halo'' of scattered light enveloped the

130: expected position of KH~15D. This was a particular problem in

131: blue-light observations (which formed the majority of cases) because

132: the bright star is very blue. In those cases, we employed the

133: technique of ``unsharp masking'' commonly used in astrophotography

134: \citep[see, e.g.,][]{mz79}. In this technique, one produces a blurred

135: image by convolving the original with a circular Gaussian tapering

136: function, and then subtracts a scaled version of the blurred image

137: from the original image. This enhances contrast by removing low

138: spatial frequencies, at the expense of an increased noise level. The

139: width of the Gaussian function, and the scaling of the blurred image,

140: are adjusted to achieve optimal results. Here the method works because

141: the image of HD~47887 is very broad and flat-topped, whereas images of

142: fainter stars such as KH~15D are more sharply peaked (see Figs.~1b and

143: 1c). Star KH~24D, being brighter than KH~15D and slightly further away

144: from HD~47887, provided a useful test of our ability to detect stars

145: close to the bright central star: if KH~24D could not be detected,

146: surely no useful information about KH~15D could be obtained from that

147: plate.

148: 

149: In this manner we determined that 40 of the plates had sufficient

150: sensitivity and resolution for KH~15D to have been detected when

151: uneclipsed. Among these, we securely detected KH~15D in 36 cases. In

152: the other 4 cases, we believe KH~15D is present, but we are not as

153: confident in the detection because it is right at the plate limit

154: (i.e.\ brighter comparison stars were detected, but fainter comparison

155: stars were not detected). We summarize the results as follows: in 90\%

156: of the cases for which detection of KH~15D was possible, it was

157: securely detected, and there were no secure non-detections.

158: 

159: We have not attempted accurate photometry because of the contaminating

160: influence of HD~47887. However, judging from the comparison stars, we

161: conclude that the detection of KH~15D on a given plate rules out an

162: eclipse deeper than 1~mag at that epoch.  Thus, the data are

163: consistent with no eclipses deeper than 1~mag.

164: 

165: \section{Analysis}

166: \label{sec:analysis}

167: 

168: Fig.~2 shows the time coverage of the 40 high-quality plates, and also

169: includes the POSS image (1951 December 1) and the 2 epochs of the

170: DPOSS-2 images\footnote{The Digitized Sky Surveys were produced at the

171: Space Telescope Science Institute under U.S. Government grant NAG

172: W-2166. The images of these surveys are based on photographic data

173: obtained using the Oschin Schmidt Telescope on Palomar Mountain and

174: the UK Schmidt Telescope. The plates were processed into the present

175: compressed digital form with the permission of these institutions.}

176: (1989 November 7 and 1992 December 21), on which KH~15D appears

177: uneclipsed. The data fall naturally into four groups in time, which we

178: have distinguished by arbitrary colors and symbols: black triangles,

179: for 1913--1915; red circles, for 1924--1939; green squares, for

180: 1946--1955; and blue diamonds, for 1978--1992. Solid symbols are

181: secure detections and open symbols are the ambiguous cases.

182: 

183: We use this data to place limits on the ``eclipse fraction'' $f$,

184: defined as the fraction of time that KH~15D is eclipsed by

185: $>$1~mag. The simplest approach is to ignore the modern ephemeris and

186: assume that the phase of each epoch is random. We also assume,

187: conservatively, that the 4 lower-confidence detections represent

188: eclipses. Then, using binomial statistics with a uniform prior on $f$,

189: the observation that KH~15D is uneclipsed in 39/43 cases implies

190: $f<0.20$ with 95\% confidence. The result changes to $f<0.24$ when

191: only the 1924--1939 data are used, and $f<0.17$ when only the

192: 1946--1955 data are used.

193: 

194: A more sophisticated approach is to use the modern ephemeris to

195: convert the time of each observation into $\phi$, the phase of the

196: photometric period. \citet{hhs+01} estimated the period to be

197: $48.34\pm 0.02$ days. After obtaining additional data, \citet{hhv+02}

198: revised the period to 48.36 days but did not give an error estimate.

199: We adopt the ephemeris of \citet{hhv+02},

200: \begin{equation}

201: {\rm JD (mideclipse)} = 2,452,352.26 + 48.36E,

202: \end{equation}

203: and an uncertainty of 0.02~days in the period. The phase coverage of

204: our observations is shown in Fig.~3, for the choices $P=48.34$, 48.36,

205: and 48.38 days. The dashed lines show the modern eclipse duration.

206: The dotted lines show the approximate phases of the mideclipse

207: re-brightenings, during which the star has sometimes been observed at

208: its uneclipsed brightness (or even slightly brighter).

209: 

210: The phase coverage is fairly complete and can be used to rule out

211: eclipses with the present characteristics. Suppose, for example, that

212: the past eclipses were nearly symmetric about mideclipse, as they are

213: today. We define the ``eclipse duty cycle'' $f'$ as the fraction of

214: the photometric period between the 1~mag levels of ingress and

215: egress. In general, $f'\geq f$ because of the possibility of

216: mideclipse re-brightenings. We determine the maximum phase of the

217: eclipses, $\phi_{\rm max}$, as the minimum $|\phi|$ among all the

218: secure detections, not counting detections with $|\phi| < 0.05$ (where

219: re-brightenings are expected). The resulting limit on the eclipse duty

220: cycle is $f'< 2\phi_{\rm max}$. To deal with the period uncertainty,

221: we determine $\phi_{\rm max}$ for all possible values of $P$, and then

222: compute a weighted histogram of the results, using a weighting

223: function that is Gaussian in $P$ with mean 48.36~days and

224: $\sigma=0.02$~days.

225: 

226: Using all the data, the result is $f'<0.14$ with 95\% confidence.

227: Using only the data from 1924--1939, the result is $f'<0.18$.  With

228: only the 1946--1955 data, the result is $f'<0.24$.  Neither the

229: earliest data nor the most recent data are constraining by themselves.

230: 

231: \section{Summary and discussion}

232: \label{sec:discussion}

233: 

234: We conclude that the deep eclipses of KH~15D observed today did not

235: occur in the first half of the 20th century with their present

236: characteristics.  The eclipses may have been shorter in duration or

237: shallower than 1~mag, or perhaps the mideclipse re-brightenings

238: extended to larger phases. Of course, a combination of these effects

239: or a more complex time history are also possible.

240: 

241: The simplest explanation is that the eclipses have grown in duration,

242: as has been observed over the last decade.  Fig.~4 shows the upper

243: limits on the eclipse duty cycle derived at the end of the previous

244: section, along with modern values that we estimated from the light

245: curves of \citet{hhv+02} and from unpublished data from 2002-3

246: obtained by a group led by one of the authors (P.G.).  Apparently a

247: change is occurring on decadal timescales, in either the features of

248: the obscuring material surrounding the KH~15D system, or in the

249: alignment of those features with our viewing angle.

250: 

251: What could explain both the eclipses and their recent onset?

252: \citet{hhs+01} and \citet{hhv+02} offered two hypotheses to explain

253: the eclipses: (1) KH~15D has a nearly edge-on protoplanetary disk, and

254: a warp or ridge in the disk periodically blocks the star. (2) KH~15D

255: has a higher-mass companion that is hidden by an edge-on circumstellar

256: disk. The observed star is eclipsed as its orbit carries it through

257: the projected plane of the disk. \citet{gt02} offered another

258: hypothesis: (3) KH~15D has a low-mass companion, and the resulting

259: accretion and outflow of circumstellar material creates an asymmetric

260: common envelope, with a large and dense region near the low-mass

261: companion. The orbital motion of the secondary causes periodic

262: occultations of the primary.

263: 

264: Hypothesis (2) seemed unlikely even before this work, because

265: \citet{hhv+02} and \citet{hhm+03} measured only a small radial

266: velocity shift ($3.3\pm 0.6$~km~s$^{-1}$) between widely separated

267: phases. Those authors also expressed concern about systematic effects

268: that may allow the true radial velocity shift to be zero. In addition,

269: we do not know how to make this hypothesis compatible with the recent

270: onset of the eclipses.

271: 

272: The low-mass companion of hypothesis (3) could be small enough to be

273: allowed by the current radial velocity measurements. \citet{gt02} do

274: not make any specific statements regarding the expected timescale of

275: changes in eclipse duration, but they do claim that their model can

276: explain decadal variations seen in other young photometrically active

277: stars such as UX Ori stars. This makes hypothesis (3) viable and

278: worthy of further investigation.

279: 

280: Hypothesis (1) is not only consistent with a small radial velocity

281: shift, but also naturally involves a physical timescale that is

282: compatible with our result. The viscous timescale at the orbital

283: distance of 0.2~A.U.\ (corresponding to the 48-day period) is 10--100

284: years. If protoplanet/disk interactions cause perturbations in the

285: disk, then one would expect these perturbations to evolve on this

286: timescale. All other timescales in the disk at 0.2~AU are much

287: shorter: the dynamical and cooling timescales are $\sim$2 days, while

288: a single-impulse perturbation would be damped in less than 1 month

289: \citep{dch+99}. Thus, protoplanet/disk interactions are an appealing

290: explanation of the KH~15D phenomenon and the observed evolution.

291: 

292: For the sake of providing a concrete and falsifiable example of such

293: an interaction, we indulge in further speculation, elaborating on a

294: suggestion by \citet{hhv+02}. Models of planet formation at larger

295: orbital distances predict density perturbations in the form of two

296: spiral waves: a leading wave extending to smaller orbital distances,

297: and a lagging wave extending to larger distances \citep[see,

298: e.g.,][]{blo+03}. The spiral waves extend more than a radian in

299: azimuth, and for a Jupiter-mass planet the density is enhanced by a

300: factor of 3-5 even at one pressure scale-height, $h$, above the

301: midplane. For a typical T Tauri disk, the optically thick portion of

302: the disk extends to a height of $H_{s}\approx 4h$ above the midplane

303: \citep{dch+99,js03}, corresponding to 0.02~A.U.\ at the estimated

304: orbital distance (0.2~A.U.) of the occulting feature. This thickness

305: is approximately equal to the diameter of the young star ($2R_s

306: \approx 4R_{\odot} = 0.02$~A.U.). The spiral shock waves induced by a

307: Jupiter-sized protoplanet are capable of increasing $H_{s}$ by more

308: than a factor of 2 and creating an elongated ridge along our line of

309: sight that is thick enough to block the central star

310: entirely. Moreover, as previously suggested \citep{hhv+02} and

311: recently demonstrated with three-dimensional computational models

312: \citep{blo+03}, the ridge will have a local depression at the position

313: of the protoplanet, which would explain the mid-eclipse

314: re-brightenings.

315: 

316: If this is correct, our na\"{\i}ve prediction is that the eclipse duty

317: cycle will increase to $\approx$0.75 and remain there for the next

318: decade, because in the simulations, the spiral waves cover a maximum

319: of $\approx$75\% of the orbit. Furthermore, if the true period is

320: actually 97 days, as suggested by \citet{hhv+02} based on small

321: differences between alternate eclipses, this scenario would be

322: wrong. We emphasize that this speculation was inspired by simulations

323: at much larger orbital distances; we are not aware of any detailed

324: simulations of protoplanet/disk interactions at 0.2~A.U.

325: 

326: Finally, we note that although the Harvard collection does not happen

327: to include many photographs of this field after 1960, many such

328: photographs undoubtedly exist. Because of its placement in a

329: scientifically interesting young cluster, and near a photogenic

330: nebula, there should be many images from 1960 onward that are of

331: sufficient quality to measure KH~15D.  A compilation of these data

332: should reveal the onset of the modern eclipse behavior.

333: 

334: \acknowledgments The authors wish to thank Alison Doane and Doug Mink

335: for valuable assistance with the plate archive. This material is based

336: upon work supported by the National Science Foundation under Grant

337: No.\ 0104347.

