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

2: \documentclass[onecolumn]{emulateapj}

3: %\usepackage{graphicx}

4: \newcommand{\vdag}{(v)^\dagger}

5: \newcommand{\myemail}{gokhale@baton.phys.lsu.edu}

6: \renewcommand{\d}{{\rm d}}

7: \newcommand{\sgn}{{\rm sgn}}

8: \renewcommand{\thefootnote}{\fnsymbol{footnote}}

9: 

10: 

11: \shorttitle{Evolution of Double White Dwarfs}

12: \shortauthors{Gokhale et al.}

13: 

14: \begin{document}

15: \title{Evolution of Close White Dwarf Binaries}

16: \author{Vayujeet Gokhale, Xiao Meng Peng, Juhan Frank}

17: \affil{Department of Physics and Astronomy,\\ Louisiana State

18: University, Baton Rouge, LA 70803-4001.\\ \myemail.}

19: 

20: \begin{abstract}

21: We describe the evolution of double degenerate binary systems,

22: consisting of components obeying the zero temperature mass radius

23: relationship for white dwarf stars, from the onset of mass transfer

24: to one of several possible outcomes including merger, tidal

25: disruption of the donor, or survival as a semi-detached AM CVn

26: system. We use a combination of analytic solutions and numerical

27: integrations of the standard orbit-averaged first-order evolution

28: equations, including direct impact accretion and the evolution of

29: the components due to mass exchange. We include also the effects of

30: mass-loss during super-critical (super-Eddington) mass transfer and

31: the tidal and advective exchanges of angular momentum between the

32: binary components. We cover much the same ground as \citet{Maet04}

33: with the additional effects of the advective or consequential

34: angular momentum from the donor and its tidal coupling to the orbit

35: which is expected to be stronger than that of the accretor. With the

36: caveat that our formalism does not include an explicit treatment of

37: common envelope phases, our results suggest that a larger fraction

38: of detached double white dwarfs than what has been hitherto assumed,

39: survive the onset of mass transfer, even if this mass transfer is

40: initially unstable and rises to super-Eddington levels. In addition,

41: as a consequence of the tidal coupling, systems that come into

42: contact near the mass transfer instability boundary undergo a phase

43: of oscillation cycles in their orbital period (and other system

44: parameters). Unless the donor star has a finite entropy such that

45: the effective mass-radius relationship deviates significantly from

46: that of a zero temperature white dwarf, we expect our results to be

47: valid. Much of the formalism developed here would also apply to

48: other mass-transferring binaries, and in particular to cataclysmic

49: variables and Algol systems.

50: \end{abstract}

51: 

52: \keywords{accretion, accretion disks --- binaries: close ---

53: gravitational waves --- stars: novae, cataclysmic variables ---

54: stars: white dwarfs}

55: 

56: \section{Introduction}

57: \label{Intro} Binary white dwarfs which are close enough to be

58: driven together by angular momentum losses due to gravitational

59: radiation, and perhaps additional mechanisms, will undergo a phase

60: of mass transfer during which the ultimate fate of the binary is

61: decided. There are many reasons to revisit these dynamical phases of

62: the evolution of double degenerate binaries today. Chief among these

63: is that a survey of the published literature on the subject reveals

64: that many pieces of the puzzle have been explored in different

65: degrees but we still lack a uniform theoretical understanding of the

66: fate of these binaries for all possible mass ratios, orbital

67: parameters and origins. Ideally, one would like our theoretical

68: understanding to be such that given a white dwarf binary of

69: arbitrary masses and compositions at the time that the less massive

70: component gets into contact, one could reconstruct the previous

71: evolutionary pathways and the subsequent evolution to merger, tidal

72: disruption or stable mass transfer.

73: 

74: The reason for this lack of uniformity is that understanding

75: double white dwarf binaries (DWD) in this rapid phase of

76: interaction can be quite complex and demanding, necessitating

77: detailed hydrodynamics, nuclear physics, radiative transfer and

78: stellar structure. Therefore it is natural that different

79: assumptions and techniques are used when one attempts to answer a

80: question relevant to different classes of phenomena. For example,

81: the approaches adopted in studying the putative progenitors of

82: supernovae of type Ia, or the sources of gravitational waves for

83: LISA, or the progenitors of AM CVn binaries, are all very

84: different. While we shall not attempt to cover all the rich

85: physics that may be ultimately necessary to have a uniform and

86: reliable treatment of all possible outcomes of the interaction, we

87: will describe these interactions within a single semi-analytic

88: framework, making connections where appropriate to well-known

89: results already in the literature.

90: 

91:  Our particular motive for having developed the

92: understanding described in this paper is that our group has been

93: improving and running a 3-D numerical hydrocode \citep{MTF, DMTF}

94: which is currently capable of following self-consistently the

95: evolution of model white dwarf binaries through these rapid phases

96: of mass transfer, tracking mass and angular momentum with high

97: accuracy for over 30 orbital periods. In the course of numerous

98: evolutions in which we controlled the rate of driving by angular

99: momentum losses, or prescribed an arbitrary rate of pseudo-thermal

100: expansion of the donor, it became clear that we needed some simpler

101: insight into the behavior of the models. This led us to extend the

102: analytic solution of \citet{WebIb} (WI) for the time-dependent

103: behavior of the mass transfer by relaxing most of the assumptions

104: made to render the problem tractable. Here, we retain the

105: simplifying assumption of Roche geometry, but allow all the binary

106: parameters to vary self-consistently. Thus we investigate

107: numerically a system of first-order evolution equations for a

108: variety of cases and recover the WI solution when appropriate

109: conditions are applied. We then extend our insight to more general

110: cases, and comment on the applicability and limitations of our

111: approach. While we are still far from the more ambitious goal

112: described above, we think that the present investigation may also

113: provide a useful framework for other workers in the field of binary

114: evolution and especially for those interested in large-scale

115: numerical simulations of these binaries. Many of the techniques and

116: theoretical insights in this paper are applicable to other

117: semi-detached binaries and may have some relevance to contact

118: binaries.

119: 

120: \section{Basic Equations}

121: \label{Eqns} Consider a binary having nearly spherical components of

122: masses $M_1$ and $M_2$ and separation $a$. Without loss of

123: generality, we shall assume that as this binary evolves from a

124: detached to a semi-detached state, star 2 is the one that fills its

125: critical Roche lobe and becomes the donor. We define the mass ratio

126: of the binary to be $q=M_2/M_1$.  Assuming for simplicity that the

127: spin axes of the individual components of the binary are

128: perpendicular to the orbital plane, and that the orbit is circular,

129: the total angular momentum of the system is given by:

130: \begin{eqnarray}

131: \nonumber J_{\rm tot} &=& J_{\rm orb} + J_1 + J_2 \\

132:                   &=& M_1 M_2 \left(G a/M \right)^{1/2} +

133:                   k_1 M_1 R_1^2~ \omega_1 + k_2 M_2 R_2^2

134:                   ~\omega_2 ,

135:                   \label{Jtot}

136: \end{eqnarray}

137: where $k_i$ are dimensionless constants depending on the internal

138: structure of the components and $\omega_i$ are the angular spin

139: frequencies. The first term in eq. (\ref{Jtot}) represents the

140: orbital angular momentum of the components, and the two other terms

141: represent the spin angular momenta of the stars. The form of the

142: orbital angular momentum term adopted above assumes the binary

143: revolves at the Keplerian frequency $\Omega = (GM/a^3)^{1/2}$, which

144: is a good approximation if the stars are centrally condensed.

145: 

146: For a fully synchronous configuration, it is easy to show that the

147: total angular momentum and the total energy have a minimum at the

148: same separation $a_{\rm min}= [3(I_1+I_2)/\mu]^{1/2}$, where the

149: $I_i = k_i M_i R_i^2$ are the moments of inertia of the components,

150: and $\mu$ is the reduced mass. Even if the orbital frequency and the

151: spin frequency are not synchronized but remain proportional to one

152: another ($\omega_i=f_i \Omega$), there is a minimum of $J_{\rm tot}$

153: at some $a_{\rm min}$. Also the total energy of the system will have

154: a minimum but in general it will not occur at $a_{\rm

155: min}$.\footnote{LRS2 show that for Riemann-S and Roche-Riemann

156: sequences $\d E = \Omega\d J + \Lambda\d{\cal C}$, where $\Lambda$

157: is the angular velocity of internal motions and ${\cal C}$ is the

158: equatorial circulation. Thus, as a binary evolves driven by

159: gravitational wave radiation, circulation is conserved and the

160: minima of $E$ and $J$ will coincide. However tidal dissipation does

161: not preserve ${\cal C}$ and thus in general the minima will {\em

162: not} coincide in the presence of tidal spin-orbit coupling.}

163: 

164:  For our purposes it will be sufficient to work with the

165: approximate form of eq. (\ref{Jtot}) given above. A more complete

166: and thorough discussion of the secular and dynamical stability of

167: polytropic binaries has been presented in two well-known series of

168: papers by Lai, Rasio \& Shapiro (LRS1-LRS5) and further developed

169: with SPH simulations in papers by Rasio \& Shapiro (RS1-RS3), who

170: also addressed the role of mass-transfer.

171: 

172: If the donor has a relatively soft equation of state and if $q\ne

173: 1$, the binary tends to become semi-detached and mass-transfer

174: occurs before it falls prey to the tidal instability mentioned

175: above. Mass transfer changes the initial configuration as the system

176: evolves and may either drive the system to smaller separations and

177: thus closer to the onset of the tidal instability or to larger

178: separations and toward stability. Similarly, mass loss from the

179: system can affect the dynamic stability of the system. We

180: investigate the behavior of such systems (in particular DWD

181: binaries) in the cases of conservative and non-conservative mass

182: transfer, driven by gravitational wave radiation (GWR).

183: 

184: \subsection{Orbital and spin angular momentum}

185: \label{AML}

186: 

187: A system of two point masses orbiting around each other, in circular

188: orbits, radiates gravitationally \citep{LaLi}. The loss of orbital

189: angular momentum as a result is given by

190: \begin{equation}

191: \Big(\frac{\dot J}{J}\Big)_{\rm GWR} = ~ -\frac{32}{5}

192: \frac{G^3}{c^5} \frac{M_1 M_2 M}{a^4} \label{JdotGWR}

193: \end{equation}

194: Although GWR is likely to be the only important mode of angular

195: momentum loss from DWD binaries, for any general form of systemic

196: angular momentum loss $\dot J_{\rm sys}$, we can re-arrange eq.

197: (\ref{Jtot}) to obtain

198: \begin{eqnarray}

199: J_{\rm orb} = J_{\rm tot} -(J_1+J_2) \Rightarrow \dot J_{\rm orb} =

200: \dot J_{\rm sys} -(\dot J_1 + \dot J_2)\, , \label{Jorb}

201: \end{eqnarray}

202: allowing for the possibility of spin-orbit coupling of the angular

203: momenta. The rate of change of the spin angular momentum of each

204: individual star can be given as the sum of the advected or {\it

205: consequential} angular momentum transport plus the effect of the

206: tidal torques

207: \begin{eqnarray}

208: \dot J_1 = \dot M_1 j_{1} + \dot J_{\rm 1, tid}\ , \label{J1dot}

209: \\

210: \dot J_2 = \dot M_2 j_{2} + \dot J_{\rm 2, tid}\ . \label{J2dot}

211: \end{eqnarray}

212: The term denoting the advected component from the donor has not been

213: included in previous treatments concerning white dwarf donors

214: \citep{Maet04}, but has been discussed in the case of evolved donors

215: (see for example, \cite{PratStritt76} and \cite{Sav78}). For this

216: reason, a few remarks clarifying the meaning of the consequential

217: terms are in order. In the above equations $j_1$ and $j_2$ indicate

218: the specific angular momenta of the matter {\em arriving} at the

219: accretor and the matter {\em leaving} the donor respectively. In a

220: conservative system these will refer to the specific angular

221: momentum of the {\em same} material with respect to the center of

222: mass of each star, but at {\em different} times. We assume that as

223: the stream leaves the donor there is no back torque that could

224: modify the spin of the donor. Therefore $j_2$ is entirely determined

225: by the instantaneous conditions at the donor. However, the material

226: traveling in the stream experiences a time varying torque due to the

227: binary that changes $j_1$ at the expense of the orbital angular

228: momentum alone i.e. not torquing the spins of either component.

229: These are the assumptions underlying the standard calculations of

230: $j_1$ and the estimates of the circularization radius which go back

231: to \citet{Fla75} and \citet{LuSh75}. We will discuss later what are

232: the appropriate values for $j_1$ and $j_2$, and write down just the

233: general expressions here. Substituting eqs. (\ref{J1dot}) and

234: (\ref{J2dot}) in eq. (\ref{Jorb}) and rearranging

235: \begin{eqnarray}

236: \dot J_{\rm orb} = \dot J_{\rm sys} - \left[- \dot M_2(j_1-j_2) +

237: \dot J_{\rm 1, tid}+ \dot J_{\rm 2, tid}\right]\ , \label{JorbDot}

238: \end{eqnarray}

239: where we have assumed conservative mass transfer.

240: Also, we can write the tidal torque as

241: \begin{eqnarray}

242: \dot J_{\rm 1, tid} = \frac{k_1 M_1 R_1^2}{\tau_{\rm s_1}} (\Omega -

243: \omega_1) \label{J1dottid}

244: \\

245: \dot J_{\rm 2, tid} = \frac{k_2 M_2 R_2^2}{\tau_{\rm s_2}} (\Omega -

246: \omega_2) \label{J2dottid}

247: \end{eqnarray}

248: where $\tau_{\rm s_1}$ and $\tau_{\rm s_2}$ are the synchronization

249: timescales of the accretor and donor respectively (See Section

250: \ref{cycles}). Eq. (\ref{JorbDot}) becomes

251: \begin{eqnarray}

252: \dot J_{\rm orb} = \dot J_{\rm sys} +\dot M_2 (j_1-j_2) - \frac{k_1

253: M_1 R_1^2}{\tau_{\rm s_1}} (\Omega - \omega_1) - \frac{k_2 M_2

254: R_2^2}{\tau_{\rm s_2}} (\Omega - \omega_2) \label{JorbDot a}

255: \\

256: = \dot J_{\rm sys} + \frac{\dot M_2}{M_2}\Big[M_2(j_1-j_2) + k_1 M_1

257: R_1^2 \frac{\tau_{{M_2}}}{\tau_{\rm s_1}} (\Omega - \omega_1) + k_2

258: M_2 R_2^2 \frac{\tau_{{M_2}}}{\tau_{\rm s_2}}(\Omega-\omega_2)\Big]

259: \label{JorbDot b}

260: \end{eqnarray}

261: where $\tau_{M_2} = - M_2/\dot{M_2}$ is the mass transfer time

262: scale. In eq. (\ref{JorbDot b}) the tidal torques have been placed

263: inside the brackets although, strictly speaking, they are not

264: consequential. In most cases encountered in cataclysmic variables

265: (CVs) and low-mass X-ray binaries (LMXBs), $\tau_{\rm

266: s_1}\gg\tau_{M_2}$ and $\tau_{\rm s_2}\ll\tau_{M_2}$, and thus

267: usually $\omega_1\gg\Omega$, while $\vert (\Omega - \omega_2)/\Omega

268: \vert \sim \tau_{\rm s_2}/\tau_{M_2}$. However, in double degenerate

269: binaries there is considerable uncertainty about the synchronization

270: timescales, and in numerical simulations of mass transfer

271: $\tau_{M_2}$ is orders of magnitude shorter than the typical

272: timescales one encounters in long-term accreting binaries. Therefore

273: we will retain both tidal terms and investigate their effect on

274: dynamical evolutions. Note that our eq. (\ref{JorbDot a}) is

275: equivalent to eq. (1) of \citet{Maet04}, with two extra terms

276: arising from the advective and tidal contributions from the donor.

277: 

278: \subsection{Binary separation}

279: \label{binsep}

280: We now derive the equations for the evolution of the

281: binary separation, the radius of the donor  and the Roche lobe

282: radius using the above equations. From the functional form of the

283: orbital angular momentum, we know that for a conservative system

284: \begin{eqnarray*}

285: \Big(\frac{\dot J}{J}\Big)_{\rm orb} = \frac{\dot M_2}{M_2}(1-q)

286: +\frac{1}{2} \frac{\dot a}{a} .

287: \end{eqnarray*}

288: Comparing this with eq. (\ref{JorbDot a}) and rearranging, we obtain

289: \begin{eqnarray}

290: \frac{\dot a}{2a} = \frac{\dot J_{\rm sys}}{J_{\rm orb}} - \frac{

291: \dot M_2}{M_2} \Big[1 - q - M_2\frac{j_1-j_2}{J_{\rm orb}}\Big] -

292: \frac{k_1 M_1 R_1^2}{J_{\rm orb}\tau_{\rm s_1}} (\Omega - \omega_1)

293: - \frac{k_2 M_2 R_2^2}{J_{\rm orb}\tau_{\rm s_2}} (\Omega-\omega_2)

294: \label{adot}

295: \end{eqnarray}

296: The sign of the quantity in brackets determines whether mass

297: transfer tends to expand the system and thus oppose the effect of

298: angular momentum losses, or lead to enhanced contraction. The change

299: of sign will generally occur at some mass ratio $q_a$ whose value we

300: discuss below. Symbolically we may write

301: \begin{eqnarray}

302: \frac{\dot a}{2a} &=& \frac{\dot J_{\rm sys}}{J_{\rm orb}}

303: -\frac{\dot J_{\rm 1, tid}+ \dot J_{\rm 2, tid}}{J_{\rm orb}} -

304: \frac{\dot M_2}{M_2} [q_a - q] \label{qa}

305: \\

306: q_a &\equiv& 1- M_2\frac{j_1-j_2}{J_{\rm orb}} , \label{qadef}

307: \end{eqnarray}

308: Thus (remembering that $\dot M_2$ is intrinsically negative) if

309: $q>q_a$, mass transfer will contribute to reducing the separation

310: and, as we will see below, tend to make the binary more unstable to

311: mass transfer. On the other hand, if $q<q_a$, mass transfer will

312: oppose the effects of driving and will tend to stabilize the binary.

313: As $q$ decreases on mass transfer, it is possible that a system

314: which started life with $q>q_a$ may evolve to a more stable

315: configuration, if it does not fall prey to the tidal instability. In

316: eq. (\ref{qa}) the tidal synchronization torques may be considered

317: additional contributions to the driving and they may subtract or add

318: angular momentum to the orbit depending on the case. Note that $q_a$

319: should be interpreted as the mass ratio at which the last term in

320: eq. (\ref{qa}) changes sign. Its value can be estimated from the

321: initial mass ratio since it is a slowly varying function of $q$ in

322: the conservative case, but in general it is obtained

323: self-consistently as the binary evolution is followed numerically.

324: 

325: The second term in the definition of $q_a$ represents the effects of

326: the net consequential transfer of angular momentum from orbit to

327: spins. We introduce the symbol $\zeta_{\rm c} = M_2(j_1-j_2)/J_{\rm

328: orb}$ for this term, noting that it stands for the consequential

329: contribution to  $-\d\log{J_{\rm orb}}/\d\log{M_2}$. For direct

330: impact accretion $j_1=j_{\rm circ} (\approx b_1^2 \Omega)$ is the

331: specific angular momentum carried by the stream as it hits the

332: accretor. The approximate value in parenthesis is valid for a

333: synchronous donor and $b_1=a(0.5-0.227\log{q})$ is the distance from

334: the center of mass of the accretor to the inner Lagrangian point

335: $L_1$ \citep{FKR}. With the standard definition of circularization

336: radius $r_h= R_{\rm circ}/a$ \citep{Fla75, LuSh75, VeRa, Maet04},

337: $M_2 j_{\rm circ}/J_{\rm orb} = [(1+q)r_h]^{1/2}$. Note, however,

338: that the tidal coupling of the donor spin to the orbit was neglected

339: by \citet{Maet04}, and the consequential term proportional to

340: $j_2\approx R_2^2 \omega_2$ was not included either. The precise

341: value of $j_2$ depends on the details of the flow in the vicinity of

342: $L_1$ in a non-synchronous donor (See \cite{Krus63} and

343: \cite{CsatSko05}). For the purposes of this investigation we will

344: simply adopt $j_2=R_2^2 \omega_2$. With these definitions we may

345: write

346: \begin{equation}

347: q_a = 1 - \zeta_{\rm c} = 1

348: -[(1+q)r_h]^{1/2}(1-\frac{R_2^2\omega_2}{\sqrt{G M_1 R_{\rm circ}}}) .

349: \label{explicitqa}

350: \end{equation}

351: The net effect of the consequential redistribution of angular

352: momentum in the binary depends on the sign of $\zeta_{\rm c}$. For

353: DWD binaries, $j_1>j_2$ during the direct impact stage, and this holds

354: even after the onset of disk accretion, when $j_1=\sqrt{G M_1 R_1}$ and

355: $\zeta_{\rm c}$ becomes smaller but remains positive.

356: In cataclysmic variables and low-mass X-ray binaries, the accretion

357: disk returns via tides most of the angular momentum advected by the

358: stream, the donor is almost synchronous, and the tidal coupling of

359: the accretor to the orbit is very weak. However, in this case $R_1\ll a$ and thus it is

360: more likely that $j_2>j_1$. While all the additional terms in

361: eqs. (\ref{qa}, \ref{qadef}) are relatively small, yielding $q_a\approx 1$,

362: in some cases $\zeta_{\rm c}<0$, and thus

363: $q_a\ga 1$ making these systems

364: slightly more stable.

365: 

366: The expression (\ref{explicitqa}) obtained above proves very useful

367: in the interpretation of results of large-scale numerical

368: hydrodynamic simulations of the dynamical evolution of binaries

369: undergoing mass transfer, with and without driving by angular

370: momentum losses \citep{DMTF}(see also \S\ref{hydrocomp}).

371: 

372: The Roche lobe radius for the donor is accurately given by the

373: formula due to \citet{Egg83}

374: \begin{eqnarray}

375: r_{\rm L}\equiv\frac{R_{\rm L}}{a} = \frac{0.49 q^{2/3}}{0.6 q^{2/3}

376: + \ln (1+q^{1/3})} \label{Egg form}

377: \end{eqnarray}

378: and so with the notation of \citet{Maet04}

379: \begin{eqnarray*}

380: \nonumber \frac{\dot R_{\rm L}}{R_{\rm L}} = \zeta_{r_{\rm L}}

381: \frac{\dot M_2}{M_2} + \frac{\dot a}{a}\, ,

382: \end{eqnarray*}

383: where $\zeta_{r_{\rm L}}\approx 1/3$ is the logarithmic derivative

384: of $r_{\rm L}$ with respect to $M_2$\footnote{In the range $0<q\le

385: 1$, the function $\zeta_{r_{\rm L}}$ takes values between 0.32 and

386: 0.46, and is well approximated by $\zeta_{r_{\rm L}}\approx

387: 0.30+0.16 q$ for $0.1\le q\le 1$}. Collecting results, we get

388: \begin{eqnarray}

389: \frac{\dot R_{\rm L}}{R_{\rm L}}= \frac{2 \dot J_{\rm sys}}{J_{\rm

390: orb}}-2\frac{\dot J_{\rm 1, tid}+ \dot J_{\rm 2, tid}}{J_{\rm orb}}

391: -\frac{2 \dot M_2}{M_2}[q_a-\frac{\zeta_{r_{\rm L}}}{2} -q]

392: \label{RLdot}

393: \end{eqnarray}

394: Generalizing the meaning of the symbols introduced by \citet{WebIb}

395: to include tidal and consequential terms, we can write the

396: equivalent expressions

397: \begin{eqnarray} \frac{\dot R_{\rm L}}{R_{\rm

398: L}} &=& \nu_{\rm

399: L}+\zeta_{\rm L}\frac{\dot M_2}{M_2}\\

400: \nu_{\rm L} &=& \frac{2 \dot J_{\rm sys}}{J_{\rm orb}}-2\frac{\dot

401: J_{\rm 1, tid}+ \dot J_{\rm 2, tid}}{J_{\rm orb}}\\

402: \zeta_{\rm L} &=& -2 q_a+\zeta_{r_{\rm L}}+2q, \label{RLdotarr}

403: \end{eqnarray}

404: where the symbol $\nu$ stands for driving terms and the $\zeta$ for

405: logarithmic derivatives with respect to donor mass. In the same

406: spirit we write the logarithmic time derivative of the donor radius

407: $R_2 \equiv R_2 (M_2,t)$ as:

408: \begin{equation}

409: \frac{\dot R_2}{R_2} = \nu_2 +\zeta_2 \frac{\dot M_2}{M_2}

410: \label{R2dot}

411: \end{equation}

412: where $\nu_2$ = $(\partial \ln R_2/\partial t)_{M_2}$ represents the

413: rate of change of the donor radius due to intrinsic processes such

414: as thermal relaxation and nuclear evolution, whereas $\zeta_2$

415: usually describes changes resulting from adiabatic variations of

416: $M_2$ \citep{DMR}. More generally, since the radial variations due

417: to the above mentioned effects operate on different timescales, it

418: is more appropriate to think of $\zeta_2$ as the effective

419: mass-radius exponent, averaged over the characteristic timescale of

420: mass transfer. If the donors are degenerate as in the WD case, or if

421: thermal relaxation is sufficiently rapid, $\zeta_2$ is simply

422: obtained from the equilibrium mass-radius relationship for the

423: donor. But in non-degenerate donors, if the thermal relaxation

424: timescale $\tau_{\rm th}$ becomes comparable to the mass transfer

425: timescale $\tau_{M_2}= M_2/\dot M_2$, as is thought to occur during

426: the evolution of CVs, the effects of thermal lag in the donor radius

427: become important, and $\zeta_2$ deviates from the equilibrium value

428: (See Appendix \ref{Appzeta}).

429: 

430: \section{Mass transfer rate} \label{masstran}

431: 

432: The following discussion of mass transfer and its stability is rooted in

433: a similar treatment of mass transfer under consequential angular momentum

434: losses \citep{KiKo95}.

435: For all types of donor star, the mass transfer rate is a strong

436: function of the depth of contact, defined here as the amount by

437: which the donor overflows its Roche lobe $\Delta R_2\equiv

438: R_2-R_{\rm L}$, suitably normalized. We adopt the following

439: expression valid for most cases of interest:

440: \begin{equation}

441: \dot M_2 = - \dot M_0(M_1, M_2, a) f(\Delta R_2), \label{M2dot}

442: \end{equation}

443: where $\dot M_0$ is a relatively gentle function of binary

444: parameters, while $f$ is a rapidly varying dimensionless function of

445: $\Delta R_2$. For example, for polytropic donors with index $n$, $f=

446: (\Delta R_2/R_2)^{n+3/2}$ \citep{PaSi}; while for a donor with an

447: atmospheric pressure scale height $H$, $f=\exp{(\Delta R_2/H)}$ is

448: appropriate \citep{Ri88}. In fact, the exact value of the

449: normalization rate $\dot M_0(M_1, M_2, a)$ is not important at all

450: in steady state because the equilibrium rate is determined by the

451: rate of driving: given a particular value of $\dot M_0$, the depth

452: of contact will adjust to yield the transfer rate sustainable by the

453: driving. In transient situations, if the mass transfer is varying

454: rapidly, or if the depth of contact becomes large, the normalization

455: becomes relevant.

456: 

457: In principle, Eqs. (\ref{J1dottid}), (\ref{J2dottid}), (\ref{qa})

458: and (\ref{RLdot})-(\ref{M2dot}) completely specify the system and

459: can be numerically integrated, and we discuss some examples of such

460: evolutions later in \S\ref{evol}. However, before proceeding, it is

461: more illuminating to analyze the general implications of eq.

462: (\ref{M2dot}). Under the assumptions mentioned above

463: \begin{equation}

464: \ddot M_2 = -\dot M_0\frac{\partial f}{\partial\Delta

465: R_2}\Big(\frac{\dot R_2}{R_2}-\frac{\dot R_{\rm L}}{R_{\rm L}}\Big)

466: , \label{M2dbldot}

467: \end{equation}

468: where we have neglected the slower variations of $\dot M_0$ with

469: system parameters and we have also assumed that the depth of contact

470: is small $\Delta R_2<<R_2$. Thus the mass transfer rate will be

471: steady whenever the size of the donor varies in step with its Roche

472: lobe. From Eqs. (17)-(20) and (\ref{M2dbldot}) we obtain the

473: equilibrium mass transfer rate

474: \begin{eqnarray}

475: \Big(\frac{\dot M_2}{M_2}\Big)_{\rm eq} = \frac{

476: \nu_{\rm L}-\nu_2} {2(q_{\rm stable}-q)}= \frac{\nu_{\rm L} -

477: \nu_2}{\zeta_2-\zeta_{\rm L}} \label{M2doteq}

478: \end{eqnarray}

479: where

480: \begin{equation}

481: q_{\rm stable} = q_a -\frac{\zeta_{r_{\rm L}}}{2} +\frac{\zeta_2}{2}

482: , \label{qstab}

483: \end{equation}

484: is the critical mass ratio for stability of mass transfer. The

485: alternative expression on the r.h.s of eq. (\ref{M2doteq}) thus has

486: the same form as in \citet{WebIb}, except that here the driving and

487: consequential terms include the effects of tidal coupling and direct

488: impact accretion. Furthermore, we will allow these terms to vary

489: self-consistently as the evolution proceeds (See \S\ref{evol}).

490: 

491: For example, if we assume that the binary is synchronized, the orbit

492: circular, the donor is a polytrope of index $n=3/2$, and $\nu_2=0$

493: (no thermal or nuclear evolution), then $\zeta_2 \sim -1/3$,

494: $\zeta_{r_{\rm L}} \sim 1/3$, and  we get

495: \begin{eqnarray*}

496: \Big(\frac{\dot M_2}{M_2}\Big)_{\rm eq} = \frac{\dot J_{\rm

497: sys}/J_{\rm orb}}{2/3 -\zeta_{\rm c}- q}

498: \end{eqnarray*}

499: which is the familiar form in the case of direct impact DWD's

500: \citep{Maet04}, except that here $\zeta_{\rm c}$ is reduced by the

501: contribution from the donor as given by eq. (\ref{explicitqa}).

502: 

503: The equilibrium mass transfer rate was obtained by demanding that

504: $\ddot M_2=0$ in eq. (\ref{M2dbldot}).  In general, if $\dot

505: M_2\neq(\dot M_2)_{\rm eq}$, one can rewrite this equation as

506: follows:

507: \begin{equation}

508: \ddot M_2 = -2\frac{\dot M_0}{M_2} \frac{\partial f}{\partial\Delta

509: R_2} (q_{\rm stable}-q)\left[\dot M_2 - (\dot M_2)_{\rm eq}\right].

510: \label{stab}

511: \end{equation}

512: The sign of the pre-factor on the r.h.s. is negative if $q<q_{\rm

513: stable}$, thus $\dot M_2$ will tend toward the equilibrium value and

514: mass transfer is stable. If $q>q_{\rm stable}$ no attainable

515: equilibrium mass transfer exists and mass transfer is unstable.

516: 

517: \section{Analytic Solutions}

518: \label{analytic} Assuming that most of the parameters characterizing

519: the binary and the donor remain constant during (the usually much

520: faster) evolution of the accretion rate, it is possible to obtain

521: analytic solutions for the evolution of the accretion rate itself.

522: In this section we generalize the result of \citet{WebIb} to include

523: donors with an arbitrary polytropic index and with an isothermal

524: atmosphere. These can be later compared to our integrations of the

525: evolution equations in which we allow all binary and donor

526: parameters to vary self-consistently and also with the results of

527: large-scale hydrodynamic simulations \citep{DMTF}.

528: 

529: Assuming that the donor in the binary can be represented by a

530: polytrope, the mass transfer rate is given by a formula derived by

531: \citet{Jedr} assuming laminar flow, and quoted by \citet{PaSi}

532: \begin{equation}

533: -\dot M_2 = \dot M_0 \Big(\frac{R_2-R_{\rm L}}{R_2}\Big)^{n+3/2}

534: \end{equation}

535: Raising both sides of the above equation to the power $2/(2n+3)$ and

536: differentiating, we obtain

537: \begin{equation}

538: \frac{d}{dt}(-\dot M_2)^{2/2n+3} = (\dot M_0)^{2/2n+3} [(\nu_2 -

539: \nu_{\rm L})+ \frac{\dot M_2}{M_2}(\zeta_2-\zeta_{\rm L})],

540: \label{dotM2power}

541: \end{equation}

542: where we have set the factor $R_{\rm L}/R_2$ to unity, given that in

543: most situations $\Delta R_2 \ll R_2$. The analytic solutions

544: discussed here assume that the driving rate $\nu_{\rm L}$, the

545: intrinsic radial variation rate $\nu_2$ (which includes the

546: intrinsic thermal and nuclear evolution), and radial reaction

547: exponents $\zeta_2, \zeta_{\rm L}$ remain constant while the depth

548: of contact changes. This is only approximately true; and a

549: self-consistent solution will require numerical integrations. It is

550: interesting first to look at the implications of equation

551: (\ref{dotM2power}) when no driving is present, because the solution

552: is immediate and instructive. This is a situation we encounter in

553: some large-scale hydrodynamic simulations of mass transfer in

554: polytropic binaries \citep{DMTF}. Defining a positive dimensionless

555: mass transfer $y=(-\dot M_2/\dot M_0)^{2/(2n+3)}$, and a

556: characteristic time scale $\tau=M_2/\dot M_0$, eq.

557: (\ref{dotM2power}) becomes

558: \begin{equation}

559: \frac{\d y}{\d t} = -\frac{\zeta_2-\zeta_{\rm L}}{\tau} y^{n+3/2}.

560: \end{equation}

561: The solution can be easily inverted to yield

562: \begin{equation}

563: y(t) = y(0)\left[1+y(0)^{n+1/2}(n+1/2)(\zeta_2-\zeta_{\rm

564: L})t/\tau\right]^{-\frac{2}{2n+1}}\, , \label{nodrivsol}

565: \end{equation}

566: where $y(0)$ is the initial mass transfer rate, normalized as above.

567: This solution illustrates explicitly the role of $\zeta_2-\zeta_{\rm

568: L}=2(q_{\rm stable}-q)$. In the stable case, $\zeta_2>\zeta_{\rm

569: L}$, the mass transfer decays asymptotically to zero over a

570: characteristic time $\tau_{\rm

571: chr}=\tau/[(n+1/2)y(0)^{n+1/2}(\zeta_2-\zeta_{\rm L})]$, whereas in

572: the unstable case, $\zeta_2<\zeta_{\rm L}$, it will blow up in

573: finite time equal to $\tau_{\rm chr}$. Thus the essence of the

574: stability of mass transfer in a binary is already contained in the

575: simple case of no driving. The presence of driving exacerbates the

576: natural instability or, in the stable case, the mass transfer

577: settles asymptotically to a non-zero stable value. This is what we

578: observe, for example, in CVs, AM CVns and LMXBs.

579: 

580: Returning now to eq. (\ref{dotM2power}), with the same definitions

581: as above for $y$ and $\tau$, we obtain for the general case in which

582: driving is present,

583: \begin{equation}

584: \frac{\d y}{\d t} = -\frac{\zeta_2-\zeta_{\rm L}}{\tau}

585: (y^{n+3/2}-y_{\rm eq}^{n+3/2})\, , \label{dydtpo}

586: \end{equation}

587: where $y_{\rm eq}^{n+3/2}\equiv-(\dot M_2)_{\rm eq}/\dot

588: M_0=(\nu_2-\nu_{\rm L})\tau/(\zeta_2-\zeta_{\rm L})$ is the

589: equilibrium value normalized to $\dot M_0$. Note that in the stable

590: case, this value is positive; while it is negative in the unstable

591: case. Before we attempt to solve the above differential equation, it

592: is clear from its form and the signs just discussed, that it

593: describes a stable solution in which $y\rightarrow y_{\rm eq}$ when

594: $q<q_{\rm stable}$. If, however, $q>q_{\rm stable}$, the r.h.s is

595: positive even if the mass transfer vanishes initially, and it just

596: gets bigger as the mass transfer grows. Since $y$ diverges for the

597: no-driving case in a finite time, the driven case diverges even

598: sooner.

599: 

600: In order to obtain an analytic solution to eq. (\ref{dotM2power})

601: which can be compared to the solution of \citet{WebIb}, it is

602: necessary to cast it in a slightly different form using the

603: equilibrium rate defined in eq. (\ref{M2doteq}), and modifying

604: appropriately the definitions of the integration variable and

605: characteristic time. With the definitions $y_*\equiv [\dot M_2/(\dot

606: M_2)_{\rm eq}]^{2/(2n+3)}$, and $1/\tau_*\equiv (\nu_2-\nu_{\rm L})

607: (\dot M_2/\vert(\dot M_2)_{\rm eq}\vert)^{2/(2n+3)}$, the

608: differential equation for the evolution of mass transfer becomes

609: \begin{equation}

610: \frac{\d y_*}{\d t} = \sgn(y_*)\frac{1}{\tau_*} (1- \sgn(y_*)\vert

611: y_*\vert^{n+3/2})\, ,

612: \end{equation}

613: where $\sgn(y_*)$ is the sign of $y_*$. Thus, for the stable case

614: $y_*>0$, while $y_*<0$ for the unstable case, and $\tau_*$ is

615: defined positive. The general analytic solution comprising both the

616: stable and the unstable case can be given in terms of the

617: hypergeometric function, as follows

618: \begin{equation}

619: \frac{t}{\tau_*} = y_*{}_2F_1(1,

620: \frac{1}{n+3/2};1+\frac{1}{n+3/2},\sgn(y_*)\vert y_*\vert^{n+3/2})\,

621: . \label{genanalsol}

622: \end{equation}

623: While this solution can be easily plotted numerically, it is not

624: possible in general to invert it to obtain $y_*(t)$. In a few cases,

625: simpler analytic forms can be obtained. For example, the case

626: $n=3/2$ yields the solution of \citet{WebIb}, and the case $n=1/2$

627: is particularly simple:

628: \begin{eqnarray}

629: y_* &=& -\tan{(t/\tau_*)} \qquad y_*<0 \qquad {\rm unstable}\cr y_*

630: &=& \tanh{(t/\tau_*)}\ \qquad y_*>0 \qquad {\rm stable}

631: \end{eqnarray}

632: 

633: In the case of isothermal atmospheres, the mass transfer rate

634: \citep{Ri88} is given by :

635: \begin{equation}

636: \dot M_2  = - \dot M_0 ~ e^{(R_2 - R_{\rm L})/H}

637: \end{equation}

638: where, $H$ is the scale height. This form of the mass transfer

639: equation is much simpler to integrate than the one for polytropes

640: considered above. With the same approximations and notation as in

641: the steps leading to eq. (\ref{dydtpo}), and defining $y=-\dot

642: M_2/\dot M_0 = \exp{((R_2-R_{\rm L})/H)}$, we obtain

643: \begin{equation}

644: \frac{1}{y} \frac{dy}{dt}=  - \frac{\zeta_2 - \zeta_{\rm

645: L}}{\tau}\frac{R_2}{H} (y-y_{\rm eq})\, . \label{dydtis}

646: \end{equation}

647: This can be easily integrated to obtain:

648: \begin{equation}

649: y =  \frac{y_{\rm eq}}{1-(1- y_{\rm eq}/y_0) e^{-t/\tau_{\rm iso}}}

650: \label{yiso}

651: \end{equation}

652: where $\tau_{\rm iso} \equiv H/R_2 (\nu_2-\nu_{\rm L})$ is the

653: timescale required for the driving to change the depth of contact by

654: $\sim H$, and $y_0$ is the initial value, always positive for

655: physically meaningful cases. In the stable case $y_{\rm eq}>0$, and

656: $y\rightarrow y_{\rm eq}$, while $y_{\rm eq}<0$ for the unstable

657: case and $y$ diverges in a finite time $t_{\rm div} = \tau_{\rm iso}

658: \ln{(1-y_{\rm eq}/y_0)}$. If no driving is present, we may set

659: $y_{\rm eq}=0$ and integrate eq. (\ref{dydtis}) for an isothermal

660: donor. The result is again simple and instructive

661: \begin{equation}

662: y = \frac{y_0}{1+(\zeta_2-\zeta_{\rm L}) y_0 \frac{R_2}{H}

663: \frac{t}{\tau} }\, .

664: \end{equation}

665: In the stable case, for any initial mass transfer, the system will

666: detach and mass transfer will tend to zero. In the unstable case,

667: any non-zero initial mass transfer will grow and diverge in a finite

668: time.

669: 

670: 

671: \section{Numerical Integration Results} \label{evol}

672: 

673: We now relax some of the constraints imposed in the previous section

674: and integrate the evolution equations allowing the binary parameters

675: to adjust self-consistently. Specifically, we compute the changes in

676: the masses of the components (assuming conservative mass transfer),

677: allow the binary separation to change as a result of any driving

678: present, and compute the values of $\zeta_2$ and $\zeta_{\rm L}$ as

679: they evolve. The values of $\zeta_2$ depend on the adopted

680: mass-radius relationship for the donor. Here we use Eggleton's

681: interpolated zero-temperature mass-radius relationship cited by

682: \citet{VeRa} and \citet{Maet04}, which is a good approximation for

683: systems containing old, cold white dwarfs. It is possible that the

684: systems emerge from a common envelope evolution with a massive

685: hydrogen atmosphere around the donor \citep{DAnt06}, or that the

686: donor has a finite entropy such that $\zeta_2$ deviates

687: significantly from the value obtained from the zero temperature

688: mass-radius relationship for white dwarfs \citep{DelBil05}. In such

689: situations, the radial variation rate $\nu_2$ is non-zero, and in

690: fact can reduce the \textit{net driving rate} $\nu_L-\nu_2$, leading

691: to a lower value for the equilibrium mass transfer (see Eq.

692: \ref{M2doteq}). Also, a mass radius exponent ($\zeta_2$)

693: significantly different from $\approx -1/3$ clearly affects the

694: stability and evolution of the systems at the onset of mass

695: transfer, and can lead to shrinking orbits even if the mass transfer

696: is stable with $\ddot M_2\approx 0$. In our subsequent analysis,

697: where we are concerned about the long term integrations of the OAE,

698: we set $\nu_2$ = 0 and use the zero-temperature mass-radius

699: relationship. We note that for the study of the onset of mass

700: transfer in finite entropy systems like the ones addressed by

701: \cite{DelBil05} \& \cite{DAnt06}, a more realistic model for the

702: donor is required.

703: %\clearpage

704: \begin{figure}[!t]

705: \centering \epsscale{0.9}\plotone{f1} \figcaption{Comparison of

706: integrations with the analytic solution by Webbink and Iben. The

707: mass transfer rate normalized to the {\em initial} equilibrium rate

708: as a function of time in units of the {\em initial} $\tau$: Analytic

709: (black curve) and numerical -- green ($q=0.663$), orange (0.613),

710: cyan (0.563 same as WI), blue (0.543), magenta (0.523) and red

711: (0.513). \label{WIcomp}}

712: \end{figure}

713: %\clearpage

714: In order to calculate $\zeta_{\rm L}$, we need to either assume or

715: determine from other assumptions how the mass and angular momentum

716: are re-distributed in the binary during mass transfer. As we have

717: seen in \S\ref{masstran}, this depends on the mode of accretion

718: appropriate for the binary considered: does the stream impact the

719: accretor or is an accretion disk present; is the mass transfer

720: sub-Eddington and conservative, or are mass and angular momentum

721: being ejected from the system following super-Eddington mass

722: transfer. For most of the numerical integrations that follow, we use

723: the appropriate rate of driving by gravitational wave radiation

724: $\nu_{\rm L}$, wherever necessary we assume a constant driving rate.

725: However, if one is interested in the relatively rapid phases of mass

726: transfer that follow contact and onset, then the qualitative

727: properties of our integrated evolutions do not depend strongly on

728: these assumptions.

729: 

730: 

731: \subsection{Surviving unstable mass transfer}

732: \label{surviving}

733: 

734: We integrate numerically the evolutionary equations for a sample of

735: double degenerate binaries whose mass ratios at the onset of mass

736: transfer exceed the stable value. In order to mimic the conditions

737: assumed by \citet{WebIb} in their pioneering analysis, we assume a

738: constant rate of driving, that no mass loss from the system occurs

739: even during super-Eddington phases, and that a tidally truncated

740: accretion disk is present at all stages. For these choices, $q_{\rm

741: stable}\approx 0.49$, and thus we expect unstable mass transfer in

742: binaries whose initial mass ratio exceeds this value. In Fig. 1 we

743: present the evolution of mass transfer for a selection of unstable

744: binaries undergoing conservative disk accretion, including the

745: specific value $q=0.563$ used by Webbink \& Iben (hereafter WI) to

746: illustrate their analytic solution. These integrations show that the

747: mass transfer in an initially unstable binary grows at first

748: rapidly, peaks, and then evolves asymptotically toward an

749: equilibrium rate which is also evolving as $q$ changes. The analytic

750: WI solution diverges in a finite time, while all numerical solutions

751: reach a peak and then return to stability. The peak transfer

752: decreases as the initial $q$ approaches $q_{\rm stable}$. The

753: equations we integrate are orbit-averaged evolution equations (OAE)

754: in the sense that the rates of change of the orbital parameters are

755: averaged over one orbital period. This approximation is valid as

756: long as the evolution is not too rapid, and the eccentricity of the

757: orbit remains negligible. Furthermore, our formulation does not

758: include the full effects of tidal distortion and instability

759: discussed by LRS. Our results suggest that an initially unstable

760: binary may survive the onset of unstable mass transfer as long as

761: the mass transfer does not get too big and that the separation does

762: not get too small. See \S\ref{supercrit} for a discussion of the

763: limits of validity of the OAE under unstable mass transfer.

764: 

765: It is also interesting to compare the analytic solution given by eq.

766: (\ref{yiso}) for an isothermal donor in the unstable case, with

767: numerical integrations of the OAEs for various initial mass ratios.

768: This comparison is shown in Fig. \ref{isocomp}, with the analytic

769: unstable solution plotted as the single divergent black solid line.

770: We elected to plot the natural logarithm of $y$, which is simply the

771: depth of contact $R_2 - R_{\rm L}$ in units of the pressure scale

772: height $H$. All the integrations were started from the same initial

773: depth of contact corresponding to an initial mass transfer of

774: $10^{-5}$ of the reference rate.

775: %\clearpage

776: \begin{figure}[!t]

777: \centering \epsscale{0.9}\plotone{f2} \figcaption{Comparison of

778: integrations with the unstable isothermal analytic solution. The

779: natural logarithm of the mass transfer rate normalized to the {\em

780: initial} equilibrium rate for the case with $q=0.663$, is shown as a

781: function of time in units of $\tau_{\rm iso}$: Analytic (black

782: curve) and numerical -- green ($q=0.663$), orange (0.613), cyan

783: (0.563 same as WI), blue (0.543), magenta (0.523) and red (0.513).

784: \label{isocomp}}

785: \end{figure}

786: %\clearpage

787: \subsection{Super-Eddington mass transfer}

788: \label{supercrit}

789: Another effect may come into play as discussed by \citet{HaWe} when

790: the mass transfer exceeds the critical Eddington rate and the evolution

791: becomes non-conservative. We can

792: incorporate this effect into our numerical integrations, allowing

793: the excess mass to be blown out of the binary as a wind, carrying away a specific

794: angular momentum $j_{\rm w}$. We calculate the accreted fraction $\beta$ following

795: \citet{HaWe}, and modify the evolution equations as follows:

796: \begin{eqnarray}

797: \dot J_1 &=& -\beta\dot M_2 j_{1} + \dot J_{\rm 1, tid}\ , \label{J1dotb}

798: \\

799: \dot J_2 &=& \dot M_2 j_{2} + \dot J_{\rm 2, tid}\ , \label{J2dotb}

800: \\

801: \dot J_{\rm orb} &=& \dot J_{\rm sys} - \left(- \dot M_2[\beta j_1-j_2 +(1-\beta)j_{\rm w}]

802: + \dot J_{\rm 1, tid}+ \dot J_{\rm 2, tid}\right)\ , \label{JorbDotb}

803: \\

804: \frac{\dot a}{2a} &=& \frac{\dot J_{\rm orb}}{J_{\rm orb}} - \frac{

805: \dot M_2}{M_2} \left(1 - \beta q - \frac{1-\beta}{2(1+q)}\right)\ . \label{adotb}

806: \end{eqnarray}

807: During the direct impact phase, it is reasonable to assume that the

808: material blown out carries away the characteristic specific angular

809: momentum of the stream $j_{\rm w}=j_1$. After a disk forms, the

810: angular momentum advected by the wind will depend on details of the

811: flow beyond the scope of the present study. For example, if the

812: material is lost mostly from the vicinity of the accretor, it will

813: carry approximately the specific orbital angular momentum of the

814: accretor. Since we do not know exactly how $j_{\rm w}$ varies, we

815: set $j_{\rm w}=j_1$ throughout. In this case eq. (\ref{JorbDotb})

816: simply reduces to eq. (\ref{JorbDot}). While it would be possible to

817: cast the evolution equation for the binary separation (\ref{adotb})

818: in the same form as eq. (\ref{qa}) by defining

819: \begin{equation}

820: q_a \equiv 1+ (1-\beta)q -\frac{1-\beta}{2(1+q)}

821: - M_2\frac{\beta j_1-j_2+(1-\beta)j_{\rm w}}{J_{\rm orb}}\ , \label{qadefb}

822: \end{equation}

823: the explicit appearance of $q$ above makes it obvious that $q_a$

824: must be calculated self-consistently during the evolution. Note that

825: when $\beta=1$, the above expression reduces to eq. (\ref{qadef}),

826: as it should.

827: %\clearpage

828: \begin{figure}[!t]

829: \centering \epsscale{0.7}\plottwo{f3a}{f3b} \figcaption{Various

830: parameters for super-Eddington accretion with direct impact with $q

831: = 0.25$, just above $q_{\rm stable}$ and $q=0.35$, well above

832: $q_{\rm stable}$. The panels show from the top down: the accreted

833: fraction $\beta$; the logarithm of the mass transfer rate (magenta)

834: and the corresponding Eddington rate (dashed red) in $M_\odot$/yr;

835: the mass ratio $q$; and the total mass normalized to the initial

836: value. The abscissa shows times in years from the time of first

837: contact. The lower $q$ accretes at super-Eddington rates for less

838: time as compared to the higher $q$; and it does so more gradually

839: losing less mass (last panel). Even for the initially more unstable

840: mass transfer (larger $q$), the fraction of mass lost is below 3\%.

841: \label{evexa}}

842: \end{figure}

843: %\clearpage

844: Fig. \ref{evexa} shows two examples of evolutions with a

845: super-Eddington mass-transfer phase. Because the OAE do not include

846: tidal distortions of the components or dissipative effects, arising

847: for example from friction during a common envelope phase, they

848: always predict survival, no matter how high the mass transfer gets

849: during an unstable phase. The only exception to this rule occurs if

850: during the evolution the binary separation falls below the value

851: $a_{\rm min}$ at which the angular momentum and the energy of the

852: binary reach a minimum. In that case, further loss of angular

853: momentum inevitably breaks the synchronism and the system is

854: secularly unstable. As the binary frequency increases, the spin

855: frequencies of the components fall behind, tidal synchronization

856: torques further reduce the orbital angular momentum, and finally a

857: dynamical instability leads to a rapid merger. However, for DWD

858: binaries the lower mass component almost always fills its Roche lobe

859: at a ``contact separation" $a_{\rm c}$ well before this minimum is

860: reached. Then mass transfer commences and soon the second term in

861: eq. (\ref{adot}) rises enough to drive the binary apart saving it

862: from a merger. If initially $q>q_{\rm stable}$ mass transfer rises

863: even more quickly and causes the binary to expand and recover from

864: the instability. Clearly the OAE break down if during the transient

865: the transfer is so high that a significant fraction of the donor

866: overflows in one orbit, and the dissipative effects mentioned above

867: may promote a merger even before the mass transfer gets that high.

868: 

869: Suppose that during the transient, the mass transfer rises to $\xi$ times

870: the critical Eddington rate.

871: We can estimate the fraction of the donor mass transferred in a single orbit

872: using the standard assumptions about the donor filling its Roche lobe, and

873: taking for simplicity a mass-radius relationship $R_i \approx 5\times 10^8

874: (M_\odot/M_i)^{1/3}$ cm. We find

875: \begin{equation}

876: \frac{-\dot M_2 P}{M_2}=\frac{P}{\tau}\approx 10^{-9} \xi

877: \left(\frac{M_1}{0.5 M_\odot}\right)^{-1/3}\left(\frac{M_2}{0.1 M_\odot}\right)^{-2}\, .

878: \label{fracc}

879: \end{equation}

880: In most of the parameter space this fraction is so small that we can

881: trust the OAEs to describe approximately the correct behavior within

882: the limits of the physical effects included in their derivation.

883: Looking at Fig. 3 of \citet{HaWe}, we see that $\xi\la 10$ for most

884: cases with the exception of very rare binaries in which the accretor

885: is already very close to the Chandrasekhar limit. Thus we expect a

886: certain fraction of the binaries that come into contact with

887: $q>q_{\rm stable}$ to survive unstable mass transfer, even if it

888: happens to be super-Eddington. However, to estimate the fraction

889: that actually make it through, one would have to deal with the

890: common envelope phase which is beyond the aims of the present study.

891: 

892: \section{Applications}

893: In the above sections we have developed the basic framework for

894: investigating the evolution of close DWD systems. By imposing the

895: appropriate constraints on the OAE, we have compared our OAE

896: integrations with analytic solutions illustrating the limitations of

897: the latter. In what follows we investigate the consequences of the

898: OAE when they are applied with all effects included. One immediate

899: consequence of the OAE is the phenomenon of tidally induced

900: oscillations which we describe first. Next, we apply the OAE to a

901: grid of systems with different initial component masses in the $M_2

902: - M_1$ space and study the evolutionary outcome of each of these

903: systems. Finally, we follow the evolution of a single system in

904: order to achieve a better qualitative understanding of the results

905: obtained for this system using a full 3-D hydro-code.

906: %\clearpage

907: \begin{figure}[!t]

908: \centering \epsscale{0.6} \plotone{f4} \figcaption{Evolution of the

909: period normalized to the period of first contact and onset of mass

910: transfer, for a binary with initial values $q = 0.28$, and

911: $M_2=0.125 M_\odot$. This choice yields an orbital period of

912: $\approx300$ s at onset.  The curves shown correspond to different

913: {\it initial} tidal synchronization times; $\tau_{\rm s_1}$ = 500

914: yrs (green), 1000 yrs (red), 1500 yrs (blue) and 2000 yrs (magenta).

915: Tidal timescales are evolved according to eq. (\ref{tidal}). The

916: longer the tidal timescale, there are more oscillations with higher

917: amplitude. \label{evcyc1}}

918: \end{figure}

919: %\clearpage

920: \subsection{Tidally induced cycles}

921: \label{cycles} As a system gets into contact, the binary separation

922: decreases at first until the mass transfer rate is high enough to

923: reverse the trend in $a$. In unstable or near unstable cases, during

924: this phase the mass transfer timescale $\tau_{\rm M_2}$ decreases

925: rapidly and becomes shorter than the synchronization timescale of

926: the accretor $\tau_{\rm s_1}$, allowing efficient spin-up and

927: building a large asynchronism. As the separation increases, the mass

928: transfer rate begins to fall and correspondingly $\tau_{\rm M_2}$

929: increases rapidly (See Fig. \ref{WIcomp}). The synchronization

930: timescale does not evolve as rapidly as the mass transfer rate and

931: eventually $\tau_{M_2}\ga\tau_{\rm s_1}$. Now the angular momentum

932: stored in the spin of the accretor is efficiently returned to the

933: orbit. If enough asynchronism has been built up in the accretor

934: during the spin-up phase, the additional injection of spin angular

935: momentum to the orbit can cause the separation -- and consequently

936: the Roche lobe radius of the donor -- to increase at a rate faster

937: than the radius of the donor. On the other hand, the donor radius

938: increases at a slower rate due to the decreasing mass transfer rate.

939: This leads to detachment -- the radius of the star cannot keep up

940: with the increase in the Roche lobe radius. In general, whenever the

941: effective driving $\nu_L -\nu_2>$ 0, the systems detach. Eventually,

942: when tides synchronize the spins again, the driving will shrink the

943: binary back into contact and accretion recommences, slowly at first,

944: and accelerating as the separation $a$ shrinks and the cycle

945: repeats. However, as the cycles repeat, the deviations from the

946: equilibrium rate decrease until the system settles to a steadily

947: expanding behavior with the mass transfer following the equilibrium

948: rate closely. In the absence of tidal effects, one would expect the

949: orbital period to increase monotonically with time once equilibrium

950: mass transfer has been established for AM CVn type systems. However,

951: in the presence of appropriate tidal coupling, the systems may

952: detach and this leads to oscillations in the orbital period --

953: increasing when the system gets into sufficient contact, and

954: decreasing when it detaches and the GWR dominates over the tidal

955: terms. This behavior is evident from Figs. \ref{evcyc1} and

956: \ref{evcyc2}, in which we plot the orbital period of the binary as a

957: function of time for different mass ratios and tidal synchronization

958: timescales. Notice that the number of oscillations the system goes

959: through is a function of both the tidal timescale and the mass

960: ratio. For a given tidal timescale, the higher the mass ratio, the

961: larger the number of oscillations a system is likely to go through.

962: This is because the higher mass ratio implies that the system is

963: more unstable which leads to a higher mass transfer rate and thus a

964: higher degree of asynchronism between the accretor and the orbit. On

965: the other hand, for a given mass ratio, a longer tidal timescale

966: allows for a higher spin-up of the accretor, leading to more

967: oscillations and higher amplitudes. There are, however, limits to

968: how high or low the tidal timescale (as compared to the mass

969: transfer timescale) can be in order to observe this behavior. If the

970: timescale is too high, the spins and orbit are affectively decoupled

971: whilst if the timescale is too low, the coupling is too efficient to

972: allow for any significant asynchronism. Thus, in these extremes we

973: do not observe any oscillations.

974: 

975: During the evolution of the binary, the tidal synchronization times

976: change because the masses and stellar radii relative to the orbital

977: separation are changing. We assume that the synchronization

978: timescales evolve as in \citet{Camp84}:

979: \begin{eqnarray}

980: \tau_{\rm s_1} &\propto& \left({M_1\over M_2}\right)^2\left({a\over R_1}\right)^6\cr

981: \tau_{\rm s_2} &\propto& \left({M_2\over M_1}\right)^2\left({a\over R_2}\right)^6

982: \label{tidal}

983: \end{eqnarray}

984: and choose a normalization ($\tau_{s_0}$) that yields the desired

985: initial timescales.

986: %\clearpage

987: \begin{figure}[!t]

988: \centering \epsscale{1} \plottwo{f5a}{f5b}\figcaption{\textit{Left

989: Panel}: Evolution of a binary with $q = 0.28$ with $\tau_{\rm s_1}$

990: = 2500 yrs (Magenta) and 500 yrs (Green). \textit{Right Panel}: Same

991: for $q = 0.24$; $\tau_{\rm s_1}$ = 2500 yrs (Magenta) and 500 yrs

992: (Green). The time is set to 0 when the binary gets into contact the

993: first time. The higher the mass ratio, the system goes through a

994: larger number of oscillations which are of greater

995: amplitude.\label{evcyc2}}

996: \end{figure}

997: %\clearpage

998: As has been mentioned above, tidally induced detachment has

999: implications for ultra-compact DWD systems \footnote{The higher the

1000: donor mass, the shorter the period at initial contact. Also, a

1001: higher donor mass makes it more likely that the system has an

1002: unstable mass ratio. Thus, in general, oscillations are more likely

1003: in short period systems.}; in particular RX J0806 and V407. In these

1004: systems it is observed that the orbital period is decreasing at a

1005: rate consistent with GWR, but mass transfer is obviously underway in

1006: these systems (\cite{Stroh02} and \cite{Hakala03}). This is at odds

1007: with the theoretical expectation that the orbital period should

1008: increase. Building on ideas discussed in \citet{LaMe}, and

1009: complementing cases considered by \citet{MaNe05}, we have

1010: investigated the possibility of tidally induced detachment. We see

1011: that it is possible for the binary to detach, especially in the case

1012: of unstable, direct impact systems.

1013: \begin{table}[!t]

1014:   \caption{Time spent in different regimes during the oscillation phase as a function of the mass ratio

1015: $q$ and tidal timescale $\tau_{\rm s_1}$.}

1016:   \label{OsctimesTab}

1017:   \begin{center}

1018:     \leavevmode

1019:     \begin{tabular}{lcccccc} \hline \hline

1020:   $q$  & $\tau_{\rm s_1}$ & $T_{osc}$ & $T_1$ & $T_2$ & $T_3$  & $ N_{osc}$ \\

1021:        & (yrs)            & (yrs)     & (yrs) & (yrs) & (yrs)  &  \\

1022:        \hline

1023:   0.28 & 1000 & 3200 & 1260 & 1000  & 936 & 1 \\

1024:   0.28 & 2500 & 23000 & 4450 & 4125 & 14425 & 4 \\

1025:   0.28 & 5000 & 64000 & 4025 & 11655 & 48320 & 5 \\

1026:   0.26 & 1000 & 0 & 985 & $\forall t > T_1$ & 0 & 0 \\

1027:   0.26 & 2500 & 5400 & 1750 & 810 & 1840 & 1 \\

1028:   0.26 & 5000 & 15200 & 1600 & 4100 & 9510 & 2 \\

1029:   0.24 & 1000 & 0 & 710 & $\forall t > T_1$ & 0 & 0 \\

1030:   0.24 & 2500 & 3600 & 1130 & 1820 & 650 & 1 \\

1031:   0.24 & 5000 & 5500 & 1110 & 2300 & 2100 & 1 \\ \hline

1032:       \end{tabular}

1033:   \end{center}

1034: \end{table}

1035: The system parameters like the mass of the donor, accretor and the

1036: various angular momentum loss mechanisms are not accurately known

1037: for the known AM CVn systems. We have calculated the amount of time

1038: the tidal oscillations are in effect in the case of a DWD system

1039: after it first gets into contact by loss of GWR for different system

1040: parameters. In Table \ref{OsctimesTab} we have tabulated the

1041: relevant timescales as a function of the mass ratio $q$ and the

1042: tidal timescale $\tau_{\rm s_1}$ of the accretor. $T_{\rm osc} = T_1

1043: + T_2 + T_3$ is the timescale for which the oscillations last after

1044: initial contact. $T_1$ represents the time when the binary is in

1045: contact but the separation is decreasing, $T_2$ is the time when the

1046: system is in contact and the separation is increasing, whilst $T_3$

1047: represents the time for which the system is out-of-contact. $ N_{\rm

1048: osc}$ represents the number of oscillations a system encounters

1049: during its evolution. We see that for a given mass ratio, a system

1050: tends to spend an increasing amount of time out of contact with

1051: increasing tidal timescales. Moreover, the more unstable the mass

1052: ratio, the larger the number of oscillations and the timescale for

1053: which the oscillations last.

1054: 

1055: A binary can spend a considerable amount of time in which tides are

1056: effective, especially in the case of unstable mass transfer. In

1057: fact, since the systems also spend quite a significant fraction of

1058: time out-of-contact, there should be many more systems with short

1059: periods than can be observed. However, even in the most favorable

1060: case, a given system spends less than 30\% of its time in a regime

1061: where the system is in contact and the orbit is shrinking. Thus it

1062: is unlikely that tidally induced oscillations are responsible for

1063: the observations of $\dot P < 0$ for RX J0806 and V407.

1064: Nevertheless, the probability that we catch a system in contact with

1065: $\dot P < 0$ is enhanced significantly as compared to the case when

1066: there are no oscillations \citep{Kalo05}. For example, for the $q$ =

1067: 0.26 case the system does not undergo any oscillations and $T_1 \sim

1068: $ 1000 yrs for $\tau_{\rm s_1}$ = 1000 yrs. A more promising idea is

1069: the recent proposal by \citet{DAnt06}, according to which the

1070: behavior of these systems can be understood if the donor possesses a

1071: substantial non-degenerate atmosphere which allows the donor to

1072: shrink as mass transfer proceeds. Under these conditions mass

1073: transfer tends to be more stable, at least initially, and cycles are

1074: unlikely.

1075: 

1076: In the next section we follow a large number of evolutions

1077: with different initial conditions by integrating the OAEs for $10^9$ yr,

1078: in order to address the question of which kind of systems, and under

1079: which conditions, are likely to experience the cycles described here.

1080: 

1081: \subsection{Exploring Evolutionary Outcomes}

1082: \label{popsyn}

1083: 

1084: We explore the parameter space for DWD binaries by investigating the

1085: different types of binary evolution that can occur under a variety

1086: of initial assumptions. Our procedure consists of selecting the

1087: initial masses for binary components and evolving them under driving

1088: by gravitational radiation for $10^9$ yr. In the most general case

1089: we include tidal coupling between the orbit and both components, we

1090: allow mass loss in super-Eddington cases, and we include the

1091: transition from direct impact to disk accretion. We also choose the

1092: tidal synchronization timescales for either the accretor alone, as

1093: in \citet{Maet04} or for both components allowing even the donor to

1094: become non-synchronous. As a check, we have evolved a grid of models

1095: suppressing the tidal and advective terms from the donor and

1096: choosing the same synchronization timescale as \citet{Maet04}. Under

1097: these assumptions our results are indistinguishable from theirs.

1098: 

1099: Before presenting the results of the evolutions, we investigate the

1100: expected behavior of the systems, focussing on the stability limits

1101: and whether the mass transfer is super-Eddington or not. In Fig.

1102: \ref{SuperEexp} we plot the stability limits for two cases: one with

1103: the donor spin properly accounted for (blue line) and the other with

1104: the donor effects ignored (magenta line). In addition for each case,

1105: we plot the locus of points for which $\dot M_{2_{\rm{eq}}} \approx

1106: \dot M_{\rm{Edd}}$, i.e., the locus of points that defines a

1107: transition from super-Eddington accretion to sub-Eddington

1108: accretion. Note that these curves represent the equilibrium mass

1109: transfer and Eddington rates of the system \textit{at initial

1110: contact} where we assume that the systems are synchronous and thus

1111: the tidal terms are zero.

1112: %\clearpage

1113: \begin{figure}[!h]

1114: \centering \epsscale{0.7} \plotone{f6}\figcaption{Mass-transfer

1115: stability limits (dash-dot lines) and super (above) and

1116: sub-Eddington (below) accretion boundaries (solid lines) with (blue)

1117: and without (magenta) the donor terms included. The thin dashed

1118: black line divides direct impact systems from disk accretion systems

1119: \citep{MaSt02}. Because the transition from direct impact to disk

1120: accretion makes mass transfer more stable, both the stability limits

1121: and Eddington accretion rate boundaries follow closely the locus of

1122: that transition toward higher donor masses (see text for details).

1123: \label{SuperEexp}}

1124: \end{figure}

1125: %\clearpage

1126: For direct impact systems the accretion rate is always

1127: super-Eddington if $M_2 >0.21 M_\odot$ (0.17 $M_\odot$ if donor

1128: terms are neglected). This is because of two reasons -- i) in

1129: general, the higher the donor mass, the higher the mass ratio and

1130: thus the systems are closer to instability, which in turn implies a

1131: higher $\dot M_{2_{\rm{eq}}}$ and ii) the higher the accretor mass,

1132: the lower the threshold for super-Eddington accretion \citep{HaWe}.

1133: However, as can be seen from Fig. \ref{SuperEexp}, the transition of

1134: the systems from direct impact to disk leads to changes in the

1135: stability properties of these systems -- they tend to be slightly

1136: more stable because the loss of orbital angular momentum to the

1137: accretor spin in disk systems is smaller than in the direct impact

1138: case. This is reflected in the slight upturn in the stability curve

1139: around $M_1 \sim$ 1 $M_\odot$ (0.85 $M_\odot$ if donor terms are

1140: neglected). The result of this is that the equilibrium mass transfer

1141: rate for the disk systems is lower than it would have been for that

1142: same system if it were a direct impact system. Consequently these

1143: systems tend to undergo sub-Eddington accretion, and the locus of

1144: the transition between super and sub-Eddington accretion systems

1145: follows the stability curve for both cases. However the disk

1146: transition `saves' only a few systems because DWD binaries with $M_2

1147: > 0.25 M_\odot$, (0.23 $M_\odot$ if donor terms are neglected) are

1148: all super-Eddington initially.

1149: 

1150: We present now the results of the numerical integrations of the OAE

1151: for a grid of models in the $M_2-M_1$ parameter space. In Fig.

1152: \ref{popsynus}, we plot the results for two extreme values of the

1153: accretor tidal synchronization timescales -- one with an extremely

1154: large tidal timescale ($10^{15}$ yrs) corresponding to inefficient

1155: coupling and one for a short timescale of 10 yrs corresponding to

1156: highly efficient spin-orbit coupling. Firstly, we note a

1157: significantly increased region (as compared to when the donor terms

1158: are ignored) of parameter space corresponding to sub-Eddington mass

1159: transfer and probable survival as a long-term, stable mass transfer

1160: binary of the AM CVn type. Also, from our discussion above, we

1161: expect that the case with inefficient tidal coupling (left panel in

1162: Fig. \ref{popsynus}) to match the curves in Fig. \ref{SuperEexp}.

1163: The transition from super to sub-Eddington accretion overlaps the

1164: stability boundary until $M_2 \approx$ 0.21 $M_\odot$ after which it

1165: follows the stability curve defined by the direct impact to disk

1166: transition, in accordance with our expectations. On the other hand,

1167: since the tides almost always act to stabilize the system by

1168: effectively reducing the driving rate, a simplistic analysis based

1169: on comparing the \textit{initial} equilibrium mass transfer and

1170: Eddington rates has only partial validity. While the threshold for

1171: super-Eddington accretion remains unaffected by the tidal coupling,

1172: the mass transfer rate is significantly lowered due to the reduced

1173: driving rate, and can be below the Eddington accretion rate. Systems

1174: which have super-Eddington accretion rates when the tidal coupling

1175: is inefficient, can thus accrete at sub-Eddington rates if the tidal

1176: coupling is strong. As a consequence we expect the domain of

1177: sub-Eddington accretion to extend to higher donor masses, and on

1178: comparing the left and right panels of Fig. \ref{popsynus} this is

1179: what we observe.

1180: %\clearpage

1181: \begin{figure}[!h]

1182: \centering \epsscale{1} \plottwo{f7a}{f7b} \figcaption{Evolution for

1183: $10^9$ yrs for an initial $\tau_{s_1}$ of $10^{15}$ yrs (left panel)

1184: and for $\tau_{s_1}$ of 10 yrs (right panel). This synchronization

1185: time is for the initial configuration and evolves according to eq.

1186: (\ref{tidal}). $\tau_{s_2}$ is calculated based on whatever

1187: $\tau_{s_0}$ is required to obtain the desired value of

1188: $\tau_{s_1}$. The red symbols represent super-Eddington accretion,

1189: the green are sub-Eddington, the pluses ($+$) and hollow diamonds

1190: ($\lozenge$) represent systems with total mass below and above the

1191: Chandrasekhar limit, respectively. Among the latter, asterisks over

1192: diamonds indicate systems in which the accretor does not reach the

1193: Chandrasekhar limit in $10^9$ yr. The blue dash-dot line is the

1194: initial stability boundary with all donor effects included. The

1195: magenta line is the stability boundary without these effects, shown

1196: here for comparison. Note the transition in the super and

1197: sub-Eddington accretion rate around an $M_2$ of 0.2 $M_\odot$ to the

1198: latter stability curve. \label{popsynus}}

1199: \end{figure}

1200: %\clearpage

1201: Finally, the locus of systems that undergo oscillations in their

1202: separation and binary period is illustrated in Fig. \ref{popsyncyc}.

1203: Whether a system undergoes oscillations or not, depends primarily on

1204: two factors: the accretor tidal synchronization timescale $\tau_{\rm

1205: s_1}$ and the mass ratio $q$. As can be seen from Fig.

1206: \ref{popsyncyc}, oscillations are seen to occur only in a certain

1207: domain around the stability curve, and that domain decreases with

1208: increasing tidal synchronization times. This is due to the fact

1209: that, for higher donor masses, the mass transfer timescale

1210: $\tau_{M_2}$ is quite short and if the tidal timescale $\tau_{s_1}$

1211: is much longer than $\tau_{M_2}$, the spin and the orbit are

1212: effectively decoupled. As a result there is minimal return of the

1213: spin angular momentum to the orbit and consequently, there are no

1214: oscillations for these long timescales. In fact, referring to Fig.

1215: \ref{popsyncyc}, we observe that systems with high donor mass ($M_2

1216: > 0.35 M_\odot$) do not undergo any oscillations at all. Again, this

1217: is because the mass transfer rates are high in this domain and thus

1218: the tidal timescales (considered here) are much longer than the mass

1219: transfer timescales. Consequently, the radius of the donor keeps up

1220: with the increase in the Roche lobe radius throughout the evolution,

1221: and the systems stay in contact.

1222: 

1223: On the other hand, systems to the bottom right (the ones with low

1224: donor mass and high accretor mass) are stable systems and are more

1225: likely to be disc-systems. Thus the accretor is not spun up as much

1226: as in the case of less stable or unstable systems. Moreover in the

1227: disc systems, tides are extremely efficient in returning angular

1228: momentum back to the orbit. Both these factors conspire to decrease

1229: the level of asynchronism achieved by the accretor and consequently,

1230: these systems do not undergo oscillations.

1231: 

1232: We have also studied the effect of the donor's tidal synchronization

1233: timescale on the domain over which systems undergo tidally induced

1234: oscillation cycles. In most cases, the level of the donor's

1235: asynchronism is rather low. Thus the magnitude of the term

1236: associated with the tidal effects of the donor in the driving rate

1237: $\nu_L$ (eq. (\ref{RLdotarr})) is much smaller than the

1238: corresponding term for the accretor. As a result, the donor

1239: synchronization timescale has a limited impact on the domain over

1240: which the systems oscillate.

1241: 

1242: %\clearpage

1243: \begin{figure}[!h]

1244: \centering \epsscale{0.7} \plotone{f8} \figcaption{Systems which

1245: undergo `oscillations' at least once during their evolution for

1246: $\tau_{s_1}$ = 500 yrs (pluses, $+$), 2500 yrs (crosses, $\times$)

1247: and 5000 yrs (open circles, $\circ$). Red indicates super-Eddington

1248: transfer during any part of the evolution, whereas green indicates

1249: sub-Eddington transfer throughout the $10^9$ yr evolution. The

1250: magenta and blue lines are the stability limits as in previous

1251: figures. $\tau_{s_2}$ is held at a constant 100 yrs.

1252: \label{popsyncyc}}

1253: \end{figure}

1254: %\clearpage

1255: \subsection{Comparison with hydrodynamic simulations}

1256: \label{hydrocomp} One of the stated objectives of this paper is to

1257: develop a theoretical framework to interpret and analyze results of

1258: large-scale, self-consistent, 3-D hydrodynamic simulations of

1259: binaries undergoing mass transfer \citep{DMTF}. To this end, we

1260: apply the same initial conditions to our OAE as have been to the

1261: various runs carried out by D'Souza et al. We have the tidal

1262: normalization factor ($\tau_{s_0}$, see eq. (\ref{tidal})) and the

1263: mass transfer rate scaling ($\mathfrak{m}$) -- such that $\dot M_2 =

1264: -\mathfrak{m} \dot M_0 f(\Delta)$ -- as `free parameters', which we

1265: adjust so as to obtain as close a match to the hydro-runs as we can.

1266: In the particular run shown in Fig. \ref{numsim}, $\tau_{s_0}$ =

1267: 0.75 and $\mathfrak{m}$ = 35.0. This choice is not unique -- indeed,

1268: we get reasonable `fits' even for other combinations of these `free

1269: parameters'.

1270: %\clearpage

1271: \begin{figure}[!h]

1272: \epsscale{.85}\plotone{f9} \figcaption{Comparison with some of the

1273: numerical simulations ($q = 0.5$ run) in \citet{DMTF}. Three

1274: simulations were performed for the same initial conditions but the

1275: binary was driven by angular momentum losses at the level of 1\% per

1276: orbit for different times in order to achieve increasing depths of

1277: contact: Q0.5-a (blue; driven initially for 2.7 orbits), Q0.5-b

1278: (green; driven for the first 5.3 orbits) and Q0.5-c (red; driven

1279: throughout). The solid black curves show the accretion rate in donor

1280: masses per period, the binary mass ratio, the separation normalized

1281: to the initial separation, and the spin angular momenta as predicted

1282: by the OAEs, while the dotted color curves show the same quantities

1283: as derived from the results of the simulations. We arbitrarily

1284: change $\tau_{\rm s_i}$ of the donor and accretor to match the

1285: Q0.5-a run and then predict the outcomes of the other simulations.

1286: Here, $\tau_{\rm s_1} \sim 150 P$ and $\tau_{\rm s_2} \sim 3.5 P$

1287: initially.\label{numsim}}

1288: \end{figure}

1289: %\clearpage

1290: It should be noted that in the numerical simulations of \citet{DMTF}

1291: the minimum resolvable mass transfer was on the order of $\sim

1292: 10^{-5} M/P$, which translates for short period AM CVn binary

1293: parameters to $\sim 10^4 \dot M_{\rm Edd}$! Also, in the case of the

1294: hydro-runs one observes severe distortion of the accretor and the

1295: formation of an accretion belt around the accretor towards the end

1296: of the evolution. These features cannot be easily incorporated in

1297: the OAEs, and so the later stages of evolution especially in the

1298: case of Q0.5-c, cannot be properly represented in the OAE. An added

1299: complication is that for the hydro-runs, the driving was cut off

1300: after the systems were thought to have reached a deep enough

1301: contact. Since the effective density levels, especially near the

1302: edge of the stars in the 3-D numerical model differs from an ideal

1303: $\rm{n=3/2}$ polytrope due to the finite numerical resolution of the

1304: code, the depth of contact achieved after a certain amount of

1305: driving for a certain period of time is not necessarily the same for

1306: the hydro-runs and the OAE runs. This is especially true for Q0.5-a,

1307: because it is the one which is most sensitive to the depth of

1308: contact at the instant the driving is cut-off. For runs Q0.5-b and

1309: Q0.5-c, the depth achieved is deep enough to make small differences

1310: between the hydro-runs and the OAE unimportant. We are working on

1311: another set of runs in which the driving is not cut off and the

1312: systems are driven at slightly lower (and more realistic) rates. We

1313: hope that this will eliminate another source of discrepancy between

1314: the hydro-runs and the OAE.

1315: 

1316: Despite the above mentioned shortcomings, the OAEs do reproduce

1317: reasonably well the behavior of the binaries we have studied. One

1318: notices that the hydro-runs have a gentler slope initially which

1319: progressively gets steeper as compared to the OAE runs. This, we

1320: believe, is a consequence of the complicated fluid flow around the

1321: $L_1$ point and the distortion of the donor star. Moreover, during

1322: the initial stages of the evolutions, the 3-D hydrodynamic

1323: simulations are subject to numerical noise which is not the case for

1324: our numerical integrations. The relative significance of this noise

1325: diminishes as the mass transfer rate increases during the evolution.

1326: Also, the epicyclic motion that one encounters in the hydro-runs

1327: (see \cite{DMTF} for details) cannot be reproduced in our results,

1328: since we assume circular orbits. Thus the behavior of the OAE is

1329: much smoother than the numerical hydro-runs with no abrupt changes

1330: in slopes of the various parameters.

1331: 

1332: We conclude from the above discussion and Fig. \ref{numsim}, that

1333: the OAEs confirm that: a) tidal effects play an important role in

1334: the numerical simulations of the binary, b) direct impact accretion

1335: is an important effect and can lead to significant spin-up of the

1336: accretor at the expense of orbital angular momentum, and c) the OAE

1337: prediction that systems that are initially unstable can indeed

1338: survive mass transfer seems to bear out in the hydro-runs despite

1339: the rather extreme conditions to which the binaries in the

1340: hydro-runs are subjected.

1341: 

1342: In addition to the Q0.5 runs presented above, \cite{DMTF} also

1343: present other runs which result in the disruption of the binary. As

1344: can be seen from their Fig. 3, 6 and 10, the donor star becomes

1345: increasingly distorted towards the end of these simulations. The

1346: OAE, at least in their present form, are not capable of accounting

1347: for the distortion of the components and the consequence of these

1348: distortions on the fate of the binary. Additional hydro-runs with

1349: different values of the mass ratio $q$ and lower driving rates $\nu$

1350: (though these rates are still orders of magnitude higher than

1351: realistic values) are being carried out \citep{Motletal06}. In

1352: principle this can help, for example, in determining a limit on the

1353: mass ratio $q$ ($>~ q_{\rm stable}$) above which the tidal

1354: disruption of the binary is likely. However, other physical effects

1355: not yet included in the 3-D simulations, such as radiation and the

1356: formation of a common envelope, are more likely to determine the

1357: fate of such systems.

1358: 

1359: The tidal time scales that most closely account for the behavior

1360: observed in the simulations will serve as a measure of the numerical

1361: dissipation present in the simulations. We will present elsewhere a

1362: more comprehensive study of the numerical dissipation and its

1363: dependence on spatial resolution.

1364: 

1365: 

1366: \section{Discussion}

1367: \label{discuss} We have re-examined the standard circular

1368: orbit-averaged equations (OAEs) that describe the evolution of mass

1369: transferring binaries allowing for advective and tidal exchange of

1370: angular momentum between the components. We found that the mass

1371: transfer stream issuing through the $L_1$ point has two effects in

1372: the internal redistribution of angular momentum in the binary: 1) it

1373: delivers a specific angular momentum $j_{\rm circ}$ to the accretor

1374: spinning it up at the expense of the orbital angular momentum; and

1375: 2) it reduces the spin angular angular momentum of the donor by a

1376: specific amount $\sim R_2^2\omega_2$, and ultimately couples to the

1377: orbit via tides. In the cases examined in this paper, the effect of

1378: this additional term is mildly stabilizing. For example, with

1379: parameters thought to be appropriate for the two short-period,

1380: direct impact binaries V407 Vul and RX J0806, the additional donor

1381: term is $\sim 0.1$ (see eq. (\ref{qstab})), so that the net effect

1382: of the consequential terms (accretor plus donor) $\approx -0.3$, is

1383: still de-stabilizing but less than estimated by \citet{Maet04}, and

1384: $q_a\approx 0.7$. Its effect on the evolution of other systems

1385: remains to be explored further, but in systems driven by magnetic

1386: braking, it is likely to be relatively minor since it is relatively

1387: smaller and would be masked by the magnetic torques acting directly

1388: on the donor.

1389: 

1390: We have extended analytically and numerically our understanding of

1391: the evolution of stable and unstable mass transfer in semi-detached

1392: binaries, with special emphasis on DWD binaries. In particular we

1393: have extended the analytic solutions of the type discussed by

1394: \citet{WebIb} to other polytropic indexes and the isothermal case.

1395: The analytic solutions predict that if $q

1396: > q_{\rm stable}$ at contact, the mass transfer rate diverges in a finite

1397: time implying the catastrophic merger of the two components. The

1398: OAEs on the other hand, always predict that a binary undergoing

1399: unstable mass transfer would survive after a phase of rapid mass

1400: transfer which in many cases would reach super-Eddington levels.

1401: This is because, unlike the analytic case, in the numerical

1402: integration of the OAE we allow the binary parameters to evolve

1403: self-consistently throughout the evolution. Thus, even if initially

1404: $q > q_{\rm stable}$, at some point in the evolution $q < q_{\rm

1405: stable}$, and the systems settle to the equilibrium mass transfer

1406: rate corresponding to the current values of the system parameters.

1407: While we do incorporate mass loss due to super-Eddington accretion,

1408: following the treatment along the lines of \citet{HaWe}, clearly the

1409: OAEs must break down if a common envelope forms. The treatment of

1410: common envelope evolution is beyond the scope of the present paper,

1411: but we suspect that many systems previously considered doomed to a

1412: merger would actually survive provided that the mass transfer peak

1413: is not too high. This point clearly needs further investigation and

1414: probably requires 3-D hydrodynamic simulations with the inclusion of

1415: radiative effects.

1416: 

1417: One interesting consequence of the tidal coupling of the accretor to

1418: the orbit is the appearance of mass transfer oscillation cycles

1419: occurring for a limited time after the onset of mass transfer. The

1420: likelihood of these cycles is higher in situations where the mass

1421: transfer is high, or rises to a high value rapidly. Therefore they

1422: tend to occur near and around the stability boundaries. Note that

1423: all the systems that undergo oscillations are direct impact systems,

1424: at least for the system parameters and synchronization timescales we

1425: have considered. Thus it is highly unlikely that systems that have

1426: accretion disks will undergo the tidally induced oscillation cycles

1427: in the case of DWD binaries.

1428: 

1429: The presence of a massive ($\sim 0.01 M_\odot$) non-degenerate

1430: atmosphere on the donor at the onset of mass transfer \citep{DAnt06}

1431: can affect the stability and evolution of these systems. For

1432: example, a large and positive $\zeta_2$ would imply $q_{\rm

1433: stable}>1$, and has significant implications for cycles, stability

1434: boundaries and super-Eddington mass transfer. The full consequences

1435: of these circumstances remain to be explored further.

1436: 

1437: DWD systems are some of the most common compact systems in the

1438: galaxy and are of particular importance for the space based

1439: gravitational wave detector LISA. AM CVn systems are guaranteed

1440: sources for LISA and the knowledge of possible evolutionary

1441: trajectories is valuable. The framework we have outlined in this

1442: paper can be used to generate templates for short period DWD's in

1443: general and AM CVn systems in particular. Similar work has been done

1444: already (see for example, \cite{RavToh06} and \cite{SVN05}); but the

1445: effects of the tidal couplings and the advective term associated

1446: with the donor spin remain to be incorporated into future studies.

1447: 

1448: The LSU theory group has performed a number of large-scale 3-D

1449: numerical simulations of interacting binaries with polytropic

1450: components \citep{MTF, DMTF}. These simulations did not include

1451: enough physics to tackle the common envelope evolution, but they

1452: should be viewed as the first steps toward that goal. In the

1453: meantime, we have used the OAEs with suitably adjusted tidal

1454: coupling time scales to analyze and interpret the results of some of

1455: the simulations described by \citet{DMTF}. The mass transfer rates

1456: that these simulations can resolve are much higher than the

1457: Eddington critical rate and probably much higher than the rates

1458: likely to arise during the onset of mass transfer in most realistic

1459: cases. Nevertheless they describe correctly the dynamical aspects of

1460: the mass transfer and tidal interactions under these conditions.

1461: Comparing the predictions of the OAEs with the simulations, we were

1462: pleased to find that they predict the outcome of the simulations

1463: reasonably well.

1464: 

1465: \acknowledgements This work has been supported in part by NASA's ATP

1466: program grants NAG5-8497 and NAG5-13430.  We thank Joel E. Tohline,

1467: Patrick M. Motl and Ravi Kopparapu for helpful discussions. Finally,

1468: we would like to acknowledge the referee for constructive comments.

1469: 

1470: \appendix

1471: 

1472: 

1473: 

1474: \section{The Effective Mass-Radius Exponent $\zeta_2$}

1475: 

1476: \label{Appzeta}

1477: 

1478: We derive here a simple analytic approximation to the effective

1479: mass-radius exponent when the response of the donor is a combination

1480: of the adiabatic and thermal adjustments to mass loss. Our starting

1481: point is the same as in \S 3 of \citet{DMR}, namely

1482: \begin{equation}

1483: \frac{\d\ln{R_2}}{\d t} = \left(\frac{\partial\ln{R_2}}{\partial

1484: t}\right)_{\rm th}+ \left(\frac{\partial\ln{R_2}}{\partial

1485: t}\right)_{\rm nuc}+\zeta_s\frac{\dot M_2}{M_2}\, , \label{AppStart}

1486: \end{equation}

1487: where the first two terms on the r.h.s. represent the effects of

1488: thermal relaxation and nuclear evolution, and $\zeta_s$ is the

1489: purely adiabatic mass-radius exponent. The thermal relaxation term

1490: may arise as a result of initial conditions, nuclear evolution and

1491: mass transfer, and it is usually not possible to disentangle these

1492: effects if they operate on similar timescales. However, an

1493: approximate description of the radial evolution can be obtained by

1494: viewing it as the result of the superposition of thermal relaxation

1495: of initial conditions and nuclear evolution, plus a thermal

1496: adjustment to mass loss, as follows

1497: \begin{equation}

1498: \frac{\d\ln{R_2}}{\d t} = \nu_2 + \nu'_2+\zeta_s\frac{\dot

1499: M_2}{M_2}\, , \label{Appnus}

1500: \end{equation}

1501: where $\nu_2$ is the superposition of intrinsic thermal and nuclear

1502: evolution, while $\nu'_2$ stands for the rate of thermal radial

1503: reaction to mass loss. Our goal is to combine the last two terms on

1504: the r.h.s. by absorbing the thermal adjustment to mass loss into an

1505: effective mass-radius exponent. We write

1506: \begin{equation}

1507: \zeta_2\frac{\dot M_2}{M_2} = \nu'_2 +\zeta_s \frac{\dot M_2}{M_2}

1508: \label{AppR2dot}

1509: \end{equation}

1510: where $\zeta_2$ is the effective mass-radius exponent we seek. With

1511: the above interpretation, the term $\nu_2$ in eq. (\ref{R2dot})

1512: represents intrinsic thermal and nuclear processes which may operate

1513: on timescales that differ from the standard thermal relaxation rate.

1514: This approach is consistent only if $\nu_2<<\zeta_2(\dot M_2/M_2)$.

1515: Therefore, in what follows, we shall only consider thermal

1516: relaxation in response to mass loss, and set $\nu_2=0$. As a

1517: consequence of the mass loss, the donor's radius $R_2$ will differ

1518: from the equilibrium radius corresponding to its instantaneous mass

1519: $R_{\rm eq}(M_2)$. With these definitions we write

1520: \begin{equation}

1521: \frac{\dot R_2}{R_2} = \frac{R_{\rm eq}(M_2)-R_2}{R_2\tau'} +\zeta_s

1522: \frac{\dot M_2}{M_2} \label{AppR2dot2}

1523: \end{equation}

1524: where we have adopted a simple linear approximation for the thermal

1525: reaction term and $\tau'$ is the corresponding timescale. The

1526: secular evolution of the binary takes place on a mass-transfer

1527: timescale $\tau_{M_2}=-M_2/\dot M_2$. Differentiating  eq.

1528: (\ref{AppR2dot2}) with respect to time, we get

1529: \begin{equation}

1530: \frac{\d^2\ln{R_2}}{\d t^2} =  \frac{1}{\tau'}\frac{R_{\rm eq}}{R_2}

1531: \left(\frac{\zeta_{\rm eq}}{\tau_{M_2}} -

1532: \frac{\zeta_2}{\tau_{M_2}}\right)-\frac{\zeta_s}{\tau_{M_2}^2}

1533: \label{AppR2ddot}

1534: \end{equation}

1535: If the effective mass-radius exponent is to have any meaning, it

1536: must not change much over the evolutionary phase one is considering.

1537: Thus, we require

1538: \begin{equation}

1539: \frac{\d^2\ln{R_2}}{\d t^2} = -\zeta_2/\tau_{M_2}^2 \, .

1540: \end{equation}

1541: Finally, setting $R_{\rm eq}=R_2$ in eq. (\ref{AppR2ddot}), and

1542: solving for $\zeta_2$, we obtain

1543: \begin{equation}

1544: \zeta_2=\frac{\zeta_{\rm eq} +

1545: \zeta_s\tau'/\tau_{M_2}}{1+\tau'/\tau_{M_2}}.

1546: \end{equation}

1547: This expression shows that if the evolution is much slower than the

1548: thermal relaxation ($\tau'\ll\tau_{M_2}$), the donor radius follows

1549: the equilibrium radius closely, whereas if mass transfer occurs

1550: rapidly, the donor reacts adiabatically. The above discussion may be

1551: applied to cataclysmic variables to describe approximately how the

1552: donor becomes increasingly oversized as the orbital period decreases

1553: because $\nu_2\approx 0$. If relaxation from initial conditions or

1554: significant nuclear evolution is taking place on timescales

1555: comparable to either $\tau'$ or $\tau_{M_2}$, the above discussion

1556: is strictly not valid.

1557: 

1558: 

1559: \begin{thebibliography}{}

1560: 

1561: \bibitem[Campbell(1984)]{Camp84} Campbell, C.G. 1984, MNRAS, 207,

1562: 433

1563: 

1564: \bibitem[Csataryova \& Skopal(2005)]{CsatSko05} Csataryova, M. \&

1565: Skopal, A. 2005, Contrib. Astron. Obs. Skalnate Pleso, 35, 17

1566: 

1567: \bibitem[D'Antona et al.(2006)]{DAnt06} D'Antona, F., Ventura,P.,

1568: Burderi, L. \& Teodorescu, A., 2006, preprint (astro-ph/0606577)

1569: 

1570: \bibitem[D'Antona, Mazzitelli \& Ritter(1989)]{DMR} D'Antona, F., Mazzitelli, I. \& Ritter, H. 1989, A\&A, 225, 391

1571: 

1572: \bibitem[Deloye, Bildsten \& Nelemans(2005)]{DelBil05} Deloye, C.,

1573: Bildsten, L. \& Nelemans, G. 2005, ApJ, 624, 934

1574: 

1575: \bibitem[D'Souza et al.(2006)]{DMTF} D'Souza, M.C., Motl, P.M., Tohline, J.E. \& Frank, J. 2006, ApJ,

1576: 643, 381

1577: 

1578: \bibitem[Eggleton(1983)]{Egg83} Eggleton, P. P. 1983, ApJ, 268, 368

1579: 

1580: \bibitem[Flannery(1975)]{Fla75}

1581: Flannery, B. 1975, MNRAS, 170, 325

1582: 

1583: \bibitem[Frank et al.(2002)]{FKR} Frank, J., King, A. R., \& Raine, D. J. 2002, Accretion Power in Astrophysics (3d ed; Cambridge: Cambridge Univ. Press)

1584: 

1585: \bibitem[Hakala et al.(2003)]{Hakala03} Hakala, P. and Ramsay, G. and Wu, K. and Hjalmarsdotter, L. and

1586:     J\"arvinen, S. and J\"arvinen, A. and Cropper, M., 2003, MNRAS,

1587:     343, L10-L14

1588: 

1589: 

1590: \bibitem[Han \& Webbink(1999)]{HaWe}

1591: Han, Z. \& Webbink, R.F. 1999, A\&A, 349, L17

1592: 

1593: \bibitem[J\c edrzejec(1969)]{Jedr}

1594: J\c edrzejec, E. 1969, MS thesis, Warsaw University

1595: 

1596: \bibitem[King \& Kolb(1995)]{KiKo95}

1597: King, A.R. \& Kolb, U. 1995, \apj, 439, 330

1598: 

1599: \bibitem[Kopparapu \& Tohline(2006)]{RavToh06} Kopparapu, R. \&

1600: Tohline J.E., 2006, in preparation

1601: 

1602: \bibitem[Kruszewski(1963)]{Krus63} Kruszewski, A. 1963, Acta

1603: Astr., 13, 106

1604: 

1605: \bibitem[Lai et al.(1993a)]{LRS1}

1606: Lai, D., Rasio, F.A., \& Shapiro, S.L. 1993, \apjl, 406, 63

1607: 

1608: \bibitem[Lai et al.(1993b)]{LRS2}

1609: Lai, D., Rasio, F.A., \& Shapiro, S.L. 1993, \apjs, 88, 205

1610: 

1611: \bibitem[Lai et al.(1994a)]{LRS3}

1612: Lai, D., Rasio, F.A., \& Shapiro, S.L. 1993, \apj, 420, 811

1613: 

1614: \bibitem[Lai et al.(1994b)]{LRS4}

1615: Lai, D., Rasio, F.A., \& Shapiro, S.L. 1993, \apj, 423, 344

1616: 

1617: \bibitem[Lai et al.(1994c)]{LRS5}

1618: Lai, D., Rasio, F.A., \& Shapiro, S.L. 1993, \apj, 437, 742

1619: 

1620: \bibitem[Lamb \& Melia(1987)]{LaMe}

1621: Lamb, D.Q. \& Melia, F. 1987, ApJ, 321, L133

1622: 

1623: \bibitem[Landau \& Lifshitz(1975)]{LaLi}

1624: Landau, L.D. \& Lifshitz, E.M. 1975, The Classical Theory of Fields,

1625: Pergamon Press, Oxford

1626: 

1627: \bibitem[Lubow \& Shu(1975)]{LuSh75}

1628: Lubow, S.H. \& Shu, F.H. 1975, ApJ, 198, 383

1629: 

1630: \bibitem[Marsh \& Steeghs(2002)]{MaSt02}Marsh, T.R. \& Steeghs, D. 2002, MNRAS, 331, L7

1631: 

1632: \bibitem[Marsh et al.(2004)]{Maet04}Marsh, T.R., Nelemans, G. \& Steeghs, D. 2004, MNRAS, 350, 113

1633: 

1634: \bibitem[Marsh \& Nelemans(2005)]{MaNe05}Marsh, T.R. \& Nelemans, G. 2005, MNRAS, 363, 581

1635: 

1636: \bibitem[Motl et al.(2002)]{MTF}

1637:   Motl, P. M., Tohline, J. E. \& Frank, J. 2002, ApJS, 138, 121 (MTF)

1638: 

1639: \bibitem[Motl, Tohline \& Frank(2006)]{Motletal06} Motl, P., Tohline,

1640: J. \& Frank, J., 2006, \textit{In preparation}

1641: 

1642: \bibitem[Paczy\'nski \& Sienkiewicz(1972)]{PaSi} Paczy\'nski B. \& Sienkiewicz R.,1972, Acta Astr., 22, 73.

1643: 

1644: \bibitem[Pratt \& Strittmatter(1976)]{PratStritt76} Pratt, J.P. \&

1645: Strittmatter, P.A., 1976, ApJ, 204, L29

1646: 

1647: \bibitem[Rasio \& Shapiro(1992)]{RS92}

1648:   Rasio, F. A. \& Shapiro, S. L. 1992, ApJ, 401, 226 (RS92)

1649: 

1650: \bibitem[Rasio \& Shapiro(1994)]{RS94}

1651:   Rasio, F. A. \& Shapiro, S. L. 1994, ApJ, 432, 242 (RS94)

1652: 

1653: \bibitem[Rasio \& Shapiro(1995)]{RS95}

1654:   Rasio, F. A. \& Shapiro, S. L. 1995, ApJ, 438, 887 (RS95)

1655: 

1656: \bibitem[Ritter(1988)]{Ri88} Ritter H., 1988, A\&A, 202, 93

1657: 

1658: \bibitem[Savonije(1978)]{Sav78} Savonije, G.J., 1978, A\&A, 62, 317

1659: 

1660: \bibitem[Stroeer et al.(2005)]{SVN05} Stroeer A., Vecchio, A. \&

1661: Nelemans G. 2005, ApJ, 633, L33

1662: 

1663: \bibitem[Strohmayer(2002)]{Stroh02} Strohmayer, T.E. 2002, ApJ, 581,

1664: 577

1665: 

1666: \bibitem[Verbunt \& Rappaport(1988)]{VeRa}

1667:  Verbunt, F. \& Rappaport, S. 1988, ApJ, 332, 193

1668: 

1669: \bibitem[Webbink \& Iben(1987)]{WebIb}

1670: Webbink, R.F., \& Iben, I., Jr. 1987, Proceedings of the

1671: $2^\mathrm{nd}$ Conference on Faint Blue Stars, p.~445

1672: 

1673: \bibitem[Willems \& Kalogera(2005)]{Kalo05} Willems, B. \& Kalogera, V. 2005,

1674: preprint (astro-ph/0508218)

1675: 

1676: \end{thebibliography}

1677: 

1678: \end{document}

1679: