1
Supplementary Materials for
Isotopic evidence of long-
lived volcanism on Io
Katherine de Kleer
et al.
Corresponding author: Katherine de Kleer,
dekleer@caltech.edu
DOI: 10.1126/science.adj0625
This PDF file includes:
Materials and Methods
Supplementary Text
Figs. S1 to S6
Tables S1 to S5
References
2
Materials and Methods
Observations
Observations were made with ALMA of Io’s leading and trailing hemispheres on UT 2022 May
24 and 2022 May 18 (respectively) through program 2021.1.00849.S. QSO B0420-0127 and
QSO B2251+155 were observed for flux and bandpass calibration. The time on source was
1h40m for the leading hemisphere observation, and 50 minutes for the trailing hemisphere
observation, because better weather
conditions allowed
the desired SNR to be obtained in a
shorter time. Parameters
of the observations and Io’s geometry at the time
are given in Table S1
.
For the leading hemisphere observation, Io rotated 28 degrees during the observation, which
smears the data by
0
′′
.23
in the rotation direction
; this is more than half the spatial
resolution, but
does not affect our interpretation because all presented spectra are averaged over regions larger
than the spatial resolution
.
The o
bservations were conducted while ALMA was in its C
-4 antenna configuration, which
provides a spatial resolution of about a quarter to a third of Io’s diameter
(~1000 km) at the
frequencies of observation
. The maximum baselines were 780 and 740 m for the leading and
trailing hemisphere
observations respectively.
The o
bservations used
ALMA’s Band 8 receivers, operating between 416 and 432 GHz. Within
this frequency coverage w
e selected
13 spectral windows, each with a
frequency resolution of
244 kHz (corresponding to a velocity resolution of 170 m s
-1
) and a bandwidth of either 235 or
118 MHz, for a total recorded bandwidth of 1.5 GHz. Band 8 covers rotational lines of all 4
targeted species
(SO
2
, SO, NaCl, KCl
) and their isotopologues in a single spectral set-up (
Table
S2
). ALMA’s
frequency tunings
were set to track the changing line-of-sight velocity of Io during
the observations, so
there is no spectral smearing provided the tracking uses an ephemeris
sampled at sufficient precision
. For these observations the ephemeris
(
35
)
was sampled at 10
minute intervals; the resulting spectral smearing is
< 2 kHz, which is a small fraction of the
spectral resolution. The full width at half maxima for the observed lines are in the range of 0.6 to
1.0 MHz for the sulfur-bearing species and 0.95-1.15 MHz for the chlorine-
bearing
species. A
tuning error resulted in some of the spectral windows being tuned incorrectly for the leading
hemisphere observation such that two SO
2
lines (430.229 and 430.232 GHz) were not covered by
that observation. These lines are not used in our analysis.
Data r
eduction
The data were processed through the standard ALMA pipeline (
36
) to produce a calibrated
measurement set (MS) containing the interferometric visibilities, which are the amplitude and
phase of the
cross
-correlat
ed signal between each pair of antennas. The line-
free spectral
channels across all spectral windows were split out to produce a continuum MS. This continuum
MS was
self
-calibrated (
37
) and
imaged with
the
C
ASA
software
(
38, 39
) using an iterative
procedure. First, a continuum limb-darkened disk the size of Io at the time of observation was
produced and converted to visibilities; a limb darkening parameter of 0.2 was used, but the final
image is only weakly sensitive to the choice of limb darkening parameter
. The data were then
phase calibrated using the limb
-darkened disk model visibilities
, and imaged using the same
3
model as a starting point for the CLEAN deconvolution algorithm (
40
). The self
-calibration and
CLEAN deconvolution process was then repeated, using the output of the previous self-
calibration and imaging round as the starting point for the subsequent it
eration. In each round,
the self
-calibration was performed using an increasingly shorter solution interval, and the
CLEAN algorithm was employed using an increasingly deep threshold. We us
ed three iterations
of self
-calibration and deconvolution; further iterations did not improve the SNR
of our images.
We appl
ied
the phase calibration derived from the continuum data (described above) to the
continuum-
subtracted spectral line data, then
image
d each
channel in the spectral line data
channel separately
. This produced a spectral data
cube for each spectral window with ~0
′′
.28
resolution and a spectral sampling of 244 kHz, and
a frequency
-averaged continuum image at
~0
′′
.28 resolution. Figure
s S1 and S2
show the continuum-
subtracted
image integrated over each
spectral line
listed
in Table S2. We find
that each spectral line from a single species,
including
both isotopologues, has the same spatial distribution
, which indicates that the measured spatial
distributions of the species are not biased by artifacts.
The 1D spectrum in each spectral window was extracted using an aperture that includes all pixels
that are 5% of the peak continuum level or higher in the continuum image. This produces an
aperture whose diameter is 1.3
×
the diameter of Io
, i.e., extending one resolution element beyond
the edge of Io.
The
flux density uncertainties were
estimated in two ways. First, a 1D spectrum was extracted in
a region absent of sources, in exactly the same way as for the source and using the same aperture
size; the standard deviation of the spectrum across each spectral window
was used as an estimate
of the noise in that
spectral window. Second, the standard deviation of the spectrum of Io in
spectral regions without apparent spectral lines was calculated, again providing a noise estimate
per spectral window. We adopt the
more conservative uncertainties derived from the latter
method, because it
incorporates
the thermal noise from Io’s continuum. The first method gave
uncertainties that were
a factor of 1
to 3 lower, which we regard as under
-estimates
. The degree
of bandpass calibration noise varies between spectral windows, and is non-negligible in some
windows. The bandpass calibrator spectrum was smoothed using a frequency width of 7.8 MHz.
This smoothing reduces the noise introduced by the bandpass calibrator, and although it can
introduce spectral artifacts
, such artifacts would be similar in width to the smoothing window
and hence much broader than Io’s emission lines. The uncertainties on the datapoints incorporate
this noise, which is particularly high in the spectral windows containing the lines at 419.640,
428.298, 429.863, 429.952, and 420.887 GHz. The 1
σ
noise is
shown in Figures 2 and 3
and
incorporated into the maximum-likelihood calculations (see below)
.
Overview of modeling and retrievals
To determine the isotope ratios for sulfur and chlorine, we found best-
fit ting
model parameters
by fitting a forward model to the observations as follows. Model spectra we
re generated
using a
radiative
-transfer model for the atmosphere of Io (
8, 9
), which we
updated to add additional
species. Our model includes opacity from SO
2
, SO, NaCl, and KCl, including the
32
S and
34
S
isotopes of sulfur and the
35
Cl and
37
Cl isotopes of chlorine. Models were fitted to the data using
a Nelder
-Mead minimization algorithm as implemented in the
optimize
package in the S
CIPY
software (
41
). Once the best
-fit ting
solution was determined, we used the E
MCEE
Markov chain
4
Monte Carlo (MCMC) Ensemble sampler (
42
) to explore the model parameter space and
determine the uncertainties on the best-
fit parameters. Additional detail
s on each of these steps
are given below.
The parameter values that correspond to the maximum likelihood values output by the MCMC
simulations match the best-
fitting values found by the minimization, with differences well below
1
σ
. We report the resulting parameter uncertainties as the 1
σ
range measured from the posterior
probability distributions output by the MCMC simulations. Table 1
reports the best-
fitting
parameters and MCMC
-derived uncertainties for all free parameters in the models.
Emission line selection and treatment
For the radiative-transfer modeling, the line frequencies and strengths were adopted from the
Cologne Database for Molecular Spectroscopy [
CDMS
(
43, 44
)
]. We also tested
line lists from
the JPL Molecular Spectroscopy repository (
45
), but found they did not match the observed line
positions in our datasets for any species except
32
SO
2
and
34
SO
2
. The CDMS frequencies for the
lines of the sulfur-bearing molecules agree with our observations after accounting for the line-of-
sight velocity of Io. For
NaCl
, KCl
, and their isotopologues, we observed an additional velocity
shift
that we
ascribe to bulk motion of the gas
and
included as a free parameter in our model
fitting
. The observed ~100 m s
-1
velocity shift is the same across all chlorine
-bearing species
within each dataset
; it is smaller than
the velocities of
Io’s largest class of plumes [500-1000 m s
-
1
(
46
)]. A velocity shift parameter is not included in the SO
2
model because the line positions
match
the expected frequencies
from CDMS
. Th
e velocity
difference between the chlorine- and
sulfur-bearing species
could arise because SO
2
and SO are more uniformly distributed, such that
gas velocity components produce line broadening rather than a frequency shift. The broadening
of the SO
2
and SO lines due to Io’s rotational motion is included in the model (see below
). T
he
velocity shifts for the chlorine-bearing gasses are unlikely to be due to line list errors
, because
the shift is the
same across NaCl and KCl lines (per dataset), whereas
line list errors introduce
offsets that differ between lines but are the same between
datasets
(per line)
. T
he frequency
errors
that would be introduced by using the JPL line lists are much larger than the observed
frequency shifts due to gas velocity.
The strengths of the emission lines are sensitive to temperature, and the temperature profile in
Io’s atmosphere is poorly known. The isotope ratio also
varies
with altitude
due to gravitational
stratification
. To determine the isotope ratio, we therefore used only the two lines of
32
SO
2
(418.815 and 429.863 GHz) that are sensitive to the same low atmospheric altitudes as the
34
SO
2
lines.
The other
32
SO
2
lines covered by our data, as well as those used for the previous isotope
ratio measurement (
15
),
have higher line opacities such that emission arises predominantly from
higher altitudes than the
34
SO
2
lines are primarily sensitive to.
This is particularly true near the
limb where the path length through the atmosphere is longest. This is illustrated in Figure S3
,
which
shows the contribution functions calculated from the model opacities at line center
for the
32
SO
2
and
34
SO
2
lines targeted in our observations, compared to
those used for the previous
isotope ratio measurement
.
Particularly near
the limb, all the
32
SO
2
lines used in the past work,
and most of the
32
SO
2
lines covered by our data, are sensitive to different altitudes than the
34
SO
2
lines.
5
The SO
2
gas temperature is tightly constrained by our observations because the SO
2
model fitting
includes six emission lines from
high to low excitation
(Table S2
). Temperature affects both the
line width
s and the relative strengths of the lines.
If each line w
ere
fitted
independently, the
temperature would be degenerate with column density and with the velocity distribution of the
gas, leading to much larger uncertainties. However, because we fit multiple lines simultaneously
,
the best
-fit ting
temperature is constrained
by the relative line strengths. The resulting best
-fit ting
temperature, in combination
with Io’s rotation, then determines the model line widths
. T
he
imperfect match between the model line widths and some of the observed spectra (Fig. 2) could
arise from velocity components (e.g. from winds or plumes) that a
re not included in our model.
Model atmosphere geometry and line opacity
To determine a disk-
integrated
model spectrum for comparison with the data, w
e model
ed the
emission from Io’s atmosphere as a function of latitude and longitude, accounting for the
dependence of atmospheric path length on emission angle and
the Doppler shift corresponding
to Io’s solid-body rotation at each
latitude and longitude. The model was output with
a range of
spatial resolutions
, from which we selected
the coarsest model resolution
that did
not result in a
disk-integrated spectrum that differed substantially from that produced by higher resolution
models. Based on this criterion, we
selected models generated with a spatial resolution of 0
′′
.06
(about 6% of Io’s diameter), which were then spatially integrated to produce a disk-
integrated
model spectrum
. Changing the spatial resolution of the model causes
minor changes in the
derived column densities, but does not affect the derived isotope ratios.
For the
34
S/
32
S model fitting
, we assumed that
Io’s
atmosphere is homogeneous in latitude and
longitude. Figure 1 indicates that the observed fractional coverage (
fraction of the surface area of
Io above which there is SO
2
gas
) is
closer to ~50%. If the lines are optically thin, there is a linear
trade-off between column density and fractional coverage: if column density is increased and
fractional coverage decreased proportionally, the model line strength remains the same. This is
the case p
rovided the fractional coverage is above ~15%. Our assumption of uniform coverage
therefore does not impact the derived isotope ratio, but it does affect the derived column
densities.
For NaCl and KCl, the fractional coverage is un
clear from the images
. Figure S3
shows
contribution functions that have the same disk-integrated column for the cases of 20% and 5%
fractional coverage, demonstrating that if these species exhibit a lower fractional coverage and
higher column density, the observed emission is coming from higher altitudes than if the species
are more uniformly distributed across Io.
If the fractional coverage is below ~10% there is
enough opacity in the Na
35
Cl line that it no lo
nger traces emission from the same altitudes as the
Na
37
Cl line. This necessitates our inclu
sion of fractional coverage as a free parameter in the
chlorine model fit
ting; if a broad range of fractional coverages is allowed by the data, the effect
is to increase the derived uncertainties on all parameters that are correlated with fractional
coverage. In our
analysis
, including this free parameter primarily increases the uncertainties on
the gas column densities, because the best
-fit ting
models are in a region of
the parameter space
where opacity is low and the derived isotope ratio is not strongly impacted.
6
As a
n additional check
of whether
opacity effects may bias the derived isotope ratios
, we
extracted
spectra from localized regions on Io’s disk with
both low and high path lengths (disk
center and low
-latitude limbs
, respectively
) then applied the sulfur model fit
ting
. For the chlorine
model fitting, we performed the same check using spectra from both fainter and bright
er
emission regions. Th
ese test
s used
circular apertures
with
0
′′
.3 diameter
s, shown in Fig. S4, to
extract the spectra.
The best
-fit ting
parameters for each region are given in Table S3. This test
assume
s a coverage fraction of 1.0 within the 0
′′
.3 aperture. Some of the NaCl and KCl emission
occurs very close to the limb. We find that the derived column densities are
sensitive to the exact
emission angle used in the models, so should be interpreted with caution, but the derived isotope
ratios are not sensitive to this parameter. The
results of this test show that the
34
S/
32
S and
37
Cl/
35
Cl ratios derived from all spectra extracted
within
a given hemisphere differ by
<1.5
σ
from
the values derived from the disk-
integrated spectra. The lack of systematic differences in
the isotope ratios derived for low and high emission regions indicates that local opacities do not
bias our derived isotope ratios. For the best-
fit ting
atmospheric parameters and model resolution
adopted above, the optical depths of all
lines are
<0.1 over most of the surface, and a
lways
<0.5
for the sulfur-
bearing species and <0.7 for the chlorine-
bearing species (with
the highest values
being for the strongest lines at the limbs).
The lines of Na
35
Cl and Na
37
Cl are detected at a
much higher SNR
than K
35
Cl and K
37
Cl, due to
the higher abundance of NaCl relative to KCl. Therefore the chlorine model fitting is
dominated
by NaCl; using the Na
35
Cl and Na
37
Cl lines alone
gives the same
37
Cl/
35
Cl ratio, within the 1
σ
uncertainties, as using both NaCl and KCl. The Na/K ratios for both hemispheres, including the
disk-
integrated and local
analyses
, are all in the range 3 to 10, consistent with previous studies
(
14
).
Examples of derived uncertainties
Figure S5 shows a random selection of model spectra drawn from the posterior probability
distribution
, compared to the observed sulfur data for the trailing hemisphere, to illustrate the
variation
s in the spectra produced by varying the parameter
s within their
uncertainties.
Figure S6
shows models in which the SO
2
column density i
s fixed at its best-
fit ting
value but the
34
S/
32
S
ratio is varied from 0.040 (below the Solar System average) to 0.080 (above our best-
fit ting
value)
.
Supplementary Text
Patera co
-located with chlorine
-bearing gasses
Our models indicate that
NaCl and KCl have high gas temperatures, and
Fig. 1 shows they are
localized to discrete locations. As discussed in the main text, we interpret
them as
only present in
volcanic plumes. For
each of the two dates of observation, we determined the position of each
source in the NaCl image
s and convert
ed it to a latitude and longitude on Io using the geometry
at the time of observation
(Table S1
). The latitudes and longitudes of the sources marked in Fig.
1 are given in Table S
4. The u
ncertainties on the latitude
s and longitude
s were determined
by
7
calculating the latitude and longitude of every pixel within a 10
×
10 pixel box (roughly one
resolution element on each side) surrounding the determined source center, then taking the
standard deviation within the box. We investigated whether these correspond to known surface
features
(
47
). Table S
4 lists the most likely patera as well as
all paterae that fall within the 1
σ
uncertainties.
Volcanic activity at Kurdalagon Patera (
consistent with location
1 in Figure 1), is thought to have
been at least partially responsible for the massive brightening of Jupiter’s sodium nebula in early
2015 (
48
); our identification of NaCl gas at the location of the patera is consistent with
this
connection because it suggests that the style of volcanism taking place at Kurdalagon Patera
produces Na-bearing gas
. However, i
nfrared
images taken simultaneously with our ALMA
observations on 2022 May 24 (
49
) do not show thermal emission at the latitudes and longitudes
where we observe NaCl
and KCl
gasses. If the regions of high NaCl
and KCl gas
density are
indeed volcanic plumes, they originate from volcanic centers that are not actively extruding large
volumes of lava.
Rayleigh distillation m
odel
We use a Rayleigh distillation model to relate the present
-day ratio of two isotopes to the fraction
of material that has been lost from th
e system over time. This relationship is quantified through
the Rayleigh equation, which is
푅푅
=
푅푅
0
푓푓
α
loss
34
−1
34
34
for the
34
S and
32
S sulfur isotopes. We use
this
to determine the value of
f
, the fraction of Io’s original sulfur inventory remaining at present
day
, based on our measured
34
R
.
Rayleigh fractionation entails the progressive and irreversible removal of material from a system
(referred to as a reservoir
). W
e assume an initial
34
S/
32
S (
34
R
0
) for our system and a value for the
fractionation factor (
34
α
loss
), which describes the instantaneous isotopic partitioning between the
modelled system and each packet of material removed at each time step. The Rayleigh
distillation framework assumes
that i) the material is being removed continuously, and ii) the
residue is well mixed.
For the case of sulfur on Io, we propose that the well-mixed system consists of all of Io’s sulfur
that is not in the moon’s
core. It is also possible that a smaller shallow system, consisting just of
Io’s atmosphere and crust (and perhaps some portion of the upper mantle), is mixed more
rapidly, due to shallow re-melting of surface frosts recycled into the crust.
In such a scenario,
this smaller well-mixed sulfur reservoir would become isotopically fractionated more rapidly
than the full mantle plus crust system
. However, to maintain the crust-
atmosphere reservoir
,
mantle material would need to be continuously injected to balance Io’s mass loss rate of 1000 to
3000 kg s
-1
. This injection of essentially unfractionated mantle material would buffer the near
-
surface
system such that the atmospheric
34
S/
32
S could not become highly fractionated: the
atmospheric
34
S/
32
S would take values between those of steady
-state and Rayleigh fractionation
scenarios
(
20
). Fractionation in the shallower, smaller system could only produce a highly
fractionated atmospheric
34
S/
32
S if there is no addition of unfractionated mantle. We consider
that
scenario
highly unlikely, given the observed mantle-derived volcanism and the amount of
sulfur input into the crust that is required to balance Io’s mass loss.
8
By assuming the
well
-mixed reservoir consists of
all Io’s sulfur that is not in the core, our model
calculation also requires
that there is no sulfur
exchange between the core and mantle.
The sulfur
isotopic fractionation factor between metal and silicates is close to 1 (
50
)
, so we expect
that
the
formation of
Io’s core
left it with the moon’s initial sulfur isotope composition. If Io’s core
suppl
ies
sulfur
to the mantle, this provides a source at Io’s initial
34
S/
32
S ratio and therefore
lowers the average
34
S/
32
S ratio of the sulfur
that is available for loss. This would therefore
require even greater sulfur
loss to explain
our measurement.
The derived fraction of sulfur
lost from Io depends
on our adopted initial isotope ratio
34
R
0
and
34
α
loss
. As discussed in the main text, potential deviations in the
34
S/
32
S ratio of Io
-forming
material from the S
olar S
ystem average are expected to be four
orders of magnitude smaller (in
δ
34
S
VCDT
) than our observed fractionation. Adopting the Solar System average for
34
R
0
is
therefore not a large source of uncertainty. The value we adopt for
34
α
loss
assumes
that all loss
takes place
at or above an exobase located at 600 km altitude. T
he altitude of the exobase is
uncertain and
could be as low as 100 to 200 km (
51, 52
). During Io’s
night time, the exobase
might
be
at the surface itself
. If we adopted
an exobase at a lower altitude than 600 km
, it
would
result in an
34
α
loss
value closer to 1
. This would put our loss fraction derived from the Rayleigh
equation at the upper end of our
reported range but does not qualitatively change our
conclusions.
In our implementation of th
e Rayleigh model, we assume a constant fractionation factor,
34
α
loss
.
On Io, however,
34
α
loss
probably changes on diurnal, seasonal, and stochastic timescales
, as Io’s
exobase altitude changes in response to changing atmospheric densities. We made the
simplifying assumption of a fixed altitude because we expect the exobase is typically at or below
our adopted value, such that any deviation from our assum
ption would result in the same
conclusion, or even
greater sulfur loss.
9
Fig. S
1.
Distributions of all observed
molecular emissions from Io’s leading hemisphere.
Same as Figure 1, but for all the lines we observed (listed in Table S2) on the leading hemisphere
for (
A
-G
)
32
SO
2
, (
H-K
)
34
SO
2
, (
L
)
32
SO, (
M
)
34
SO, (
N
) Na
35
Cl, (
O
) Na
37
Cl,
(P
) K
35
Cl, and (
Q-R
)
K
37
Cl. A tuning error led to no recorded data for
the SO
2
lines at (
F
) 430.229 and (
G
) 430.232
GHz. C
olorbars are in the intensity units given in Panel R.
Fig. S
2.
D
istributions of all observed molecular emissions from Io’s trailing hemisphere.
Same as Fig. S1, but for the trailing hemisphere.
10
Fig. S3.
Contribution functions
for emission lines from this and previous work
.
Contribution
functions for all observed
lines of the sp
ecies used to derive the sulfur
and chlorine isotope
ratios
. The lines used in previous work (
15
) are shown for comparison. (
A,C,E
) C
alculated
values for the center of the disk and (
B,D,F
) the limb,
for the species indicated on the panels.
The SO
2
contribution functions assume an SO
2
column density of 1x10
16
cm
-2
and gas
temperature of 240 K, corresponding to the leading hemisphere best fitting values in Table 1
. For
32
SO
2
, only the lines at 418.816 and 429.864 GHz are used to derive the isotope ratio because
they are sensitive to the same atmospheric altitudes as the
34
SO
2
lines, especially
near the limb
where much of the emission appears
. T
he NaCl and KCl
contribution functions assume a
temperature of 800 K and 20% fractional coverage unless otherwise indicated; some contribution
functions assum
e an alternative 5% coverage (for an equivalent disk-integrated column) to show
how much the fractional coverage can impact the relative altitudes the isotopologues are
sensitive to
.
11
Fig S4.
Locations used for local model fitting
.
Images of
(
A-B
) the leading and (
C-D
) trailing
hemisphere in example emission line
s of (
A,C
)
32
SO
2
and (
B,D
) Na
35
Cl (frequencies labeled
above each panel)
. Red circles indicate the apertures used for our tests (see text). For
32
SO
2
, the
identified regions are the e
ast l
imb (EL), w
est limb (WL), and disk c
enter (DC). For Na
35
Cl, the
identified regions are the three brightest emission locations in each observation. Other symbols
are the same as in Figure 1.
12
Fig S
5.
Visualization of parameter uncertainties.
Same as Fig. 2
I-K and
M-O but with
multiple models (gray curves) corresponding to 150 parameter combinations randomly selected
from the joint posterior probability distribution determined by the MCMC simulation
.
Fig S
6.
Visualization of different isotope ratios.
Same as Fig. S5, but with model
s
corresponding to different isotope ratios
(see legend)
. The models shown all adopt the best
-
fit ting
32
SO
2
column density for this observation. The
34
R
values
shown for
34
SO
2
were chosen to
be ±
5
σ
from
the best fitting value
.
13
Table S1. Observing parameters for the two observations presented in this paper.
Hemisphere
Leading
Trailing
Date/Time [UT]
2022
-
05
-
24
12:48 to 16:07
2022
-
05
-
18
10:28 to 12:02
Time on Source
1h40m
50m
Precipitable water vapor [mm]
0.7 to 0.8
0.5
Angular
resolution
0
′′
.23
×
0
′′
.35
0
′′
.27
×
0
′′
.29
Angular Diameter
0
′′
.937
0
′′
.924
Sub
-
obs longitude
[°W]
73 to 101
27 to 287
Sub
-
obs latitude
[°N]
2.1
2.1
North pole angle
335°
335°
14
Table S2. Molecular data for all emission lines detected in our observations.
Quantum
numbers (QN) are given for the total rotational quantum number (J), the total rotational angular
momentum (N)
, and the projections of N onto the A and C inertial axes (K
a
and K
c
respectively),
for the upper and lower state
. The lower state energy (E
L
) is given in cm
-1
. Data from CDMS (
43,
44
,
53-
61
).
Species
Frequency [
M
Hz]
E
L
[cm
-1
]
Line Strength at
300 K
[cm
-1
/(mol
ecule
×
cm
-2
)]
QN
Detected
Leading
Detected
Trailing
32
SO
2
416825
.
55
76
±
0.0019
289.053
6.87986
×
10
-
22
J=28
K
a
=4
←
5
K
c
=24
←
23
Y
Y
32
SO
2
418815
.
8002
±
0.0020
192.736
1.03748
×
10
-
22
J=
18
←
1
7
K
a
=7
←
8
K
c
=1
1
←
10
Y
Y
32
SO
2
419019
.
03
78
±
0.0019
331.436
5.58444
×
10
-
22
J=31
K
a
=3
←
4
K
c
=29
←
28
Y
Y
32
SO
2
429863
.
84
67
±
0.0019
659.183
1.91637
×
10
-
22
J=44
K
a
=3
←
4
K
c
=41
←
40
Y
Y
32
SO
2
430193
.
70
70
±
0.0015
180.632
1.05604
×
10
-
21
J=23
←
24
K
a
=2
←
1
K
c
=22
←
23
Y
Y
32
SO
2
430228
.
64
87
±
0.0016
168.493
1.66180
×
10
-
21
J=23
←
24
K
a
=1
←
0
K
c
=23
←
24
N/A
Y
32
SO
2
430232
.
31
26
±
0.0017
138.228
1.02560
×
10
-
21
J=20
←
21
K
a
=1
←
2
K
c
=19
←
20
N/A
Y
34
SO
2
419070
.
9
415
±
0.0056
222.079
5.31348
×
10
-
22
J=25
←
26
K
a
=3
←
2
K
c
=23
←
24
Y
Y
34
SO
2
428537
.
943
5
±
0.0065
167.768
1.63448
×
10
-
21
J=23
←
24
K
a
=1
←
0
K
c
=23
←
24
Y
Y
34
SO
2
429952
.
4
205
±
0.0075
188.460
8.54829
×
10
-
22
J=22
K
a
=4
←
5
K
c
=18
←
17
Y
Y
34
SO
2
431498
.
35
74
±
0.0061
179.871
1.07468
×
10
-
21
J=23
←
24
K
a
=2
←
1
K
c
=22
←
23
Y
Y
32
SO
431808
.
1
96
±
0.020
67.731
8.78778
×
10
-
21
J=9
←
10
N=10
←
11
Y
Y
34
SO
419640
.
353
±
0.014
68.080
6.61728
×
10
-
21
J=9
←
10
N=8
←
9
Y
N
15
Na
35
Cl
428518
.
551
2
±
0.0040
229.071
2.99626
×
10
-
19
J=32
←
33
Y
Y
Na
37
Cl
419381
.
126
4
±
0.0044
224.178
2.86536
×
10
-
19
J=32
←
33
Y
Y
K
35
Cl
428297
.
81
76
±
0.0027
393.948
1.58008
×
10
-
19
J=55
←
56
Y
Y
K
37
Cl
416185
.
800
3
±
0.0027
382.778
1.52012
×
10
-
19
J=55
←
56
Y
N
K
37
Cl
430886
.
5999
±
0.0028
410.788
1.47360
×
10
-
19
J=57
←
58
Y
N
Table S3. Best
-fitting model parameters for local analysis.
Same as Table 1, but for the local
regions shown in Fig S4, with 1
σ
uncertainties from MCMC simulations.
Sulfur
-
bearing molecules
Region
SO
2
column density
[cm
-2
]
T
gas
[K]
34
SO
2
/
32
SO
2
Leading
:
E
L
(1.
306
±
0.03
4
)
×
10
16
238
.8
±
2.9
0.05
79
±
0.002
0
Leading:
W
L
(
7.19
±
0.
34
)
×
10
15
23
6.8
±
5
.5
0.057
4
±
0.00
38
Leading: DC
(1.
971
±
0.0
90
)
×
10
16
198
.4
±
4
.0
0.05
26
±
0.00
28
Trailing: EL
(3
.59
±
0.
19
)
×
10
15
260
.5
±
7.9
0.06
59
±
0.005
1
Trailing: WL
(4.
22
±
0.3
1
)
×
10
15
240
.3
±
9.4
0.066
0
±
0.006
1
Trailing
: DC
(5.
61
±
0.5
5
)
×
10
15
210
.2
±
9.7
0.06
82
±
0.00
81
Chlorine
-
bearing molecules
Region
NaCl column
density [cm
-2
]
KCl column
density [cm
-2
]
T
gas
[K]
37
Cl/
35
Cl
Velocity [m s
-1
]
Leading: 1
(
4.18
±
0.2
3
)
×
10
12
(7
.79
±
0.7
7
)
×
10
11
9
34
±
59
0.4
05
±
0.02
0
19
±
1
1
Leading: 2
(4.
99
±
0.2
3
)
×
10
12
(5
.3
±
1
.4
)
×
10
11
72
7
±
49
0.3
48
±
0.02
1
5
2
±
1
1
Leading: 3
(2.
82
±
0.4
0
)
×
10
11
(4
.0
±
1
.1
)
×
10
10
13
2
0
±
18
0
0.
399
±
0.0
44
2
75
±
2
2
Trailing: 1
(4.
38
±
0.1
3
)
×
10
12
(5
.4
±
1
.1
)
×
10
11
6
39
±
28
0.4
19
±
0.0
18
-
9
0.3
±
6
.4
Trailing: 2
(
7.12
±
0.3
1
)
×
10
11
(7
.5
±
1
.3
)
×
10
10
79
6
±
4
2
0.4
17
±
0.0
16
-
150
±
1
1
Trailing: 3
(
6.21
±
0.6
2
)
×
10
11
(
3.7
±
2
.1
)
×
10
10
12
9
0
±
1
3
0
0.3
50
±
0.0
29
-
2
4
±
1
5
16
Table S4. NaCl and KCl
source locations and possible identifications with paterae
.
L
ocation
numbers correspond to Figure 1. The most likely patera is identified if there is a clear, isolated
patera at the identified latitude and longitude. All other named paterae within the uncertainties on
the latitudes and longitudes are also listed. The latitudes and longitudes of leading hemisphere
location 1 and trailing hemisphere location 3 are consistent within 1
σ
of one another so might be
the same gas source.
Location
Latitude
Longitude
Most likely p
atera
Other
paterae within
uncertainties
Leading hemisphere
1
27±
10
ºS
2
0
±
14
ºW
Kanehekili Fluctus,
Cataquil Patera, Uta Patera,
Angpetu Patera
2
66±
11
ºS
136±1
7
ºW
3
22±
10
ºN
159±1
4
ºW
Thomagata
Patera, Reshef
Patera, Surya Patera, Chaac
Patera
Trailing hemisphere
1
58±
10
ºS
218±
17
ºW
Kurdalagon Patera
Gabija Patera
2
30±
11
ºN
211±1
3
ºW
Isum Patera
Susanoo Patera
3
24±
10
ºS
355±
1
5ºW
many
4
33±
12
ºN
330±1
5
ºW
Fuchi Patera
Manua Patera
5
59±
11
ºS
341±1
5
ºW
Creidne Patera
Hiruko Patera, Inti Patera
17
Table S5.
Data sources for previous isotope measurements shown in Fig 4.
Sample
Reference
Sulfur
Solar wind
(
62
)
Bulk
Silicate Earth (BSE)
(
63
)
Earth sediments
(
64
)
Earth volcanic
gasses
(
65
)
Lunar melt inclusions (MIs)
(
66
)
Lunar mare basalts
(
67
)
Gale crater sediments
(
68
)
Ordinary chondrites (OC)
(
2
8
)
Comet Hale
-
Bopp
H
2
S
(
69
)
Comet Hale
-
Bopp CS
(
70
)
Comet
67P/Churyumov
-
Gerasimenko
(
2
9
)
Comet
C/2014 Q2
(Lovejoy)
(
71
)
Comet
C/2012 F6 (Lemmon)
(
71
)
Interstellar medium (ISM)
(
72
)
Galactic cosmic rays
(
73
)
Chlorine
Venus, HCl gas
(
74
)
Moon, basalts and soils
(
75
,
76
)
Mars, HCl gas
(
77
)
Vesta, from apatite in eucrites
(
78
)
Comet 67P/
Churyumov
-
Gerasimenko, from HCl gas
(
79
)
18
References and Notes
1. S. J. Peale, M. H.
Lee, A primordial origin of the Laplace relation among the Galilean
satellites
.
Science
298
, 593–597 (2002). doi:10.1126/science.1076557 Medline
2. K.
Batygin, A.
Morbidelli, Formation of giant planet satellites
.
Astrophys. J.
894
, 143 (2020).
doi:10.3847/1538-4357/ab8937
3. H.
Hussmann, T.
Spohn, Thermal
-orbital evolution of Io and Europa.
Icarus
171
, 391–410
(2004). doi:10.1016/j.icarus.2004.05.020
4. C. B.
Phillips
,
Voyager and Galileo SSI V
iews of Volcanic Resurfacing on Io and the Search
for Geologic Activity on Europa,
t
hesis
(The University of Arizona, 2000);
https://repository.arizona.edu/handle/10150/289119
.
5. T. V.
Johnson, A. F.
Cook II,
I. I. C.
Sagan
, L. A.
Soderblom, Volcanic resurfacing rates and
implications for volatiles on Io
.
Nature
280
, 746–750 (1979). doi:10.1038/280746a0
6. V.
Dols
, P. A.
Delamere, F.
Bagenal
, A multispecies chemistry model of Io’s local interaction
with the plasma torus
.
J. Geophys. Res.
113
, A09208 (2008). doi:10.1029/2007JA012805
7. Materials and methods are available as supplementary materials.
8. S.
Luszcz-Cook, K.
de Kleer
, kdekleer/Io_mm_RT: Code version for models published in de
Kleer et al.
,
Zenodo
(2024); https://doi.org/10.5281/zenodo.10794511
9. I.
de Pater
, S.
Luszcz-Cook, P.
Rojo, E.
Redwing, K.
Kleer
, A. Moullet, ALMA observations
of Io going into and coming out of eclipse.
Planet. Sci. J.
1
, 60–85 (2020).
doi:10.3847/PSJ/abb93d
10. E.
Lellouch
, D. F.
Strobel, M. J. S. Belton
, M. E.
Summers
, G.
Paubert
, R. Moreno,
Detection of sulfur monoxide in Io’s atmosphere.
Astrophys. J.
459
, L107–L110 (1996).
doi:10.1086/309956
11. M. E.
Summers
, D. F.
Strobel, Photochemistry and vertical transport in Io’s atmosphere.
Icarus
120
, 290–316 (1996). doi:10.1006/icar.1996.0051
12. Y.
Endo, Y.
Sekine,
Y. Ueno
, Sulfur mass
-independent fractionation during SO
2
photolysis
in low-
temperature/pressure atmospheres
.
Chem. Geol.
609
, 121064 (2022
).
doi:10.1016/j.chemgeo.2022.121064
13. A.
Moullet, M. A.
Gurwell
, E.
Lellouch
, R.
Moreno, Simultaneous mapping of SO
2
, SO,
NaCl in Io’s atmosphere with the Submillimeter Array
.
Icarus
208
, 353–365 (2010).
doi:10.1016/j.icarus.2010.02.009
14. E.
Redwing,
I. de Pater
, S.
Luszcz
-Cook, K.
de Kleer
, A. Moullet, P. M. Rojo, NaCl and KCl
in Io’s atmosphere.
Planet. Sci. J.
3
, 238 (2022).
doi:10.3847/PSJ/ac9784
15. A.
Moullet, E.
Lellouch
, R.
Moreno, M.
Gurwell
, J. H.
Black
, B.
Butler
, Exploring Io’s
atmospheric composition with APEX: First measurement of
34
SO
2
and tentative detection
of KCl
.
Astrophys. J.
776
, 32 (2013). doi:10.1088/0004-637X/776/1/32
16. T.
Ding
, S.
Valkiers
, H. Kipphardt, P.
De Bièvre, P. D. P. Taylor, R.
Gonfiantini, R.
Krouse,
Calibrated sulfur isotope abundance ratios of three IAEA sulfur isotope reference
19
materials and V
-CDT with a reassessment of the atomic weight of sulfur.
Geochim.
Cosmochim. Acta
65
, 2433–2437 (2001). doi:10.1016/S0016-7037(01)00611-1
17. F.
Bagenal
, V.
Dols
, The space environment of Io and Europa.
J. Geophys. Res. Space Phys.
125
, e27485 (2020). doi:10.1029/2019JA027485
18. E. M.
Sieveka, R. E.
Johnson, Ejection of atoms and molecules from Io by plasma-ion
impact
.
Astrophys. J.
287
, 418–426 (1984). doi:10.1086/162701
19. J. I.
Moses, M. Y.
Zolotov, B.
Fegley
, Photochemistry of a volcanically driven atmosphere
on Io: Sulfur and oxygen species from a Pele-type eruption.
Icarus
156
, 76–106 (2002).
doi:10.1006/icar.2001.6758
20. E.
Hughes, K.
de Kleer
, J.
Eiler
,
et al.
, Io’s sulfur cycle: Tracking tidal heating by modelling
the evolution of sulfur isotopes.
JGR Planets
10.1029/2023JE008086 (2024).
21. J.
Saur
, F. M.
Neubauer
, D. F.
Strobel, M. E.
Summers
, Three dimensional plasma
simulation of Io’s interaction with the Io plasma torus: Asymmetric plasma flow
.
J.
Geophys. Res.
104
, 25105–25126 (1999). doi:10.1029/1999JA900304
22. S. J. Peale, P.
Cassen
, R. T.
Reynolds, Melting of io by tidal dissipation
.
Science
203
, 892–
894 (1979). doi:10.1126/science.203.4383.892 Medline
23. T. C.
O’Reilly
, G. F.
Davies
, Magma transport on Io: A mechanism allowing a thick
lithosphere.
Geophys. Res. Lett.
8
, 313–316 (1981). doi:10.1029/GL008i004p00313
24. O. L.
Kuskov, V. A.
Kronrod, Core sizes and internal structure of Earth’s and Jupiter’s
satellites
.
Icarus
151
, 204–227 (2001). doi:10.1006/icar.2001.6611
25. G.
Dreibus, H.
Palme
, B.
Spettel
, J.
Zipfel
, H.
Wanke, Sulfur and selenium in chondritic
meteorites
.
Meteoritics
30
, 439–445 (1995). doi:10.1111/j.1945-5100.1995.tb01150.x
26. W. B.
McKinnon, “Formation and early evolution of Io” in
Io After Galileo
, R. M. C. Lopes,
J. R. Spencer
, Eds. (Springer-Praxis Books, 2007), pp. 61–88. doi:
10.1007/978-3-540-
48841-5_4
27. C. J. Bierson
, F.
Nimmo
, Explaining the galilean satellites’ density gradient by
hydrodynamic escape.
Astrophys. J. Lett.
897
, L43
(2020).
doi:10.3847/2041-
8213/aba11a
Medline
28. X.
Gao
, M. H.
Thiemens
, Variations in the isotopic composition of sulfur in enstatite and
ordinary chondrites.
Geochim. Cosmochim. Acta
57
, 3171–3176 (1993).
doi:10.1016/0016-7037(93)90301-C
29. U.
Calmonte
, K.
Altwegg
, H.
Balsiger
, J.
-J. Berthelier
, A.
Bieler
, J.
De Keyser
, B.
Fiethe
, S.
A. Fuselier
, S.
Gasc, T. I.
Gombosi, L.
Le Roy
, M.
Rubin, T.
Sémon
, C.
-Y. Tzou
, S. F.
Wampfler
, Sulphur isotope mass
-independent fractionation observed in comet
67P/Churyumov-Gerasimenko by Rosetta/ROSINA.
Mon. Not. R. Astron. Soc.
469
(Suppl_2), S787–S803 (2017). doi:10.1093/mnras/stx2534
30. J. D.
Anderson, R. A.
Jacobson, E. L.
Lau
, W. B.
Moore, G.
Schubert, Io’s gravity field and
interior structure
.
J. Geophys. Res.
106
, 32963–32969 (2001).
doi:10.1029/2000JE001367
20
31. A.
Godon, N.
Jendrzejewski
, H. G. M.
Eggenkamp, D. A.
Banks, M.
Ader
, M. L.
Coleman
,
F. Pineau
, A cross
-calibration of chlorine isotopic measurements and suitability of
seawater as the international reference material
.
Chem. Geol.
207
, 1–12 (2004).
doi:10.1016/j.chemgeo.2003.11.019
32. M.
Küppers, N. M.
Schneider
, Discovery of chlorine in the Io torus.
Geophys. Res. Lett.
27
,
513–516 (2000). doi:10.1029/1999GL010718
33. C. T.
Ewing
, K. H.
Stern
, Equilibrium vaporization rates and vapor pressures of solid and
liquid sodium chloride, potassium chloride, potassium bromide, cesium iodide, and
lithium fluoride
.
J. Phys. Chem.
78
, 1998–2005 (1974). doi:10.1021/j100613a005
34. J. I.
Moses, M. Y.
Zolotov, B.
Fegley
, Alkali and chlorine photochemistry in a volcanically
driven atmosphere on Io.
Icarus
156
, 107–135 (2002). doi:10.1006/icar.2001.6759
35. R. S. Park
, W. M.
Folkner, J. G.
Williams
, D. H.
Boggs, The JPL planetary and lunar
ephemerides DE440 and DE441.
Astron. J.
161
, 105 (2021).
doi:10.3847/1538-
3881/abd414
36. T. R.
Hunter, R.
Indebetouw, C. L.
Brogan, K.
Berry
, C.
-S. Chang, H.
Francke,
V. C.
Geers
,
L. Gómez, J. E.
Hibbard, E. M.
Humphreys, B. R.
Kent
, A. A.
Kepley
, D.
Kunneriath, A.
Lipnicky, R. A.
Loomis, B. S. Mason
, J. S. Masters
, L. T.
Maud
, D.
Muders, J.
Sabater
,
K. Sugimoto, L.
Szűcs
, E.
Vasiliev
, L.
Videla
, E.
Villard
, S. J. Williams
, R.
Xue
, I. Yoon,
The ALMA interferometric pipeline heuristics
.
Publ. Astron. Soc. Pac.
135
, 074501
(2023). doi:10.1088/1538-3873/ace216
37. T.
Cornwell, E. B.
Fomalont, “
Self
-calibration
” in
Synthesis Imaging in Radio Astronomy II
,
G. B.
Taylor, C. L.
Carilli, R. A.
Perley
, Eds., Astronomical Society of the Pacific
Conference Series
, vol. 180 (1999), pp. 187–199.
38. M. P. McMullin
, B.
Waters
, D.
Schiebel
, W.
Young, K.
Golap
, “CASA architecture and
applications
” in
Astronomical Data Analysis Software and Systems XVI
, R. A.
Shaw
, F.
Hill, D. J.
Bell
, Eds., Astronomical Society of the Pacific Conference Series
, vol. 376
(2007), pp. 127–130.
39. B.
Bean
, S.
Bhatnagar
, S.
Castro
, J. D.
Meyer
, B. Emonts, E.
Garcia, R.
Garwood
, K.
Golap
,
J. G.
Villalba
, P.
Harris
, Y. Hayashi
, J.
Hoskins,
M.
Hsieh
, P.
Jagannathan
, W.
Kawasaki
,
A. Keimpema
, M.
Kettenis
, J.
Lopez, J.
Marvil
, J.
Masters
, A.
McNichols
, D.
Mehringer,
R. Miel
, G.
Moellenbrock, F.
Montesino, T.
Nakazato
, J.
Ott
, D.
Petry
, M.
Pokorny, R.
Raba, U.
Rau
, D.
Schiebel
, N.
Schweighart, S.
Sekhar
, K.
Shimada, D.
Small
, J.
-W.
Steeb
, K.
Sugimoto, V.
Suoranta, T.
Tsutsumi,
I. M.
van Bemmel
, M.
Verkouter, A.
Wells
, W.
Xiong, A.
Szomoru
, M.
Griffith
, B.
Glendenning, J.
Kern
; The CASA Team
,
CASA, the Common Astronomy Software Applications for radio astronomy.
Publ.
Astron. Soc. Pac.
134
, 114501 (2022). doi:10.1088/1538-3873/ac9642
40. J. A.
Högbom, Aperture synthesis with a non-
regular distribution of interferometer baselines
.
Astron. Astrophys.
15
(suppl.), 417–426 (1974).
41. P.
Virtanen
, R.
Gommers
, T. E.
Oliphant, M.
Haberland
, T. Reddy, D.
Cournapeau, E.
Burovski, P.
Peterson
, W.
Weckesser
, J.
Bright, S. J. van der Walt
, M.
Brett
, J.
Wilson
,
K. J.
Millman
, N.
Mayorov, A. R. J. Nelson
, E.
Jones, R.
Kern
, E.
Larson
, C. J. Carey
,
İ.
Polat
, Y.
Feng
, E. W.
Moore
, J.
VanderPlas
, D.
Laxalde, J.
Perktold, R.
Cimrman
, I.