Claimed detection of PH 3 in the clouds of Venus is consistent with mesospheric SO 2
Andrew P. Lincowski, Victoria S. Meadows, David Crisp, Alex B. Akins, Edward W. Schwieterman, Giada N. Arney, Michael L. Wong, Paul G. Steffes, M. Niki Parenteau, Shawn Domagal-Goldman
DD RAFT VERSION J ANUARY
26, 2021Typeset using L A TEX twocolumn style in AASTeX63
Claimed detection of PH in the clouds of Venus is consistent with mesospheric SO A NDREW
P. L
INCOWSKI ,
1, 2 V ICTORIA
S. M
EADOWS ,
1, 2 D AVID C RISP ,
2, 3 A LEX
B. A
KINS , E DWARD
W. S
CHWIETERMAN ,
2, 5, 6 G IADA
N. A
RNEY ,
2, 7 M ICHAEL
L. W
ONG ,
1, 2 P AUL
G. S
TEFFES , M. N
IKI P ARENTEAU ,
2, 9
AND S HAWN D OMAGAL -G OLDMAN
2, 7 Department of Astronomy and Astrobiology Program, University of Washington, Box 351580, Seattle, Washington 98195, USA NASA Nexus for Exoplanet System Science, Virtual Planetary Laboratory Team, Box 351580, University of Washington, Seattle, Washington 98195, USA Jet Propulsion Laboratory, California Institute of Technology, Earth and Space Sciences Division, Pasadena, California 91011, USA Jet Propulsion Laboratory, California Institute of Technology, Instruments Division, Pasadena, California 91011, USA Department of Earth and Planetary Sciences, University of California, Riverside, CA 92521 USA Blue Marble Space Institute of Science, Seattle, WA, USA NASA/Goddard Space Flight Center, Greenbelt, MD 20771, USA School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0250, USA MS 239-4, Space Science Division, NASA Ames Research Center, Moffett Field, CA, USA (Received October 21, 2020; Revised January 18, 2021; Accepted January 21, 2021)
Submitted to ApJLABSTRACTThe observation of a 266.94 GHz feature in the Venus spectrum has been attributed to PH in the Venusclouds, suggesting unexpected geological, chemical or even biological processes. Since both PH and SO arespectrally active near 266.94 GHz, the contribution to this line from SO must be determined before it can beattributed, in whole or part, to PH . An undetected SO reference line, interpreted as an unexpectedly low SO abundance, suggested that the 266.94 GHz feature could be attributed primarily to PH . However, the low SO and the inference that PH was in the cloud deck posed an apparent contradiction. Here we use a radiativetransfer model to analyze the PH discovery, and explore the detectability of different vertical distributions ofPH and SO . We find that the 266.94 GHz line does not originate in the clouds, but above 80 km in the Venusmesosphere. This level of line formation is inconsistent with chemical modeling that assumes generation ofPH in the Venus clouds. Given the extremely short chemical lifetime of PH in the Venus mesosphere, animplausibly high source flux would be needed to maintain the observed value of 20 ±
10 ppb. We find thattypical Venus SO vertical distributions and abundances fit the JCMT 266.94 GHz feature, and the resultingSO reference line at 267.54 GHz would have remained undetectable in the ALMA data due to line dilution.We conclude that nominal mesospheric SO is a more plausible explanation for the JCMT and ALMA data thanPH . INTRODUCTIONGreaves et al. (2020a) recently attributed a 266.94 GHz(1.123 mm) line observed in the Venus spectrum to ∼
20 ppbof phosphine (PH ) absorbing above 56 km altitude, in theupper clouds. In the strongly-oxidizing Venus atmosphere,PH formation is disfavored and its destruction is enhanced,leading Greaves et al. (2020a) to argue that its presence inthe clouds points to unknown geological, chemical or evenbiological processes. The discovery team identified no vi-able abiotic production mechanism for PH in the Venus at- Corresponding author: Victoria S. [email protected] mosphere (Greaves et al. 2020a; Bains et al. 2020), and soa biological origin was considered. PH has been proposedas a potential biosignature in terrestrial planet atmospheres(Sousa-Silva et al. 2020) due to its association with decayingorganic matter (Glindemann et al. 2005), and significant—presumed biological—fluxes from marine environments onEarth (Zhu et al. 2007). However, the specific mode of bio-logical production of PH remains uncertain and is still vig-orously debated (Roels & Verstraete 2001), with no knowndirect metabolic pathway (Roels et al. 2005).The identification of PH in the Venus clouds was madeusing multiple observations of a single spectral feature at266.94 GHz, where both PH (266.944 GHz) and SO (266.943 GHz) have absorption lines (Greaves et al. 2020a).After the initial detection using coadded spectra from the a r X i v : . [ a s t r o - ph . E P ] J a n James Clark Maxwell Telescope (JCMT), which were takenover 5 nights between 2017 June 9–16, follow-up observa-tions were made with the Atacama Large Millimeter Array(ALMA) on 2019 March 5. The latter dataset included si-multaneous narrow-band (0.1171875 GHz) and wide-band(1.875 GHz) observations, centered on the Venus rest-framePH frequency. The 266.94 GHz line, seen in the JCMT dataat a S/N of 4.3 (Greaves et al. 2020a; although this detec-tion significance has been subsequently called into question,Thompson 2020), was also detected in the ALMA narrow-band and wideband datasets at higher significance than in theJCMT data (Greaves et al. 2020a), although a subsequent re-analysis of the ALMA data also suggests a less significant de-tection, with a correspondingly lower inferred abundance ofPH (Greaves et al. 2020b). Assuming a uniform mixing ra-tio for the PH , Greaves et al. (2020a) derive an abundance of20 ppb from the JCMT observations, and calculate an emis-sion weighting function peaked at 56 km. They thereforeconclude that the PH absorption feature was sourced pri-marily from within the Venus clouds. However, as Greaveset al. (2020a) point out, with a FWHM of 4–5 km s − , thisline could potentially contain contributions from both PH and SO , as the SO line center is only +1.3 km s − from thePH line center.Consequently, the PH line identification is strongly de-pendent on accurately estimating and excluding a potentially-significant contribution from SO , which, after the bulk at-mospheric gases CO and N , is the third most abundantgas in the Venus atmosphere. Greaves et al. (2020a) at-tempted to quantify the SO contribution to the observed266.94 GHz feature by searching the ALMA wide-band ob-servations for the nearby, stronger SO J K a , K c = , ← , ,based on potentially large spectral “ripples”, artifacts in thedata induced by interferometric response to Venus as a bright,extended source. Greaves et al. (2020a) also noted that the ≤
10 ppb value was comparable to a 346.652 GHz ALMAVenus SO measurement of 16 . ± . at 85 km altitude(Piccialli et al. 2017) in the Venus mesosphere (which ex-tends from 65–120 km), and not to the middle/upper clouddeck (53–61 km). The ≤
10 ppb constraint derived from thenon-detection implied a maximum 10% contribution fromSO to the 266.94 GHz absorption band depth, and a shiftin the observed line centroid of no more than 0.1 km s − .Greaves et al. (2020a) concluded that SO had been ruled outas a significant contaminant for the putative PH line. Con-versely, they argued that the 266.94 GHz line could not be ex-plained solely by SO , because the corresponding referencelines would be significantly stronger than the − . by Greaves et al.(2020a) supports a corresponding low inferred abundance,and a low contamination fraction for the 266.94 GHz line,it is the key piece of evidence supporting the PH line iden-tification at 266.94 GHz— and so it warrants closer scrutiny.There is an apparent contradiction between the inferred alti-tudes that the PH feature probed, and the SO abundanceconstraint. If the putative PH (266.94 GHz) absorptionis sensitive to altitudes near 56 km, and thus probes theVenus middle and upper cloud, then the 267.94 GHz SO reference line should also originate from this altitude range,since it has similar line strength and amount of underlyingcontinuum absorption. Data and modeling estimates placethe SO abundance near 1–5 ppm at 60 km in the uppercloud, which should increase with depth to match the higher ∼
130 ppm measured below the cloud deck (Zasova et al.1993; Krasnopolsky 2012; Zhang et al. 2012; Belyaev et al.2012; Marcq et al. 2008; Arney et al. 2014; Encrenaz et al.2019). Previous measurements therefore suggest that the in-ferred disk-averaged <
10 ppb of SO is anomalously low,especially if the observations probe within the clouds. As-suming similar spatial distribution of the two gases, for aninferred SO abundance at 56 km of 10ppm, and the 10 ppbPH abundance of Greaves et al. (2020a), the SO contribu-tion to the observed line would exceed that from PH by twoorders of magnitude (Krasnopolsky 2020).If the observations were instead sensitive to the meso-spheric levels above the clouds, as is the case for higher fre-quency ALMA observations (Sandor et al. 2010; Encrenazet al. 2015), then the inferred Venus SO abundance wouldbe closer to, but still lower than previously measured levels(Sandor et al. 2010; Encrenaz et al. 2015; Piccialli et al. 2017;Vandaele et al. 2017). While the abundance of SO abovethe clouds is known to vary significantly over time (Espositoet al. 1988; Encrenaz et al. 2012, 2019) with a minimum ob-served around 10-100 ppb at ∼
80 km, the abundances in themesosphere have been measured to be in the range 10 ppbto 10 ppm (Krasnopolsky 2010; Belyaev et al. 2012; Van-daele et al. 2017). A planet-wide decrease from a highercloud-top SO abundance in 2006 to a low in 2014 of 30ppb was also observed, but more recent observations from2016 through September 2018, which span the Greaves et al.(2020a) JCMT observation, show a strong increase to typicalcloud-top values of several hundred ppb of SO (Encrenazet al. 2019).While line absorption occurring predominantly within themesosphere would make the non-detection and inferred lowabundance of SO more plausible, it would also suggest thatthe line attributed to PH was formed at mesospheric levels.Consequently, the 266.94 GHz line would not be sensitive to, − − − − − − Mixing Ratio [mol/mol]10 P r e ss u r e [ P a ] A,B,CB,C AA,BC
VIRA 45 ° lat. Temp.H O 0.2 ppm meso.SO
10 ppb constantSO
10 ppb + VIRAPH
20 ppb constantPH photochemical A l t i t ud e [ k m ]
200 400 600Temperature [K] a − − − − − − Mixing Ratio [mol/mol]10 P r e ss u r e [ P a ] Case D
VIRA-2 Temp.H OOCSSO Belyaev+2012Encr+2019 A l t i t ud e [ k m ]
200 400 600Temperature [K] b − − − − Extinction coefficient at λ = 0.6 µ m ( d τ/ dz )10 P r e ss u r e [ P a ] Particle Mode m1m2m2‘m3 A l t i t ud e [ k m ] c Figure 1.
Atmospheric structures for Venus used in our spectral modeling cases.
Panel a : Temperature and vertical profiles for Cases A–C,which use parameters assumed/derived by Greaves et al. (2020a). Temperature profile (black line) is for VIRA 45 ◦ latitude (Seiff et al. 1985),and vertical profiles are shown for PH (solid, Cases A, B; dashed Case C), SO (green), and H O (blue).
Panel b : Temperature and verticalprofiles for our Case D best fit to the JCMT 266.94 GHz line. The temperature profile (black line) is from VIRA-2 (Moroz & Zasova 1997).The nominal gas mixing ratios for H O (blue line) is based on VIRA values (von Zahn & Moroz 1985) updated for the lower atmosphere(De Bergh et al. 2006) but have also been modified slightly as described in § 2. For OCS (red line), the profile is constructed based on recentmeasurements by Krasnopolsky (2010) and Arney et al. (2014), and by the surface abundance in the lower atmosphere model by Krasnopolsky(2013). For SO (green line), we fit the 266.94 GHz line guided by the vertical profile of Belyaev et al. (2012) in the mesosphere and uppercloud, and consistent with a suite of SOIR and SPICAV UV SO measurements taken from 2007–2008 (green shaded region). This profilepasses through the cloud-top SO measurement (200–350 ppb) obtained by Encrenaz et al. (2019) in July 2017 (green data point), one monthafter the Greaves et al. (2020a) JCMT observations. We use 130 ppm in the lower atmosphere (Marcq et al. 2008) and generated a profilebetween the lower atmosphere and cloud tops. Panel c : optical depth extinction profiles (optical depth per meter at a wavelength of 0.6 µ m)for the Venus cloud particle modes: m1 (haze), m2, m2’, and m3 (Crisp 1986). The clouds are defined via optical depth considerations to spanapproximately 48–70 km (3 × –1 . × Pa). and so not able to confirm, the presence of PH in the Venusclouds—potentially weakening support for a biological ori-gin. The presence of 20 ppb of mesospheric PH would re-quire an extremely large source flux due to photolysis andreactions with radical species, including Cl and H, that resultin a sub-second lifetime for PH in the Venus mesosphere(Bains et al. 2020, their Fig. 2). Indeed, the vertical distri-bution predicted using photochemical-kinetics studies witha cloud source of PH indicates a sharply reduced meso-spheric abundance of PH ( < .
001 ppb) alongside signifi-cant ( >
100 ppb below 95 km) SO (Greaves et al. 2020a,extended data figure 9; Bains et al. 2020).To explore the potential contradictions posed by theGreaves et al. (2020a) PH observations, and to verify thesource region for the 266.94 GHz absorption, here we usea radiative transfer model of the Venus atmosphere to simu-late the impact on the Venus millimeter-wavelength spectrumof different abundances and vertical distributions of PH andSO , including those proposed by Greaves et al. (2020a) andBains et al. (2020). METHODSTo generate synthetic millimeter-wavelength spectra ofVenus, we use SMART (Spectral Mapping AtmosphericRadiative Transfer), a 1D line-by-line, multi-stream, fully multiple-scattering radiative transfer model (Meadows &Crisp 1996; Crisp 1997). SMART has been validated againstobservations of Solar System planets, with heritage modelingthe Venus atmosphere (Meadows & Crisp 1996; Arney et al.2014; Robinson & Crisp 2018).Our spectral simulations consist of Cases A–C, for whichwe generate spectra based on the mixing ratios and verticalprofiles used and derived by Greaves et al. (2020a), and ourbest fit model, Case D, which does not contain PH and usesconstraints from additional Venus observations (Figure 1).Cases A–C include CO , SO , H O, and PH and use theVIRA 45 ◦ latitude temperature profile (Seiff et al. 1985). Tomatch the H O estimate of Greaves et al. (2020a), we use theDe Bergh et al. (2006) H O profile but reduced to 0.2 ppmabove 68 km. For SO , we use the De Bergh et al. (2006)compilation below 100 km for cases B and C, but reducedto 10 ppb above 70 km, and for case A we maintain 10 ppbdown through the cloud deck to 53 km. For PH , we use auniformly mixed 20 ppb profile for cases A and B, and thephotochemical profile from Greaves et al. (2020a) (their fig-ure ED7) for case C.For our best-fit scenario, Case D, we do not include PH and use the De Bergh et al. (2006) update to the VIRA below100 km and more recent observations where available. Weuse the VIRA-2 temperature profile (Moroz & Zasova 1997).For H O, we use 30 ppm below the cloud deck (De Berghet al. 2006, and references therein), and we assume 3 ppmabove the cloud deck (Krasnopolsky et al. 2013; Cottini et al.2015; Piccialli et al. 2017). For SO , we use 130 ppm be-low the cloud deck (Gelman et al. 1979; Bezard et al. 1993;De Bergh et al. 2006; Marcq et al. 2008; Arney et al. 2014),decreasing with increasing altitude to the July 2017 obser-vation of ∼
275 ppb at 64 km (Encrenaz et al. 2019), whichwas measured within a month of the Greaves et al. (2020a)JCMT data. In the mesosphere, we fit the SO profile to theobserved feature at 266.94 GHz guided by the vertical pro-file fit to 2007–2008 data from Belyaev et al. (2012), whichis consistent with the cloud-top SO abundance observed inJuly, 2017 (see Encrenaz et al. 2019). Long-term monitoringhas shown that 2007–2008 and 2017–2018 were similar max-imum periods of global mesospheric SO abundance (En-crenaz et al. 2019), although short-term temporal variabil-ity within these secular changes can be orders of magnitude(Belyaev et al. 2017). We prescribe the OCS profile guidedby recent measurements (Krasnopolsky 2010; Arney et al.2014) and models (Zhang et al. 2012; Krasnopolsky 2012,2013; Lincowski et al. 2018). We adopt the same aerosolproperties, modes, and optical depth profiles as Arney et al.(2014), which originate from Crisp (1986). Temperature andgas profiles, and aerosol optical depths, are shown in Fig-ure 1.Absorption cross-sections associated with vibrational-rotational transitions are calculated using a line-by-linemodel, LBLABC (see Meadows & Crisp 1996; Crisp 1997,for details), with the HITRAN2016 line database (Gordonet al. 2017) for all gases except CO , which is calculated fromthe extensive Ames line database (Huang et al. 2017). Be-cause these line lists assume terrestrial isotopic abundance,we use the methods described in Lincowski et al. (2019) toadjust the line list isotopologue abundances for H O to 200times the D/H abundance compared to Earth, the standardvalue used in the literature for the Venus mesosphere (Encre-naz et al. 2015). Collision-induced absorption data is usedfor CO -CO (Gruszka & Borysow 1997).Data on the foreign broadening of gases by CO is notwell-characterized, compared to broadening by air, but ismore appropriate for Venus simulations. To reproduce theresults of Greaves et al. (2020a), we use their foreign broad-ening parameter for PH of 0.186 cm − atm − , which theyused to estimate PH as 20 ±
10 ppb in the JCMT data. Be-cause their broadening treatment for gases other than PH is not specified, we use the default HITRAN air broaden-ing for cases A–C. To fit the 266.94 GHz detection fea-ture with SO in case D, we employ data for broadeningby CO , as available. For SO and OCS, we use data forbroadening by CO available in HITRAN (Wilzewski et al.2016; Gordon et al. 2017). Although the SO broadening data are derived from a single line experiment (Chandra &Chandra 1963), the parameters in the frequencies of inter-est are consistent with recent laboratory results by Bellotti& Steffes (2015). The broadening values for our SO linesof interest are approximately 1.8–2.0 × air broadening (i.e. γ CO (cid:39) . − .
19 cm − atm − ). For HDO, we multiply theHITRAN air foreign broadening parameters by 2.4, which isconsistent with this frequency range (Sagawa et al. 2009).To better visualize individual line signal and compare tothe published data, we processed our flux spectra to nor-malize the continuum. Because we are processing noiselessmodel results, we mask spectral intervals for individual linesand linearly interpolate the continuum across the interval.The line:continuum (l:c) spectra were determined by divid-ing the original model spectrum by the continuum and sub-tracting one.As an additional validation of our radiative transfer modeland fit to the Greaves et al. (2020a) JCMT data, we appliedour model to simulate the line shape and peak intensity of the346.65 GHz late-2011 observation of Encrenaz et al. (2015),using their SO profile of 10 ppb from 86–100 km, and ob-tained an excellent fit to the data (see Fig. 2). − L i n e / C o n t i nuu m –1 Our ModelEncrenaz+2015 .
64 346 .
65 346 .
66 346 . Frequency [GHz]10 − − SO Figure 2.
We demonstrate the validity of our model by fitting the346.652 GHz SO line observed by Encrenaz et al. (2015, Fig. 19)in 2011 using their best-fit profile of no SO from 70–85 km and10 ppb above 85 km, with all other modeling parameters specifiedas for our Case D (but also including CO from the De Bergh et al.(2006) compilation). SO absorption line strength in the bottompanel is given in units of cm − /(molecule cm − ). This compari-son shows our model and associated parameters are consistent withprevious sub-mm observations of SO .3. RESULTSTo explore the spectral impacts of different abundancesand vertical profiles for PH and SO , we simulated spec-tra of Venus from 266 to 268 GHz. This spectral range in-cludes the HDO, PH and SO line positions discussed inGreaves et al. (2020a), as well as OCS, which includes a − L i n e / C o n t i nuu m –1 Our ModelGreaves et al. [2020] JCMT − −
20 0 20 40Velocity [km/s]10 − − SO PH − − HDOSO OCSPH B r i g h t n e ss T e m p e r a t u r e [ K ] − − − − L i n e / C o n t i nuu m –1
10 ppb Cloud & Mesospheric SO + 20 ppb PH + 0.2 ppm H O A − L i n e / C o n t i nuu m –1 Our ModelGreaves et al. [2020] JCMT − −
20 0 20 40Velocity [km/s]10 − − SO PH − − HDOSO OCSPH B r i g h t n e ss T e m p e r a t u r e [ K ] − − − L i n e / C o n t i nuu m –1 VIRA Cloud + 10 ppb Mesospheric SO + 20 ppb PH + 0.2 ppm H O B − L i n e / C o n t i nuu m –1 Our ModelGreaves et al. [2020] JCMT − −
20 0 20 40Velocity [km/s]10 − − SO PH − − HDOSO OCSPH B r i g h t n e ss T e m p e r a t u r e [ K ] − − − L i n e / C o n t i nuu m –1 VIRA Cloud + 10 ppb Mesospheric SO + photochemical PH + 0.2 ppm H O C velocity [km/s] − L i n e / C o n t i nuu m –1 Our Model1 σ ModelsGreaves et al. [2020] JCMT − −
20 0 20 40Velocity [km/s]10 − − SO − − HDO SO OCS B r i g h t n e ss T e m p e r a t u r e [ K ] − − − − L i n e / C o n t i nuu m –1 Best fit SO + 3 ppm H O D Figure 3.
Venus spectral simulations for different PH and SO abundances and vertical profiles, including brightness temperature spectra (grey lines) to showthe continuum source, and absorption line strengths (lower panels in each case) in units of cm − /(molecule cm − ). For each case, the left hand panel shows thecorresponding fit to the 266.94 GHz line, the right panel shows the 266–268 GHz spectrum, including the SO reference line at 267.54 GHz. Case A : modifiedVIRA temperature and gas profiles with uniformly mixed 20 ppb PH and 10 ppb SO down through the cloud deck (c.f. Figure 1 Panel a). Case B : Case Abut with the VIRA SO profile in the cloud deck up to 70 km instead of evenly mixed at 10 ppb. Case C : VIRA and SO profile as in Case B, but using thephotochemically self-consistent profile for PH from Greaves et al. (2020a) (ED Fig. 9). Case D : VIRA-2 temperature profile, no PH , and using a vertically-resolved SO profile derived from a suite of spacecraft and ground-based measurements, with a mesospheric profile that increases from 30 ppb at 78 km to400 ±
150 ppb at 100 km (see Figure 1, panel b). Cases A and B demonstrate similar fits for PH to the the 266.94 GHz line as in Greaves et al. (2020a), and showlack of sensitivity to the vertical distribution of SO in the clouds. Case C demonstrates that the PH profile generated assuming a source in the Venus cloudsis inconsistent with the observed 266.94 GHz line. Case D shows that we can fit the detection feature with no PH but with a typical Venus SO abundance,although this produces SO reference line features that are over 10 times stronger than the other cases. transition at 267.530 GHz. We simulated spectra for caseswith the abundances determined by Greaves et al. (2020a)and vertical profiles determined by previous measurementsof the Venus atmosphere (Figure 1). Line-to-continuum (l:c)spectra generated at 0.0001 cm − (3 MHz) resolution areshown in Figure 3, along with the emission brightness tem-perature in grey. The brightness temperatures demonstratethe effective altitude of continuum emission, and are directlycorrelated with SO abundance in the cloud deck between54–57 km, depending on the case. Lower cloud SO abun-dance (10 ppb evenly-mixed) yields higher continuum emis-sion from deeper in the atmosphere.3.1. Simulated Spectra
For our Case A spectral simulation (Figure 3A) we as-sumed updated VIRA-derived profile (our Figure 1, see vonZahn & Moroz 1985; De Bergh et al. 2006) for all con-stituents except SO and PH . Following Greaves et al.(2020a), we assumed an evenly mixed abundance of 20 ppbPH and 10 ppb SO above 52 km altitude (near the base ofthe Venus cloud deck; green and purple dotted lines in Fig-ure 1a). We also assumed their foreign broadening parameterfor PH of 0.186 cm − atm − . Our model produces a com-parable fit to Greaves et al. (2020a) for the 266.94 GHz line(c.f. their Figure 1). Additionally, with the evenly-mixed10 ppb of SO , we also confirm that the 267.54 GHz SO line is below the spectral-ripple-inferred maximum limit onthe l:c ratio ( − . down through the cloud deck, we usedthe VIRA-derived profile such that the SO abundance in-creased with cloud depth (green dashed line in Figure 1b). Atthe 56 km level, the SO abundance is now closer to 20 ppm.The increased SO opacity raises the emission layer to coolerlevels of the atmosphere, as shown in the the brightness tem-perature difference between Cases A and B. This producesa small change in the SO continuum, which results in onlymarginal differences in the intensities of the 266.94 GHz PH line and the 267.54 GHz SO line, and the latter is still con-sistent with the maximum limit in sensitivity due to spectralripple. Thus the observed line intensities are largely insensi-tive to SO abundance within the clouds.In our Case C simulation (Figure 3C), we again used10 ppb SO in the mesosphere, increasing through the clouddeck (green dashed line in Figure 1a). However, instead ofPH evenly mixed throughout the atmosphere (as in Cases Aand B), we used the photochemical profile for PH used tointerpret the 266.94 GHz detection, as provided in Greaveset al. (2020a, their ED Figure 9), and Greaves et al. (2020a,reproduced as the purple dashed line in our Figure 1a). Thisdistribution is derived from the assumption that PH produc-tion is concentrated within the cloud deck with abundance dropping rapidly in the upper cloud deck and mesosphere,and more slowly towards the surface. The small absorp-tion line present here at 266.94 GHz is due to SO —no PH absorption is visible in this spectral simulation. This indi-cates that the line core observation is not sensitive to PH inthe cloud, and demonstrates that the assumed profile in theGreaves et al. (2020a) and Bains et al. (2020) photochemicalsimulations are inconsistent with the JCMT observations.In our Case D simulation (Figure 3D), we removed PH from our atmosphere and fit the JCMT detection feature at266.94 GHz using SO alone. As described in §2, we usedparameters for HDO, SO , and OCS foreign broadening byCO . We guided the mesospheric data fit for SO usingVenus Express UV/IR occultation data from Belyaev et al.(2012). This profile is consistent with cloud-top SO abun-dances measured by Encrenaz et al. (2019) within a monthof the Greaves et al. (2020a) JCMT observations. Our best-fit SO profiles, fitting the observed line (black) and ± σ about the line (grey) are shown in Figure 1b (green curves),with SO increasing from 30 ppb at 78 km to 400 ±
150 ppb at100 km. These abundance profiles are well within the rangeof measurements compiled in Belyaev et al. (2012) and Van-daele et al. (2017). This simulation provides an excellent fitto the JCMT detection line without PH , and predicts a pairof SO reference lines that have l:c ratios a factor of ∼ Spectral Line Sensitivity
To confirm the altitudinal sensitivity of the 266.94 GHzline for key PH and SO vertical profiles, we calculated ra-diance Jacobians, i.e. the increase in top-of-atmosphere ra-diance as a function of perturbations to the abundances forSO and PH at each layer of our model atmosphere (Fig-ure 4). The outgoing radiance will be most sensitive to re-gions of the atmosphere that contribute most to the spectralfeature. The Jacobians show that the observed line cores forboth gases originate from atmospheric pressures only as deepas ∼
400 Pa, corresponding to altitudes of ≥
80 km, in themesosphere. This absorption feature cannot be generated atlevels within the cloud deck, where the background contin-uum emission originates. It must be generated well abovethis layer, where the absorbing gas is cooler and thereforeabsorbs more efficiently than it emits. The narrow width ofthe absorption line also suggests that it was formed at pres-sures substantially less than those of the cloud top (70 km, ∼ ALMA Line Dilution
While the non-detection of prominent SO spectral fea-tures in the ALMA wideband data could indicate a low abun-dance, as argued by Greaves et al. (2020a), the estimationof this abundance was done without correcting for line dilu-tion as a result of the ALMA observing geometry (Greaves P r e ss u r e [ P a ] − − − − − − − − − SO Radiance Jacobian [d I ν /dSO ] ∆ =50 -0.00010.0000 L i n e / C o n t i nuu m –1 − − SO − SO [ppm] 30405060708090100 A l t i t ud e [ k m ]
250 500Temperature [K] P r e ss u r e [ P a ] − − − − − − − − − PH Radiance Jacobian [d I ν /dPH ] ∆ =15 − L i n e / C o n t i nuu m –1 − − SO PH − − PH [ppb] 30405060708090100 A l t i t ud e [ k m ]
250 500Temperature [K]
Figure 4.
Radiance mixing ratio Jacobians ( dI ν / dr mix ) as a function of layer pressure for the radiance streams at 21 degrees zenith an-gle, for the 266.94 GHz feature for SO (left) or PH (right). Absorption line strengths (lower panels in each case) are given in units ofcm − /(molecule cm − ). The continuum originates from SO at ∼ × Pa ( ∼
54 km, within the cloud deck), while the line cores for eitherspecies do not originate in the clouds (48-70 km) but at over 400 Pa (over 80 km) in the mesosphere. In the right panels in both plots, thetemperature structure is given as a black line, while the colored lines denote SO (green) or PH (purple) mixing ratios. On the right, theevenly-mixed 20 ppb PH profile is shown with a solid line and the photochemical PH profile is shown with a dashed line. − L i n e / C o n t i nuu m –1 − − SO − − L i n e / C o n t i nuu m –1 Our ModelALMA imaged long baselines − − SO OCS − − Figure 5.
Modeled line dilution for the ALMA observations: disk-averaged line/continuum ratios for our nominal Case D atmosphere modelcontaining SO uniformly distributed over the Venus disk, at 0.00003 cm − (1 MHz) resolution (black lines), for the detection frequency (leftpanel) and reference lines frequencies (right panel). Absorption line strengths (lower panels) are given in units of cm − /(molecule cm − ). Theorange lines show the same spectral model as imaged using the ALMA antenna configuration of the Greaves et al. (2020a) observations. Theinset shows the suppression of the 267.54 GHz reference line to an l:c of close to − . absorption lines in the wideband data. et al. 2020a). Significant line dilution is likely, especiallyconsidering the global distribution of SO in the Venus atmo-sphere, and the exclusion of the short baseline ALMA mea-surements. Greaves et al. (2020a) estimated line dilution (fil-tering losses) of 60–92% depending on position on the disk.To determine the disk-averaged line dilution for the SO ref-erence line search, we simulated observations of Venus usingthe ALMA configuration of Greaves et al. (2020a) by im-posing an appropriate resolution spectrum (0.00003 cm − , 1 MHz) of our Case D atmospheric model over a limb-darkened disk model. The Fourier Transform of this modelwas re-sampled to match the ALMA configuration and re-imaged using the imaging routines of Greaves et al. (2020a),as provided in their Supplementary Software 3. As shownin Figure 5, line dilutions on the order of 95% at the linecore are observed for the full disk. We observe similar dilu-tions when the spectrum is only imposed on one hemisphere( ∼ reference features producedby our best fit SO distribution (Case D) would be heavilysuppressed by line dilution in the ALMA data, which wouldcause them to mimic smaller features below the ripple detec-tion limit of − . DISCUSSIONThe claim that PH has been detected in the Venus cloudsis currently supported by observations of a single absorp-tion line at a frequency that also coincides with absorptionfrom SO , a known and relatively common Venus gas, andbased on an emission weighting function that peaks at 56 km(Greaves et al. 2020a). However, our radiative transfer anal-ysis indicates that the line at 266.94 GHz does not measureabsorption within the Venus clouds. Our explicit calcula-tion of radiance Jacobians confirms the assessment that both266.94 GHz PH and SO line core absorption would beproduced well above the Venus cloud deck at altitudes ex-ceeding 80 km. Arguments for a mesospheric origin for the266.94 GHz line core, based on the observed narrow widthof the line, are also provided in a recent commentary by Vil-lanueva et al. (2020). This mesospheric contribution is in-consistent with a vertical abundance profile that concentratesPH in the middle and upper clouds, as used by Greaves et al.(2020a) and Bains et al. (2020) to interpret their discovery.Our spectral simulation using this photochemical PH profilealso shows that it is not consistent with the strength of the ob-served 266.94 GHz line. However, the presence of PH in theVenus clouds is not conclusively ruled out either, a point alsomade by Greaves et al. (2020b), because the Greaves et al.(2020a) observations are not sensitive to absorption at clouddeck altitudes, and so can neither exclude, nor confirm, thepresence of PH in the Venus clouds.Given that we have shown that the observed 266.94 GHzline predominantly originates high in the mesosphere, at-tributing it to PH is less chemically plausible than SO .At these higher altitudes ( >
80 km) PH would be destroyedrapidly, while SO is photochemically regenerated (Sandoret al. 2010; Belyaev et al. 2012; Zhang et al. 2012). Be-tween 82 km and 96 km (70–300 Pa, where the line coreabsorption originates, Fig. 4) PH has a sub-second lifetime,due to the destruction by Cl and H radicals and UV photol-ysis (Bains et al. 2020). To balance this rapid destructionrate and maintain a mesospheric concentration of 20 ppb,an extremely large flux of PH is required, potentially aslarge as 3 . × molecules cm − s − . For comparison,this production rate is about ∼
100 times the flux of O pro-duced by Earth’s global photosynthetic biosphere (Field et al.1998), the dominant metabolism on our planet. Greaves et al.(2020a), assuming the 266.94 GHz absorption was from PH in the clouds, calculated a significantly smaller productionrate of 10 molecules cm − s − , due to the lower destruction rate within the clouds. However, the assumption of this in-cloud production rate results in a PH mixing ratio that effec-tively falls to zero at >
80 km altitude (Greaves et al. 2020a,Fig. 5b), which is inconsistent with our analysis that the ob-served line is sourced in the mesosphere. Although a recentreanalysis of the ALMA data by Greaves et al. (2020b) hasgreatly reduced the significance of the 266.94 GHz line de-tection, their assignment of 1 ppb of PH in the mesospherewould still require a production rate significantly higher thanthe Earth’s photosynthetic biosphere, and the larger 20 ppbPH value inferred from the JCMT data still stands.These challenges to mesospheric production rate are notrelevant if the observed 266.94 GHz line is instead attributedto SO , which is known to increase in abundance with al-titude in the mesosphere (Belyaev et al. 2012; Mills et al.2018). A combination of infrared observations that probe theupper cloud and lower mesosphere, and UV occultation mea-surements that probe the upper mesosphere, has been usedto map the vertical distribution of mesospheric SO (Belyaevet al. 2012, 2017). This distribution drops from the cloud topsto a minimum just below 80 km, but increases substantiallyfrom 80–100 km to typically several hundred ppb (Belyaevet al. 2012; Vandaele et al. 2017).Assuming that the Venus atmosphere does not containPH , we find that a realistic vertical profile for SO fitsthe JCMT 266.94 GHz detection. Because the JCMT ob-servations were single dish, any SO contribution to the266.94 GHz line would not have been suppressed, as was thecase for the ALMA data, and so should be sensitive to the truemesospheric SO abundance. We used a mesospheric SO profile that is based on the profile observed in 2007–2008 byBelyaev et al. (2012), which is likely a good fit to similarhigher values seen in 2016–2018, a time span that includesthe Greaves et al. (2020a) JCMT observation. This profileis also consistent with cloud top values of 200–350 ppb ob-served in the mid-infrared within a month of the Greaveset al. (2020a) JCMT observations (Encrenaz et al. 2019). TheEncrenaz et al. (2019) observations support the validity ofour SO vertical profile, and suggest that the Venus meso-sphere was unlikely to be experiencing a period of anoma-lously low SO abundance at the time of the JCMT obser-vations. Using this vertical abundance profile and a CO -broadened SO line profile, we can fit the width and shapeof the 266.94 GHz line using SO alone, without needing anadditional PH component. The SO is also a better fit to theline centroid than the PH (cf. Fig. 3 A/B, D). This excellentfit counters the argument of Greaves et al. (2020b) that SO alone would be too narrow to fit the observed line. Greaveset al. (2020b) also recently argued that the SO abundancerequired to fit the JCMT 266.94 GHz line (evenly-mixed150 ppb for their fit, and 100 ppb for Villanueva et al. 2020) isunrealistically large, given previous mm-wave observations,which have returned lower values for mesospheric SO (San-dor et al. 2010; Encrenaz et al. 2015). However, mm-waveobservations do not have as long, or as well sampled, a base-line as dedicated Venus spacecraft observations of the meso-sphere (Belyaev et al. 2012, 2017; Vandaele et al. 2017), andmesospheric SO abundance has been observed to vary by anorder of magnitude on daily to yearly timescales, with valuesat 90–95 km altitude between 10 to 300 ppb. There is alsoevidence for longer-term secular changes in mesospheric andcloud-top SO abundances, with maxima in 2007–2008 and2016–2018, and a minimum in 2012–2014 (Belyaev et al.2017; Encrenaz et al. 2019). We note that the model thatwe used to fit the JCMT 266.94 GHz line assuming a higherabundance of SO also produced an accurate fit to the lowerabundance observation of Encrenaz et al. (2015) (see our Fig.2), which was observed near an SO minimum.We also find that strong ALMA line dilution allows the ver-tical abundance profile of SO that fits the JCMT 266.94 GHzobservations to still be consistent with the non-detection ofthe SO ALMA reference lines—which are likely poor in-dicators of the impact of SO on the JCMT observations.Spectral simulations using our Case D SO vertical distri-bution predict SO lines at 267.54 and 267.72 GHz with l:cratios that are close to a factor of 10 larger than the nomi-nal ALMA non-detection limit of − . ≤
33 m ALMA baselines would havelikely resulted in at least 90–95% line dilution (factor of 10–20 suppression) for spatially-uniform SO gas. Therefore,taking the sensitivity of the two telescopes into account, ourJCMT fit does not need to be adjusted, but our modeled SO l:c ratios should be divided by at least ∼
20, if the SO isuniform across the disk, to approximate the ALMA detec-tion for that set of baseline configurations. In doing so, ourpredicted SO reference line values fall below the “10 ppb”( − . -only model with up to several hundred ppbof SO in the mesosphere can fit the JCMT data, and stillbe consistent with the non-detection of SO in the ALMAwide-band data. Moreover, this strong line dilution, with thecorresponding loss of sensitivity to even high levels of SO ,suggests that the ALMA wide-band SO reference observa-tions were likely poor indicators that SO was low enough tobe ruled out as a significant source of the JCMT 266.94 GHzline—thereby significantly weakening the argument that thisline was instead due primarily to PH .In addition to explaining the JCMT single-dish detectionof the 266.94 GHz line, and the suppression of the SO ref-erence lines in the ALMA data, our SO -only hypothesis would also predict that the 266.94 GHz ALMA line wouldbe, like the SO reference lines, strongly suppressed by linedilution and potentially non-detectable. While this was notthe case in the original Greaves et al. (2020a) paper, this isnow consistent with recent significant challenges to the de-tection confidence of the 266.94 GHz ALMA line. Theseinclude reanalyses of the Greaves et al. (2020a) narrowbandALMA discovery data by both Snellen et al. (2020) andVillanueva et al. (2020) who concluded that the feature at-tributed to PH could not be detected with statistical signifi-cance. Our own further analysis of the Greaves et al. (2020a)ALMA data, including testing the robustness of the detec-tion at 266.94 GHz, comes to a similar conclusion, and ispresented in Akins et al. (accepted). Additionally, a recentreanalysis of high-resolution, S/N ∼ transition near10.47 µ m, but it was not detected, setting a stringent upperlimit of 5 ppb above the Venus clouds (Encrenaz et al. 2020).Finally, the recent Greaves et al. (2020b) communication an-alyzing a reprocessing of the ALMA data suggests that the266.94 GHz feature in the narrow-band whole-planet ALMAdata is now significantly reduced in detection significancefrom the original discovery paper (4.8- σ vs 13.3- σ ), withan l:c of − × − , consistent with 1 ppb of PH . However,this much-reduced 266.94 GHz feature would also be consis-tent with line-diluted SO , which in our model would havel:c of − − × − at this frequency, for line dilution inthe range 95–97%—which is likely well within the range ofpotential line dilution (Akins et al. accepted).Although the SO hypothesis self-consistently explains ourcurrent understanding of the detection and non-detections inthe JCMT and ALMA data, additional analyses and obser-vations will be needed to more definitively discriminate be-tween PH and SO as the source of the 266.94 GHz JCMTline. Re-observing Venus at 266.94 GHz will likely still beneeded to independently confirm the discovery observation,and detection of an additional PH absorption feature wouldprovide a much stronger case for its presence in the Venus at-mosphere. Future observations to confirm the PH J = ← abundance is critical to the PH identifica-tion for the ALMA data, we recommend that future attemptsto confirm the ALMA PH observations should also obtainnear-simultaneous SO measurements. The narrowband cor-relator configuration can be tuned to 266.94 GHz and to thefrequencies of two nearby, stronger SO lines (near 267.54and 267.72 GHz). To mitigate the spectral ripple features thatcompromised measurement of the line intensities (Greaveset al. 2020a), these observations should occur when the ap-0parent angular diameter of Venus is smaller and therefore lessresolved by the ALMA antennas.Ultimately, the claimed detection of PH in the atmosphereof Venus has underscored the necessity of identifying and as-sessing the context of the environment within which we findpotential biosignatures. The identification of the 266.94 GHzline as due to PH , and its plausibility as a potential biosigna-ture, is inextricably intertwined with the physical and chem-ical environment of the Venus cloud and above-cloud atmo-sphere. This initial, controversial detection has highlightedjust how much we still need to understand about our sisterplanet, and how important that knowledge is in interpretingthis discovery. If the 266.94 GHz line is confirmed, and con-clusively attributed to PH , its presence in the mesospherewould require additional observations to understand potentialsources and sinks, and the attendant (and as yet unknown)phosphorous chemistry that enables its persistence at thesehigh altitudes. Moreover, if PH is being generated abioti-cally, especially at these high altitudes, this would have nega-tive implications for the robustness of PH and other reducedgases to serve as biosignatures in oxidizing terrestrial atmo-spheres. Regardless of the outcome, additional targeted ob-servations will reveal processes on a terrestrial planet thatinforms our understanding of our own world, and potentiallya large number of exoplanets that may share a similar evolu-tionary path and current environment. CONCLUSIONSWe simulated millimeter-wavelength Venus spectra to ex-plore the vertical distribution and detectability of PH andSO in the Venus atmosphere. We find that the observa-tions of the 266.94 GHz absorption line are insensitive to theabundance of PH and SO within the cloud deck. Instead,the observed absorption at this wavelength originates fromthe mesosphere at altitudes above 80 km. At these altitudes,PH would be rapidly destroyed, such that 20 ±
10 ppb ofPH would require a flux of PH to the Venus mesospherethat is ∼
100 times higher than the global production rateof photosynthetically-generated O on Earth. Because PH and SO both absorb within the width of the line detectedat 266.94 GHz, we emphasize that the identification of this absorption line as due to PH in both the ALMA and JCMTdata relies heavily on the apparent low abundance of SO inferred from the non-detection of an SO reference line at267.54 GHz in the ALMA data. However, we show thatSO absorption is likely heavily suppressed in the ALMAdata. Using SO vertical profiles within the range of previ-ous observations (from 30 ppb at 78 km to 400 ±
150 ppb at100 km)—including SO observations taken within a monthof the JCMT data—our model can fit the depth and widthof the 266.94 GHz feature without PH . We also show thatALMA line dilution suppresses the values for nominal Venusmesospheric SO to below the corresponding detectabilitylimit set by Greaves et al. (2020a). Given the mesospheric al-titude range, short chemical lifetime of PH , and consistencywith existing mesospheric SO abundances observed within amonth of the JCMT observations, we argue that SO providesa more self-consistent explanation for the 266.94 GHz fea-ture than PH . Single dish observations optimized for Venusand used to assess the PH detection and SO abundance inthe Venus upper mesosphere should be prioritized to discrim-inate between PH or SO as the source of the 266.94 GHzline. ACKNOWLEDGEMENTSThis work was performed by the Virtual Planetary Labo-ratory Team, a member of the NASA Nexus for ExoplanetSystem Science, and funded via NASA Astrobiology Pro-gram Grant No. 80NSSC18K0829. Part of this work wasconducted at the Jet Propulsion Laboratory, California In-stitute of Technology, under contract with NASA. Govern-ment sponsorship acknowledged. This work made use of theadvanced computational, storage, and networking infrastruc-ture provided by the Hyak supercomputer system at the Uni-versity of Washington. We thank Jacob Lustig-Yaeger, KevinZahnle, and the anonymous reviewers for helpful comments.
Software:
LBLABC (Meadows & Crisp 1996), SMART(Meadows & Crisp 1996), CARTA (Comrie et al. 2020),CASA (McMullin et al. 2007), Matplotlib (Hunter 2007),Numpy (van der Walt et al. 2011), GNU Parallel (Tange2011), WebPlotDigitizer (Rohatgi 2018).REFERENCES
Akins, A., Lincowski, A. P., Meadows, V. S., & Steffes, P. G.accepted, Complications in the ALMA Detection of Phosphineat VenusArney, G., Meadows, V., Crisp, D., et al. 2014, Journal ofGeophysical Research: Planets, 119, 1860,doi: 10.1002/2014JE004662Bains, W., Petkowski, J. J., Seager, S., et al. 2020, arXiv e-prints,arXiv:2009.06499. https://arxiv.org/abs/2009.06499 Bellotti, A., & Steffes, P. G. 2015, Icarus, 254, 24,doi: 10.1016/j.icarus.2015.03.028Belyaev, D. A., Montmessin, F., Bertaux, J.-L., et al. 2012, Icarus,217, 740, doi: 10.1016/j.icarus.2011.09.025Belyaev, D. A., Evdokimova, D. G., Montmessin, F., et al. 2017,Icarus, 294, 58, doi: 10.1016/j.icarus.2017.05.002Bezard, B., de Bergh, C., Fegley, B., et al. 1993,Geophys. Res. Lett., 20, 1587, doi: 10.1029/93GL013382