0803.3548/ms.tex
1: % \voffset=-0.7truein   % needed for thorstensen's system, remove before sub
2: %\documentclass[manuscript]{aastex}
3: %\documentclass[preprint2]{aastex}
4: \documentclass[preprint]{aastex}
5: % \documentclass{aastex}
6: 
7: \def\simgr{\,\hbox{\hbox{$ > $}\kern -0.8em \lower 1.0ex\hbox{$\sim$}}\,}
8: \def\simle{\,\hbox{\hbox{$ < $}\kern -0.8em \lower 1.0ex\hbox{$\sim$}}\,}
9: \begin{document}
10: \title{SDSS 1507+52: A HALO CATACLYSMIC VARIABLE?
11: \footnote{Based on observations obtained at the MDM Observatory, operated by
12: Dartmouth College, Columbia University, Ohio State University, Ohio University, and
13: the University of Michigan.}
14: }
15: 
16: \author{Joseph Patterson}
17: \affil{Department of Astronomy, Columbia University, New York, NY 10027;
18:     jop@astro.columbia.edu}
19: \author{John R. Thorstensen}
20: \affil{Department of Physics and Astronomy, Dartmouth College, 6127 Wilder
21: Laboratory, Hanover, NH 03755-3528}
22: \author{Christian Knigge}
23: \affil{School of Physics \& Astronomy, University of Southampton, Highfield, 
24: Southampton SO17 1BJ, United Kingdom}
25: 
26: \begin{abstract} 
27: We report a photometric and spectroscopic study of
28: the peculiar cataclysmic variable SDSS 1507+52.  The star shows very
29: deep eclipses on the 67 minute orbital period, and those eclipses are
30: easily separable into white-dwarf and hot-spot components.  This leads
31: to tight constraints on binary parameters, with $M_1=0.83(8)
32: M_{\odot}$, $M_2=0.057(8) M_{\odot}$, $R_1=0.0097(9) R_{\odot}$,
33: $R_2=0.097(4) R_{\odot}$, $q=0.069(2)$, and $i=83.18(13)$ degrees.
34: Such numbers suggest possible membership among the WZ Sge stars, 
35: a common type of dwarf
36: nova.  The spectroscopic behavior (strong and broad H emission,
37: double-peaked and showing a classic rotational disturbance during
38: eclipse) is also typical.  But the star's orbital period is shockingly
39: below the ``period minimum" of $\sim$77 min which is characteristic of
40: hydrogen-rich CVs; producing such a strange binary will require some
41: tinkering with the theory of cataclysmic-variable evolution.
42: The proper motion is also remarkably high for a star of its distance,
43: which we estimate from photometry and trigonometric parallax as 230
44: $\pm$ 40 pc.  This suggests a transverse velocity of 164 $\pm$ 30 km/s
45: - uncomfortably high if the star belongs to a Galactic-disk
46: population.  These difficulties with understanding its evolution and
47: space velocity can be solved if the star belongs to a Galactic-halo
48: population. 
49: \end{abstract}
50: 
51: \keywords{binaries: close; binaries: eclipsing; novae, cataclysmic variables;
52: stars: individual (SDSS J150722.30+523039.8); white dwarfs}
53: 
54: \section*{}
55: \begin{quote}
56: \textit{``\ldots of a most insignificant smallness, and a swift
57: wanderer among the stars."} -- \citet{kepler1619}, \textit{Harmonices Mundi}
58: \end{quote}
59: 
60: \section{Introduction}
61: 
62: Among the $\sim150$ cataclysmic variables so far discovered in the
63: Sloan Digital Sky Survey, one of the most peculiar is the 67-minute
64: eclipsing binary SDSS J150722.30+523039.8 (hereafter SDSS1507). 
65: \citet[][hereafter S05]{szkody05} presented a light curve 
66: showing sharp
67: eclipses, and a spectrum with the characteristic absorption lines of a
68: DA white dwarf.  The orbital period is very peculiar, far below the
69: well-known ``period minimum" of $\sim77$ min for hydrogen-rich
70: binaries.  This is quite difficult to understand.  However, binaries
71: with sharply defined eclipses offer good diagnostics for measuring the
72: individual masses, and so we carried out an extensive photometric,
73: spectroscopic, and astrometric study during 2005-7.  Here we report
74: the results of that study.  Another study is presented by 
75: \citet[][hereafter L07]{littlefair07}.
76:    
77: \section{Astrometry, Distance, and Velocity}
78: 
79: We have accumulated a two-year series of astrometric images of
80: SDSS1507 with the 2.4m Hiltner telescope at MDM Observatory.  Table 1
81: is a journal of the astrometric observations.
82: 
83: The astrometric images contain some photometric information.  While
84: nearly all observations were in the $I$-band, we also took four $V$-band
85: images on three different photometric nights, in order to
86: calibrate the field and program star magnitudes and colors for the 
87: astrometric reduction.  The $VI$ image pairs, reduced using 
88: \citet{landolt92} standards and corrected for orbital variation, 
89: gave average values $V$ = 18.34 and $V-I$ = +0.23.
90: 
91: We also derived differential magnitudes from the $I$-band parallax
92: images and folded them using the eclipse ephemeris derived from the
93: high-speed photometry (discussed below).  The resulting light curve
94: appears similar to the high-speed light curve, but with coarser time 
95: resolution because of the relatively long (120 to 150 s) integration 
96: times.  This fine agreement, obtained from many short visits spread 
97: over two years, emphasizes the reproducibility of the light curve.
98: 
99: The parallax and proper motion analysis followed procedures detailed
100: by \citet{thorstensen03}.  On each $I$-band image we measured 57
101: stars, 28 of which we used to define the reference frame.  We found a
102: relative parallax $\pi_{\rm rel} = -0.2 \pm 1.3$ mas, where the
103: uncertainty is estimated from the goodness of fit.  Correcting for the
104: estimated mean distance of the reference frame stars, we found an
105: absolute parallax $\pi_{\rm abs} = +0.7$ mas.  The scatter in the
106: fitted parallax for stars in the field with brightness similar to
107: SDSS1507 is 1.9 mas, so we adopt 2.5 mas as a conservative estimate on
108: the external error of $\pi_{\rm abs}$. Thus we find $\pi_{\rm abs} =
109: 0.7 \pm 2.5$ mas.  The parallax is consistent with zero; our somewhat
110: conservative upper limit implies a distance $d > 175$ pc with over
111: 90\% confidence.
112: 
113: Given its  distance, SDSS1507 has a strikingly large proper motion:
114: $[\mu_X, \mu_Y] = [-149, +58]$ mas yr$^{-1}$.  A proper motion this
115: large is typical of disk-population stars only a few tens of parsecs
116: distant, so the space velocity is clearly very large.
117: 
118: Combining the proper motion with the parallax, interpreting the proper
119: motion using the assumed CV space-velocity model discussed by
120: \citet{thorstensen03}, and accounting for the Lutz-Kelker parallax
121: bias, we obtain a purely astrometric distance $d = 280 (+72,-62)$ pc
122: (68\% confidence).  This is much smaller than $1/\pi$ because the high
123: proper motion drags the estimate to smaller distances.  The velocity
124: assumptions (that space velocities of CVs are generally low, with a
125: small high-velocity tail) basically quantify the implausibility of the
126: huge space velocity required by high proper motion at the simple
127: $1/\pi$ distance.
128: 
129: There is an independent constraint available from the time-series
130: photometry (reported below).  Eclipse ingress and egress of the white
131: dwarf (hereafter WD) are easily seen in the light curve, and show that
132: a small object of magnitude $V=18.9 \pm 0.15$ reappears when the WD
133: does.  The photometry discussed below reveals a WD radius of $0.0097
134: \pm 0.0009\ \rm{R}_\odot$, and the SDSS colors, along with our
135: scattered $BVI$ photometry, suggests $T_{\rm eff} = 12000 \pm 1200$
136: K.  Assuming that the line of sight permits full hemispheric
137: visibility of the WD near the contact phases of the eclipse, this
138: implies an absolute magnitude $M_V = 12.3 \pm 0.4$.  This in turn
139: yields a ``white dwarf parallax" distance of $215 \pm 45$ pc.  The
140: $T_{\rm eff}$ estimate ($11000 \pm 500$ K) of
141: L07 might be superior, since it is
142: based on full multicolor eclipse light curves; that implies $M_V= 12.5
143: \pm 0.3$, and thus $190 \pm 40$ pc.  Splitting the difference yields
144: $202 \pm 40$ pc.  Combining this with the astrometric result, we
145: % need to propagate the resulting changes here ... xxx
146: obtain a final estimate $d = 233(+42,-36)$ pc, which we shall
147: characterize as $230 \pm 40$ pc.
148: 
149: At this distance, the transverse velocity of SDSS1507 (corrected
150: to the LSR) is $164 \pm 30$ km s$^{-1}$
151: -- a speedy traveler through local skies!  To judge its
152: significance, we examined a catalogue of all CVs of known orbital
153: period (an expanded version of Table 1 of \citealt{patterson84}), and
154: calculated transverse velocities for all 354 CVs of known $P_{\rm
155: orb}$, distance, and proper motion.  We excluded stars known or
156: strongly suspected to be helium-rich (although these stars are very
157: few and mostly ineligible anyway, since few of them offer distance
158: constraints).  The distance estimates used all the various clues
159: available (discussed by \citealt{patterson84}), with highest weight
160: accorded to parallaxes and detections of the secondary.  The proper
161: motions came mainly from the UCAC-2 \citep{zacharias04}, or from
162: published studies of individual stars, or from the Dartmouth parallax
163: program. No correction for solar motion was applied.
164:  
165: The upper frame of Fig.~\ref{fig:vtrans} shows the distribution of transverse
166: velocities.  The velocities average 33 km s$^{-1}$, a plausible value
167: for 1 M$_{\odot}$ stars in the thin-disk population.  There are two
168: outliers: SDSS1507 at 174 km s$^{-1}$ (ignoring the solar motion
169: correction boosts its transverse velocity slightly),
170: and BF Eri at $\sim 400$ km
171: s$^{-1}$ \citep[][not shown here, but discussed in Section 9
172: below]{sheets07}.  The lower frame shows the distribution with $P_{\rm
173: orb}$, and includes 482 stars, since the only admission credential is
174: a known $P_{\rm orb}$.  This illustrates the familiar 2-3 hr period
175: gap, the pile-up of stars at the 1.3 hr minimum period $\ldots$ and
176: SDSS1507, a solitary intruder sitting aloof at 1.1 hr.  These two
177: oddities are the motivation for this paper.
178: 
179: Just for the record, we can use our measurement of the emission-line
180: mean velocity $\gamma$ ($-46$ km s$^{-1}$, discussed below) as an
181: indicator of the star's systemic radial velocity; at face value this
182: gives a space velocity (referred to the LSR) of $167 \pm 30$ km s$^{-1}$.
183: Very speedy.  The components are $[U,V,W] = [+135, -93, +31]$ km s$^{-1}$. 
184: 
185: 
186: \section{Spectroscopy}
187: 
188: During 2006 Jun 16-19, we obtained 53 spectra of 300 s each, with a
189: total exposure equivalent to 3.9 orbital periods.  We used the Hiltner
190: 2.4 m telescope, modular spectrograph, and a thinned SITe CCD
191: detector, yielding 2.0 \AA\ per pixel and 3.5 \AA\ FWHM resolution in
192: the range 4300-7500 \AA , with vignetting near the shorter
193: wavelengths.  The transparency was good throughout, but the spectra
194: varied in quality due to seeing and guiding variations.  For
195: wavelength calibration we used a comparison spectrum taken during
196: twilight, shifted to match the zero point derived from night-sky lines
197: in the individual exposures; a cross-check using OH bands in the far
198: red showed this procedure to be accurate to typically 5 km s$^{-1}$.
199: For reductions we mostly followed standard IRAF-based procedures, but
200: in place of the IRAF {\it apsum} routines we used an original coding of
201: the optimal extraction algorithm published by \citet{horne86}.  We
202: flux-calibrated the spectra using numerous observations of standard
203: stars obtained during twilight on clear nights throughout the run.
204: 
205: Fig.~\ref{fig:spec} shows the average flux-calibrated spectrum, and
206: Table 2 gives measurements of spectral features.  The blue
207: continuum and absorption around H$\beta$~indicates that the WD
208: contributes strongly to the total light.  The H$\alpha$~and H$\beta$β
209: emission lines are double-peaked, with the H$\alpha$~peaks
210: separated by 1320 km s$^{-1}$.  The continuum is noisy, because of the
211: star's faintness -- synthetic photometry using the \citet{bessell90}
212: passband gives $V = 18.4$, in good agreement with the filter
213: photometry.  This average spectrum is very similar to the SDSS
214: spectrum (S05, L07).   
215: 
216: We measured radial velocities of H$\alpha$ using a double-Gaussian
217: convolution method \citep{schneider80,shafter83}, with Gaussians of
218: 370 km s$^{-1}$ FWHM separated by 2200 km s$^{-1}$.  In effect, this
219: measured the steep sides of the line profile.  The observations were
220: not distributed to optimize period finding, since the period was
221: already known from eclipses; even so, a period search of the
222: resulting time series recovers the correct orbital period.  There is
223: no evidence for any other period in the velocities.  Fig.~\ref{fig:foldvel} shows
224: the velocities folded on the eclipse ephemeris, together with the
225: best-fitting sinusoid. 
226: % (Table 3 gives its parameters).  -- put the few params in fig capt  
227: The velocity
228: reaches maximum redshift at $\phiφ=0.90 \pm 0.03$, although the WD
229: must be in maximum recession at $\phiφ= 0.75$.  This is very commonly
230: seen in cataclysmic variables; the disk's rotational velocities are
231: large compared to any plausible orbital motion of the WD, so a slight
232: periodic asymmetry in the disk's line emission can create a large
233: spurious radial velocity signal.  It is famously cursed as the
234: ``phase-shift problem" (\citealt{stover81, hessman87, thorstensen91};
235: Sec.  2.7.6 of \citealt{warner95}).  There is now depressing evidence
236: that the phase-shift problem can even contaminate the ultraviolet
237: absorption lines, thought to arise from near the WD \citep{steeghs07}.
238: 
239: To make weaker features visible, we combined our spectra into a
240: two-dimensional greyscale representation using procedures explained by
241: \citet{taylor99}.  The individual spectra were rectified before
242: combining (suppressing continuum variations), and cosmic rays were
243: edited out by hand.  Fig.~\ref{fig:trail} shows the region around H$\alpha$ and
244: He I $\lambda6678$.  A prominent `S-wave' is seen in both lines, and
245: is especially clean in the He I line.  A similar S-wave is seen in HeI
246: $\lambdaλ5876$, and less prominently in H$\beta$.  We measured the
247: positions of these by eye in an image display; sinusoidal fits to
248: these measures gave nearly identical results.  The half-amplitude of
249: the S-wave motion is 840 $\pm$ 30 km s$^{-1}$, and the S-wave
250: velocities cross from blue to red at $\phi =0.86 \pm 0.01$.  
251: 
252: The amplitudes and phases of these radial-velocity variations merit
253: comparison to WZ Sge, the granddaddy of the short-period systems.  In
254: SDSS1507 the centroid of the broad emission lags the WD by 0.15 $\pm$
255: 0.03 cycles, and in WZ Sge the lag is 0.12 $\pm$ 0.02 \citep{spruit98,
256: gilliland86}.  Their respective S-waves reach maximum radial-velocity
257: at $\phiφ=0.11$ and 0.13.  This suggests great similarity in the
258: emission-line structures.  In particular, it underlines that
259: phase-shift woes significantly affect both stars, and likely for the
260: same (unknown) reason, preventing -- as usual -- any direct
261: interpretation of $K_1$ as the dynamical motion of the WD.
262:           
263: 
264: \section{High-Speed Photometry}
265: 
266: \subsection{Data Taking}
267: 
268: Time-series differential photometry, relative to a nearby comparison
269: star, was obtained on 20 nights in 2005-7, with a CCD photometer on
270: the 1.3 m and 2.4 m telescopes of MDM Observatory.  We typically chose
271: an integration time of 15 s, and usually employed a ``wide blue"
272: filter to better define the deep eclipses, and to maximize sensitivity
273: to weak signals.  The observations spanned 55 binary orbits, plus a
274: few orbits in $B$, $V$, and $I$ filters.  Outside eclipse, the star's
275: mean brightness was $V = 18.3$.
276: 
277: 
278: \subsection{Light Curves and Eclipses}
279: 
280: A typical night's light curve is shown in the top frame of 
281: Fig.~\ref{fig:onenight}; and indeed,
282: every detail of the light curve repeated almost exactly on
283: every night of observation.  The main features are periodic humps and
284: sharp dips, which recur every 67 minutes.  The mean 2006 light curve
285: is shown in the middle frame, which has been converted to an intensity
286: scale, where the comparison star is arbitrarily scaled as 1000 counts.
287: This shows that the dip consists of two distinct parts: a WD eclipse
288: which is about 70\% of the mean light, and a ``bright-spot" (BS)
289: eclipse which is about 15\% of the mean light.  The bottom frame shows
290: a close-up view of the eclipse region.  This too is a mean orbital
291: light curve, but we achieved a higher effective by calculating
292: the exact orbital phase at each mid-exposure, and using
293: 800 bins per orbit.  This was only possible because the light curve
294: repeats so exactly, and because we observed for many orbits.
295: 
296: We measured mid-ingress and mid-egress times of the WD eclipse (the
297: sharp transitions at $\phi = \pm 0.0203$) for each eclipse, and
298: consider the midpoint between these precisely measurable events as the
299: time of mid-eclipse.  These times are presented in Table 3, and shown
300: in the O-C diagram of Fig.~\ref{fig:ominusc} (along with the L07 timings,
301: which are given triple weight since they show a lower dispersion).  