338: 

339: \begin{thebibliography}{}

340: 

341: \bibitem[Bate et al.(2003)]{blo+03} Bate, M.R., Lubow, S.H., Ogilvie,

342: G.I., \& Miller, K.A.\ 2003, \mnras, 341, 213

343: 

344: \bibitem[Bertout(1989)]{b89} Bertout, C.\ 1989, \araa, 27, 351

345: 

346: \bibitem[D'Alessio et al.(1999)]{dch+99} D'Alessio, P., Cant\'{o}, J.,

347: Hartmann, L., Calvet, N., \& Lizano, S.\ 1999, \apj, 511, 896

348: 

349: \bibitem[Grinin \& Tambovtseva(2002)]{gt02} Grinin, V.~P.\ \&

350: Tambovtseva, L.~V.\ 2002, Ast.\ Lett.\ 28, 601

351: 

352: \bibitem[Hamilton et al.(2001)]{hhs+01} Hamilton, C.M., Herbst, W.,

353: Shih, C., \& Ferro, A.J.\ 2001, \apj, 554, L201

354: 

355: \bibitem[Hamilton et al.(2003)]{hhm+03} Hamilton, C.~M., Herbst, W.,

356: Mundt, R., Bailer-Jones, C.~A.~L., \& Johns-Krull, C.~M.\ 2003,

357: preprint [astro-ph/0305477]

358: 

359: \bibitem[Hazen(1994)]{h94} Hazen, M.~L.\ 1994, IAU Symp.~161:

360: Astronomy from Wide-Field Imaging, 161, 365

361: 

362: \bibitem[Herbst et~al.(2002)]{hhv+02} Herbst, W.\ et al.\ 2002, \pasp,

363: 114, 1167

364: 

365: \bibitem[Jang-Condell \& Sasselov(2003)]{js03} Jang-Condell, H.\ \&

366: Sasselov, D.\ 2003, \apj, in press [astro-ph/0304330]

367: 

368: \bibitem[Kearns et al.(1997)]{keh+97} Kearns, K.~E., Eaton, N.~L.,

369: Herbst, W., \& Mazzurco, C.~J.\ 1997, \aj, 114, 1098

370: 

371: \bibitem[Kearns \& Herbst(1998)]{kh98} Kearns, K.~E.\ \& Herbst,

372: W.~H.\ 1998, \aj, 116, 261

373: 

374: \bibitem[Malin \& Zealey(1979)]{mz79} Malin, D.~F.~\& Zealey, W.~J.\

375: 1979, Sky \& Telescope, 57, 354

376: 

377: \end{thebibliography}

378: 

379: \newpage

380: 

381: \begin{figure}

382: \epsscale{1.0}

383: \plotone{f1.abbr.eps}

384: \caption{ The 4.8 arcminute region surrounding HD~47887 from (a) the

385: POSS, (b) a typical blue plate, and (c) the same plate after unsharp

386: masking. In all cases, north is up and east is left. }

387: \label{fig1}

388: \end{figure}

389: 

390: \begin{figure}

391: \epsscale{1.0}

392: \plotone{f2.eps}

393: \caption{ Time coverage of 40 high-quality plates and also the DPOSS

394: and DPOSS-2 images. In order to separate the points vertically, the

395: $y$-axis was taken to be the number of days since the nearest

396: Jan.~1. }

397: \label{fig2}

398: \end{figure}

399: 

400: \begin{figure}

401: \epsscale{0.7}

402: \plotone{f3.eps}

403: \caption{ Phase coverage of detections of KH~15D, for three choices of

404: period $P$. The symbols and color scheme are the same as in Fig.~2. }

405: \label{fig3}

406: \end{figure}

407: 

408: \begin{figure}[ht]

409: \epsscale{1.0}

410: \plotone{f4.eps}

411: \caption{ Determinations of the eclipse duty cycle ($f'$) from this

412: work, from our estimates using the light curves of Herbst et al.\

413: (2002), and from unpublished data by Garnavich et al.\ (2003). }

414: \label{fig4}

415: \end{figure}

416: 

417: \end{document}

418: