1: %%%%%% ms3a.tex , March 28, 2003, BP
2:
3: %\documentstyle[12pt,aaspp4]{article}
4: \documentclass[12pt,preprint]{aastex}
5: \usepackage{graphicx}
6:
7: \newcommand{\invday}{\,\rm{d}^{-1}}
8: \newcommand{\porb}{P_{\rm{orb}}}
9:
10: \begin{document}
11:
12: \title{Ellipsoidal Variability in the OGLE Planetary Transit Candidates}
13:
14: \author{E. Sirko \& B. Paczy\'nski}
15: \affil{Princeton University Observatory, Princeton, NJ 08544--1001, USA}
16: \affil{E-mail: esirko@astro.princeton.edu}
17: \affil{E-mail: bp@astro.princeton.edu}
18:
19: \begin{abstract}
20: We analyze the photometry of 117 OGLE stars with periodic transit events
21: for the presence of ellipsoidal light variations, which indicate the presence
22: of massive companions. We find that $ \sim 50\% $ of objects may
23: have stellar companions, mostly among the short period systems. In our
24: Table 1 we identify a coefficient of ellipsoidal variability for
25: each star, $a_{c2}$, which can be used to select
26: prime candidates for planetary searches. There is a
27: prospect of improving the analysis, and the systems with smaller ellipsoidal
28: variability will be identified, when the correlations in the OGLE photometry
29: are corrected for in the future, thereby providing a cleaner list of systems
30: with possible planets.
31: \end{abstract}
32:
33: \keywords{planetary systems -- surveys -- techniques: photometric}
34:
35: \section{Introduction}
36:
37: The first extra solar planet discovered to exhibit photometric transits was
38: HD 209458 (Charbonneau et al. 2000, Henry et al. 2000), but its orbit had
39: been first determined spectroscopically (Mazeh et al. 2000). Massive efforts
40: to detect photometric transits on their own put at least 23 teams into
41: the competition (Horne 2003). By far the largest list of periodic transit
42: candidates was published by the OGLE team (Udalski et al. 2002a,b,c), with
43: a total of 121 objects, with all photometric data available on the Web.
44: The first confirmation that at least one of these has a `hot
45: Jupiter' planet was obtained by Konacki et al. (2003). They also found that
46: a large fraction of OGLE candidates were ordinary eclipsing binaries
47: blended with a brighter star which was not variable, and its light diluted
48: the depth of the eclipses, so they appeared as shallow transits.
49:
50: It was clear from the beginning that many transits may be due to red dwarfs
51: or brown dwarfs, and that some of these massive companions may give rise to
52: ellipsoidal light variability (Udalski et al. 2002a). In fact in that
53: paper OGLE-TR-5 and OGLE-TR-16 were noted to exhibit such a variability,
54: indicating that
55: the companion mass had to be substantial. Ellipsoidal variability
56: is a well known phenomenon among binary stars (cf. Shobbrook et al. 1969,
57: and references therein). Tidal effects are responsible for making a star
58: elongated toward the companion. Ellipsoidal variability is due to the
59: changes in the angular size of the distorted star and to gravity
60: darkening, with half of the orbital period. Obviously,
61: the more massive the companion, and the closer it is to the primary,
62: the stronger the
63: effect. It can be used to remove from the OGLE sample objects which have
64: massive, and therefore not planetary, companions (Drake 2003).
65:
66: The aim of this paper is to extend the work of Drake (2003) to the new
67: list of OGLE transit candidates (Udalski 2002c) and to provide a more
68: realistic error analysis, which is needed in order to assess the reality
69: of ellipsoidal variability.
70:
71: \section{Data Analysis}
72:
73: We took all photometric data from the OGLE Web site:
74:
75: \centerline{http://sirius.astrouw.edu.pl/\~{}ogle/}
76:
77: \centerline{http://bulge.princeton.edu/\~{}ogle/}
78:
79: \noindent
80: There are three data sets: two provide a total of 59 objects in the field
81: close to the Galactic Center (Udalski et al. 2002a,b), and the third
82: provides 62 objects in a field in the Galactic Disk in Carina (Udalski
83: et al. 2002c). There are typically 800 data points for OGLE-TR-1 -
84: OGLE-TR-59, and 1,150 data points for OGLE-TR-60 - OGLE-TR-121. The
85: Carina data set had somewhat longer exposures and the field was much
86: less crowded, so the photometric accuracy is somewhat better than in
87: the Galactic Center field.
88:
89: To analyze the light curves for ellipsoidal variability the data points
90: within the transits must first be removed. Our algorithm was to
91: start at mid-transit, working outwards, and to reject data points
92: until the third data point
93: for which the baseline magnitude was within its photometric error bars.
94: %whose error placed it within or brighter than the baseline magnitude.
95: We then obtained a five parameter fit to the rest of the data:
96: $$
97: I_k = <I> + a_{c1} \cos p_k + a_{s1} \sin p_k
98: + a_{c2} \cos 2 p_k + a_{s2} \sin 2 p_k ,
99: \eqno(1)
100: $$
101: where $ I_k $ is data point number $ k $, and $ p_k $ is its
102: phase calculated with the orbital period provided by Udalski et al.
103: (2002a,b,c). The phase is $ p = 0 $ at mid transit. The values of all five
104: parameters: $ <I> $, $ a_{c1} $, $ a_{s1} $, $ a_{c2} $, and $ a_{s2} $
105: were calculated with a least squares method. The formal errors for all
106: the sinusoidal coefficients for a given star were practically the same.
107: Fig. 1 shows period-folded light curves of several objects, with the
108: best-fit five-parameter function (eq. 1) over plotted. Ellipsoidal
109: variability is clearly visible in objects such as OGLE-TR-16 as a
110: sinusoidal component with half the period of the binary and with
111: the phase of minimum light flux being the same as the phase of the transit.
112:
113: In the five parameter fit given with eq. (1) the term $ a_{c2} $
114: corresponds to the ellipsoidal light variations. Tidal effects make
115: a star the brightest at phases 0.25 and 0.75, and the dimmest at phases
116: 0.0 and 0.5. This corresponds to $ a_{c2} > 0 $, as the $ I $ magnitude is
117: largest when the star is at its dimmest. The values of $ a_{c2} $ with their
118: nominal error bars are shown in Fig. 2. Note that for a number of stars
119: the coefficient is negative, many standard deviations smaller than zero.
120: This clearly shows that the nominal errors are not realistic. This is
121: understandable, as there are strong correlations between
122: consecutive photometric data points (Kruszewski \& Semeniuk 2003).
123:
124: The rms deviation between
125: the five parameter fit and the data is a measure of the photometric
126: errors for individual measurements. The rms errors and the average of
127: the formal errors
128: of the four sinusoidal coefficients of the fit are shown in Fig. 3
129: with circles and triangles, respectively.
130: The formal errors for the fitted parameters are approximately
131: $ N^{1/2} $ times smaller than the rms values, as expected;
132: $ N $ is the number of photometric data points for a given star.
133: Open symbols refer to the stars in the Galactic Center field, and filled
134: symbols to the stars in Carina. The crosses are error estimates
135: by Drake (2003, Table 1) for stars in the Galactic Center field.
136: The squares will be explained later.
137:
138: It is clear that the errors increase with the $ I $ magnitude, as expected.
139: The errors are smaller for the Carina stars, as expected. Two stars,
140: OGLE-TR-24 and OGLE-TR-58, show anomalously large rms. Inspection of their
141: light curves shows they exhibit long term changes of about 0.015
142: magnitudes. The nature of this variability is not known, and we exclude
143: these two objects from the error histogram shown in Fig. 5, discussed below.
144: The orbital periods are not known for OGLE transits 43, 44, 45,
145: and 46, so these stars are not considered in this paper.
146: The following analysis was done for the remaining 117 objects.
147: Note that objects 8 and 29 are the same star, but they are treated
148: separately in the OGLE database, so we treat them separately here as well.
149:
150: The planetary transit candidates have orbital periods in the range
151: $ 0.57 - 9.2 $
152: days, which corresponds to the frequency of ellipsoidal variability in
153: the range $ 0.22 - 3.5 \invday $. We calculated power spectra for every
154: star as follows:
155: $$
156: p_{\nu} = a_{c, \nu }^2 + a_{s, \nu}^2 , \hskip 1.0cm
157: \nu = { i \over 250 ~ {\rm day } } , \hskip 1.0cm
158: i = 1, 2, 3, .... 1000,
159: \eqno(2)
160: $$
161: where
162: $$
163: a_{c, \nu } =
164: ~ { \sum _{k=1}^N (I_k-<I>) \cos \left( 2 \pi \nu t_k \right) \over
165: \sum _{k=1}^N \cos ^2 \left( 2 \pi \nu t_k \right) } ,
166: \hskip 1.0cm
167: a_{s, \nu } =
168: ~ { \sum _{k=1}^N (I_k-<I>) \sin \left( 2 \pi \nu t_k \right) \over
169: \sum _{k=1}^N \sin ^2 \left( 2 \pi \nu t_k \right) } ,
170: \eqno(3)
171: $$
172: where the summation
173: is done over all photometric data points for a given star, excluding the
174: data points within the transits.
175:
176: An example of the power spectrum is shown in Fig. 4 for OGLE-TR-5.
177: Also shown is a power law fit to the spectrum:
178: $$
179: p_{ \nu ,f} = p_0 ~ \nu ^{p_1} ,
180: \eqno(4)
181: $$
182: where the two parameters $ p_0 $ and $ p_1 $ were calculated for each
183: star using a least squares method. The binary period for OGLE-TR-5 is given
184: by Udalski et al. (2002a) as 0.8082 days. The arrows
185: correspond to the orbital frequency $ 1 / \porb = 1.237 \invday $,
186: and the expected frequency of ellipsoidal variability
187: $ 2 / \porb = 2.475 \invday $.
188:
189: We calculated power spectra for all stars, and each was fitted with its
190: own power law. For every star and for every frequency we calculated
191: the coefficients $ a_{c, \nu } $ and $ a_{s, \nu } $, as defined
192: by eq.~(3), and divided each by
193: $ b_{ \nu} \equiv ( p_{ \nu , f} / 2 )^{1/2} $ to normalize it.
194: A histogram of these normalized terms is shown in Fig. 5.
195: Also shown is Gaussian distribution with unit variance; it is
196: fairly similar to the histogram, which implies that $ b_{ \nu } $
197: provides a reasonable estimate of the statistical error.
198:
199: Our power spectrum formula (eqs. 2 and 3) can be derived by a least squares
200: method at each frequency. It is similar to the Lomb-Scargle (LS)
201: periodogram formula (Lomb 1976, Scargle 1982, Press et al. 1992), but
202: differs in several respects. Perhaps most importantly for this work,
203: the LS periodogram does not distinguish between the sine and cosine terms.
204: Therefore, a phase offset $\tau$ is necessary in the LS periodogram
205: formula, but not in our power spectrum since we fit $a_{c,\nu}$ and
206: $a_{s,\nu}$ separately. One strength of the LS periodogram is a prescription
207: to determine the significance level of any peak in the power spectrum.
208: However, this prescription is most useful when the resonant frequencies
209: are unknown beforehand, and requires a knowledge of the number of
210: effectively independent frequencies $M$ in the power spectrum, because
211: as $M$ increases, the probability that spurious noise looks like
212: genuine signals increases. However, in our case, we know the expected
213: frequency of ellipsoidal variability beforehand, so the value of $M$ is
214: irrelevant for our simple analysis. As shown in Fig.~5,
215: we are able to define an error $b_{\nu}$ for each value of $a_{c,\nu}$
216: and $a_{s,\nu}$ which is approximately Gaussian, so that the usual
217: (68, 95, 99.7)\%, etc. confidence values `approximately' apply for values of
218: $a_{c,\nu}$ and $a_{s,\nu}$ within (1, 2, 3)$\sigma$, etc. In other words,
219: a value of $a_{c2}$ that is $3\sigma$ away from zero indicates
220: ellipsoidal variability at a 99.7\% confidence level, depending on how
221: much faith one puts into the Gaussian nature of Fig. 5. Note that
222: the tails of the distribution are exaggerated on the logarithmic axis,
223: and that some contribution to the tails should be from genuine signal.
224:
225: For every star we calculated the $ a_{c1} $ and $ a_{c2} $ terms of eq. (1)
226: and estimated their errors $b_1$ and $b_2$ as $b_{ \nu } $ evaluated at the
227: corresponding frequencies. The errors for the $a_{c2}$ term, $b_2$,
228: are shown in Fig. 3 as squares.
229: It is clear that these statistical errors are much larger than the
230: errors shown as triangles, which were based on the assumption that
231: all photometric data points are uncorrelated. The `triangle' error
232: bars were used in our Fig. 2, which provided the first hint that they are
233: underestimates of the true errors, as so many values of the ellipsoidal
234: variability parameter $ a_{c2} $ were negative with very small `triangle'
235: error bars.
236:
237: The two amplitudes $ a_{c2} $ and $ a_{c1} $ are shown in Fig. 6 and Fig. 7,
238: respectively, as a function of orbital period. Negative values of $ a_{c2} $
239: are physically meaningless, and presumably these are $ a_{c2} \approx 0 $
240: which were `scattered' to negative values by statistical errors. Indeed,
241: there are 29 negative values, with 20 within one $ \sigma $ of zero, and
242: 9 outside of one $ \sigma $, with the ratio 20/9 close to that
243: expected for a Gaussian distribution.
244: This also implies that our error estimate is realistic; here and in
245: the following we take $\sigma = b_1$ for $a_{c1}$ and $a_{s1}$,
246: $\sigma = b_2$ for $a_{c2}$ and $a_{s2}$.
247:
248: We expect that about the same number, 29, of positive $ a_{c2} $ values
249: is due to errors, while the remaining $ 117 - 29 - 29 = 59 $ are real
250: ellipsoidal variables. This estimate suggests that $ \sim 50\% $ of all
251: OGLE transit candidates have massive companions, not planets.
252: Tidal effects responsible for ellipsoidal
253: variations increase strongly with reduced orbit size, and therefore they
254: are expected to be more common at short orbital periods. Indeed, this
255: appears to be the case in Fig. 6.
256:
257: Ellipsoidal variability scales with the mass ratio, and for low mass
258: companions like planets it becomes much smaller than our errors. Hence,
259: all OGLE transit stars with measurable ellipsoidal variability should
260: be excluded from the list of planetary candidates. Note that blending
261: with non-variable stars dilutes the variable component, and may suppress
262: the amplitude of the apparent ellipsoidal variability.
263:
264: Some stars vary with the orbital period. Positive values of $ a_{c1} $,
265: as shown in Fig. 7, may indicate a heating (reflection)
266: effect of the companion by the primary, as in the well known case of Algol,
267: or they may indicate that the true orbital period is twice longer than
268: listed by OGLE, and the variability is ellipsoidal. In the former case
269: nothing can be said about the companion mass; in the latter case the companion
270: is too massive to be a planet.
271:
272: Note that there are 44 stars in Fig. 7 with negative values of $ a_{c1} $,
273: and 27 of these are within their $ 1 \sigma $ error bars of zero. This
274: is consistent with no credible negative $ a_{c1} $. As the errors are
275: expected to be symmetric there must be $ \sim 44 $ stars which nominally
276: have positive values of $ a_{c1} $, but in fact are consistent with
277: $ a_{c1} = 0 $. Unfortunately,
278: we cannot point to the remaining 29 stars which may have real
279: `reflection' effects, except for OGLE-TR-39,
280: which has the orbital period given as 0.8 days, and it obviously has
281: $ a_{c1} > 0 $. This star shows large values
282: (in terms of their errors)
283: of both terms: $ a_{c1} $ and $ a_{c2} $. This indicates that the
284: orbital period is correct, the companion has its hemisphere facing the
285: primary noticeably heated, and its mass is large enough to induce ellipsoidal
286: light variations. The reflection effect is expected to be stronger
287: for binaries with small separations, i.e. short orbital period. There
288: is some evidence for this effect in Fig. 7.
289:
290: Table 1 lists all OGLE transit candidates with known orbital periods,
291: giving their OGLE-TR number, average magnitude $ <I> $, orbital period
292: $ \porb $ in days, the $ a_{c1} $ and $ a_{c2} $ coefficients in
293: milli-magnitudes, and
294: the number of transits detected by OGLE, $N_{\rm{tr}}$.
295: Objects with few detected transits are more likely to have
296: incorrect periods. Therefore, the ellipsoidal variability of the objects
297: with $N_{\rm{tr}} \sim 2$ is more likely to go undetected in our
298: analysis (but possibly could be detected in the $a_{c1}$ term if
299: the incorrect period happened to be half the true period).
300: The $ a_{c1} $ and $ a_{c2} $ coefficients are expressed in terms of their
301: errors in Table 1, in the format $b_i(a_{ci}/b_i \pm 1)$,
302: so that objects with larger values of $a_{c2}/b_2$ are more
303: likely to be ellipsoidal variable, and those with smaller or negative
304: values are better planetary system candidates.
305: It should become possible to recognize more objects with bona fide
306: ellipsoidal light variations when the systematic errors are reduced
307: by a more thorough analysis carried out by Kruszewski \& Semeniuk (2003).
308:
309: It is interesting to plot $ a_{c1} $ versus $ a_{c2} $, as shown in Fig. 8.
310: It is clear that while there are several possibly real positive $ a_{c1} $
311: terms, there are considerably more positive $ a_{c2} $ terms.
312:
313: In a simple model where the presence of a companion may give rise to
314: ellipsoidal light variations and the `reflection' effect the terms
315: $ a_{s1} $ and $ a_{s2} $ in eq. (1) should be zero. Fig. 9 presents
316: these coefficients in units of their errors. There are 29 stars with
317: $ a_{s2} $ having absolute value larger than one $ \sigma $, with
318: 88 values smaller than one $ \sigma $. The corresponding numbers for
319: $ a_{s1} $ are 41 and 76, respectively. This is close to the ratio
320: expected if the true values of both coefficients were zero, and their
321: errors were Gaussian. The average and the rms values are:
322: $ < a_{s1} /b_1 >\ = -0.27 $, $ < a_{s2} / b_2 >\ = +0.10 $,
323: $ <(a_{s1}/b_1)^2>^{1/2}\ = 1.17 $, and $ <(a_{s2}/b_2)^2>^{1/2}\ = 0.93 $.
324: All this implies that our error estimate is reasonable. Note that
325: OGLE-TR-68, with $ a_{s1}/b_1 = -4.4 $, may have a real variability,
326: possibly induced by a spot on the star.
327:
328: By treating the duplicate objects OGLE-TR-8 and OGLE-TR-29 separately
329: in this analysis, we are able to verify the consistency of the algorithm.
330: As can be seen from Table 1, for these two objects
331: the cosine coefficients $a_{c1}$ and
332: $a_{c2}$ are consistent within error. Furthermore, for OGLE-TR-8,
333: $a_{s1} = -0.006 \pm 1.47, a_{s2} = 0.643 \pm 1.32$, and for OGLE-TR-29,
334: $a_{s1} = -0.550 \pm 1.71, a_{s2} = 0.664 \pm 1.55$, so the sine
335: coefficients are also consistent within error.
336:
337: \section{Discussion}
338:
339: It is interesting to compare our errors, given in Table 1,
340: with those estimated by Drake (2003, Table 1, Galactic Center region only).
341: Our errors are also shown in Fig. 3 as squares, while Drake's errors are
342: shown with crosses. For the faintest stars
343: systematic errors are comparable to the photon noise, hence there is
344: only a small difference, approximately a factor of 2. For the bright stars
345: photon noise is negligible compared to systematic errors, and our estimate
346: is up to 5 times larger than Drake's (OGLE-TR-16). We stress that our
347: estimate is realistic, as demonstrated with Fig.~8 and Fig.~9.
348: Consider as an example OGLE-TR-5, with its spectrum
349: shown in Fig. 4. The peak corresponding to the ellipsoidal variability
350: at $ 2/\porb = 2.48 \invday $ is strong and certainly real. The
351: amplitude is $ a_{c2} = 7.5 \,\rm{mmag} $
352: in our Table 1, and 7.2 mmag in Drake's
353: Table 1. However, the errors are 1.3 mmag and 0.4 mmag, respectively. The
354: power spectrum presented in Fig. 4 clearly shows a high level of noise, which
355: is used for our error estimate.
356:
357: In the case of OGLE-TR-5 the ellipsoidal
358: variability is highly significant with either of the two error estimates.
359: It is not so with OGLE-TR-40, which is listed by Drake as being ellipsoidal
360: variable at the $ 3.5 \sigma $ level, while our analysis puts it at just a
361: one $ \sigma $ level, i.e. nothing definite can be said about this case.
362:
363: The star with the first planetary
364: companion confirmed spectroscopically, OGLE-TR-56 (Konacki et al.
365: 2003), should not have a measurable value of $a_{c2}$,
366: and reassuringly we do not detect a significant
367: ellipsoidal variability (cf. Table 1). The planetary disk covers
368: $ \sim 2 \times 10^{-4} $ of the sky as seen from the star, i.e. the
369: reflection effect has to be small, $ a_{c1} < 0.1 ~ \rm{mmag} $. The
370: measured value is $ a_{c1} = 1.08 \pm 0.46 ~ \rm{mmag} $, and presumably
371: it is not significant.
372:
373: A thorough analysis of various systematic effects apparent in the
374: photometry of tens of thousands of variable and non-variable stars measured
375: in the OGLE fields is currently being done by Kruszewski \& Semeniuk
376: (2003). Preliminary results indicate that various systematic errors
377: may be reduced considerably. This may allow a detection of smaller
378: ellipsoidal effects than we could find, and may provide a
379: cleaner list of systems which are likely to have planetary companions.
380: At this time
381: spectroscopists may use our Table 1 to select stars for their planetary
382: search, eliminating objects with measurable ellipsoidal variability,
383: as it implies a large mass ratio, and most likely a red dwarf companion.
384:
385: If the objective of this work was to identify stars with definite
386: ellipsoidal variability, we would select stars with their $ a_{c2} $ terms
387: positive at the several $ \sigma $ level. But our objective is different: we
388: identify stars which are the best candidates to have planetary
389: companions, i.e. stars without ellipsoidal variability. This can be done
390: in a statistical sense only. The best candidates are those for which
391: the $ a_{c2} $ term is either negative or positive but small. However,
392: even if we had perfect statistical information we could only assign a
393: probability that a given star does or does not have ellipsoidal variability.
394: This would always be in the sense that the smaller the $ a_{c2} $ term
395: relative to its error, the more likely the star is a planetary system.
396: Given limited access to big telescopes needed for spectroscopic follow-up
397: it is best to study the stars with the smallest (i.e., negative)
398: $ a_{c2} $ terms first, and gradually move `up' in $a_{c2}/b_2$ from Table 1.
399:
400: There are several developments which will improve the situation gradually.
401: First, OGLE continues its `planetary campaigns' and new candidate stars
402: will be published periodically. Second, a highly improved analysis of
403: photometry is under way (Kruszewski \& Semeniuk 2003), which will reduce the
404: correlation of consecutive photometric data points considerably.
405: This will allow a detection of smaller
406: amplitude ellipsoidal variations, and a much better rejection of stars
407: with non-planetary companions. It will also be possible to identify
408: stars with even smaller depth of transits, extending the current list.
409:
410: It is a pleasure to acknowledge many useful suggestions and discussions with
411: A. Kruszewski, R. Lupton, S. Ruci\'nski and A. Udalski. We thank the
412: referee for many useful suggestions and for identifying the star with
413: two names. This research was supported
414: by the NASA grant NAG5-12212 and the NSF grant AST-0204908.
415:
416: %REFERENCES
417: \begin{references}
418:
419: \reference{} Charbonneau, D., Brown, T. M., Latham, D. W., \& Mayor, M. 2000,
420: ApJ, 529, L45
421: \reference{} Drake, A. J. 2003, astro-ph/0301295
422: \reference{} Henry, G. W., Marcy, G. W., Butler, R. P., \& Vogt, S. S. 2000,
423: ApJ, 529, L41
424: \reference{} Horne, K. 2003, astro-ph/0301250
425: \reference{} Konacki, M., Torres, G., Jha, S., \& Sasselov, D. D. 2003,
426: Nature, 421, 507
427: \reference{} Kruszewski, A., \& Semeniuk, I. 2003, in preparation
428: \reference{} Lomb, N.~R.\ 1976, \apss, 39, 447
429: \reference{} Mazeh, T., Naef, D., Torres, G., Latham, D. W., Mayor, M. et al.
430: 2000, ApJ, 532, L55
431: \reference{} Press, W.~H., Flannery, B.~P., Teukolsky, S.~A., \&
432: Vetterling, W.~T. 1992, Numerical Recipes.
433: Cambrdige Univ. Press, Cambridge
434: \reference{} Scargle, J.~D.\ 1982, \apj, 263, 835
435: \reference{} Shobbrook, R. R., Herbison-Evans, D., Johnston, I. D., \&
436: Lomb, N. R. 1969, MNRAS, 145, 131
437: \reference{} Udalski, A., Paczy\'nski, B., Zebru\'n, K., Szyma\'nski, M.,
438: Kubiak, M. et al. 2002, AcA, 52, 1
439: \reference{} Udalski, A., Zebru\'n, K., Szyma\'nski, M., Kubiak, M.,
440: Soszy\'nski, I. et al. 2002, AcA, 52, 115
441: \reference{} Udalski, A., Szewczyk, O., Zebru\'n, K., Pietrzy\'nski, G.,
442: Szyma\'nski, M. et al. 2002, AcA, 52, 317
443: %\reference{} Wilson, R. E. \& Devinney, E. 1971, ApJ, 166, 605
444:
445:
446: \end{references}
447:
448:
449:
450: \begin{deluxetable}{lrrrrr}
451: \tablewidth{0pt}
452: \tablecaption{List of transiting objects}
453: \tablecomments{We present the $a_{c1}$ and $a_{c2}$ coefficients in the
454: format $b_i(a_{ci}/b_i \pm 1)$ to more easily identify the ratio of
455: the coefficients to their errors.
456: *OGLE-TR-8 and OGLE-TR-29 are the same
457: object, but recorded as two separate events by OGLE. We treat them
458: separately throughout this paper.}
459: \tablehead{\colhead{Object name} & \colhead{$<I>$} & \colhead{$\porb$} &
460: \colhead{$a_{c1}$} & \colhead{$a_{c2}$} & \colhead{$N_{\rm{tr}}$} }
461: \startdata
462: \input{transit_table.tex}
463: \enddata
464: \end{deluxetable}
465:
466: \begin{figure}
467: \includegraphics[width=\textwidth]{f1.eps}
468: \caption{Light curves of five transiting objects, folded with the transit
469: at zero phase, are shown as examples.
470: Photometric data points within the transits, indicated
471: with circles, are not used in fitting. The best-fit five-parameter
472: function (eq. 1) is also plotted.
473: Objects OGLE-TR-16 and TR-61 have ellipsoidal variability at
474: about the 8$\sigma$ level, while OGLE-TR-101 has ellipsoidal variability
475: at about a 4$\sigma$ level (cf. Table 1). OGLE-TR-39 has a strong
476: $a_{c1}$ term, indicative of surface heating (cf. Fig. 7 and 8), as
477: well as a strong ellipsoidal variability term $a_{c2}$.
478: OGLE-TR-68 has a strong $a_{s1}$ term (cf. Fig. 9), indicating a probable
479: spot activity.}
480: \end{figure}
481:
482: \begin{figure}%[t]
483: \includegraphics[width=\textwidth]{f2.eps} %{8cm}{90}{50}{50}{190}{0}
484: \caption{
485: The values of the parameter corresponding to ellipsoidal light variations
486: are shown as a function of average stellar magnitude $ <I> $. Open and
487: filled symbols refer to the data sets OGLE-TR-1 to TR-59 and TR-60 to TR-121,
488: respectively. Formal errors obtained
489: with the least squares fit are also shown. Note a number of stars with
490: negative, i.e. not physical, values of $ a_{c2} $, which are many formal
491: standard deviations below zero. This indicates that the errors are not
492: realistic.
493: }
494: \end{figure}
495:
496: \begin{figure}%[t]
497: \includegraphics[width=\textwidth]{f3.eps} % {8cm}{90}{50}{50}{190}{0}
498: \caption{
499: Photometric errors are shown as a function of $ <I> $ magnitude. Circles
500: refer to the errors of individual data points, defined as an rms deviation
501: from a 5 parameter fit to the light curves outside of the transits
502: (cf. eq. 1). The two outliers near $ <I>\ \approx 14.8 $, stars TR-24 and
503: TR-58, have a long term variability at the level of 0.015 mag.
504: Triangles are the average of the formal errors of the 5 parameter fit
505: obtained with the least squares solution of eq. (1). Squares are
506: the errors of the $a_{c2}$ term determined with our spectral analysis - note
507: they are considerably larger than the formal errors, as the photometric data
508: have a strong time correlation. Open and filled symbols refer to
509: the OGLE-TR-1 to TR-59 and TR-60 to TR-121 data sets, respectively.
510: Errors listed by Drake (2003) in his Table 1
511: are shown with crosses for stars in the Galactic Center region.
512: }
513: \end{figure}
514:
515: \begin{figure}%[t]
516: \includegraphics[width=\textwidth]{f4.eps} %{8cm}{90}{50}{50}{190}{0}
517: \caption{
518: The power spectrum is shown for OGLE-TR-5. The thick solid
519: line is the least squares power law fit to the spectrum. The two arrows
520: correspond to the orbital frequency $ 1 / \porb = 1.237 \invday $,
521: and the frequency of expected ellipsoidal variability
522: $ 2 / \porb = 2.475 \invday $. Notice a very strong peak at
523: the `ellipsoidal' frequency, and the two aliases at $ 1.475 \invday $ and
524: at $ 3.475 \invday $.
525: }
526: \end{figure}
527:
528: \begin{figure}%[t]
529: \includegraphics[width=\textwidth]{f5.eps} %{8cm}{90}{50}{50}{190}{0}
530: \caption{
531: The distribution of normalized amplitudes of sine and cosine terms for
532: all sampled frequencies and for all transit candidates except OGLE-TR-24
533: and TR-58. The ordinate axis is logarithmic to bring out the tails
534: in the distribution. For comparison
535: a Gaussian distribution with unit variance is also shown.
536: }
537: \end{figure}
538:
539: \begin{figure}%[t]
540: \includegraphics[width=\textwidth]{f6.eps} %{8cm}{90}{50}{50}{190}{0}
541: \caption{
542: The values of the parameter $ a_{c2} $, corresponding to ellipsoidal light
543: variations, are shown as a function of orbital period, $ \porb $. Open
544: and filled symbols refer to the stars OGLE-TR-1 to TR-59, and TR-60 to TR-121,
545: respectively. The errors are based on the limited spectral analysis
546: presented in this paper. None of the negative values are significant.
547: Up to $ \sim 50\% $ of all stars may have significant ellipsoidal
548: variability, indicating a massive, not planetary, companion.
549: }
550: \end{figure}
551:
552: \begin{figure}%[t]
553: \includegraphics[width=\textwidth]{f7.eps} %{8cm}{90}{50}{50}{190}{0}
554: \caption{
555: The same as Fig. 6, but for the $ a_{c1} $ parameter, which is indicative
556: of possible `heating' (reflection) effects of the companion's hemisphere
557: facing the primary. Only OGLE-TR-39 shows the effect strongly, but there
558: may be several other stars for which the effect is real.
559: }
560: \end{figure}
561:
562: \begin{figure}%[t]
563: \includegraphics[width=\textwidth]{f8.eps} %{8cm}{90}{50}{50}{190}{0}
564: \caption{
565: A relation between the ellipsoidal variability parameter $ a_{c2} $, and the
566: heating variability parameter $ a_{c1} $ (cf. eq. 1), both
567: normalized by their errors. The stars with the
568: strongest effects are labeled with their OGLE names. The negative
569: values of either parameter are consistent with them being due to errors.
570: }
571: \end{figure}
572:
573: \begin{figure}[t]
574: \includegraphics[width=\textwidth]{f9.eps} %{8cm}{90}{50}{50}{190}{0}
575: \caption{
576: A relation between the coefficients of the sine terms in eq. (1),
577: normalized by their errors.
578: %The ratio of the number of stars with the values of either coefficient
579: %being within one $ \sigma $ of zero, to the number outside of one $ \sigma $
580: %is approximately 3/1, not very different from the ratio 2/1 expected
581: %if their true values were zero and their distributions were Gaussian.
582: The ratio of the number of stars having $a_{s1}$ within one $\sigma$
583: of zero to the number outside of one $\sigma$ is approximately 2/1,
584: and for $a_{s2}$ the ratio is approximately 3/1. We expect these ratios
585: to be 2/1 if the true values are zero and the distributions are Gaussian.
586: We conclude that our error estimate is reasonable, and we have no
587: coefficients $ a_{s1} $ or $ a_{s2} $ that are measurably
588: non-zero, with the possible exception of OGLE-TR-68, the filled
589: circle with $ a_{s1} / b_1 = -4.4 $. Visual inspection of its light curve
590: clearly shows that the system is approximately 15 mmag brighter at phase
591: 0.25 than it is at phase 0.75.
592: }
593: \end{figure}
594:
595: \end{document}
596:
597: