This is a repository copy of Transmission spectroscopy of the inflated exoplanet
WASP-52b, and evidence for a bright region on the stellar surface.
White Rose Research Online URL for this paper:
http://eprints.whiterose.ac.uk/116318/
Version: Accepted Version
Article:
Kirk, J., Wheatley, P.J., Louden, T. et al. (5 more authors) (2016) Transmission
spectroscopy of the inflated exoplanet WASP-52b, and evidence for a bright region on the
stellar surface. Monthly Notices of the Royal Astronomical Society, 463 (3). pp. 2922-2931.
ISSN 0035-8711
https://doi.org/10.1093/mnras/stw2205
eprints@whiterose.ac.uk
https://eprints.whiterose.ac.uk/
Reuse
Unless indicated otherwise, fulltext items are protected by copyright with all rights reserved. The copyright
exception in section 29 of the Copyright, Designs and Patents Act 1988 allows the making of a single copy
solely for the purpose of non-commercial research or private study within the limits of fair dealing. The
publisher or other rights-holder may allow further reproduction and re-use of this version - refer to the White
Rose Research Online record for this item. Where records identify the publisher as the copyright holder,
users can verify any specific terms of use on the publisher’s website.
Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing eprints@whiterose.ac.uk including the URL of the record and the reason for the withdrawal request.
Mon. Not. R. Astron. Soc. 000, 000–000 (0000) Printed 1 September 2016 (MN L
A
T
E
X style file v2.2)
Transmission spectroscopy of the inflated exoplanet
WASP-52b, and evidence for a bright region on the stellar
surface
J. Kirk
1⋆
, P. J. Wheatley
1
†, T. Louden
1
, S. P. Littlefair
2
, C. M. Copperwheat
3
,
D. J. Armstrong
1,4
, T. R. Marsh
1
and V. S. Dhillon
2,5
1
Department of Physi cs , University of Warwic k , Coventry, CV4 7AL, UK
2
Department of Physi cs and Astronomy, University of Sheffield, Sheffield, S3 7RH, UK
3
Astrophysics Research Institute, Liverpool John Moores University, Liver pool, L3 5RF, UK
4
ARC, School of Mathem a ti cs and Physics, Queen’s University Belfast , University Road, Belfast BT7 1NN, UK
5
Instituto de Astrof´ısica de Canarias, V´ıa L´actea s/n, 38205, La Laguna, Spain
1 September 2016
ABSTRACT
We have measured the transmission spectrum of the extremely inflated hot Jupiter
WASP-52b using simul tan eou s photometric observations in SDSS u
′
, g
′
and a filter
centred on the sodium double t (NaI) with th e ULTRACAM instrument mounted on
the 4.2m William Herschel Telescope. We find that Rayleigh scattering is not the
dominant source of opacity within the planetary atmosphere and find a transmission
spectrum more con si s t ent with wavelength-i nd ependent opacity such as from clouds.
We detect an in-t r an si t anomaly that we attribute to the presence of stellar activi ty
and find that this feature can be more simply modelled as a bright region on the stellar
surface akin to Solar faculae rather than spots. A spot model requires a significantly
larger planet/star radius ratio than that found in previous studies. Our results high-
light the precision that can be achieved by ground-based photometry with errors in
the scaled planetary radii of less than one atmospheric scal e height, comparable to
HST observations.
Key words: stars: planetary systems – stars: individual: WASP-52 – stars: starspots
1 INTRODUCTION
Transmission spectroscopy using HST/STIS was used to de-
tect the first spectral feature of any exop la n et , that of the
narrow core of atomic sodium originating from the upper at-
mosphere of HD 209458b (Charbonneau et al. 2002). Broad-
band op ac i ty sources were later detected in this planet’s
atmosphere in the form of th e pressure broadened wings of
NaI (Sing et al. 200 8 ) and a blueward slope in the trans-
mission spectrum interpreted as Rayleigh scattering by H
2
(Lecavelier Des Etangs et al. 2008b). This feature was also
seen in HST data fo r HD 189733b (Pont et al. 2008; Sing
et al. 2011b), however the absence of the broad wings of
NaI suggested the presence of s il ic a t e conde n sa t es in the up-
per atmosphere of the planet (Lecavelier Des Etangs et al.
2008a). Recent studies of hot Jupiters with HST have re-
vealed a diverse array of atmospheres from clear to cloudy
⋆
James.Kirk@warwick.ac.uk
† P.J.Wheatley@warwick.ac.uk
(e.g. Sing et al. 2016). These studies have included the de-
tections of Rayleigh scattering by high altitude hazes also
in WASP-12b (Sin g et al. 2013), additional unknown optical
absorbers in HAT-P-1b (Nikolov et al. 2014 ) , a cloud deck
masking the majority of the NaI absorption and near in-
frared water features in WASP-31b (Sing et al. 2015) and a
clear atmosphere consistent w it h solar and sub-solar metal-
licities in WASP-39b (Sing et al. 2016; Fischer et al. 2016).
Ground-based measurements of hot Jupiter transmis-
sion s ig n a ls have also found success, beginning with the
detection of the narrow component of NaI in HD 189733b
(Redfield et al. 2008) and HD 209458b (Snellen et al. 2008;
Langland-Shula et al. 2009). More recent results have in-
cluded further detections of sodium (e.g. Wood et al. 2011;
Zhou & Bayliss 2012), potassium (e.g. Sing et al. 2011a;
Wilson et al. 2015), blueward scattering slopes (e.g. Jord´an
et al. 2013; Stevenson et al. 2014), and atmospheres domi-
nated by clouds or hazes (e.g. Gibson et al. 2013a; Mallonn
& S t ra ssmei er 2016).
Although there has been success with ground-based
c
0000 RAS
arXiv:1608.08993v1 [astro-ph.EP] 31 Aug 2016
2 J. Kirk et al.
spectroscopic observation s they are more susceptible to sys-
tematics arising from, for example, differential slit losses
(e.g. Sing et al. 2012). A simpler approach, which avoids
this problem, is to use ph ot o m et ers to perform simultane-
ous broadband multi-wavelength measurements. Th e se stud-
ies have included evidence for large blueward slopes (e.g.
Southworth et al. 2012; S o u t hworth et al. 2015; Mancini
et al. 2016a), enhanced absorption around the alkali metal
lines (e.g. Mancini et al. 2013; Bento et al. 2014) and trans-
mission spectra most consist ent with cloud s (e.g. Mallonn
et al. 2015a). By observing the planetary radius at c a re-
fully selected wavelengths we can probe for the existence of
Rayleigh scattering and the broad a b so rpt io n wings of the
NaI doublet that are expected to be present in the cloud
free a t mo sph e res of hot Ju p i te rs (Seager & Sasselov 2000).
Transmission spectroscopy is ofte n fo cu s sed on inflated
hot Jupiters; planets with very low density due to their large
radii and relatively low masses. As a result of their low densi-
ties, low surface gravities and high temperatures these plan-
ets have large atmospheric scale heights, H, given by
H =
kT
gµ
(1)
where k is Boltzmann’s constant, T is the temperature of
the planet, g is the acc el era t io n due to gravity and µ is the
mean molecular mass. The outermost ∼ 5 scale heights may
account for up to 10% of the cross sectional area of a hot
Jupiter, where different atmospheric species can affect the
observed transmission spectrum (Brown 2001) . The relative
size of the planet’s scale height to the stellar disc governs
the amplitude of the transmission sign a l .
WASP-52b (H´ebrard et al. 2013) is an extremely in -
flated hot Jupi t er, with a mass of 0.46 M
J
and radius of
1.27 R
J
giving it a mean density of 0.299 g/cm
3
. It orbits its
0.87 M
⊙
K2V host star with a period of 1.75 d ays. Due to the
combination of the infl a te d planetary radius with th e small
radius of the host (0.79 R
⊙
), it shows a deep transit in the
WASP photometry (2.7 %). From the table of system param-
eters (Table 2; H´ebrard et al. 2013), and assumi n g a Jupiter
mean molecular mass of 2.3 times the mass of a proton,
the scale height of WASP-52b is calculated to be 731 km.
This makes WASP-52b an exceptional target for transmis-
sion spectroscopy as the difference in the transit depth cor-
responding to one atmosp h e ric sca le he ig ht is 4.4 ×10
−4
, at
least three times stronger than that of HD 189733b.
In this paper we present multi-wavelength observations
of WASP-52b taken using the high speed multi-band pho-
tometer ULTRACAM (Dhillon et al. 2007).
2 OBSERVATIO NS
WASP-52b was observed on the night of th e 7th September
2012 using t h e ULTRACAM (Dhillon et al. 2007) instru-
ment on the 4.2m William Herschel Telescope (WHT), La
Palma. ULTRACAM is a high speed triple beam CCD pho-
tometer. I n c o mi n g light is split into t h ree bandpasses, using
two dichroics, and re-imaged onto three CCDs at a resolu-
tion of 0.3” per pixel, with a field of view of 5’.
ULTRACAM is partic u la rl y useful for the ground-based
application of transmission spectroscopy as it simultane-
ously takes measurements at three different wavelengths,
enabling the transit depth to be mea su red as a function
of wavelength (e.g. Copperwheat et al. 2013; Bento et al.
2014). The use of frame transfer CCDs allow for h i g h frame
rates (up to 300 Hz) with little dead time (24 ms), which is
useful to avo id saturation when observing bright stars and
enables many more sky flats to be taken.
The observations were made using SDSS u
′
(λ
central
=
3557
˚
A, FWHM = 599
˚
A) and g
′
(λ
central
= 4825
˚
A, FWHM
= 13 7 9
˚
A) filters and a filter centred on the NaI doublet
(λ
central
= 5912
˚
A, FWHM = 312
˚
A). These filters were se-
lected to probe for Rayleigh sca tt eri n g by observing the dif-
ference in transit depth between the u
′
and g
′
bands and to
search for the broad wings of t h e sodium doublet with the
NaI filter.
The observations were performed with moderate tele-
scope defocussing (∼ 3 arcsec) in windowed mode with ex-
posure times of 0.7 6 seconds in the red and green channels
and a cadence of 0.79 seconds. Due to the reduced pho-
ton count in the blue channel multiples of 10 frames were
averaged on-chip before readout, leading to a 7.9 second ex-
posure time in th i s channel.
The m oon was at 54% illum in a t i on on the nig ht of our
observations and we analysed the data with an airmass vary-
ing from 1.41 → 1.06 → 2.00.
All the data were reduc ed using the ULTRACAM data
reduction pipeline
1
with bias su b t ra ct i on and flatfielding
performed in the standard way. Aperture p h o t o met ry was
performed for all frames u sin g a fixed a perture. Initially,
many reductions were performed with a variety of aperture
sizes and the signa l- t o -n o i se ratio was calculated for each
using the ratio of aperture counts to aperture errors. The
optimal aperture rad iu s was found to be 18 pixels with a
sky annulus of inner radius 23 pixels a n d outer radius 27
pixels.
Some of the observations of WASP-52 were taken
through c lo u d , seen as dips in transmiss io n in the raw light
curves, which had to be removed before analysi s. These were
well defined, discrete events, with g ood quality data taken
between the clouds. The clou d y data were removed using an
iterative process. An array of running medians was calcu-
lated over a sliding box of 600 data points in the red and
green arms and 60 points in the blue. This array of running
medians was then subtracted from the raw data to flatten it.
The median absolute deviation (MAD) of this subtracted ar-
ray was calculated and sigma clipping performed to remove
those data points lying at ≥ 6σ from the MAD. This pro-
cess was then repeated with a smaller sliding box of 400
data points in the red and green and 40 in the blue with
a final sigma cut at 5σ from the MAD. The sigma clipping
was performed on both the target and comparison star in-
dependently and only those frames that passed the sigma
cut for both were kept, resulting in the removal of 20 % of
the data. After the sigma clipping and binning of the red
and green channels to the cadence of the blue channel, there
were 2494 data points i n each of the three light curves at
a cadence of 7.9 seconds giving us excellent sampling even
after the cleaning of the data.
Differential photometry was subsequently performed to
1
http://deneb.astro.warwick.ac.uk/phsaap/software/ultracam/
html/index.html
c
0000 RAS, MNRAS 000, 000–000
Transmission spectroscopy of WASP-52b 3
remove the worst effects of telluric extinction using a com-
parison star with similar magnitude and colour to WASP-52
(BD+08 5023, 23:14:12.026 +08:50:56.28). This star has a
V ma g n i tu d e of 10.59 a nd B- V col o u r of 0.86, whilst WASP-
52 has a V magnitude of 12.22 and B-V colour of 0.82. The
comparison star was checked and found t o be photometri-
cally stable. Combinations of fainter stars in the field were
tested as compari so n stars but led to more scatter in the
differential light curve than division by the single, bright
comparison.
3 DATA ANALYSIS
3.1 Light curve fitting with analytic model
We initially fitted the differential light curves wit h analytic
limb-darkened transit light curves (Mandel & Agol 2002) us-
ing a Markov Chain Monte Carl o (MCMC) algorithm, im-
plemented through the emcee (Foreman-Mackey et al. 2013)
Python package. A quadratic limb darkening law was used
and fit simultaneously with a long t im e- sc a le trend so as not
to bias the derived radii. I n the red and green channels, this
trend was fit with a second order polynomial whilst in the
blue it was fit as a function of airmass since it was clearly
related to extinction. In order to fit the airmass term in the
blue channel, an extinction coefficient that varied quadrati-
cally in time was used, which replicated the trend well. An
extinction coefficient that varied linearly in time was also
tested but co u ld not fit the sharp downt u rn in the blue light
curve at the end of the night (Fig. 1).
The scaled semi major axis, a/R
∗
, inclination of the
orbit and the time of mid transit, T
0
, were tied across the
three light curves when fitting. The p a ra met ers that were fit
individually in each of the channels were the ratio of planet
to star radius R
P
/R
∗
, the second limb darkening co effi ci ent
u2, and the parameters defining the long time scale trend.
The first limb darkening coefficient, u1, was held fixed in
the fitting as there is a degeneracy between the limb dark-
ening parameters which can affect the lig ht curve solution
(Southworth 200 8 ). The limb darkening coefficients and pri-
ors were chosen from the tables of Claret & Bloemen (2011).
Uniform priors were adopted for all th e model parameters,
with the MCMC walkers started at th e values from H´ebrard
et a l . (2013).
The resulting fits of the analytic limb-darken e d tran-
sit light curves are shown in Fig. 1. The strongest residual
across the whole lig ht curve in all three wavelengths is seen
during transit and is consistent across the three bands. This
residual is akin to the planet occulting areas of stellar activ-
ity. We considered the possibility that this anoma ly c o u l d
have been associated with the use of incorrect limb darken-
ing coefficients but no choice of coefficients could replicate
this feature.
3.2 Fitting of star spot model
The presenc e of the in-transit anomaly after the fitting of
limb-darkened analytic light curves motivated the use of
spot models. Star spot occultations have been seen in th e
transit light curves of several planets, including HD 189733b
(Pont et al. 2007; Sing et al. 2011b), TrES-1b (Rab u s et al.
2009; Dittmann et al. 2009), CoRoT-2b (Wolter et al. 2009;
Huber e t al. 20 0 9 ; Silva-Valio et al. 2010), HAT-P-11b
(Sanchis-Ojeda & Winn 2011; Deming et al. 2011; B´eky
et al. 2014), WASP-4b (Sanchis-Ojeda et al. 2011; Hoyer
et al. 2013), WASP-19b (Mancini et al. 2013; Tregloan-Reed,
Southworth & Tappert 2013; Huitson et al. 2 0 1 3 ; Mandell
et al. 2013; Sedaghati et al. 2015), HATS-2b (Mohler-F i scher
et al. 2013), Kepler-63b (Sanchis-Ojeda et al. 2013), Qatar-
2b (Ma n c in i et al. 2014), and HAT-P- 3 6 b (Mancin i et al.
2015).
If a plan et occu l t s a spot (a region cooler than the sur-
rounding photosphere) it blocks less of the stellar flux than
compared with its transit across the hotter pristine stellar
disc. This results in a bump during transit and therefore a
smaller derived planetary radius. Star spot ac ti vi ty is n o t
unexpected for WASP-52 since H´ebrard et al. (2013) found
modulations in its light curve and chromospheric emission
peaks in the Ca II H+K lines. Using these modulations they
calculated the rotation period of WASP-52 to be 16.4 days.
Spot crossing events have been modelled with a variety
of techniques, such as prism (Tregloan-Reed, Southworth &
Tappert 2013), soap-t (Oshagh et al. 2013) and spotrod
(B´eky, Kipping & Holman 2014). For this analysis we used
spotrod to generate the spot affected light curves and w ro te
an MCMC wrapper around these generated light curves to
fit them simultaneously across the bands and wit h the same
long time-scale trends as before. We chose to use spotrod
due to t h e sp eed of its integration, which uses polar coordi-
nates in the projection plane. The integration with respect
to the polar coordinate is done analytically so that only the
integration with respect to the radial coordinate needs to
be performed numerically. To calculate the projection of the
planet on the stellar surface, spotrod calculates the a rrays
of planar orbital elements ξ and η, using the formalism of
P´al (2009), and assumes the sa m e limb darkening law for
the spot as for the star.
The system parameters were again fit across the three
light curves simultaneously (a l th o u g h this time fitting for
impact parameter, b, rathe r than the inclination, as required
by spotrod) and with the addition of the pa ra met ers defin-
ing the spots. The fittin g of one spot was tested but was
unable to fit both bumps on either side of the transit mid-
point, therefore two spots were used in fu rt h er analysis. The
parameters defining each spot were the longitude, latitude ,
radius ratio of spot to star, and ratio of the spot flux to
stellar flux (with 1 being a spot with the same flux as the
pristine photosphere and 0 being a spot with zero flux). We
held u1 fixed as before but now also put Gaussian priors on
u2 with means equal to the values from Claret & Bloemen
(2011) and standard deviations from the propagated errors
in the effective temperature and surface gravity of the host
star. This prior was necessary as the limb darkening and
spot models can play off each other in trying to fit th e tran-
sit shape.
The MCMC was initiated wit h the system parameters
equal to those in H´ebrard et al. (2013) and was run for 10000
steps in burn in and another 10000 steps in the production
run. There were 31 fitted parameters with 124 walkers. The
parameters g overning a spot’s characteristics are correlated
with one another. The correlation between spot contrast and
size has been shown previously by e.g. Pont et al. (200 7 ) ,
Wolter et al. (2009), Tregloan-Reed, Southworth & Tappert
c
0000 RAS, MNRAS 000, 000–000
4 J. Kirk et al.
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1.00
Normalised flux
4 3 2 1 0 1 2
Time from mid-transit (hours)
0.030
0.025
0.020
0.015
0.010
0.005
0.000
Residuals
Figure 1. MCMC fits of analytic quadratic limb-darkene d transit light curves (Mandel & Agol 2002) revealing the presence of an
in-transit anomaly. The upper panel shows the fits to each of the 3 wavelengths; NaI (top), g
′
(middle) and u
′
(bottom) with the latter
two light curves each offset by -0.02 for clarity. The lower panel shows the residuals from these fits with the latter two offset by -0.01
and -0.02, respectively.
(2013) and B´eky, Kipping & Ho lm a n (2014). We initiated
the MCMC startin g positions with spo t s at various, ran-
domly selected, latitudes on the stellar surface so as not
to bias the fits and test for convergence. After the burn in
phase t h e error bars in the data points were rescaled to give
a re d u ced χ
2
of u n i ty.
After the first MCMC chain, a second MCMC was run
but this time with the parameters that were tied across chan-
nels fixed to the results from the first run (a/R
∗
, b, T
0
and
the spot sizes and positions). Correlations with t h ese pa ra m-
eters cause R
P
/R
∗
to move up and down together across the
three wavelengths, contributing to the uncertainty in the ab-
solute planetary radius in each of the bands. Since we are
concerned with the shape of the transmission spectrum, we
are interested in the relative radii between the bands and
not the absolute planetary radius, thus motivating the sec-
ond run of the MCMC with fixed system parameters.
We present the best fitting spot model in Fig. 2, af-
ter the second MCMC run, with the results in Table 1 and
transmission spectrum in Fig. 6 (blue squares).
With the sizes and contrast s calculated from spotrod,
we were able to create a schematic of the stellar surface
(Fig. 3, left panel) and consider what filling factor would
reproduce the derived contrasts (section 4.1.1 and Fig. 3,
right panel). This figure displays the large regions of stellar
activity along the transit chord. The second spot crossing
event compri ses of a smaller region of higher contrast (0.2
in the g
′
band, Table 1). The error in the contrast of this
dark spot also allowed for a larger, less dark sp o t.
3.3 Fitting of bright region model
As an alternative interpretation, the in-transit anomaly was
also modelled as a bright feature analogous to Solar fac-
ulae. Solar faculae are bright regions on the solar photo-
sphere which display limb brightening behaviour (e.g. Un-
ruh, So la n k i & Fligge 1999). They are often co-spatial with
sunspots but not perfectly so (e.g. Haywood et al. 2016).
The effects of occultations of bright regions in exoplanet
transits have been d i sc u s sed by Oshagh et al. (2014) and
could lead to an observable an o ma l y in transit data. There
has not yet been any conclu s ive evidence of a facula occul-
tation in a t ra n si t light curve although Mohler-Fischer et al.
(2013) found evi d en c e for a h ot spot in GROND photometry
of HATS- 2 b . They detected a bright feature i n the Sloan-g
band, which covered the Ca II lines, that was consistent with
a chromospheric plage occultation.
spotrod was also used to model the facula scenario but
instead of two individual spot s with flux ratios < 1, a single
feature was modelled with a flux ra t io of > 1. This fitting
method produced the fits seen in Fig. 4 with the results in
Table 1. This model was able to reproduce the in-transit
anomaly wit h an equally good fit as the two spot model but
with six fewer parameters. The transmission spectrum re-
sulting from the facula model is also shown in Fig. 6 (red
triangles). In contrast to the flat transmission s pectrum re-
sulting from the fit t in g of spots, the fitting of a facula led
to a slope in the planet a ry radius increasing toward s the red
(Fig. 6).
4 DISCUSSION
4.1 Spots or faculae?
It is difficult to distinguish between the spot and facula mod-
els of sections 3.2 and 3.3 using the quality of the fits alone.
Application of the Bayesian Information Criterio n (BIC;
given by BIC = χ
2
+ k ln N with k free parameters and
c
0000 RAS, MNRAS 000, 000–000