Complications in the ALMA Detection of Phosphine at Venus

A recent paper by Greaves et al. (2020a) reports the identification of PH₃ in the atmosphere of Venus via radio telescope observations of the PH₃ 1-0 transition line near 266.94 GHz (1.123 mm). A spectral feature was first observed near this frequency using the James Clerk Maxwell Telescope (JCMT), a single dish millimeter/submillimeter radio telescope operating in Maunakea, H i.⁵ This spectral feature, however, could not be uniquely identified as PH₃ due to the presence of the SO₂ 30_9,21–31_8,24 line within 1.5 km s⁻¹ of the PH₃ 1-0 line. Since SO₂ is an abundant trace gas in the Venus atmosphere at all altitudes (see Marcq et al. 2018 and references therein), it could be the case that the detected line was that of SO₂ instead of PH₃. To assess this uncertainty, the JCMT observations were followed up by interferometric observations of Venus near 266.94 GHz using the Atacama Large Millimeter Array (ALMA). The ALMA Band 6 correlator configurations for these observations included wideband (1.875 GHz bandwidth, 976.562 kHz spectral resolution) and narrowband (117.1875 MHz bandwidth, 61.035 kHz spectral resolution) setups centered on both the PH₃ 1-0 line of interest and the nearby HDO 2_2,0–3_1,3 line for validation. The wideband configuration centered on the 266.94 GHz spectral feature can also be used to estimate the intensities of two stronger SO₂ tracer lines (13_3,11–13_2,12 near 267.54 GHz and 28_4,24–28_3,25 near 267.72 GHz). These tracer lines could then be used by the authors to determine if the spectral feature at 266.94 GHz could be attributed to PH₃.

Following data reduction and imaging, Greaves et al. (2020a) report a nondetection of the SO₂ 13_3,11–13_2,12 tracer line, and they suggest an upper concentration limit of 10 ppb SO₂ based on the amplitude of the spectral ripple observed in the wideband data. Assuming this SO₂ abundance, the authors determine that the contamination from the weaker SO₂ 30_9,21–31_8,24 line is not sufficient to explain the observed spectral feature, and therefore that the 266.94 line can be attributed to 20 ppb PH₃ (see Figure 4 in Greaves et al. 2020a). They also report detection of the 266.94 GHz spectral feature in both the wideband and the narrowband ALMA data, as well as identifying the expected HDO line in their other spectral configurations.

Due to the weighting function of 266.94 GHz continuum emission peaking in the middle cloud layer of Venus (53–61 km above the mean surface), Greaves et al. (2020a) propose that 20 ppb PH₃ is present in the clouds of Venus. Given the chemistry of the Venus atmosphere and the PH₃ formation and destruction rates from the simulations of Bains et al. (2020), it is not expected that such a high concentration of PH₃ can exist with long-term stability without a significant, unaccounted for source (see Figure 5 in Greaves et al. 2020a). After ruling out several formation mechanisms based on the current understanding of Venus atmospheric chemistry, Greaves et al. (2020a) suggest that the origin of PH₃ could be biogenic. A recent paper by Limaye et al. (2018) reviews hypotheses for biological activity within the Venus cloud deck in the context of observed UV absorption, and another recent paper by Seager et al. (2020) proposes a life cycle for such life through cycling of H₂SO₄ cloud aerosols. The identification of PH₃ in the cloud region, therefore, could be in favor of such a hypothesis. However, several recent papers call into question the detection of a spectral feature at 266.94 GHz (Snellen et al. 2020; Thompson 2021; Villanueva et al. 2020) and its interpretation as being due to PH₃ (Encrenaz et al. 2020; Lincowski et al. 2020; Villanueva et al. 2020). The authors of the original study have also performed a reanalysis of their ALMA data following the identification of errors in their calibration procedures (Greaves et al. 2020b). Their reanalysis again reports detection of PH₃ with the ALMA data, but they report a lower abundance of 1 ppb. Greaves et al. (2020b) also identify image regions with apparently stronger absorption compared with the full disk average.

Given the implications of the identification of PH₃ as a trace gas in the Venus mesosphere, the observations reported by Greaves et al. (2020a) must be carefully considered and independently verified by other research groups. In this paper, we discuss our independent analysis of the Greaves et al. (2020a) ALMA observations, which is complementary to the recent reanalyses of Snellen et al. (2020), Villanueva et al. (2020), and Greaves et al. (2020b). We re-reduce and image the ALMA data using both the initial calibration methods of Greaves et al. (2020a) and the revised calibration methods presented in their follow-on letter (Greaves et al. 2020b). We also perform model simulations to assess how different spatial extents of absorbing gases affect the apparent spectral line dilution (suppression of an absorption feature) that results from the observing geometry and configuration of the ALMA antennas. While we are able to observe a 266.94 GHz feature in the ALMA narrowband data using the initial calibration procedures, we fail to detect it following the revised calibration. No 266.94 GHz spectral feature is observed in the wideband data using either set of calibration procedures. We determine the line dilution incurred by the ALMA observation geometry admits SO₂ concentrations that are considerably higher than the 10 ppb limit suggested by Greaves et al. (2020a) if the gas is globally or hemispherically distributed. While we are unable to reproduce the detection of PH₃ in the ALMA data, we encourage new observations of Venus in the vicinity of PH₃ spectral lines to further investigate the possibility of PH₃ at Venus.

ALMA observed Venus in the C-2 configuration (15–314 m baselines) on 2019 March 5 over two observing sessions, and the observation data were obtained from the ALMA archive. The Common Astronomy Software Applications (CASA) package was used for all data reduction and imaging of the ALMA data, as well as for simulations described in this section. The spectra of the resulting images of Venus were then analyzed using the Cube Analysis and Rendering Tool for Astronomy (CARTA).

2.1. Reduction and Imaging of ALMA Data

2.2. Simulation of ALMA Line Dilution

3.1. ALMA Data

Figure 1 shows narrowband and wideband spectra obtained from the ALMA data of Greaves et al. (2020a) binned to a common spectral resolution. The spectra shown in the top row of Figure 1 result from the application of the initial calibration procedures, and those shown in the bottom row result from the use of the revised calibration procedures. Spectra are shown where all ALMA antenna baselines are included and where baselines shorter than 33 m in length are excluded, following Greaves et al. (2020a). For all spectra, only a first-order polynomial is subtracted from the data set, and a best-fit twelfth-order polynomial with a 5 km s⁻¹ exclusion window is superimposed on the narrowband spectra. While a distinct spectral feature would be apparent following the subtraction of the twelfth-order polynomial from the narrowband spectra obtained with the initial calibration, no such feature is apparent upon correction of the calibration procedures.

Zoom In Zoom Out Reset image size

Figure 2 shows the ALMA spectra determined with revised calibration averaged over the equatorial, mid-latitude, and polar regions. As per Greaves et al. (2020a) Table 1, both north and south mid-latitude and polar regions are averaged to form their respective spectra, and the regional spectra are vertically offset for clarity. The narrowband LB spectra in Figure 2 are directly comparable to Greaves et al. (2020a) Figure 2 and Extended Data Figure 2.

Zoom In Zoom Out Reset image size

3.2. Baseline-dependent Effects and Line Dilution

The dependence of disk-averaged spectral line dilution on the spatial scale of spectral features was studied by re-imaging simulated Venus models enhanced with the SO₂ spectra from Lincowski et al. (2020, see their Figure 2(D)) over a variety of spatial scales and distributions. The apparent dilution of the 267.54 GHz line is discussed here, although there is no change in line dilution over the range of frequencies relevant to the ALMA observations. The resulting brightness distributions of the re-imaged continuum-subtracted models are illustrated in Figure 3. These model images were re-imaged using the LB data set baseline coverage and are shown at line center. The regions where SO₂ is added to the the starting models are shown in white. Depending on the spatial scale of the spectral features, spatial ripples can be observed across the disk.

Zoom In Zoom Out Reset image size

The apparent disk-averaged line dilutions for both LB and AB data sets are shown in Figure 4. The resulting line intensities are normalized by the disk-averaged intensity of their respective starting models. For both cases, the model SO₂ spectral lines are diluted considerably, many by over 90%. The images excluding short antenna baselines suffer greater line dilution, on the order of 95% for larger scale distributions. For some cases, the disk-averaged spectral lines appear in emission instead of in absorption due to the ripples resulting from the limited short-baseline coverage.

Zoom In Zoom Out Reset image size

To justify their detection of PH₃ in the Venus atmosphere, Greaves et al. (2020a) use the following arguments, listed in the order of their subsequent discussion. First, they report the detection of a spectral feature at 266.94 GHz from Venus observations using JCMT. Second, they report a repeated detection of this feature from narrowband ALMA observations of Venus. Third, they report a simultaneous detection of the 266.94 GHz spectral feature in their ALMA wideband data. Finally, they rule out the possibility that the spectral feature observed can be associated with SO₂ by estimating its abundance using a stronger SO₂ transition within their wideband ALMA spectrum at 267.54 GHz. Given the results of our independent data reduction and line dilution modeling results, we can address each of these arguments.

First, Greaves et al. (2020a) report the detection of a spectral feature at 266.94 GHz from JCMT observations. The authors implicitly suggest that this feature is associated with PH₃ and not SO₂ due to the nondetection of SO₂ lines in the wideband ALMA data observed 2 yr later. We find that differences in observation methods between ALMA and JCMT complicate this argument. Since ALMA is a radio interferometer, the MRS and the spatial resolution are determined via the spacing between correlated antenna pairs. Due to the configuration of ALMA at the time, the observed spectrum of gases distributed globally with hemispherical variations, like SO₂, will suffer from significant line intensity dilution on the order of 95%. This problem is compounded by the exclusion of baselines shorter than 33 m from the analysis of Greaves et al. (2020a), as shorter baselines are required to measure flux over larger spatial scales. Unlike ALMA, JCMT is a single dish radio telescope measuring the total flux emitted by sources in the direction of antenna pointing independently of the spatial scale (as long as the emission originates from a region observed by the antenna beam). Thus, the JCMT observations do not suffer from scale-dependent spectral line dilution, and so the inferred abundance of SO₂ from the Greaves et al. (2020a) ALMA wideband data is not applicable here. Recent analyses by Lincowski et al. (2020) and Villanueva et al. (2020) find that nominal abundances of SO₂ in the Venus mesosphere are capable of producing the spectral feature at 266.94 GHz as detected by JCMT. In their response paper, Greaves et al. (2020b) argue that the 266.94 GHz feature measured by JCMT could not be SO₂ due to the required abundance of 150 ppb, which is considerably higher than millimeter-wavelength measurements (Sandor et al. 2010) from 2004 to 2008 (which are sensitive to emission from higher altitudes). The vertical abundance of SO₂, however, is possibly variable up to several hundreds of ppb over longer timescales (Vandaele et al. 2017). SO₂ observations with TEXES/IRTF within a month of the Greaves et al. (2020a) observations also reported cloud top abundances of hundreds of ppb (Encrenaz et al. 2019).

Second, Greaves et al. (2020a) report the detection of the spectral feature at 266.94 GHz in their ALMA narrowband data. As shown in our Figure 1, use of the initial, erroneous calibration procedures and subtraction of a twelfth-order polynomial (with a 5 km s⁻¹ exclusion window) from the spectral baseline would produce a distinct spectral feature at 266.94 GHz. Following the use of the revised calibration procedures, this spectral feature disappears. The reanalysis of the ALMA data with CASA and AIPS by Villanueva et al. (2020), with an intermediate correction to the bandpass calibration, results in a nondetection of any spectral feature at 266.94 GHz for the full disk-averaged spectrum. Our reanalysis, using the most recently revised calibration procedures (including a third-order polynomial fit for the amplitude bandpass correction) and the same antenna baseline configuration (excluding baselines shorter than 33 m), shows no evidence of a spectral feature. This contrasts with Greaves et al. (2020b) who claim that a spectral feature remains following calibration corrections. Greaves et al. (2020b) achieve a lower noise level than our spectra due to their employment of an experimental twelfth-order polynomial fit for amplitude bandpass calibration. Correcting the bandpass using a higher order polynomial may prove to be questionable in a similar manner to the original twelfth-order polynomial fit to the spectral baseline described in Greaves et al. (2020a; and criticized by Snellen et al. 2020; Villanueva et al. 2020). All of this considered, the detection of the spectral signal at 266.94 GHz in the narrowband by Greaves et al. (2020a) appears to be highly questionable.

Third, Greaves et al. (2020a) report simultaneous detection of the 266.94 GHz spectral feature in their wideband ALMA data. Their wideband spectrum near 266.94 GHz, however, shows considerable spectral ripples, including a ripple of nearly the same peak magnitude of the detected line at roughly −20 km s⁻¹. Unlike Greaves et al. (2020a), we are unable to detect this spectral feature in the ALMA wideband data using either calibration procedure. Furthermore, the spectral ripple in our Figure 1 exhibits a lower peak-to-peak spectral ripple than Greaves et al.'s (2020a) Extended Data Figure 6. This lack of detection from simultaneous observations at comparable noise performance weakens the argument that the narrowband signal is not due to a spectral ripple.

Finally, Greaves et al. (2020a) suggest that the observed spectral feature cannot be SO₂ due to nondetection of a stronger line at 267.54 GHz in the wideband data. Considering their observed spectral ripple, Greaves et al. (2020a) assign an upper limit of 10 ppb SO₂. This is lower than expected for the Venus cloud top regions, and observations taken only a month later suggest an SO₂ abundance of hundreds of ppb (Encrenaz et al. 2019). From our simulations of spectral line dilution as shown in Figure 4, gases on global and hemispherically symmetric spatial scales can be diluted on the order of 90%–95% due to the ALMA antenna configuration and the subtraction of antenna baselines shorter than 33 m. Thus, SO₂ abundances on the order of hundreds of ppb would exhibit spectral features below the noise threshold of the ALMA spectra. The successful observation of HDO in the Greaves et al. (2020a) narrowband ALMA data, however, does contradict this assessment, as water is expected to be distributed at similar spatial scales. In their reanalysis, Greaves et al. (2020b) derive a tentative H₂O abundance of 23 ppb, which is on the order of 90% less than prior ALMA observations (Encrenaz et al. 2015). Assuming the Encrenaz et al. (2015) H₂O mesospheric abundance of 2.5 ppm, the Greaves et al. (2020b) result is consistent with our line dilution simulations and suggests large-scale spatial distribution for H₂O. Future observations of PH₃ with ALMA should also include narrowband observations of the 267.54 GHz SO₂ spectral line to obtain a more accurate constraint on SO₂ abundance.

Following reassessment of the arguments of Greaves et al. (2020a), it appears unlikely that PH₃ was detected in the Venus atmosphere. The nondetection of PH₃ at Venus using archival IR observations as reported by Encrenaz et al. (2020) provides further weight to this argument. While Greaves et al. (2020b) detect a weak feature at 266.94 GHz in their reanalysis, they admit that the strength of this feature would be consistent with SO₂ contamination. A stronger answer to this question can be found through continued ground- and space-based observation of Venus at millimeter, submillimeter, and infrared wavelengths near PH₃ spectral features. These observations should also target features of possible spectral contaminants.

We have reanalyzed the wideband and narrowband ALMA data of Greaves et al. (2020a) to investigate the robustness of their detection of PH₃. We have also simulated the spectral line dilution resulting from their ALMA configuration for several different distributions of an absorbing gas at varying spatial scales. We find that a spectral feature near 266.94 GHz in the narrowband data becomes apparent using the initial calibration procedures of Greaves et al. (2020a) and subtraction of a twelfth-order polynomial. Upon application of the revised calibration procedures, however, no such spectral feature is detected. We are also unable to detect a spectral feature in the wideband ALMA spectra. Additionally, we have used the line dilution for gases distributed at a global scale to find that the nondetection of SO₂ in the wideband ALMA data is potentially consistent with a nominal cloud top abundance far greater than the 10 ppb SO₂ upper limit suggested by Greaves et al. (2020a; see also the companion study in Lincowski et al. 2020). These findings, along with the recent papers by Encrenaz et al. (2020), Snellen et al. (2020), Lincowski et al. (2020), and Villanueva et al. (2020) undermine the reported detection of PH₃ by Greaves et al. (2020a, 2020b) and its possible biogenic origin. It is clear that further studies of Venus at millimeter, submillimeter, and infrared wavelengths with ground- or space-based observatories are necessary to further investigate the possibility of PH₃ at Venus.

We thank the authors of the original study for publicly sharing their initial calibration and imaging scripts. We also thank ESO ALMA Regional Center for their efforts during QA3 assessment of the original data set and for making the updated calibration procedures available via the ALMA archive. Venus was observed under ALMA Directors Discretionary Time program 2018.A.0023.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada), MOST and ASIAA (Taiwan), and KASI (Republic of Korea), in cooperation with the Republic of Chile. The JAO is operated by ESO, AUI/NRAO, and NAOJ. We thank J. Pearson and an anonymous reviewer for helpful comments on the manuscript. Part of this work was conducted at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. Part of this work was performed by the Virtual Planetary Laboratory Team, which is a member of the NASA Nexus for Exoplanet System Science, and funded via NASA Astrobiology Program Grant 80NSSC18K0829.

Software: CARTA (Comrie et al. 2020), CASA (McMullin et al. 2007), Matplotlib (Hunter 2007), Numpy (Harris et al. 2020).