302: The slight upward trend indicates a
303: small correction to the test period, and the straight-line fit
304: satisfies all timings to an accuracy of 2 s.  Since the original time
305: series have a time resolution of $\sim$22 s (15 s integration plus 7 s
306: readout), this is a satisfactory fit.  Thus the best orbital
307: ephemeris is 
308: \begin{equation} 
309: \hbox{Mid-eclipse} = \hbox{HJD } 2453498.892264(9) + 0.0462583411(7) E.    
310: \end{equation}
311: 
312: The lesser component (BS) of the periodic eclipse is often too
313: weak to measure reliably in individual eclipses, but is easily measured in the
314: yearly means.  In 2006, mid-ingress and mid-egress occurred at $\phi
315: = 0.0110(10)$ and 0.0735(10),  In 2005, these events occurred at
316: $\phi = 0.0118(8)$ and 0.0735(8).  In 2007, a coarser measurement
317: gave 0.010(2) and  0.074(2).  Thus all the timings were consistent
318: with 0.0114(7) and 0.0735(6).
319: 
320: Finally, the full ingress/egress durations of the WD eclipse
321: were measured to be 35 $\pm$ 4 s, by fitting the yearly light curves
322: to a model in which a tilted knife-edge (the secondary's limb) occults
323: a bright sphere with limb-darkening.  %The relevant geometry is shown
324: %in Figure 6 of \citet{robinson78}.
325: 
326: \subsection{White-Dwarf Pulsations}
327: 
328: We searched for other periodic signals by calculating power
329: from the Fourier transform of the light curve.  We did this
330: for each night with more than 4 hours observation, and then averaged
331: to find the mean nightly power spectra in 2005 and 2006, after removal
332: of eclipses (which severely contaminate the raw power spectra).  Most
333: of the significant features were merely harmonics (up to the ninth) of
334: the orbital frequency ω$\omega_o$; so we concluded that even apart
335: from the obvious eclipses, the orbital waveform is still a very strong
336: contaminant.  Hence we ``cleaned" the light curves by subtracting
337: harmonics fitted to amplitude and phase.  We then re-segregated these
338: yearly cleaned light curves, and calculated the nightly power spectra.
339: 
340: The yearly means are shown in Fig.~\ref{fig:nightlypower},  
341: with arrows flagging
342: significant features with their frequency in cycles per day ($\pm$
343: 0.5).  Signals near 76, 130, and 175 cycles d$^{-1}$ appear, though with the
344: exact frequencies somewhat unstable.  Nothing was found in the range
345: 400-1000 cycles d$^{-1}$.
346: 
347: Single-night power spectra are not a good detection tool for
348: frequencies this low: too few cycles elapse in the course of a night.
349: To increase sensitivity and accuracy, we spliced consecutive nights
350: together (JD 2453502-4 in 2005 and JD 2453878-81 in 2006), and show
351: the resultant power spectra in Fig~\ref{fig:detailedpower}. The 2005 light curve showed
352: a primary signal at 76.60 $\pm$ 0.06 cycles d$^{-1}$, with another at
353: 132.59 and a complex near 176 cycles d$^{-1}$.  As the picket-fence
354: structure in the figure suggests, each frequency has some possibility
355: of daily cycle count error.  The 2006 light curve showed apparent
356: counterparts of each signal, but with frequencies red-shifted by a few
357: percent: to 74.94, 128.05, and 171.30 (all $\pm$ 0.04) cycles d$^{-1}$.
358: The 2006 coverage was more extensive, and daily cycle count error is
359: very unlikely.  We assume that these trios represent basic frequencies
360: in the star, and will call them ω$\omega_1$, $\omega_2$, and
361: $\omega_3$, in order of increasing frequency.  All are incommensurate
362: with the orbital frequency $\omega_o$, but the weaker signals appear
363: to be linked to each other since $\omega_3 - \omega_2 = 2ω\omega_o$
364: within measurement error during both years.
365:    
366: Inset are the power spectra of artificial signals sampled exactly like
367: the real data; these show how a pure sinusoid at that frequency should
368: appear, in the absence of noise.  For a strong signal any deviation
369: from this window pattern signifies an underlying complexity of the
370: signal.  The strong signal at 74.94 cycles d$^{-1}$ in 2006 shows
371: significant excess power in its blue wing, which basically attests to
372: an unresolved fine structure in the signal.  The same is not evident
373: for the signals at higher frequency.  In 2005, the strongest signal
374: again is complex; that is probably also true for the weaker signals,
375: but is not quite provable amid the surrounding noise.
376:    
377: In summary, SDSS1507 shows a dominant signal near 1140 s, with at
378: least one additional component at slightly shorter period.  When we
379: attempted to parse the underlying fine structure, we could not do so
380: unambiguously.  But the components must be variable in frequency
381: and/or amplitude, and there are probably at least three of them (many
382: trials with only two components failed to give a satisfactory fit).
383: Weaker signals near 660 and 500 s are also evident; their frequencies
384: appear to be consistently separated by exactly $2ω\omega_o$.
385: 
386: The light curve and spectrum show that a WD dominates the light, so
387: these signals are very likely to be seated in the WD.  The
388: complexities observed here (multiple and incommensurate periods, with
389: fine structure) are often found in the nonradial pulsations of WDs,
390: the ZZ Ceti stars.  So SDSS1507 has joined the small club of CVs with
391: nonradial WD pulsations, which we shall call the ``GW Lib stars",
392: after its first and most famous member (for other members and a
393: review, see \citealt{mukadam07}).
394: 
395: 
396: \section{Eclipse Analysis and the Binary Parameters}
397: 
398: The timings of the WD and BS contact points can be used, essentially
399: following \citet{wood89}, to estimate the binary parameters of
400: SDSS1507.  These estimates are listed in Table 4 and have been
401: determined via the following steps.
402: 
403: First, in a binary with a Roche-lobe-filling secondary, the full width
404: at
405: % not clear what the symbol for WD eclipse width was supposed to be
406: half-intensity of the WD eclipse ($\Delta \phi_{\rmφWD}$) defines the
407: acceptable $q(i)$ solutions.  
408: Our measured $\Delta \phi_{\rm WD}$ and the method
409: described by \citet{chanan76} implies the $q(i)$ constraint seen in
410: Fig.~\ref{fig:qofi}.
411: 
412: Second, the mid-ingress and mid-egress phases of the BS eclipse
413: ($\phi_{\rm BSI}$ and φ$\phi_{\rm BSE}$) can be used to estimate $q$,
414: under the assumption that the BS lies in the orbital plane and along
415: the ballistic trajectory of the accretion stream from the L1 point.
416: We have therefore calculated stream trajectories for a range of mass
417: ratios \citep{flannery75} and worked out the ingress/egress phases of
418: a set of points along each trajectory using methods similar to that
419: described by \citet{chanan76}.  Fig.~\ref{fig:ingresspred} shows the calculated stream
420: trajectories for various mass ratios in the ($\phi_{\rm BSI},
421: \phi_{\rmφBSE}$) plane, along with the best measurement of those
422: phases.  We thus estimate $q=0.069(2)$, and Fig.~\ref{fig:qofi} then implies
423: $i=83.18(13)$ degrees.
424: 
425: Third, if the BS photometric center lies at the disk's outer edge,
426: then this method also provides an estimate of the disk's radius.
427: Fig.~\ref{fig:ingresspred} yields $R_{\rm disk}/a=0.320(5)$.
428: 
429: Fourth, the duration of the WD ingress and egress ($35 \pm 4$ s)
430: is directly related to the visible diameter of the WD (e.g.
431: \citealt{araujo03}).  Assuming that the WD is fully visible (e.g. not
432: partially blocked by the disk), we can use this to estimate a WD
433: radius for each allowed $q(i)$ pair.  Finally, we use a theoretical WD
434: mass-radius relation \citep{nauenberg72} to deduce a WD mass (and
435: therefore also $M_2$, since $q$ is known).  The situation is shown
436: graphically in Fig.~\ref{fig:qMconstraints}.  The best estimates are $M_1=0.83(8),
437: M_2=0.057(8)$ M$_{\odot}$.
438: 
439: The binary separation $a$ then follows from Kepler's Third Law:
440: $a=0.53(1)\ R_{\odot}$.  The secondary radius $R_2$ follows by
441: combining the $R_2/a$ relation given by \citet{eggleton83} with the
442: measured values for $q$ and $a$.  The result is $R_2 = 0.097(4)
443: R_{\odot}$.
444: 
445: 
446: \section{Nature and Evolution}
447: 
448: The component masses of SDSS1507 are reminiscent of dwarf novae
449: with low accretion rate and long recurrence time, sometimes called the
450: ``WZ Sagittae stars".  Most such stars do not eclipse and thus do not
451: have precisely known parameters, but WZ Sge itself appears to be
452: similar, with likely masses near $M_2 \sim 0.05$ M$_{\odot}$, $M_1 \sim
453: 1$ M$_{\odot}$ (e.g. \citealt{patterson98, skidmore00}) 
454: \footnote{A quite different value of $M_2$ (near 0.10 M$_{\odot}$) appears
455: commonly in the literature.  This comes from applying Kepler's Laws
456: with the observed $K_1$ of the emission lines.  But because these
457: lines are significantly phase-shifted, because the distortion
458: consistently increases the derived $M_2$ (see Figure 10 and Appendix A
459: of \citealt{patterson05-2155}), and because photometric evidence
460: suggests a much lower $M_2$, we favor the low-$M_2$ solution.}. Most
461: other class members are probably similar, on evidentiary grounds of
462: low $q$ and/or low accretion rate (see Table 3 of
463: \citealt{patterson05-2155}).  In SDSS1507 the eclipse suggests
464: $V=20.6(3)$ for the BS near its eclipse phase \footnote{We use this as
465: a fiducial because it is well-specified by the eclipse, and because it
466: roughly agrees with the height of the orbital hump.  But the observed
467: hump is merely the asymmetrical part of the bright-spot luminosity; a
468: slightly larger fraction of the total light could arise from the
469: bright spot.}; at 230 pc, this implies $M_V = 13.8(7)$, which is very
470: faint for dwarf novae generally.  For the total accretion light
471: averaged over orbital phase, we similarly have $V=19.3(3)$ and
472: $M_V=12.3(6)$.
473: 
474: Since WZ Sge itself appears to have a similar structure, and has 
475: a precisely known
476: distance (43 pc; \citealt{thorstensen03, harrison04}), we can use it
477: to scale estimates of SDSS1507.  The BS eclipse in WZ Sge is 11(1)\%
478: of the total light, which implies $V = 17.6(3)$ at the eclipse phase,
479: and therefore $M_V=14.4(3)$.  This substantially agrees with the
480: estimate of $M_V = 13.8(7)$ in SDSS1507.  The stars also resemble each
481: other in spectrum (strong and broad doubled emission lines, plus WD
482: absorption features).
483: 
484:       However, SDSS1507 is certainly very unusual in two respects: the high
485: space velocity and the short orbital period (Fig.~\ref{fig:vtrans}).  The high space
486: velocity is hard to explain: Galactic-disk stars of $\sim$ 1 M$_{\odot}$ 
487: normally show space
488: velocities \simle 50 km s$^{-1}$, and our estimate of 167 km s$^{-1}$ is more
489: suggestive of a Galactic-halo membership.  Then there is the 67 min orbital
490: period, a full 15\% below the 77 min period minimum.  Roche-lobe geometry
491: constrains CV secondaries to obey a $P_{\rm orb}√\sqrt{\rho_2}$ = constant 
492: relation \citep{faulkner72}, so another way
493: of putting this is ``an oddly short period implies an oddly dense secondary".
494: What would cause that?
495: 
496: Well, there are a few CVs with oddly short periods, but all have known or
497: strongly suspected high helium abundance.  This is obviously true for the
498: pure-helium secondaries in the AM CVn class \citep{faulkner72,nelemans01}, and
499: is strongly favored for a few other dwarf novae as well (V485 Cen: 
500: \citealt{augusteijn96}; 
501: EI Psc: \citealt{thorstensen02, skillman02}), on grounds
502: of very strong He I emission from the disk.  This does not much resemble
503: SDSS1507, which has helium lines of ordinary strength for low-$\dot M$ dwarf
504: novae (with He I $\lambda6678/ H\alpha$ 
505: near 0.1).  Although high He abundance is a simple and
506: effective way to explain a dense secondary, it does not seem to be supported
507: here by evidence.
508: 
509: Another way, strangely enough, is a metal-poor secondary. Cool
510: metal-poor stars lack opacity sources in the envelope; so for a given
511: mass, they radiate their luminosity at a smaller radius.  For late M
512: stars, the radii of Population II models are about 20\% smaller than
513: their Population I counterparts of equal mass \citep{dantona87}.  With
514: the $P_{\rm orb} \sqrt{\rho_2}$~relation, this moves the minimum period from 
515: 77 min to roughly 58 min, and hence leaves plenty of room for SDSS1507.  Very
516: similar results were found in the models of \citealt{stehle97} 
517: (see their Figure 1 and Table 2).
518: 
519: Population II membership also explains the high space velocity
520: in a natural way, since solar-neighborhood members of that class
521: always have high velocity.  That bodes well for a scientific theory --
522: a single hypothesis explaining the two outstanding anomalies
523: ({\it hypotheses non multiplicanda sine necesitate}; \citealt{ockham1330}).  But
524: is its a priori likelihood too low?
525: 
526:       Well, roughly 0.5\% of stars in the solar neighborhood are
527: Population II stars ``just passing through" (\citealt{gould98}) \footnote{
528: This percentage varies from 0.2 at the Galactic plane to 2 at $z=1$ kpc;
529: we estimate at $z=200 pc$, typical for CVs.}.
530: About 200 CVs of short orbital period are known, so finding
531: one Greek among the Trojans seems plausible, assuming that different
532: stellar populations are comparably talented in forming CVs.
533: 
534:      Does this prove halo membership?  Nope.  Among stars generally,
535: space velocities near 188 km s$^{-1}$ also include members of intermediate
536: populations (e.g. ``thick disk").  We do not yet have enough information
537: (and enough space velocity) to certify halo membership.  Future
538: spectroscopic study might change this, if the binary components'
539: pollution of each other's surface is not too problematic.
540: 
541: 
542: \section{Pulsations Revisited}
543: 
544:       Isolated WDs with nonradial pulsation - the ZZ Ceti stars - have
545: well-measured temperatures and masses.   Spectroscopy and colors,
546: along with measured parallax, accurately establish $\log g$ and $T_{\rm eff}$, and the
547: pulsations appear to afflict all WDs in a narrow temperature range.  For
548: garden-variety WDs of 0.6 M$_{\odot}$, that range appears to be 11000-12300 
549: K (the
550: instability strip).  However, theory suggests that the $T_{\rm eff}$ boundaries depend
551: slightly on mass.   \citet{fontaine03} calculate that the center and width
552: of the strip should increase with mass, and observations appear to support
553: this (see Figure 6 of \citealt{gianninas05}).  In particular,
554: the expected instability strip for the the 0.83 $\pm$ 0.08 M$_{\odot}$ WD mass in
555: SDSS1507 is 11400-12900 K.  Adding uncertainty for the general dearth of WD
556: variability studies for masses this high, we would characterize the ``expected"
557: range as 11300-13000 K (in the scale used by \citealt{gianninas05}; this might
558: vary slightly from author to author).
559: 
560: 
561: Fussing over these numbers is worthwhile, because we still do not know
562: (and neither does anybody else) whether the WDs in GW Lib binaries are
563: {\it merely} ZZ Cet stars, or whether accretion and/or
564: membership in a close binary plays an important role in the excitation and
565: properties of the pulsations.  On the observational side, we need to know the
566: WD temperatures, and available estimates are still crude.   Temperature
567: measures in ZZ Cet stars are typically  $\pm$ 200 K, when a fit to the full
568: spectrum can be made with no needed corrections.  For CVs, the uncertainty is
569: much larger.  Other light sources (disk plus BS) always contaminate, and this
570: is very difficult to subtract since there are no cases where it can be
571: measured in the absence of the WD.  Also, the distances to most CVs are poorly known,
572: so observers commonly fit only $T_{\rm eff}$ and merely assume a radius (in particular,
573: assume $\log g = 8$, corresponding to $M_1 = 0.6 M_{\odot}$).  Finally, most fits are fairly
574: poor, and are greatly improved by adding arbitrary second components of
575: different area and different temperature --– a little {\it deus ex machina} which
576: suggests that the model is somewhat flawed.
577: 
578: But the situation is not all bad.  Happily, the temperatures are
579: roughly in the range 10000-16000 K, where Lyman-$\alpha$ absorption is dominant and
580: temperature-sensitive.  Present constraints are summarized by 
581: % szkody02 was missing -- I guessed it was the GW Lib UV model-fitting paper.
582: \citet{szkody02, szkody07}:
583: temperatures are known for five GW Lib stars; four are near
584: 15000 K, and one is 10600 K.  The reporting papers do not contain much
585: discussion of errors, but our impression is that they are likely to be at
586: least  $\pm$ 1400 K.  Unfortunately, none of the five have WD masses usefully
587: constrained by other data (and this alone implies an error of $\pm$ 1000 K,
588: since $\log g$ and $T_{\rm eff}$ are highly correlated in absorption-line fitting).  
589: That being so, and with the oddball at 10600 K (V455 And; \citealt{araujo05})
590: the important $T_{\rm eff}$ questions for GW Lib stars appear to be still
591: unanswered: do they have an instability strip, and does it agree with the ZZ
592: Ceti strip?
593: 
594: SDSS1507 is of interest here, because its WD has a well-measured mass
595: and radius ($\log g= 8.4$).  The star has $T_{\rm eff} =11500 \pm$ 700 K (splitting
596: the difference between the estimates in Sec. 2), and the ZZ Ceti
597: instability strip for this mass is 11300-13000 K.  So for this star,
598: agreement with the ZZ Ceti strip is possible.  On the other hand, the
599: star accretes matter and may have a metal-poor composition, so the 
600: ``expected" instability strip is certainly an open question.
601: \citet{arras06} show that the theoretical 
602: instability strip is greatly affected by the WD mass and the composition 
603: of the envelope.  This challenges obervers to find ways of estimating 
604: the mass of these pulsating WDs.  (And, to spice up the wish list, the 
605: composition of accreted gas!)
606: 
607: 
608: \section{A Recently Formed CV?}
609: 
610: As this paper was being written, we received a preprint from L07
611: containing a similar discussion of SDSS1507.  Comparison of the two
612: papers might be useful.
613: 
614: The quality of the orbital light curves appears to be similar, and
615: comparison of the light curves (their Figure 1 and our Figure~\ref{fig:onenight}) shows
616: that the star was in a similar state (quiescence).  L07 present
617: filtered photometry, however, and explicitly fit a WD flux
618: distribution; this is likely to yield more accurate WD properties,
619: so their estimate of $T_{\rm eff}$ (11000 $\pm$ 500 K versus 12000 $\pm$
620: 1200) is perhaps better.  We obtained much more photometric
621: coverage (55 orbits versus $\sim$6), which perhaps explains why we found
622: the WD pulsations.       
623: 
624: The accuracy of the derived binary parameters (our Table 4 and their
625: Table 4) is worth some attention.  There is generally good agreement
626: within our uncertainties.  But L07 report very small errors,
627: about six times smaller than ours.  In part, this is probably
628: because their data are slightly better (bigger telescope,
629: faster time resolution, all multicolor).  In part it is probably
630: because their method of ``parameter estimation" tends to use all the
631: information in the light curve to constrain each component, whereas
632: our method is more piece-by-piece.  Both methods are vulnerable to
633: their untested assumptions: e.g., that the ``white dwarf" is
634: actually round and fully visible; that the bright spot's photometric
635: center occurs at the disk periphery; that the bright spot's
636: photometric center occurs at the stream/disk impact point; and that
637: the disk is circular.  And this litany of potential vulnerabilities
638: is not complete; it comes from about two minutes of passing pessimism.
639: 
640: L07 also favor a very different model for SDSS1507: a CV recently
641: formed from a detached WD - brown dwarf pair (see \citealt{politano04}
642: for a discussion of the theory).  The idea is that the “zero-age
643: assumption enables the brown dwarf to keep a relatively high density,
644: and not distend due to the loss of thermal equilibrium that
645: appears to afflict other CVs.  This formally does the trick -- oddly,
646: by invoking great youth rather than great age!  But we disfavor this
647: idea.  First, it leaves the high space velocity entirely unexplained.
648: Second, it requires an auxiliary postulate: that SDSS1507 is {\it unique}.
649: \citet{politano04} suggests that such CVs might be fairly common, because
650: brown dwarfs have long thermal timescales and can remain at their
651: natural radius for 1-2 Gyr in a CV environment (losing mass at only
652: $10^{-11}$ M$_{\odot}$ yr$^{-1}$).  But this should produce a significant 
653: ``zero-age to terminal-age" spread among the secondaries, which would be 
654: manifest by a large dispersion about the mean in mass-radius, and 
655: a quite ``soft"
656: period minimum.   Observations seem to reveal a tight mass-radius
657: relation (Figure 12 of \citealt{patterson05-epsq}, Figure 6 of 
658: \citealt{knigge06}), and a proverbial brick wall at $P = 77$ min (Figure 1).  
659: 
660: Elsewhere \citep{patterson84, patterson98, patterson01} we have speculated why the
661: observed period minimum is shifted from the pure-GR theoretical  value
662: of 70 min (because mass-transfer is slightly enhanced over the pure-GR
663: expectation), and why it is a brick wall (because the enhancement is
664: roughly the same in all CVs) \footnote{It remains
665: possible, and we earlier advocated \citep{patterson98}, that the period
666: minimum actually {\it is} soft, because the enhancement varies significantly
667: from star to star.  This is an important point often missed in the
668: literature on the subject.  It would explain the much-discussed
669: absence of a ``spike" at 77 min \citep{patterson98, king02}. 
670: But if so, then the theoretical minimum should range
671: from 77 to (say) 90 min - which still leaves 77 as the minimum, and a
672: 67 min H-rich binary still unexplained and mysterious.}.
673: But regardless of the
674: merit of those speculations, a H-rich system at $P = 67$ min is indisputably
675: odd. The L07 model can make such a binary, but has no natural way to
676: make only {\it one}; the star just doesn't have any close relatives.  A halo
677: model better survives encounter with Ockham's Razor, because it explains
678: both surprising facts, not merely the period, and because it does not
679: require any extra novelty to explain why the star is unique (the Galactic
680: halo is known to exist and to contribute $\sim$0.5\% of stars to the solar
681: neighborhood).
682: 
683: \section{Other Halo CVs?}
684: 
685: Are there other CVs which belong to the Galactic halo?  Well, they
686: must be extremely rare.  There are really only three grounds for
687: invoking halo membership: 
688: \begin{enumerate}
689: \item{Very high space velocity (decent evidence if high enough).} 
690: \item{Very low metallicity (quite good evidence).}
691: \item{Residence in a globular cluster (excellent evidence).}
692: \end{enumerate}
693: 
694: As for (1), SDSS1507 is one of just two outliers among the 354 stars
695: we considered.  We also inspected a secondary list of $\sim$150 CVs with
696: proper motion and rougher distance constraints (based on disk
697: ``standard candle" assumptions only).  None showed remarkably high
698: transverse velocity ($>$100 km s$^{-1}$).  So there are just two among
699: 500, nearly all in the solar neighborhood (within 1 kpc).
700: 
701: As for (2), no CV has ever been certified to be of low metallicity.
702: Abundance studies in CVs are still basically beyond our reach, except
703: in a few cases where the stars fairly scream at us (the helium stars).
704: 
705: As for (3), globular clusters have been closely studied for
706: binary-star members, and a few dozen likely CVs have been found.
707: Little is known about most of them.  There is one classical nova (T
708: Sco in M80, \citealt{baxendell02}), and one dwarf nova with a likely 6.7 hr
709: orbital period (V101 in M5; \citealt{neill02}).  Since CVs in globular
710: clusters probably form by capture processes, these stars are in a
711: special category.
712: 
713: To these three grounds we can add a fourth of much lower scientific
714: weight, but by far the most common argument for invoking halo
715: membership: height above the galactic plane, suggested by the galactic
716: latitude and an assumed distance.  This has revealed some odd stars,
717: but just a tiny handful at heights above 1 kpc; and even in that
718: lonely place, there appear to be plenty of Galactic-disk stars ``just
719: visiting".
720: 
721: So halo CVs are mighty rare.  The only other candidate in the  solar
722: neighborhood, BF Eri, has an even higher transverse velocity  (~400 km
723: s$^{-1}$ at the likely 800 pc distance), but presents a puzzle since
724: it shows normal-looking K-star absorption features in the spectrum
725: \citep{sheets07}.  This could be consistent with a Population II
726: origin, followed by pollution of the secondary's photosphere by ejecta
727: from the WD $\ldots$ or with an ordinary origin, followed by a violent
728: disruption event (perhaps a supernova, or debris pirated from a
729: passing galaxy).  However, in the spirit of Ockham's Razor, the
730: need for such auxiliary hypotheses could also mean that we
731: profoundly misunderstand the true meaning of these high velocities.
732:    
733: 
734: %   10. SUMMARY
735: 
736: \section{Summary}
737: 
738: \begin{enumerate}
739: \item{We report a photometric, astrometric, and spectroscopic study of the
740: peculiar CV SDSS1507, based on 2005-7 data.  The star showed deep
741: eclipses recurring with a 66.7 min orbital period -- shockingly lower than
742: the famous 77 min ``minimum period" for hydrogen-rich CVs.}
743: \item{The light curve repeated exactly during each orbit, showing a
744: double-humped wave and quite sharply defined white-dwarf and bright-
745: spot components.  We measured all the contact times and used them to
746: construct a photometric model of the system, with details in Section 5 and
747: Table 4.  Apart from the strange orbital period, the model resembles a
748: commonplace ``WZ Sge" binary, with a 0.83(8) M$_{\odot}$ white dwarf, a 
749: 0.057(8) M$_{\odot}$
750: brown-dwarf secondary, and a bright-spot at the periphery of a small disk.}
751: \item{The spectroscopy also looks fairly ordinary for an eclipsing low-$\dot M$
752: dwarf nova in quiescence: bright and broad emission lines, with typical
753: He/H line strengths, and accompanied by a strong s-wave which wheels about
754: with a semi-amplitude of 840 $\pm$ 30 km s$^{-1}$.  The line wings move
755: with $K_1 = 75 \pm 12$ km s$^{-1}$, and reach maximum recession at $\phi =0.90
756: \pm 0.03$.  As usual in CVs, these values of $K_1$ and $\phi_0$φ
757: disqualify $K_1$ as a diagnostic of actual binary motion.}
758: \item{The light curves show obvious short-period pulsations, at frequencies
759: incommensurate with the orbital frequency.  Because the signals are fast
760: (500-1200 s) and fairly coherent, and because the white dwarf dominates the
761: light, a white-dwarf origin seems likely.  And because they are
762: non-commensurate, a natural explanation is nonradial pulsation.  SDSS1507
763: should be considered a new member of the ``GW Lib" class: intrinsically
764: faint CVs with multiperiodic fast signals, probably due to nonradial
765: white-dwarf pulsation.  This is the first accurate mass determination for a
766: GW Lib star.}
767: \item{The very high proper motion of 160 mas yr$^{-1}$ suggests a very nearby star.
768: But ``white-dwarf parallax" suggests a distance of 202 $\pm$ 40 pc; and the
769: trigonometric parallax of 0.7 $\pm$ 2.5 mas also rules out a nearby
770: location.  Our Bayesian estimate includes both constraints and yields 230
771: $\pm$ 40 pc.  This implies a transverse velocity of 164 $\pm$ 30 km
772: s$^{-1}$, a space velocity 167 $\pm$ 30 km s$^{-1}$, and galactic
773: components $[U,V,W] = [135,-93,+31]$ km s$^{-1}$.  Truly a swift wanderer
774: among the stars.}
775: \item{Our binary parameters appear to be in fair agreement with those derived
776: by L07 with a somewhat different method (but also mainly based on the
777: eclipse shape).  But we favor a different interpretation: as an ``ordinary
778: Population II CV".  This readily accounts for the star's two great
779: anomalies: the period is shockingly low because Population II secondaries
780: are smaller, and the space velocity is high because that is a
781: characteristic of Population II.}
782: \item{Very few Population II CVs have yet been identified outside of globular
783: clusters; among the H-rich stars, only two seem to have acceptable
784: space-velocity credentials.  That seems a plausible number, though, because
785: Population II represents $\sim$0.5\% of all stars in the solar neighborhood, and
786: approximately 500 CVs are eligible by our criteria (known proper motion and
787: distance).  The case for SDSS1507 seems particularly good, since the oddly
788: short period confers on its secondary the extra credential of an oddly
789: small radius (``subdwarf").  Searches for metallic lines in the disk, and in
790: the white dwarf, are likely to test these ideas further.}
791: \end{enumerate}
792: 
793: \acknowledgments
794: {\it Acknowledgments.} We have received extensive help with the observations from Jonathan
795: Kemp, Eve Armstrong, Jennifer Piscionere, and Michael Shulman.
796: Discussions with Isabelle Baraffe about binary-star evolution and the
797: structure of low-mass stars were also very helpful.  This research was
798: supported by the NSF through grants AST-0406813, AST-0307413, and 
799: AST-0708810.
800: 
801: 
802: \begin{thebibliography}
803: 
804: \bibitem[Araujo-Betancor et al.(2003)]{araujo03} 
805: Araujo-Betancor, S., et al.\ 2003, \apj, 583, 437 
806: 
807: \bibitem[Araujo-Betancor et al.(2005)]{araujo05} 
808: Araujo-Betancor, S., et al.\ 2005, \aap, 430, 629 
809: 
810: \bibitem[Arras et al.(2006)]{arras06} Arras, P., Townsley, 
811: D.~M., \& Bildsten, L.\ 2006, \apjl, 643, L119 
812: 
813: \bibitem[Augusteijn et al.(1996)]{augusteijn96} Augusteijn, T., van 
814: der Hooft, F., de Jong, J.~A., \& van Paradijs, J.\ 1996, \aap, 311, 889 
815: 
816: \bibitem[Baxendell(1902)]{baxendell02} Baxendell, J.\ 1902, \aj, 
817: 22, 127 
818: 
819: \bibitem[Bessell(1990)]{bessell90} Bessell, M.~S.\ 1990, \pasp, 
820: 102, 1181 
821: 
822: \bibitem[Chanan et al.(1976)]{chanan76} Chanan, G.~A., 
823: Middleditch, J., \& Nelson, J.~E.\ 1976, \apj, 208, 512 
824: 
825: % apparently not cited
826: %\bibitem[Ciardi et al.(1998)]{ciardi98} Ciardi, D.~R., Howell, 
827: %S.~B., Hauschildt, P.~H., \& Allard, F.\ 1998, \apj, 504, 450 
828: 
829: \bibitem[D'Antona(1987)]{dantona87} D'Antona, F.\ 1987, \apj, 
830: 320, 653 
831: 
832: \bibitem[Eggleton(1983)]{eggleton83} Eggleton, P.~P.\ 1983, \apj, 
833: 268, 368 
834: 
835: \bibitem[Faulkner et al.(1972)]{faulkner72} Faulkner, J., 
836: Flannery, B.~P., \& Warner, B.\ 1972, \apjl, 175, L79 
837: 
838: \bibitem[Flannery(1975)]{flannery75} Flannery, B.~P.\ 1975, \apj, 
839: 201, 661 
840: 
841: \bibitem[Fontaine et al.(2003)]{fontaine03} Fontaine, G., 
842: Brassard, P., \& Charpinet, S.\ 2003, \apss, 284, 257 
843: 
844: \bibitem[Gianninas et al.(2005)]{gianninas05} Gianninas, A., 
845: Bergeron, P., \& Fontaine, G.\ 2005, \apj, 631, 1100 
846: 
847: \bibitem[Gilliland et al.(1986)]{gilliland86} Gilliland, R.~L., 
848: Kemper, E., \& Suntzeff, N.\ 1986, \apj, 301, 252 
849: 
850: \bibitem[Gould et al.(1998)]{gould98} Gould, A., Flynn, C., \& 
851: Bahcall, J.~N.\ 1998, \apj, 503, 798 
852: 
853: \bibitem[Harrison et al.(2004)]{harrison04} Harrison, T.~E., 
854: Johnson, J.~J., McArthur, B.~E., Benedict, G.~F., Szkody, P., Howell, 
855: S.~B., \& Gelino, D.~M.\ 2004, \aj, 127, 460 
856: 
857: \bibitem[Hessman(1987)]{hessman87} Hessman, F.~V.\ 1987, \apss, 
858: 130, 351 
859: 
860: \bibitem[Horne(1986)]{horne86} Horne, K.\ 1986, \pasp, 98, 609 
861: 
862: \bibitem[Kepler(1619)]{kepler1619} Kepler, J.\ 1619, {\it Harmonices
863: Mundi}, tr. J. Field (1997); American Philosophical Society
864: 
865: \bibitem[King et al.(2002)]{king02} King, A.~R., Schenker, K., 
866: \& Hameury, J.~M.\ 2002, \mnras, 335, 513 
867: 
868: \bibitem[Knigge(2006)]{knigge06} Knigge, C.\ 2006, \mnras, 373, 
869: 484 
870: 
871: % apparently not cited
872: %\bibitem[Kolb \& Stehle(1996)]{kolb96} Kolb, U., \& Stehle, 
873: %R.\ 1996, \mnras, 282, 1454 
874: 
875: % apparently not cited
876: %\bibitem[Krzeminski \& Kraft(1964)]{krzeminski64} Krzeminski, W., \& 
877: %Kraft, R.~P.\ 1964, \apj, 140, 921 
878: 
879: \bibitem[Landolt(1992)]{landolt92} Landolt, A.~U.\ 1992, \aj, 
880: 104, 340 
881: 
882: \bibitem[Littlefair et al.(2007)]{littlefair07} Littlefair, S.~P., 
883: Dhillon, V.~S., Marsh, T.~R., G{\"a}nsicke, B.~T., Baraffe, I., \& Watson, 
884: C.~A.\ 2007, \mnras, 381, 827 (L07)
885: 
886: \bibitem[Mukadam et al.(2007)]{mukadam07} Mukadam, A.~S., 
887: G{\"a}nsicke, B.~T., Szkody, P., Aungwerojwit, A., Howell, S.~B., Fraser, 
888: O.~J., \& Silvestri, N.~M.\ 2007, \apj, 667, 433 
889: 
890: \bibitem[Nauenberg(1972)]{nauenberg72} Nauenberg, M.\ 1972, \apj, 
891: 175, 417 
892: 
893: \bibitem[Neill et al.(2002)]{neill02} Neill, J.~D., Shara, 
894: M.~M., Caulet, A., \& Buckley, D.~A.~H.\ 2002, \aj, 123, 3298 
895: 
896: \bibitem[Nelemans et al.(2001)]{nelemans01} Nelemans, G., 
897: Portegies Zwart, S.~F., Verbunt, F., \& Yungelson, L.~R.\ 2001, \aap, 368, 
898: 939 
899: 
900: \bibitem[Ockham(1330)]{ockham1330} Ockham, W. of\ 1330, {\it De
901: Sacramento Altaris}, tr. by Birch, T.B. (Burlington, Iowa:
902: Iowa Lutheran Literary Board)
903: 
904: \bibitem[Patterson(1984)]{patterson84} Patterson, J.\ 1984, \apjs, 
905: 54, 443 
906: 
907: \bibitem[Patterson(1998)]{patterson98} Patterson, J.\ 1998, \pasp, 
908: 110, 1132 
909: 
910: \bibitem[Patterson(2001)]{patterson01} Patterson, J.\ 2001, \pasp, 
911: 113, 736 
912: 
913: \bibitem[Patterson et al.(2005)]{patterson05-2155} Patterson, J., 
914: Thorstensen, J.~R., \& Kemp, J.\ 2005, \pasp, 117, 427 
915: 
916: \bibitem[Patterson et al.(2005)]{patterson05-epsq} Patterson, J., et 
917: al.\ 2005, \pasp, 117, 1204 
918: 
919: \bibitem[Politano(2004)]{politano04} Politano, M.\ 2004, \apj, 
920: 604, 817 
921: 
922: %\bibitem[Robinson et al.(1978)]{robinson78} Robinson, E.~L., 
923: %ather, R.~E., \& Patterson, J.\ 1978, \apj, 219, 168 
924: 
925: \bibitem[Schneider \& Young(1980)]{schneider80} Schneider, D.~P., 
926: \& Young, P.\ 1980, \apj, 238, 946 
927: 
928: \bibitem[Shafter(1983)]{shafter83} Shafter, A.~W.\ 1983, \apj, 
929: 267, 222 
930: 
931: \bibitem[Sheets et al.(2007)]{sheets07} Sheets, H.~A., 
932: Thorstensen, J.~R., Peters, C.~J., Kapusta, A.~B., \& Taylor, C.~J.\ 2007, 
933: \pasp, 119, 494 
934: 
935: \bibitem[Skidmore et al.(2000)]{skidmore00} Skidmore, W., Mason, 
936: E., Howell, S.~B., Ciardi, D.~R., Littlefair, S., \& Dhillon, V.~S.\ 2000, 
937: \mnras, 318, 429 
938: 
939: \bibitem[Skillman et al.(2002)]{skillman02} Skillman, D.~R., et 
940: al.\ 2002, \pasp, 114, 630 
941: 
942: \bibitem[Spruit \& Rutten(1998)]{spruit98} Spruit, H.~C., \& 
943: Rutten, R.~G.~M.\ 1998, \mnras, 299, 768 
944: 
945: \bibitem[Steeghs et al.(2007)]{steeghs07} Steeghs, D., Howell, 
946: S.~B., Knigge, C., G{\"a}nsicke, B.~T., Sion, E.~M., \& Welsh, W.~F.\ 2007, 
947: \apj, 667, 442 
948: 
949: \bibitem[Stover(1981)]{stover81} Stover, R.~J.\ 1981, \apj, 249, 
950: 673 
951: 
952: \bibitem[Stehle et al.(1997)]{stehle97} Stehle, R., Kolb, U., \& 
953: Ritter, H.\ 1997, \aap, 320, 136 
954: 
955: \bibitem[Szkody et al.(2002)]{szkody02} Szkody, P., 
956: G{\"a}nsicke, B.~T., Howell, S.~B., \& Sion, E.~M.\ 2002, \apjl, 575, L79 
957: 
958: \bibitem[Szkody et al.(2005)]{szkody05} Szkody, P., et al.\ 
959: 2005, \aj, 129, 2386 
960: 
961: \bibitem[Szkody et al.(2007)]{szkody07} Szkody, P., et al.\ 
962: 2007, \apj, 658, 1188 
963: 
964: \bibitem[Taylor et al.(1999)]{taylor99} Taylor, C.~J., 
965: Thorstensen, J.~R., \& Patterson, J.\ 1999, \pasp, 111, 184 
966: 
967: \bibitem[Thorstensen(2003)]{thorstensen03} Thorstensen, J.~R.\ 2003, 
968: \aj, 126, 3017 
969: 
970: \bibitem[Thorstensen et al.(1991)]{thorstensen91} Thorstensen, J.~R., 
971: Ringwald, F.~A., Wade, R.~A., Schmidt, G.~D., \& Norsworthy, J.~E.\ 1991, 
972: \aj, 102, 272 
973: 
974: \bibitem[Thorstensen et al.(2002)]{thorstensen02} Thorstensen, J.~R., 
975: Fenton, W.~H., Patterson, J.~O., Kemp, J., Krajci, T., \& Baraffe, I.\ 
976: 2002, \apjl, 567, L49 
977: 
978: % reference incomplete, apparently not cited
979: %    Townsley, D. & Bildsten, L. 2004, ApJ,
980: 
981: % Apparently not cited
982: %\bibitem[van Paradijs et al.(1996)]{vanparadijs96} van Paradijs, J., 
983: %Augusteijn, T., \& Stehle, R.\ 1996, \aap, 312, 93 
984: 
985: \bibitem[Warner(1995)]{warner95} Warner, B.\ 1995, Cambridge 
986: Astrophysics Series, Cambridge, New York: Cambridge University Press,
987: |c1995
988: 
989: %\bibitem[Warner \& Woudt(2005)]{warner05} Warner, B., \& Woudt, 
990: %P.~A.\ 2005, 14th European Workshop on White Dwarfs, 334, 453 
991: 
992: \bibitem[Wood et al.(1989)]{wood89} Wood, J.~H., Horne, K., 
993: Berriman, G., \& Wade, R.~A.\ 1989, \apj, 341, 974 
994: 
995: %  Wood, M.A. 1995,  -- apparently not cited
996: 
997: \bibitem[Zacharias et al.(2004)]{zacharias04} Zacharias, N., Urban, 
998: S.~E., Zacharias, M.~I., Wycoff, G.~L., Hall, D.~M., Monet, D.~G., \& 
999: Rafferty, T.~J.\ 2004, \aj, 127, 3043 
1000: 
1001: \end{thebibliography}
1002: 
1003: \clearpage
1004: \begin{figure}
1005: \epsscale{0.81}
1006: \plotone{f1.eps}
1007: \caption{{\it Upper frame:} Distribution of CVs in transverse velocity.
1008: All 354 H-rich CVs with proper-motion and distance estimates are 
1009: shown.  The average is 33 km s$^{-1}$, and the average uncertainty is $\sim$12 
1010: km s$^{-1}$ (mostly arising from errors in distance).  SDSS1507 sits alone 
1011: with 182 $\pm$ 30 km s$^{-1}$.  One other outlier, BF Eri, is far off-scale at 
1012: 400 km s$^{-1}$ (see text for discussion).  {\it Lower frame:} distribution of 
1013: H-rich CVs with known orbital period.  Easily seen are the 2-3 hr period
1014: gap, the 77 min period minimum $\ldots$ and SDSS1507, sitting by itself.
1015: } 
1016: \label{fig:vtrans}
1017: \end{figure}
1018: 
1019: \clearpage
1020: 
1021: \begin{figure}
1022: \epsscale{0.9}
1023: \plotone{f2.eps}
1024: \caption{ The average flux-calibrated spectrum, showing the broad
1025: white-dwarf absorption, and the doubled H and He emission lines
1026: typical of an edge-on disk.}  
1027: \label{fig:spec}
1028: \end{figure}
1029: 
1030: \clearpage
1031: 
1032: \begin{figure}
1033: \epsscale{0.9}
1034: \plotone{f3.eps}
1035: \caption{Radial velocities of the H$\alpha$ wings, folded on the eclipse
1036: ephemeris.  The best-fitting sinusoid is shown; it has 
1037: semi-amplitude $K_1 = 72 \pm 14$ km s$^{-1}$, mean 
1038: $\gamma = 46$ km s$^{-1}$, and reaches 
1039: maximum radial-velocity at φ$\phi =0.90 \pm 0.03$.}
1040: \label{fig:foldvel}
1041: \end{figure}
1042: 
1043: 
1044: \clearpage
1045: 
1046: \begin{figure}
1047: \epsscale{0.9}
1048: \plotone{f4.eps}
1049: \caption{A greyscale representation showing the changes of line
1050: profile with orbital phase.  Both H$\alpha$αand He I $\lambda$6678 
1051: show S-waves, as do the other useful lines (H$\beta$ 
1052: and He I $\lambda$5876, not shown). }
1053: \label{fig:trail}
1054: \end{figure}
1055: 
1056: \clearpage
1057: \begin{figure}
1058: \epsscale{0.7}
1059: \plotone{f5.eps}
1060: \caption{{\it Top frame:} one night's light curve, rendered on
1061: an intensity scale.  {\it Middle frame:} the
1062: composite of all 2006 light curves, folded on $P_{\rm orb}$. 
1063: {\it Bottom frame:} an expanded view of the eclipse
1064: region, showing the double structure in the eclipse.  White-dwarf
1065: mid-ingress and mid-egress occur at $\phiφ= \pm 0.0203$, and bright-spot mid-
1066: ingress and mid-egress occur at $\phiφ= 0.0114(7)$ and 0.0735(6).}
1067: \label{fig:onenight}
1068: \end{figure}
1069: 
1070: \begin{figure}
1071: \plotone{f6.eps}
1072: \caption{O-C diagram of mid-eclipse timings, relative to a test
1073: period of 0.04625834 d.  The straight line shows the best fit,
1074: Eq. (1).}
1075: \label{fig:ominusc}
1076: \end{figure}
1077: 
1078: \begin{figure}
1079: \epsscale{0.79}
1080: \plotone{f7.eps}
1081: \caption{Average nightly power spectra in 2005 and 2006; significant
1082: features are labeled with their frequencies ($\pm$ 0.6 cycles d$^{-1}$).  The 
1083: nightly time series have been prewhitened by removal of the 
1084: orbital waveform.}
1085: \label{fig:nightlypower}
1086: \end{figure}
1087: 
1088: \clearpage
1089: 
1090: \begin{figure}
1091: \epsscale{0.77}
1092: \plotone{f8.eps}
1093: \caption{Detailed power spectra of the several-night time series in
1094: 2005 and 2006; significant features are labeled with their 
1095: frequencies in cycles d$^{-1}$.  Errors in 2005 and 2006, respectively, are 0.06 
1096: and 0.04 cycles d$^{-1}$.  Inset are power spectra of an artificial signal 
1097: sampled exactly like the actual data.  In 2006, the high-frequency   
1098: signals are approximately consistent with a constant amplitude and 
1099: phase, while the 75 cycles d$^{-1}$ signal is not.  In 2005, all three signals 
1100: show an intrinsic unresolved (or undeciphered) fine structure.}
1101: \label{fig:detailedpower}
1102: \end{figure}
1103: 
1104: \begin{figure}
1105: \plotone{f9.eps}
1106: \caption{The $q(i)$ dependence for φ$\Delta \phi_{\rm WD} = 0.0406$, 
1107: with a point-source eclipsed by a Roche-lobe filling secondary.  
1108: Since we use {\it mid}-eclipse, this does not depend on the white 
1109: dwarf's radius or limb-darkening.}
1110: \label{fig:qofi}
1111: \end{figure}
1112: 
1113: \begin{figure}
1114: \plotone{f10.eps}
1115: \caption{The curves show predicted phases for mid-ingress and mid-
1116: egress of the bright spot, as a function of $q$.  The observed phases
1117: [0.0114(6) and 0.0735(6)] select $q=0.069(3)$.  Heavy shading shows the
1118: curve associated with this $q$, and light shading shows the allowed 
1119: range of disk radius.}
1120: \label{fig:ingresspred}
1121: \end{figure}
1122: 
1123: \begin{figure}
1124: \plotone{f11.eps}
1125: \caption{The shaded regions express the constraints on $q$ (0.069 $\pm$ 
1126: 0.002) and the ingress/egress time of the white dwarf (35 $\pm$ 4 s).  
1127: The star should live in the black region, with $M_2 =0.057(8)$  
1128: M$_{\odot}$, $M_1=0.83(8)$ M$_{\odot}$.}
1129: \label{fig:qMconstraints}
1130: \end{figure}
1131: 
1132: \clearpage
1133: 
1134: \begin{deluxetable}{lrrrrr}
1135: \tablewidth{0pt}
1136: \tablecolumns{6}
1137: \tablecaption{Journal of Astrometric Observations}
1138: \tablehead{
1139: \colhead{UT Date} & 
1140: \colhead{$N_{\rm obs}$} & 
1141: \colhead{HA$_{\rm start}$} &
1142: \colhead{HA$_{\rm end}$} & 
1143: \colhead{$p_X$\tablenotemark{a}} &
1144: \colhead{$p_Y$} \\
1145: \colhead{} &
1146: \colhead{} &
1147: \colhead{[hh:mm]} &
1148: \colhead{[hh:mm]} &
1149: \colhead{} &
1150: \colhead{} \\
1151: }
1152: \startdata
1153: 2005 Jun 25 & 12 &  $+$0:26  &  $+$0:53  & $-0.69$ & $0.74$ \\
1154: 2006 Mar 13 &  5 &  $+$0:18  &  $+$0:36  & $0.80$ & $0.43$ \\
1155: 2006 May 20 & 14 &  $+$0:56  &  $+$2:24  & $-0.17$ & $0.95$ \\
1156: 2006 Jun 10 &  9 &  $-$0:08  &  $+$0:23  & $-0.49$ & $0.87$ \\
1157: 2006 Aug 29 &  4 &  $+$2:59  &  $+$3:13  & $-0.93$ & $-0.18$ \\
1158: 2007 Jan 28 &  8 &  $-$1:00  &  $-$0:36  & $0.93$ & $-0.27$ \\
1159: 2007 Mar 25 & 10 &  $-$0:22  &  $+$0:18  & $0.68$ & $0.59$ \\
1160: 2007 Jun 21 &  8 &  $+$0:02  &  $+$0:38  & $-0.64$ & $0.78$ \\
1161: \enddata
1162: \tablenotetext{a}{Parallax factors $p_X$ and $p_Y$ are the 
1163: parallax displacements that a star at 1 pc distance would
1164: undergo on that date, in $X$ (eastward) and $Y$ (northward).}
1165: \end{deluxetable}
1166: 
1167: %
1168: %   Ta.bli 2. Spectroscopic data. Probably EW and flux for each line,
1169: %            and parameters of fit sinusoid (maybe)
1170: 
1171: \begin{deluxetable}{lrcc}
1172: %\tabletypesize{\scriptsize}
1173: \tablewidth{0pt}
1174: \tablecolumns{4}
1175: \tablecaption{Spectral Features in Quiescence}
1176: \tablehead{
1177: &
1178: \colhead{E.W.\tablenotemark{a}} &
1179: \colhead{Flux}  &
1180: \colhead{FWHM} \\
1181: \colhead{Feature} &
1182: \colhead{(\AA )} &
1183: \colhead{(10$^{-17}$ erg cm$^{-2}$ s$^{-1}$)} &
1184: \colhead{(\AA)} \\
1185: }
1186: \startdata
1187:             H$\beta$ & 25 & 9 & 40 \\
1188:   HeI $\lambda 5876$ & 9 & 8: & 49 \\
1189:            H$\alpha$ & 87 & 15 & 51 \\
1190:   HeI $\lambda 6678$ & 12 & 1.5 & \nodata \\
1191: \enddata
1192: \tablenotetext{a}{Emission equivalent widths are counted as positive. 
1193: No attempt has been made to measure the absorption around H$\beta$, or
1194: to correct the H$\beta$ equivalent width for the underlying absorption.}
1195: \end{deluxetable}
1196: 
1197: \begin{deluxetable}{rrr}
1198: %\tabletypesize{\scriptsize}
1199: \tablewidth{0pt}
1200: \tablecolumns{3}
1201: \tablecaption{Times of Mid-Eclipse}
1202: \tablehead{}
1203: %\colhead{$t$} & 
1204: %\colhead{$t$} & 
1205: %\colhead{$t$} \\
1206: %}
1207: \startdata 
1208: 498.89228  &   498.93851  &   498.98475  \\
1209: 502.68543  &   502.73171  &   502.77795  \\
1210: 502.82422  &   502.87048  &   502.91672  \\
1211: 502.96301  &   503.84190  &   503.88814  \\
1212: 503.93441  &   504.85958  &   504.90583  \\
1213: 504.95209  &   527.66492  &   527.71120  \\
1214: 527.75746  &   527.80373  &   527.84998  \\
1215: 527.89624  &   533.67855  &   533.72482  \\
1216: 533.77104  &   534.69620  &   534.74245  \\
1217: 534.78871  &   534.83504  &   878.81200  \\
1218: 878.85828  &   878.90452  &   878.95079  \\
1219: 879.78347  &   879.87593  &   879.92224  \\
1220: 879.96851  &   880.70860  &   880.75486  \\
1221: 880.80111  &   880.84740  &   880.89360  \\
1222: 881.68008  &   881.72630  &   881.77258  \\
1223: 881.81879  &   885.84327  &   885.88953  \\
1224: 885.93583  &  1271.86909  &  1271.91539  \\
1225: 1272.84055 &  1272.88679  &  1273.81198  \\
1226: 1273.85826  \\
1227: \enddata
1228: \tablecomments{Times listed are the HJD of 
1229: mid-eclipse, minus 2453000.}
1230: \end{deluxetable}
1231: 
1232: % Table 4. Model parameters from eclipse fitting. (to be supplied)
1233: \begin{deluxetable}{lr}
1234: \tablewidth{0pt}
1235: \tablecolumns{2}
1236: \tablecaption{Model Parameters from Eclipse Fitting}
1237: \tablehead{
1238: \colhead{Quantity} & 
1239: \colhead{Value\tablenotemark{a}} \\
1240: }
1241: \startdata 
1242: Inclination $i$  &  83.18(13) deg \\
1243: Mass ratio $q$  &    0.069(2)   \\
1244: $M_1$ & 0.83(8) M$_\odot$  \\
1245: $M_2$ & 0.057(8) M$_\odot$  \\
1246: $R_1$ & 0.0097(9) R$_\odot$  \\
1247: $R_2$ & 0.097(4) R$_\odot$  \\
1248: $T_{\rm wd}$  & 11500(700) K   \\
1249: $R_{\rm disk}/a$ & 0.320(5)    \\
1250: Distance $d$ & 230(40) pc   \\
1251: \enddata
1252: \tablenotetext{a}{Uncertainties in the last digit(s) 
1253: are given in parentheses.}
1254: \end{deluxetable}
1255: \end{document}
1256: