MNRAS
472,
L25–L29 (2017)
doi:10.1093/mnrasl/slx130
Advance Access publication 2017 August 18
Accelerated tidal circularization via resonance locking in KIC 8164262
Jim Fuller,
1,2
‹
Kelly Hambleton,
3
Avi Shporer,
4
Howard Isaacson
5
and Susan Thompson
6
1
Kavli Institute for Theoretical Physics, Kohn Hall, University of California, Santa Barbara, CA 93106, USA
2
TAPIR, Walter Burke Institute for Theoretical Physics, California Institute of Technology, Pasadena, CA 91125, USA
3
Department of Astrophysics and Planetary Science, Villanova University, 800 East Lancaster Avenue, Villanova, PA 19085, USA
4
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125
5
Astronomy Department, University of California, Berkeley, CA 94720, USA
6
SETI Institute/NASA Ames Research Center, Moffett Field, CA 94035, USA
Accepted 2017 August 15. Received 2017 August 14; in original form 2017 June 14
ABSTRACT
Tidal dissipation in binary star and planetary systems is poorly understood. Fortunately,
eccentric binaries known as heartbeat stars often exhibit tidally excited oscillations, providing
observable diagnostics of tidal circularization mechanisms and time-scales. We apply tidal
theories to observations of the heartbeat star KIC 8164262, which contains an F-type primary
in a very eccentric orbit that exhibits a prominent tidally excited oscillation. We demonstrate
that the prominent oscillation is unlikely to result from a chance resonance between tidal
forcing and a stellar oscillation mode. However, the oscillation has a frequency and amplitude
consistent with the prediction of resonance locking, a mechanism in which coupled stellar and
orbital evolution maintain a stable resonance between tidal forcing and a stellar oscillation
mode. The resonantly excited mode produces efficient tidal dissipation (corresponding to an
effective tidal quality factor
Q
∼
5
×
10
4
), such that tidal orbital decay/circularization proceeds
on a stellar evolution time-scale.
Key words:
binaries: close – stars: evolution – stars: individual: KIC 8164262 – stars: kine-
matics and dynamics – stars: oscillations – stars: rotation.
1 INTRODUCTION
The mechanisms underlying tidal energy dissipation in stellar and
gaseous planetary interiors remain uncertain despite decades of
research. Observationally, tidal orbital evolution is challenging to
measure because it typically proceeds on very long time-scales.
Theoretically, tidal dissipation is difficult to calculate from first
principles because it is produced by weak friction effects that depend
on details of the stellar structure, convective turbulence, non-linear
mode coupling and other complex hydrodynamical processes.
Heartbeat stars present a new opportunity to constrain tidal dissi-
pation processes through observations of tidally excited oscillations
(TEOs). These eccentric binary stars experience tidal distortion near
periastron that produces ‘heartbeat’ signals in high-precision light
curves, and they have been studied in a number of recent works
(Welsh et al.
2011; Thompson et al.
2012; Hambleton et al.
2013;
Beck et al.
2014;Schmidetal.
2015; Smullen & Kobulnicky
2015;
Hambleton et al.
2016;Kirketal.
2016; Shporer et al.
2016;
Dimitrov, Kjurkchieva & Iliev
2017; Guo, Gies & Fuller
2017).
A fraction of heartbeat stars also exhibit TEOs that can be rec-
E-mail:
jfuller@caltech.edu
ognized because they occur at
exact
integer multiples of the or-
bital frequency (Kumar, Ao & Quataert
1995; Burkart et al.
2012;
Fuller & Lai
2012). The TEOs are produced by tidally forced stel-
lar oscillation modes, in most cases by gravity modes (g modes).
Given sufficiently accurate stellar properties, one can identify the g
modes responsible for the TEOs. From observed mode amplitudes,
one can then calculate mode energies, damping rates, tidal energy
dissipation rates and circularization/synchronization time-scales.
In this paper, we compare detailed tidal theory from a companion
paper (Fuller
2017), with observations of the heartbeat star KIC
8164262 (K81) analysed in another companion paper (Hambleton
et al.
2017). This heartbeat star contains an F-type primary star in
a highly eccentric (
e
0.89), long period (
P
=
87 d) orbit with
alowmass(
M
≈
0.36 M
) companion. The F-type primary has
M
=
1.7
±
0.1 M
,
R
=
2.4
±
0.1 R
,
T
eff
=
6900
±
100 K,
and is nearing the end of its main-sequence lifetime. Most impor-
tantly, K81 exhibits a large amplitude (relative flux variability of
L
/
L
∼
10
−
3
) TEO at exactly 229 times the orbital frequency. A
light curve and power spectrum are shown in Fig.
1. TEOs can be
excited to large amplitude if a stellar g-mode frequency happens to
be nearly equal to a tidal forcing frequency (i.e. a multiple of the
orbital frequency). We demonstrate that the prominent TEO in K81
requires a resonance so finely tuned that it is unlikely to occur by
C
2017 The Authors
Published by Oxford University Press on behalf of the Royal Astronomical Society
Downloaded from https://academic.oup.com/mnrasl/article-abstract/472/1/L25/4085195
by California Institute of Technology user
on 26 July 2018
L26
J. Fuller et al.
Figure 1.
Top left-hand panel: phased light curve of K81, in units of parts per thousand, with periastron near phase zero. The sharp variation near periastron
is the ‘heartbeat’ signal produced by the equilibrium tidal distortion, reflection and Doppler boosting. The rapid variability away from periastron
is due to the
prominent TEO at 229 times the orbital frequency. Top right-hand panel: same as top left-hand panel panel, zoomed in near periastron. Bottomleft-han
d panel:
Fourier transform of the light curve of K81. The large amplitude spike at
f
2.6 d
−
1
is produced by the dominant TEO and extends above the scale of the plot
to 1 ppt. The fuzz at lower amplitudes is composed of peaks at integer multiples of the orbital frequency. Bottom right-hand panel: Same as bottom left-
hand
panel, zoomed in to frequencies near 1.5 d
−
1
. Peaks at orbital harmonics are produced by TEOs (dynamical tide).
chance. Instead, the oscillation presents a compelling case for res-
onance locking (Witte & Savonije
1999, 2001; Fuller & Lai
2012;
Burkart, Quataert & Arras
2014), where tidal orbital decay natu-
rally maintains a finely tuned resonance with a g mode, resulting in
a large amplitude TEO.
2 TIDAL MODELS OF K81
TEOs are produced by tidally forced stellar oscillation modes. Com-
puting expectations for TEO frequencies, amplitudes, and phases
are discussed in detail in Fuller (
2017). In many cases such as in
K81, only the highest amplitude TEOs are detectable, and these
TEOs are produced by near-resonances between stellar g-mode fre-
quencies and multiples of the orbital frequency.
When a single resonant oscillation mode (labelled by
α
) domi-
nates the tidal response at a multiple
N
of the orbital frequency, it
produces a sinusoidal luminosity fluctuation of form
L
N
L
A
N
sin(
Nt
+
N
)
,
(1)
where
=
2
π
/P
is the angular orbital frequency. The amplitude
is
A
N
=
V
lm
X
Nm
|
Q
α
L
α
|
ω
Nm
√
(
ω
α
−
ω
Nm
)
2
+
γ
2
α
.
(2)
The pulsation phase
N
also contains useful information
(O’Leary & Burkart
2014), but the longitude of periastron mea-
surement in K81 from Hambleton et al. (
2017) was not precise
enough to measure the phases of its TEOs. Each term in equation
(2) is defined in Fuller (
2017) and can be calculated relatively eas-
ily. The tidal forcing amplitude
has a value of
10
−
6
for K81.
V
lm
describes the visibility of the modes based on viewing angle,
and is typically of the order of unity. Other quantities are a function
of frequency and are plotted in Fig.
2. The Hansen coefficient
X
Nm
is the strength of tidal forcing at each forcing frequency
N
and
peaks near
N
∼
40 for
m
=
2 for K81 (see Fig.
2). It has a long
tail to higher frequencies (due to the high eccentricity) allowing
oscillations up to
N
∼
300 to be observable.
The other terms in equation (2) are properties of stellar oscilla-
tion modes. To calculate them, we construct a stellar model with
parameters nearly equal to those found by Hambleton et al. (
2017)
for the primary star, using the stellar evolution code
MESA
(Paxton
et al.
2011, 2013, 2015). Our model has the same stellar/orbital
parameters quoted above, but with a 1
σ
smaller stellar radius of
2.3 R
that better matched the observed TEOs. We calculate the
non-adiabatic stellar oscillation modes of our model using the
GYRE
oscillation code (Townsend & Teitler
2013).
MESA
/
GYRE
inlists and
a discussion of the process are provided in Fuller (
2017). Next,
we calculate tidal overlap integrals
Q
α
that measure gravitational
coupling between oscillation modes and the tidal potential, and
bolometric luminosity perturbations
L
α
produced by normalized
modes at the stellar surface. Fig.
2 shows
Q
α
and
L
α
for
l
=|
m
|=
2
modes (we show below that most of the observed oscillations are
likely generated by quadrupolar,
m
=
1 and 2 oscillation modes).
Although the value of
Q
α
peaks for high-frequency (low radial or-
der) g modes, the value of
L
α
peaks for modes with
N
∼
100. Fig.
2
also shows mode damping rates
γ
α
and frequency spacings
ω
α
.
Damping rates are much larger for low-frequency (high-radial or-
der) modes, where the mode spectrum is very dense. Modes behave
like travelling waves when
γ
α
∼
ω
α
, which occurs at frequencies
N
100.
The last term in equation (2) is the resonant deturning for with
ω
α
ω
Nm
. The detuning factor can be very large for nearly resonant
MNRASL
472,
L25–L29 (2017)
Downloaded from https://academic.oup.com/mnrasl/article-abstract/472/1/L25/4085195
by California Institute of Technology user
on 26 July 2018
Resonance locking in KIC 8164262
L27
Figure 2.
Top panel: properties of normalized
l
=
m
=
2 g modes in our
model of the primary star of K81 (see the text). All values have been cal-
culated at multiples
N
of the orbittal frequency by interpolating between
neighbouring g modes. Modes within the grey shaded region have very
large radial orders and have not been calculated. The top axis shows the cor-
responding observed oscillation frequency
f
. Bottom panel: Mode damping
rate
γ
α
, and mode frequency separation
ω
α
. Note that most mode damp-
ing times are longer than 10
3
d so that TEOs maintain a nearly constant
amplitude throughout each orbit.
modes but is very sensitive to the degree of detuning. Because of this
extreme sensitivity, it is difficult to reliably calculate this quantity.
However, we can still predict mode amplitudes as a function of
frequency in a statistical fashion, as outlined in Fuller (
2017). The
median luminosity fluctuation is
A
N,
med
∣
∣
∣
∣
∣
4
L
N
ω
Nm
ω
α
∣
∣
∣
∣
∣
.
(3)
while the maximum possible fluctuation amplitude is
A
N,
max
∣
∣
∣
∣
∣
L
N
ω
Nm
γ
α
∣
∣
∣
∣
∣
,
(4)
where
L
N
=
V
lm
X
Nm
|
Q
α
L
α
|
.
With the K81 parameters from Hambleton et al. (
2017) and the
oscillation modes computed above, we can compute expected fre-
quencies and amplitudes of TEOs in K81. Fig.
3 shows both the
observed TEOs and our model results. We plot the median lumi-
nosity fluctuation at each orbital harmonic
N
(equation 3) and the
maximum luminosity fluctuation (equation 4). In addition, we plot
the luminosity fluctuations at each harmonic for our representative
stellar model. Finally, the background colour indicates the probabil-
ity density dlog
A
N
/
d
N
of observing a TEO in that part of the plot.
Most of the observed TEOs in K81 are likely to be produced by
quadrupolar
m
=
2 and 1 modes, with amplitudes above the median
amplitude of equation (3) due to chance resonances between mode
frequencies and forcing frequencies. In our model,
m
=
0 modes
are less visible due to the viewing angle relative to the stellar spin
axis.
Hambleton et al. (
2017) found evidence for spin-orbit misalign-
ment in K81, measuring an orbital inclination of
i
o
=
65
±
1
◦
and
a spin inclination of
i
s
=
35
±
3
◦
, implying a minimum obliq-
uity of 30
◦
. Such spin-orbit misalignment allows for excitation of
quadrupolar
m
=
1 modes. The model shown in Fig.
3 has
i
s
=
40
◦
and an actual (non-projected) obliquity of
β
=
30
◦
. We find mod-
els with lower values of
i
s
or higher obliquity provide a worse
fit to the data because they produce an excess of low-frequency
(
N
110)
m
=
1, 0 TEOs that are not observed. Another mod-
elling feature that improves agreement is the use of small amounts
of convective core overshoot and diffusive mixing. Our model uses
f
ov
=
0.01 and
D
mix
=
1cm
2
s
−
1
, on the low side of values inferred
from aseteroseismology (Moravveji et al.
2015, 2016; Deheuvels
et al.
2016), which are typically in the range 0.01
f
ov
0.03
and 1
D
mix
100 cm
2
s
−
1
. Larger overshoot tends to overpredict
TEO amplitudes. Our models use the Schwarzschild criterion for
convective boundaries, appropriate for main-sequence stars of this
mass (Deheuvels et al.
2016; Moore & Garaud
2016).
3 RESONANCE LOCKING IN K81
The primary TEO at
N
=
229 is not easily explained as a chance
resonance. Fig.
3 shows that this TEO lies in a region of parame-
ter space unlikely to contain TEOs, in contrast to TEOs at lower
frequencies and amplitudes. A very close resonance is required to
produce the primary TEO, which is possible but unlikely.
To quantify this statement, we compute the cumulative distribu-
tion of the expected number of TEOs to exist above a given am-
plitude
L
/
L
, and compare with the distribution of observed TEO
amplitudes. This calculation is described in Fuller (
2017), and is
performed by integrating the probability distribution in Fig.
3 over
all
N
and up to a chosen amplitude
L
/
L
. Fig.
4 shows the results.
Our model slightly overpredicts the number of low-amplitude TEOs
with
L
/
L
10
−
4
but the agreement is fairly good. However, our
model indicates the expected number of TEOs with
L
/
L
≥
10
−
3
is about 0.05. In other words, only 5 per cent of systems with prop-
erties nearly identical to K81 would be expected to exhibit such a
large amplitude TEO. We thus find it unlikely that the primary TEO
in K81 is caused by a chance resonance.
Instead, we suggest that the prominent TEO in K81 is generated
via resonance locking (see Witte & Savonije
1999, 2001; Burkart
et al.
2012, 2014; Fuller & Lai
2012). Because the g-mode frequen-
cies change as the star evolves, they pass through resonances with
tidal forcing frequencies. A resonance lock occurs when tidal dissi-
pation from a resonant mode causes orbital decay such that the tidal
forcing frequency increases at the same rate as the mode frequency,
and the mode remains resonant. Resonance locking configurations
can be stable and last for long periods of time (Burkart et al.
2014),
and they would create a single high-amplitude TEO like that
in K81.
To test the resonance locking hypothesis, we generate stellar mod-
els slightly younger and older than our model for K81, and compute
their g-mode spectra. We then compute the rates at which the mode
frequencies evolve,
t
α
=
σ
α
/
̇
σ
α
,where
σ
α
=
ω
α
+
m
s
is the mode
frequency in the inertial frame. We assume no angular momentum
loss and rigid stellar rotation. In contrast to previous assumptions
(Burkart et al.
2012; Fuller & Lai
2012), g-mode frequencies typ-
ically increase with age in intermediate-mass stars because they
MNRASL
472,
L25–L29 (2017)
Downloaded from https://academic.oup.com/mnrasl/article-abstract/472/1/L25/4085195
by California Institute of Technology user
on 26 July 2018
L28
J. Fuller et al.
Figure 3.
TEOs in the K81 system. Observed TEOs are shown as purple squares, with the prominent TEO denoted by a magenta star. Black dots are TEOs
from our representative model. Black lines are the median amplitudes of the
m
=
1 and 2 TEOs in our model, while the blue line is the maximum possible
amplitude of
m
=
1 or 2 TEOs in our model. Background colour denotes the probability density for
m
=
1 and 2 TEOs, which are more likely to exist in red
regions of the plot. Note the primary TEO lies in a very unlikely region of parameter space. Blue–green and light-green circles are expected amplitude
sfor
resonantly locked
m
=
2and1modes.
Figure 4.
Cumulative distribution of number of TEOs above an amplitude
A
N
as a function of
A
N
. We plot both observed numbers and expected
numbers from our representative model. At low amplitudes, the observed
and expected distributions overlap, indicating the model adequately explains
most of the low-amplitude TEOs. The primary TEO at
A
N
=
10
−
3
(magenta
star) is unexpected from the model and is unlikely to occur by a chance
resonance (the probability is
≈
5 per cent), and instead could be the result of
resonance locking.
become more highly stratified as they evolve. This creates increas-
ing Brunt–V
̈
ais
̈
al
̈
a frequencies and g-mode frequencies, allowing
resonance locking to occur with modes of any value of
m
.
After calculating the mode frequency evolution rates, we calcu-
late corresponding resonance locking mode amplitudes and lumi-
nosity fluctuations,
A
N,
ResLock
=
[
c
α
γ
α
t
α
]
1
/
2
V
lm
L
α
.
(5)
Here,
c
α
is a dimensionless factor of the order of 10
−
2
–10
−
3
in our
models, which is derived in Fuller (
2017). We plot these predictions
for
m
=
1 and 2 modes in Fig.
3.Wehaveaddedan
∼
3 per cent
uncertainty in frequency and
∼
30 per cent uncertainty in luminos-
ity fluctuation to account for the uncertainty in the stellar/orbital
parameters. The primary TEO has an amplitude and frequency con-
sistent with being a resonantly locked
m
=
1 oscillation mode.
A resonantly locked
m
=
2 mode may be possible but the pre-
dicted amplitude is slightly too low. Addtionally, using the system
parameters, we evaluate equation (53) of Burkart et al. (
2014)to
find that resonance locks will be stable for both
m
=
1 and 2 modes.
1
In the resonance locking scenario, both the resonantly locked
mode frequency
σ
α
and tidal forcing frequency
N
are increasing
at the same rate, set by the mode frequency evolution time-scale
t
α
∼
4Gyrfor
m
=
1 g modes in K81. We compute a corresponding
tidal orbital evolution time-scale of
t
orb
,
tide
=
E
orb
/
̇
E
orb
,
tide
∼
6Gyr,
which is somewhat longer than the stellar age in this particular case.
This tidal dissipation rate can be translated into an effective tidal
quality factor Q (Goldreich & Soter
1966) by defining
̇
E
orb
,
tide
E
orb
=
3
k
2
Q
M
M
(
R
a
peri
)
5
,
(6)
where
a
peri
=
a
(1
−
e
) is the periastron orbital separation and
k
2
3
×
10
−
3
is the Love number for our model. We calculate
Q
∼
5
×
10
4
, much lower than might be naively expected for an
F-type star. Indeed, in our model without a resonantly locked mode,
the combined energy dissipation of all TEOs leads to
Q
∼
2.5
×
10
7
,
meaning the resonantly locked mode increases the tidal energy
dissipation rate by a factor of
∼
500.
1
Equation (53) of Burkart et al. (
2014
) has a typo, the
>
sign should be
a
<
sign.
MNRASL
472,
L25–L29 (2017)
Downloaded from https://academic.oup.com/mnrasl/article-abstract/472/1/L25/4085195
by California Institute of Technology user
on 26 July 2018
Resonance locking in KIC 8164262
L29
4 DISCUSSION AND CONCLUSIONS
We have demonstrated that resonance locking in K81 is an appealing
mechanism to account for its prominent TEO. Resonance locking
correctly predicts the amplitude of the TEO, whereas producing
the TEO by a chance resonance between tidal forcing and a stellar
oscillation mode is unlikely. However, we note that K81 was chosen
for detailed analysis because of its high-amplitude TEO, which is
larger amplitude than nearly all other TEOs in catalogued heartbeat
systems. Of the
∼
175 catalogued heartbeat stars,
2
roughly 30 show
evidence for TEOs visible by eye. Thus, the chances of finding
one system amongst these with an unexpectedly large TEO (at the
2
σ
level) is of the order of unity. We therefore cannot exclude
the possibility the prominent TEO in K81 is simply an uncommon
occurrence that was selected for study due to its high amplitude.
K81 can be compared with the KOI-54 (Welsh et al.
2011) heart-
beat system. Resonance locking with
m
=
2 modes was proposed
by Fuller & Lai (
2012) to explain KOI-54’s large amplitude TEOs,
but O’Leary & Burkart (
2014) demonstrated that the TEO phases
identifies them as
m
=
0 modes. However, g-mode frequencies in-
crease as stars evolve off the main sequence and allows resonance
locking to occur with
m
=
0 modes, and hence the prominent TEOs
in KOI-54 could still be explained as a resonantly locked
m
=
0
TEO within each star. A preliminary calculation indicates this may
be possible, but the observed TEO amplitudes are a factor of a few
smaller than expected for resonance locking.
Other heartbeat systems may not exhibit resonance locking ef-
fects if they are not in an evolutionary stage where resonantly locked
modes are visible. Since TEOs are most visible in stars without thick
surface convective zones, we only expect to see large amplitude
TEOs in stars with
T
eff
6500 K. Moreover, resonance locking
amplitudes increase as the stellar evolution rate accelerates, i.e.
when stars evolve off the main sequence. Therefore, we expect to
see high-amplitude resonantly locked modes in heartbeat stars con-
taining somewhat massive (
M
1.5 M
) stars in the brief period
during which they are beginning to evolve off the main sequence,
but have not yet cooled to
T
eff
6500 K. A firm conclusion will re-
quire detailed analyses of a greater population of heartbeat systems.
If resonance locking does commonly occur, it can greatly enhance
tidal dissipation, causing orbital decay and spin synchronization to
proceed on a stellar evolution time-scale. During a resonance lock,
the tidal orbital energy dissipation rate is
̇
E
orb
/E
orb
∼
2
/
(3
t
α
).
Resonance locking is not necessarily limited to eccentric binary
star systems, and may operate in many astrophysical scenarios, in-
cluding circular (but non-synchronized) binary stars, exoplanetary
systems, inspiraling white dwarfs (Burkart et al.
2013) and out-
wardly migrating planetary moon systems Fuller, Luan & Quataert
(
2016). The observable feature of resonance locking is a larger-than-
expected variability at an integer multiple of the orbital frequency.
3
This could potentially be detected as ‘anomalous’ ellipsoidal varia-
tions (see e.g. Borkovits et al.
2014) or a perturbed gravity field. Res-
onance may not be able to operate in all scenarios, as it could some-
2
A list of catalogued heartbeat stars can be found at
http://keplerebs.
villanova.edu/search
, by filtering with the ‘HB’ flag checked.
3
Resonance locking cannot explain the pulsations in HAT-P-2 reported by
deWitetal.(
2017
) because those pulsations are larger than
A
N
,max
at the
observed pulsation frequency.
times be quenched by non-linear instabilities or be overwhelmed
by other tidal/orbital effects. When resonance locking can operate,
it generally accelerates tidal evolution to proceed on the relevant
evolutionary time-scale, for example, a stellar evolution time-scale,
magnetic braking time-scale, or gravitational radiation orbital decay
time-scale.
ACKNOWLEDGEMENTS
We thank the Planet Hunters and Dan Fabrycky for discovering
this system, and the anonymous referee for a thoughtful report. JF
acknowledges partial support from NSF under grant no. AST-
1205732 and through a Lee DuBridge Fellowship at Caltech. KH
acknowledges support through NASA ADAP grant (16-ADAP16-
0201). This research was supported by the National Science Foun-
dation under Grant No. NSF PHY11-25915, and by NASA under
grant 11-KEPLER11-0056 and the Gordon and Betty Moore Foun-
dation through grant GBMF5076.
REFERENCES
Beck P. G. et al., 2014, A&A, 564, A36
Borkovits T. et al., 2014, MNRAS, 443, 3068
Burkart J., Quataert E., Arras P., Weinberg N. N., 2012, MNRAS, 421, 983
Burkart J., Quataert E., Arras P., Weinberg N. N., 2013, MNRAS, 433, 332
Burkart J., Quataert E., Arras P., 2014, MNRAS, 443, 2957
de Wit J. et al., 2017, ApJ, 836, L17
Deheuvels S., Brand
̃
ao I., Silva Aguirre V., Ballot J., Michel E., Cunha
M. S., Lebreton Y., Appourchaux T., 2016, A&A, 589, A93
Dimitrov D. P., Kjurkchieva D. P., Iliev I. K., 2017, MNRAS, 469, 2089
Fuller J., 2017, MNRAS, preprint (
arXiv:1706.05054
)
Fuller J., Lai D., 2012, MNRAS, 420, 3126
Fuller J., Luan J., Quataert E., 2016, MNRAS, 458, 3867
Goldreich P., Soter S., 1966, Icarus, 5, 375
Guo Z., Gies D. R., Fuller J., 2017, ApJ, 834, 59
Hambleton K. M. et al., 2013, MNRAS, 434, 925
Hambleton K. et al., 2016, MNRAS, 463, 1199
Hambleton K. et al., 2017, preprint (
arXiv:1706.05051
)
Kirk B. et al., 2016, AJ, 151, 68
Kumar P., Ao C. O., Quataert E. J., 1995, ApJ, 449, 294
Moore K., Garaud P., 2016, ApJ, 817, 54
Moravveji E., Aerts C., P
́
apics P. I., Triana S. A., Vandoren B., 2015, A&A,
580, A27
Moravveji E., Townsend R. H. D., Aerts C., Mathis S., 2016, ApJ, 823, 130
O’Leary R. M., Burkart J., 2014, MNRAS, 440, 3036
Paxton B., Bildsten L., Dotter A., Herwig F., Lesaffre P., Timmes F., 2011,
ApJS, 192, 3
Paxton B. et al., 2013, ApJS, 208, 4
Paxton B. et al., 2015, ApJS, 220, 15
Schmid V. S. et al., 2015, A&A, 584, A35
Shporer A. et al., 2016, ApJ, 829, 34
Smullen R. A., Kobulnicky H. A., 2015, ApJ, 808, 166
Thompson S. E. et al., 2012, ApJ, 753, 86
Townsend R. H. D., Teitler S. A., 2013, MNRAS, 435, 3406
Welsh W. F. et al., 2011, ApJS, 197, 4
Witte M. G., Savonije G. J., 1999, A&A, 350, 129
Witte M. G., Savonije G. J., 2001, A&A, 366, 840
This paper has been typeset from a T
E
X/L
A
T
E
X file prepared by the author.
MNRASL
472,
L25–L29 (2017)
Downloaded from https://academic.oup.com/mnrasl/article-abstract/472/1/L25/4085195
by California Institute of Technology user
on 26 July 2018