of 45
A Sub
-
Neptune Exoplanet with
a
Low
-
Metallicity Methane
-
Depleted Atmosphere
and Mie
-
Scattering
Clouds
B
JÖRN
B
ENNEKE
1
,
H
EATHER
A.
K
NUTSON
2
,
J
OSH
UA
L
OTHRINGER
3
,
I
AN
J.M.
C
ROSSFIELD
4
,
J
ULIANNE
I.
M
OSES
5
,
C
AROLINE
M
ORLEY
6
,
L
AURA
K
REIDBERG
7
,
B
ENJAMIN
J
.
F
ULTON
2
,17
,
D
IANA
D
RAGOMIR
4
,1
8
,
A
NDREW
W.
H
OWARD
8
,
I
AN
W
ONG
9
,
J
EAN
-
M
ICHEL
D
ÉSERT
10
,
P
ETER
R.
M
C
C
ULLOUGH
1
1
,
E
LIZA
M.
-
R.
K
EMPTON
1
2
,1
3
,
,
J
ONATHAN
F
ORTNEY
14
,
R
ONALD
G
ILLILAND
1
5
,
D
RAKE
D
EMING
1
2
,
J
OSHUA
K
AMMER
1
6
[1]
Department of Physics
and
Insti
tute for Research on Exoplanets,
Université de Montréal, Montréal, QC, Canada
[2]
Division of Geological and Planetary Sciences,
California Institute of Technology, Pasadena, CA 91125, USA
[3]
Lunar & Planetary Lab
oratory, University of Arizona, 1629 E. University Boulevard., Tucson, AZ, USA
[4]
Department of Physics
and Kavli
Institute
of
Astronomy
, Massachusetts Institute of Technology, 77 Massachusetts Ave,
Cambridge, MA, 02139, USA
[5]
Space Science Institute, 4750
Walnut Street, Suite 205, Boulder, CO 80301, USA
[6]
Department of Astronomy, University of Texas, Austin, TX 78712, USA
[7]
Department of Astronomy, Harvard University, 60 Garden Street, Cambridge, MA 02138,
USA
[8]
Department of Astronomy,
California Institute of Te
chnology, Pasadena, CA 91125, USA
[9]
Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, 77 Massachusetts
Ave, Cambridge, MA, 02139, USA
[10]
Anton Pannekoek Institute for Astronomy, University of Amsterdam, 1090 GE Ams
terdam, The Netherlands
[11]
Department of Physics and Astronomy, Johns Hopkins University, Baltimore, MD 21218, USA
[12]
Department of Astronomy, University of Maryland, College Park, MD 20742, USA
[13]
Department of Physics, Grinnell College, 1116 8th Avenue, Grinnell,
IA 50112, USA
[14]
Department of Astronomy, University of California, Santa Cruz, CA 95064, USA
[15]
Space Telescope Science Institute, 3700 San Martin Dr., Baltimore, MD 21218, USA
[16]
Southwest Res
earch Institute, San Antonio TX, USA
[17]
IPAC
-
NASA Exoplanet Science
Institute Pasadena, CA 91125, USA
[18]
NASA Hubble Fellow
With no analogues in the Solar System,
t
he discovery
of
thousands of exo
planets with masses
and radii intermediate between Earth and Neptune was one of the big surprises of exoplanet
science.
T
hese
super
-
Earths and sub
-
Neptunes likely represent the most common outcome
of planet formation
1,2
.
Mass and radius measurements indicate a diversity in bulk
composition much wider than for gas giants
3
; however, direct spectroscopic detections of
molecular abso
rption and constraints on the gas mixing ratios have largely remained limited
to planets more massive than Neptune
4
6
.
Here, we analy
z
e a combined Hubble/Spitzer Space
Telescope dataset of 12 transits and 20 eclipses of the sub
-
Neptune GJ 3470 b,
whose mas
s of
12.6
M
places it near the half
-
way point between previously studied exo
-
Neptunes (22
-
23
M
)
5
7
and exoplanets known to have rocky densities (7 M
)
8
. Obtained over many years,
our data set provides a robust detection of water absorption (>5
σ
) and a t
hermal emission
detection from the lowest irradiated planet to date. We reveal
a
low
-
metallicity
,
hydrogen
-
dominated atmosphere similar to a gas giant, but strongly depleted in methane gas. The
low
,
near
-
solar
metallicity (O/H=0.2
-
18) sets important
constraints on the potential planet
formation processes at low masses as well as the subsequent accretion of solids. The low
methane abundance indicates that methane is destroyed much more efficiently than
previously predicted, suggesting that the CH
4
/CO t
ransition curve has to be revisited for
close
-
in planets. Finally, we also find a sharp drop in the
cloud
opacity at 2
-
3
μ
m
characteristic of Mie scattering, which enables narrow constraints on the
cloud
particle size
and makes GJ
3470b a keystone target f
or mid
-
IR characterization with JWST.
We observed the
transiting
sub
-
Neptune mass exoplanet GJ
3470b with the Hubble Space
Telescope (HST) as part of a spectral survey of atmospheres of low mass exoplanets
(GO
13665
)
.
W
ith
an orbital period of 3.3
days
an
d a mass of 12.6
M
, GJ
3470b
is a typical
member of the
intriguingly abundant class of close
-
in
sub
-
Neptunes
(Figure 1)
.
GJ
3470b
’s
low
surface gravity
combined with the
proximity and
small size of its host star mak
e
it an outstanding candidate for
detailed
atmospheric
characterization
,
especially in
the sub
-
Neptune
mass regime for which
robust
molecular detections have remained
elusive
to date
4,9,10
.
We obtained time
-
series spectroscopy
of
six transits
using HST
, inc
luding
three
from 1.1
μ
m to 1.7
μ
m using Wide Field Camera 3 (WFC3)
and
three
at optical wavelength
s
(0.55
1.0
μm) using
the
Space Telescope Imaging Spectrograph
(STIS).
We complement
the
se
HST transit observations with a total six Spitzer/IRAC transit
observations as well as a total of 20
secondary
eclipse observations
at 3.6 and 4.5
μ
m
(Figure 2
and 3)
.
We jointly analyze
all
HST and Spitzer transit data to obtain a consistent visible
-
to
-
I
R
transmission
spectrum
covering a wavelength range between 0.55 and 5.0
μ
m
(
Figure 2
)
. The details of this
analys
i
s are described
in
the Methods Section.
For each instrument, we verify that the
measurements
are consistent
over multiple epochs
by demonstra
ting repeatability among the three
transits in each channel
, setting
tight
upper
limits on the effect of star spots on the overall spectrum
(
Supplementary Figure 4
)
.
The transit depth precision with WFC3 is substantially higher than the
one obtained with S
TIS and Spitzer due to
GJ
3470’s
higher photon flux
in the WFC3 bandpass
and the substantially higher throughput of WFC3 (30
-
40%) compared to STIS (8
-
12%). The
achromaticity of the dominant WFC3 systematics further improves the transit depth precision in
the WFC3 spectroscopic channels relative to STIS.
Our
three transit obser
vations using
optimized
WFC3
spatial scans across the full sub
-
array
enable us to collect 3
0
times more photons
and
achieve 5 times
higher
precision
than
the
previously published
stare
-
mode
transit
observation
of
GJ 3470b with this same instrument
9
.
Our
t
ransmission spectrum
for
GJ
3470b
reveals a
n attenuated but statistically
significant water
absorption feature at 1.4
μ
m in the WFC3 data
,
protruding
over an otherwise
cloud
opacity
-
dominated
visible to near
-
IR
transmission spectrum
(
Figure 2
)
.
Th
e
water absorption
is detected
in
multiple
neighboring spectroscopic channels covering the 1.4
μ
m water band.
Quantitively,
retrieval models that include
molecular
absorption
by water
are favored by the Bayesian evidence
11
at
1
24
,
770
:
1
(
5.2
σ
)
and
result in
significantly better best fit
s
than models with
out
water
(see
Methods)
.
Comparisons to models show that the data are best matched by a low
-
metallicity,
hydrogen
-
dominated atmosphere (O/H
=
0.
2
18
x
solar) with water vapor absorbing above high
-
altitude
cloud
s that become optically thick
below the
1
mbar
level at 1.5
μ
m
(
Figure 4
). High
metallicity atmospheres with high mean molecular mass are ruled out
by the data
because the
associated smaller scale height would not allow for the observed
transit depth v
ariations.
This
finding is independent of the detailed atmospheric models because the observed transit depth
variations would require the cut
-
off altitude of the grazing
star
light to vary over greater than 20
atmospheric scale height
s
across the near
-
infr
ared
,
which is not realistic.
The Spitzer eclipse
observations add to the water constrain
t
s because substantially increased water opacity would not
allow for the observed contrast in thermal emission at 3.6 and 4.5
μ
m.
Intriguingly, the high
-
altitude
cloud
s on GJ
3470b are not well represented by Rayleigh hazes as
previously reported
12,13
or a simple gray cloud deck. Instead, our measurements provide direct
observational evidence for the characteristic wavelength dependent extinction of finite
-
sized Mie
sca
ttering
aerosol
particles
(Figure 2)
.
The
se aerosol clouds
become increasingly transparent at
around 2
3
μ
m enabling us
to constrain
their
effective
particle size to 0.
60±0.06
μ
m
in
the
upper
most
layers
of the
cloud
s
(Figure 4)
.
This particle size
estimate
provide
s
a
rare
direct
observational
constraint
that can
guide
the
further
development
and verification
of
physics
-
driven
cloud
and haze
models
for
exoplanets
.
Here, w
e account for the non
-
gray
cloud
opacities
in our
retrieval analysis
by modeling the
finite
-
sized particles
using Mie scattering theory
11,14
and
describing the effective particle size, the upper cloud deck pressure
, and
the cloud scale height
as
free parameters (see Methods). All posterior constraints on the atmospheric gases provided in t
his
work account for the uncert
ainties introduced by
the
cloud
s
as well
as the parameterized
“free”
temperature structure
(see
Supplementary
Figure
s
6
and
7
)
.
GJ 3470b's transmission spectrum
also show
s
a striking absence of methane
absorption
.
For
the
relatively cool
,
low
-
metallicity
atmosphere
of GJ
3470b, atmosphere models with solar carbon
-
to
-
oxygen ratio would have predicted methane to be the dominant carbon
-
bearing molecule.
However, methane absorption at 1.6
μ
m in the WFC3 bandpass and at 3.3
μ
m i
n Spitzer/
IRAC
channel 1 is not observed indicating a strong depleting of methane (Figure 2).
Th
e
low methane
abundance is independently supported by
the
twenty secondary eclipse observations, which can be
used to constrain the shape of GJ
3470b's thermal
emission spectrum in the 3.6 and 4.5
μ
m Spitzer
bands (
Figure 3
). We
detect
strong thermal flux emerging at 3.6
μ
m
at 4.7
σ
significance
(
#
=
115
)
*+
,
*-
p.p.m
)
and a tight upper limit at 4.5
μ
m
(
#
=
3
±
22
p.p.m
)
. This is
in
contrast
to the
prediction
for the fiducial solar abundance model,
but in agreement with
the low methane
abundance inferred from the transmission spectrum
(Figure 2)
.
Quantitatively
,
our retrieval
analysis shows
that the methane abundance is below 1.3x10
-
5
at greater than 99.7% confidence,
substantially below the value of 4.6x10
-
4
expected for a solar abundance atmosphere in chemical
equilibrium (
Figure 4
). The best fitting
models show a
striking
methane
depl
etion
by
three
orders
of magnitude
compared to eq
uilibrium
.
We assess the origin of the methane depletion through
state
-
of
-
the
-
art
photochemi
cal
modeling
and thermal modeling
of
GJ
3470b’s
atmosphere
(see Methods).
Consistent with
Refs
15,16
we find
that
the methane abundance
in the layers probed by the observations
should not be reduced
substantially by photochemistry in layers probed by our observations
(Figure 4, Supplementary
Figure 8)
.
Possible explanations for the unexpected lack of methane could be substantial interior
h
eating, photochemical depletion due to catalytic destruction of CH
4
in deeper atmospheric
regions
, or a
low C/O ratio as a result of the planet formation process. The interior heating
scenario
would require interior temperatures
(
T
int
)
above 300
K to push
the otherwise relatively cold mid
-
atmosphere of GJ
3470b into the CO dominated regime
17,18
. Evolution modeling of GJ
3470b
indicates that internal heat from formation should have been radiated away within a few Myr, well
below the estimated age of the syst
em
19
; however,
tidal heating due to forced eccentricity from
another unseen planet in the system, similar to the situation with Jupiter's moon Io
could be
a
possible explanation.
The residual non
-
zero eccentricity of GJ
3470b
as
independently confirmed
by
our eclipse observations
and
radial velocity
measurements
support this hypothesis.
Alternatively,
GJ
3470b's surprising lack of methane
could
potentially be the results of
photochemical depletion due to catalytic destruction of CH
4
in deeper atmospheric regions where
photolysis of NH
3
and H
2
S release large amounts of atomic
hydrogen
. The fact that ammonia is
also
depleted in comparison to expectations based on our chemical
-
kinetics modeling (Figure
4
)
is consistent with this cataly
tic
-
destruction possibility.
Ammonia is an important quenched
disequilibrium product
(
Supplementary Figure 8
)
, and we would have expected to see NH
3
absorption at 1.5
μm (Figure
2
).
In
either
scenario,
the carbon freed from methane would most
likely be
locked up in CO, which
can
be
see
n
in
the transmission spectrum
at 4.
5
μ
m (Figure
2
)
and as a suppression of thermal flux within the 4.5
μ
m Spitzer bandpass (Figure
3
).
HCN is also
one potential major sink of the carbon
in
the coupled CH
4
-
NH
3
photochemistry
if the elemental
N/C ratio is larger than solar
20,21
;
however, we also obtain an
upper bound on the HCN abundances
(
Supplementary Figure 7
).
In either case, all
CH
4
destruction
scenarios
would also
lead to the
production of CH
3
and other radicals, some fraction of which can react with other atmospheric
carbon and nitrogen species to form increasingly heavy hydrocarbons and nitriles, eventually
ending up in refractory soot
-
like
haze
particles. These photochemically produced parti
cles could
provide
an
explanation for the observed
cloud
opacity at
shortward of
3
μ
m
, although recent
experimental work has also shown that similar photochemical hazes can be formed even in the
absence of methane
22,23
.
In either case, the
30
-
90 nm particl
es
found in experiments
2
would likely
need to coagulate to form
larger
aggregates to explain the inferred wavelength dependen
ce
of the
cloud
opacity on GJ
3470b.
Overall, our spectra show directly through atmospheric observations that close
-
in sub
-
Neptunes
can have near
-
solar metallicity atmospheres likely formed by the direct accretion of primordial gas
from the protoplanetary disk onto a rock/iron or ice
-
dominated core as suggested by recent planet
formation models
19,24
.
The near
-
solar metallicity
is part
icularly intriguing because the steep
increase in planet occurrence rate
from
>
20 M
to
wards
10 M
25
suggests that sub
-
Neptunes
could have a more efficient planet formation process intrinsically different from planets more
massive than Neptune
4
7,10,17,26
.
S
ub
-
Neptune formation beyond the ice line and subsequent
migration
c
ould have led to much higher atmospheric metallicities, or even water worlds
27
, which
we do not find for GJ
3470b. Instead, our measurement of a near
-
solar water abundance favors
formati
on scenarios in which the core accreted a primordial gas envelope whose metal content was
subsequently not notably enriched by planetesimal accretion
28
, the erosion of the icy/rocky core,
or late collisions
29
. GJ
3470b’s gas envelope is also
expected
to have undergone substantial mass
loss
30
, but
this mass loss was likely in the hydrodynamic regime as
we see no evidence for
preferential loss of lighter elements.
Finally, t
he unexpected methane depletion further indicates
that our understanding of the
chemical and thermal processes on these low
-
mass planets remains
incomplete
.
Opportunely, our detection of sub
-
μ
m Mie scattering particles indicates that the
clouds
or hazes
in the atmosphere of GJ
3470b become increasingly transparent beyond 3
μ
m, making
GJ
3470b an excellent target for future JWST mid
-
IR observations
both as an archetype for the
intriguing population of sub
-
Neptunes and as a laboratory for atmospheric chemistry and cloud
formation in warm atmospheres.
Acknowledgement.
This work is based
on observations with the NASA/ESA HST, obtained at
the Space Telescope Science Institute (STScI) operated by AURA, Inc. We received support for
the analyze by NASA through grants under the HST
-
GO
-
13665 program (PI Benneke). This work
is also based in part
on observations made with the Spitzer Space Telescope, which is operated by
the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA (PIs
Knutson and Désert).
B.B. further acknowledges
financial supported by the Natural
Sciences and
Engineering Research Council (NSERC) of Canada and the Fond de Recherche Qu
é
b
é
cois
Nature et Technologie (FRQNT; Qu
é
bec).
J.M. acknowledges support from NASA grant
NNX16AC64G
,
the Amsterdam Academic Alliance (AAA) Program
, and
European Rese
arch
Council (ERC) under the programme Exo
-
Atmos (grant agreement no. 679633)
. D. D.
acknowledges support provided by NASA through Hubble Fellowship grant HST
-
HF2
-
51372.001
-
A awarded by the Space Telescope Science Institute, which is operated by the
Associ
ation of Universities for Research in Astronomy, Inc., for NASA, under contract NAS5
-
26555.
Author contributions.
B.B. led the data analysis of the HST and Spitzer transit data, with
contributions from J.L., I.W., and H.K. L.K. and J.M.D. performed indep
endent analyses of the
Spitzer transits and found consistent results. H.K. led the data analysis of the Spitzer secondary
eclipse observations. J.M. provided the chemical kinetics atmosphere models. B.B. and C.M.
provided the self
-
consistent atmospheric mo
dels and the atmospheric retrieval analysis. B.B. wrote
the manuscript, with contributions from B.F., H.K. and J.M. All authors discussed the results and
commented on the draft.
Competing interests.
The authors declare no competing financial interests.
Author information.
Correspondence and requests for materials should be addressed to B.B.
(
bbenneke@astro.umontreal.ca).
Figure 1: Planet mass versus equilibrium temperature for known low
-
mass planets with signal
-
to
-
noise ratio >3 mass measurements.
Planets
with published space
-
based spectroscopic observations for
atmospheric characterization are indicated by large circles (black and coloured) and all other planets by
small grey circles. Equilibrium temperatures are calculated for a Bond albedo of A
B
= 0.1. Vertical and
horizontal bars indicate the 1
σ
uncertainties. Among the spectroscopically studied planets, the planets with
non
-
detections of the atmosphere are shown in black, planets with detection of only transit features in blue,
planets with det
ected thermal emission deviant from black
-
body radiation in yellow, and GJ 3470 b
with
spectral features detected both in transit and eclipse measurements in red. The green dotted curves show
the equilibrium temperatures for which we expect CH
4
(below the
curve) and CO (above the curve) to be
the dominant carbon
-
bearing species in the photosphere based on self
-
consistent modelling in chemical
equilibrium. The dashed vertical line indicates the mass of Neptune for reference.
Figure
2
: Transmission spectr
um of GJ
3470b.
Black data points show transit depth measurements from
the HST/STIS, HST/WFC3, and Spitzer/IRAC observations analyzed in this
study
.
Vertical and horizontal
black bars indicate the 1
σ
transit depth uncertainties and the wavelength ranges of
the measurements,
respectively.
The best fitting model with near
-
solar water abundance, Mie scattering clouds, and strong
methane depletion is shown by the red curve, with circles indicating the bandpass integrated model. Water
absorption results in
increased transit depth at 1.4
μ
m (zoom in pane
l b
)
.
F
inite
-
size
d
Mie
-
scattering
particles
(~0.6
μ
m)
result in
a characteristic drop off in
cloud
opacity beyond 2
μ
m (red dotted curved).
Adding 100 ppm methane to the best fit model results in significant d
isagreement to the data at 1.6 and 3.6
μm
(blue curve)
. Similarly, adding 100 ppm ammonia results in disagreement at 1.5 μm (green curve
in
panel b
).
A cloud
-
free solar metallicity model (orange curve) and the best
-
fitting gray cloud model (gray
dashed
cur
ve) are shown
in panel (a)
for reference. Both provide a poor fit to the data
.
The dominant
molecular absorbers for each model are labeled at the top with colors matching the color of
the
spectra.
A
distribution of models from the joint retrieval modeling
of transit and eclipse data is
shown in
Supplementary
Figure 6
.
Previous measurements
9,12,31
34
are in statistical agreement with our data, but have
significantly larger transit depth uncertainties and are omitted here for clarity.
Figure
3
: Thermal
emission spectrum of GJ
3470
b.
Spitzer/IRAC measurements (black) are compared
to simulated model atmospheres (colored solid curves)
and black body curves
(
grey
dotted curves
)
.
Vertical
and horizontal black bars indicate the 1
σ
transit depth uncertainties
and the wavelength ranges of the
measurements, respectively.
T
he horizontal
black
dashed line represents the 3
σ
upper limit
for the 4.5
μ
m
measurement
.
Colored circles indicate the bandpass integrated model
s
. Consistent with the transit data, t
he
fiducial
methane
-
rich, chemical equilibrium model with solar metallicity (orange) results
in
a poor fit to the
data. Instead,
the
strong
flux reversal
between
3.6 μm and 4.5 μm Spitzer
/
IRAC bandpasses strongly favors
methane
-
depleted
scenarios.
The
best joint fit of transit and eclipse with near
-
solar water abundance and
strong methane depletion matches both data points within
1
σ
(red,
also see
red curve in
Figure 1)
.
Consistency between the transit and eclipse is further underscored because the bes
t fit model from the transit
data alone (
purple
) captures the
flux reversal between 3.6 μm and 4.5 μm
correctly and presents a much
better predictor of the eclipse data than
our
fiducial methane
-
rich
self
-
consistent
model (orange).
500 K
600 K
700 K
800 K
900 K
1000 K
self-consistent model
(1xsolar, C/O=0.54)
best-fit to transit only
(CH
-depleted)
best-fit to transit & eclipse
(CH
-depleted)
4
4