Collision‐Induced Absorption of CH4‐CO2 and H2‐CO2 Complexes and Their Effect on the Ancient Martian Atmosphere

Experimental measurements of collision‐induced absorption (CIA) cross sections for CO2‐H2 and CO2‐CH4 complexes were performed using Fourier transform spectroscopy over a spectral range of 150–475 cm−1 and a temperature range of 200–300 K. These experimentally derived CIA cross sections agree with the spectral range of the calculation by Wordsworth et al. (2017) however, the amplitude is half of what was predicted. Furthermore, the CIA cross sections reported here agree with those measured by Turbet et al. (2019, 2019). Additionally, radiative transfer calculations of the early Mars atmosphere were performed, and showed that CO2‐CH4 CIA would require surface pressure greater than 3 bar for a 10% methane atmosphere to achieve 273 K at the surface. For CO2‐H2, liquid water is possible with 5% hydrogen and less than 2 bar of surface pressure.

no CH 4 -CH 4 CIA cross sections in the literature, so a linear combination ends up simply as scaled CO 2 -CO 2 CIA cross section. For these reasons, experimental validation of this linear combination approximation method are still required (Karman et al., 2019).
Additional studies (Kamada et al., 2020;Ramirez, 2017;Ramirez et al., 2020) used the cross sections from Wordsworth et al. (2017) in Martian climate models and found that the inclusion of CIA between CO 2 and H 2 does result in a warm and wet early Mars that agrees with the paleopressure and climate stability constraints with only 3% hydrogen concentration. Unfortunately, computing CIA cross sections is quite challenging, and to date, the only CIA cross sections for CO 2 -H 2 and CO 2 -CH 4 complexes in the literature are limited to room temperature, a spectral range of 60-535 cm −1 , and a resolution of 1 cm −1 (Turbet et al., 2019), which found that the experimentally derived CIA cross sections are weaker than predicted by Wordsworth et al. (2017), but still stronger than N 2 -H 2 CIA. In a recent paper published at the same time as this study, Turbet et al. (2020) derived new theoretical CIA values that match their room-temperature measurements, but this still requires experimental verification of their predicted temperature dependence. This paper expands upon the experimental work of Turbet et al. (2019Turbet et al. ( , 2020, detailing the first temperature-dependent experimental measurements of CO 2 -H 2 and CO 2 -CH 4 CIA. Given that experimentally derived CIA cross sections are more reliable than predicted ones, these new experimentally derived cross sections are used in early Mars climate models to improve our understanding of the impact CIA may have had on the climate of ancient Mars.

Experimental Procedure and Data Analysis
Experiments were performed at the Far-Infrared (IR) beamline of the Canadian Light Source (CLS) Synchrotron facility. The IR absorption spectra were obtained using a Bruker IFS 125HR Fourier transform spectrometer connected to a temperature-controllable White-type cell with a path length of 7275 ± 6 cm. Experiments were performed using a globar source, mylar beam splitter, polypropylene windows, and Si bolometer detector. CO 2 -H 2 spectra were recorded at an unapodized resolution of 0.05 cm −1 (maximum optical path difference of 20 cm), while CO 2 -CH 4 spectra were recorded at an unapodized resolution of 0.1 cm −1 (maximum optical path difference of 10 cm). Each measurement consisted of 300-500 co-added spectra. The number of co-added spectra was primarily chosen due to time constraints; the experiment only had a finite amount of beamtime at the CLS to make all the required measurements, so longer measurement runs with more averaging were not possible. Temperature control was provided by a recirculating chiller. The chiller was capable of reaching temperatures as low as 203 K with the cell containing gas at atmospheric pressure.
The gas samples were commercial products from Praxair with stated high purities of greater than 99%. The cell was evacuated using a Varian turbo pump. The pressure in the gas cell was measured using a combination of a 10 and 1000 Torr MKS baratron pressure gauges. The temperature was measured with a four-wire PT100 resistance temperature detector mounted on the outside of the inner cell wall. At each temperature, measurements were made at two or three pressure combinations as detailed in the supporting information. A third pressure combination was only included if significant water vapor contamination was suspected during an experimental run. However, since water vapor lines don't affect the full spectral range and overall uncertainty is reduced with added measurements, these contaminated spectra are still included in the analysis. Empty cell scans were performed between filled-cell runs to monitor baseline stability and to account for the effect of windows and mirrors in the White cell on light intensity. Due to safety limitations when using explosive gases at the CLS, the mixing ratio of H 2 in CO 2 was restricted to a maximum 8.3%; while for CH 4 in CO 2 , the mixing ratio was restricted to maximum 20%. Methane experiments were performed by first filling the cell with CH 4 , followed by adding CO 2 . The H 2 experiments were performed using premixed H 2 -CO 2 gas cylinders as indicated in the supporting information.
The absorption of light in the White cell can be described using Beer-Lambert law: where I ( )  ν is the intensity as a function of wavenumber ( )  ν after passing through the gas sample (filled-cell measurement) and I o ( )  ν is the intensity without passing through a gas (empty-cell measurement). χ ( )  ν is the optical depth, which for a mixture of CO 2 and another gas in a cell, ignoring higher order effects (such as line broadening and pressure shifts), is given by where L is the cell length in cm, ρ is the density of the gas in mol·cm −3 , and σ(ν˜) are the absorption cross sections for either single gas species (cm 2 ·mol −1 ) or CIA of mixed species (cm 5 ·mol −2 ) depending on the subscript (x designating either H 2 or CH 4 ). The density of a gas in the cell is related to the pressure by where P o and T o are standard conditions for pressure (Torr) and temperature (Kelvin), and N L is Loschmidt's constant (mol·cm −3 ).
CIA cross sections exist for CO 2 -CO 2 (Gruszka & Borysow, 1997) and H 2 -H 2 (Abel et al., 2011) at the temperatures and spectral region investigated in this study. Measurements were conducted using only CO 2 gas in the cell to confirm the accuracy of the CIA derived using this setup by comparing CO 2 -CO 2 CIA from this work with previously published values (Gruszka & Borysow, 1997) as detailed in the supporting information. The CO 2 -CO 2 CIAs measured in this work generally agree with Gruszka and Borysow (1997); however, they can include additional features due to pressure distortions on the windows and mirrors in the White cell between full and empty cell spectra. When removing CO 2 -CO 2 CIA from CO 2 -CH 4 and CO 2 -H 2 experiments, the CO 2 -CO 2 CIAs from this work were used instead of those of Gruszka and Borysow (1997) because they may include White cell deformations that can occur between empty-cell background spectra and full-gas cell spectra, and cover the full experimental spectral range, whereas the CO 2 -CO 2 CIA from Gruszka and Borysow (1997) only extends to 250 cm −1 . H 2 -H 2 CIA from Abel et al. (2011) are used as is to remove this CIA from the spectra.
H 2 does not have significant single-molecule absorption at the concentrations and wavelengths investigated, so it was not included. Additionally, there was some water vapor contamination in the cell, which requires the subtraction of water absorption lines from the measured optical depth; however, since the concentration of water vapor is unknown, the optical depth was fit to match the absorption spectra of water vapor in contaminated regions, using the density of water as a free parameter. Single-molecule absorption cross sections exist in the literature for H 2 O (Gordon et al., 2017); however, these are primarily for air-broadened lines. To better approximate CO 2 -broadened lines, the same procedure was used as in Turbet et al. (2019): a factor of 1.14 was applied to the air-broadening coefficients for H 2 O broadening by CH 4 and using the H 2 O lines broadened by CO 2 from 200 to 900 cm −1 (Brown et al., 2007). For 150-200 cm −1 , the air-broadened lines were used. The H 2 O lines are simulated using the HITRAN Application Programming Interface (HAPI). CH 4 single-molecule and self CIA were removed by running a spectral scan when the cell was only filled with CH 4 , but before the CO 2 was added. This CH 4 -only spectra were then subtracted from the CO 2 -CH 4 optical depth.
The removal of these unwanted absorption features is not perfect, since there are small differences between the simulated lines from HAPI and the measured spectra, and CH 4 line broadening by CO 2 . A 10-cm −1 median filter was applied to remove any remaining narrow features; the filter width was determined by slowly increasing it until the residual water lines were minimized. The uncertainty associated with this is included as part of the uncertainty in removing the water lines. Lastly, the baseline was adjusted to account for fluctuations in light intensity between empty-and filled-cell measurement runs. This was achieved by comparing the difference between empty cell spectra before and after filled cell scans.
Once the optical depth has been cleaned of the unwanted absorption effects, the CIA absorption cross sections can be found via a linear fit of optical depth versus the product of the density of CO 2 and either the GODIN ET AL. 3 of 10 10.1029/2019JE006357 density of CH 4 or H 2 for a given temperature and wavenumber (plots available in the supporting information) with a forced convergence of χ = 0 for P = 0, using (4) Sources of error include subtraction of unwanted absorption features (10%), temperature fluctuations (±0.2 K), baseline adjustment (10%), and pressure readout (±0.08 Torr). These errors are propagated in the calculation of the optical depth to determine its uncertainty. The uncertainty of the optical depth is used to assign weights in the linear fit against pressure to find the absorption cross section. The final uncertainty is the sum, in quadrature, of the linear fit error, path length uncertainty (±6 cm), and sample purity error (±1.0%), expressed at a 95% confidence interval.

Collision-Induced Absorption Cross Sections
Measured CO 2 -CH 4 and CO 2 -H 2 CIA cross sections as a function of temperature are shown in Figures 1 and  2, respectively. Below 150 cm −1 , the limit of the detector begins to be reached and water vapor contamination interferes with the signal, resulting in the amplitude of the cross section becomes less reliable; above 475 cm −1 , CO 2 lines begin to saturate the detector preventing the detection of other types of absorption and the baseline fluctuations become significant. Baseline lamp fluctuations vary with wavelength and between experiment, resulting in uncertainties that can vary with wavelength. The water lines are strongest in the region of 150-300 cm −1 , resulting in large uncertainty in this spectral range. Despite using high-purity gas samples (stated water vapor content of no more than 3 ppmv) and purging the cell between experimental GODIN ET AL. runs, these lines remain an issue since even at low concentrations, as water vapor absorbs strongly in this spectral range. Since water vapor contamination in the cell varies between experimental runs, the uncertainty associated with water lines also varies in the final results shown in Figures 1. The measurements with three pressure combinations instead of only two have smaller overall uncertainties compared to the other measurements. If experimental time at the CLS was not limited, more runs could have been preformed lowering the overall uncertainty in this study.
At room temperature, the measurement from this work for CO 2 -CH 4 CIA agree with the experimental results from Turbet et al. (2020), and it is observed that the theoretical prediction of Wordsworth et al. (2017) overestimates the CIA by rAUTHOR: Please check whether the insertion of the word "ratio" in the sentence "The uncertainty associated with..." is correct.oughly a factor of 2. The uncertainty associated with Turbet et al. (2020) measurement is lower than the uncertainty in this work because the signal-to-noise ratio is larger due to White cell used in their experiment having twice the optical path length and being allowed to use higher concentrations of CH 4 and H 2 . Looking at the temperature dependence of the CIA, the predicted increase in strength with decreasing temperature is also observed. Once again, Wordsworth et al. (2017) overestimates the CIA by roughly a factor of 2 at these colder temperatures, as the original prediction is beyond the upper limit of the uncertainty. The experimental measurements in this work are consistent with the recent theoretical prediction by Turbet et al. (2020), which also agrees with our observation that the theoretical prediction of Wordsworth et al. (2017) is twice as strong as it should be.
Due to the experimental safety limitations on the amount of hydrogen gas permitted in the gas cell, it was difficult to resolve the CO 2 -H 2 CIA from the noise, especially at higher temperatures, where the CIA effect is weaker, hence the uncertainty is higher compared to the CO 2 -CH 4 CIA measurements. At 293 K, the CIA signal is so weak, that it is unlikely to be observed above the noise in the experiment; this would also GODIN ET AL. explain the unphysical negative absorption in some parts of the CO 2 -H 2 spectra. The oscillatory behavior of the CIA from 150 to 350 cm −1 corresponds with the locations of waterlines, showing the large impact they have when the CIA signal is weak. Despite this, an overestimation factor of 2 can still be reasonably assumed when comparing the prediction from Wordsworth et al. (2017) to the experimental measurements, including the measurement by Turbet et al. (2020).
Ultimately, these results agree with the prediction of Wordsworth et al. (2017) when it comes to the spectral range. However, the Wordsworth et al.'s (2017) prediction consistently overestimates the strength of the CIA by a factor of 2. At room temperature, these results agree within combined errors with the experimental cross sections of Turbet et al. (2020), strengthening our confidence in those numbers. Additionally, the temperature-dependent measurements agree with the new temperature-dependent theoretical calculations performed by Turbet et al. (2020), thus supporting the conclusion of those calculations that the absorption strength is weaker at lower temperatures than was predicted by Wordsworth et al. (2017). It should be noted, however, that given the uncertainty of the results presented here, the magnitude of predicted temperature dependence of Turbet et al. (2020) cannot be confirmed by these results.
Attempts were also made to observe the CIA above 475 cm −1 , using an MCT detector and KBr windows/ beam splitter. However, in this regime, absorption lines from CO 2 and CH 4 were so strong that in order to not saturate the detector, less than 0.05 Torr of those gas species was used. With such a small amount of gas present, there was no longer enough to produce a measurable CIA signal above the noise.

Climate Model Details
Wordsworth et al. (2017) and Ramirez (2017) investigated the effect on the ancient Martian climate of including the predicted CIA cross sections in their radiative transfer models. However, as noted in Section 2.2, there is a systematic overestimation in the predicted CIA for both CO 2 -H 2 and CO 2 -CH 4 by a factor of 2 compared to the results of this work and Turbet et al. (2020). Therefore, it is worth revisiting these climate calculations with experimentally verified CIA absorption values. Since the experimental CIA does not cover the full spectral range needed for radiative transfer modeling, it is assumed that the scaling factor of 2 needed to adjust the Wordsworth et al. (2017) CIA is consistent across the full spectral range. The recent values from Turbet et al. (2020) are not used in the following climate modeling as more value is obtained in showing how the estimate of half strength of Wordsworth et al. (2017) found in this work compares with Turbet et al. (2020). A scaled Wordsworth et al. (2017) CIA was used in a single-column cloud-free radiative climate model following the same procedure as outlined below (Kasting, 1991;Kopparapu et al., 2013;Ramirez, 2017;Ramirez et al., 2014;Ramirez & Kaltenegger, 2018): This model follows a moist adiabat at warmer temperatures and relaxes to a CO 2 adiabat when temperatures become cold enough for CO 2 to condense (Ramirez et al., 2014). The model uses the correlated-k technique to compute absorption for CO 2 , H 2 O, CH 4 , and H 2 across 38 solar intervals, 55 infrared intervals, 5 temperatures (100, 200, 300, 400, and 600 K), and eight pressures (10 −5 −10 2 bar). At the low temperatures considered here (<300 K), the HITRAN (Gordon et al., 2017) line list was utilized to compute k-coefficients at all wavelengths for CO 2 and H 2 O, and in the thermal infrared for CH 4 . The far wings of water vapor are modeled using the Baranov Paynter Serio water vapor continuum (Paynter & Ramaswamy, 2011), which allows for accurately computed water vapor absorption at warm temperatures even though the water vapor continuum has almost no contribution at the temperatures of interest here. However, the HITRAN line list is still incomplete for CH 4 at visible and near-infrared wavelengths (Kassi et al., 2008). Instead, near-infrared CH 4 k-coefficients for wavelengths under 1 micron from Karkoschka (1994) were combined with those from Irwin et al. (1996) for the spectral range between 1 and 4.5 microns. Rayleigh scattering for H 2 O, CO 2 , and CH 4 (Ramirez et al., 2014;Sneep & Ubachs, 2005) was included. CO 2 -CH 4 and CO 2 -H 2 CIA are incorporated, both from Wordsworth et al. (2017), and the CIA scaled by 0.5 to match the above experimental results. CO 2 -CO 2 CIA is modeled following the procedure in Wordsworth et al. (2010), using experimental data from Baranov et al. (2004) and Gruszka and Borysow (1997).
Overall, the new CIA yield significantly cooler mean surface temperatures as shown in Figures 3a and 3b, when comparing the results from this work (solid lines) to the previous work (dashed lines). The temperatures are still higher than those derived in Ramirez et al. (2014), using the weaker N 2 CIAs, which would require 3-4 bar total pressure to achieve the same temperatures. These results support a recent study by Ramirez et al. (2020), which showed that warming from CO 2 -H 2 CIA would likely be between the extremes of using N 2 -H 2 CIA and the CIA values from Wordsworth et al. (2017). Mean surface temperatures above 273 K can be reached with the new CIA if surface pressures exceed ∼3 bar for a 10% CH 4 atmosphere or with 5% hydrogen and less than 2 bar of surface pressure. However, current studies on geologic observations suggest that the pressure on early Mars during valley network formation was no higher than ∼2 bar (Hu et al., 2015;Kite et al., 2014;Kurokawa et al., 2017). Hu et al. (2015) specifically states that for warm atmospheres such as are considered here and the assumption that there is a substantial hydrologic cycle and carbonate deposition, CO 2 pressures could have been as high as 1.8 bar; for a scenario in which there were no large water bodies and little carbonate deposition on the surface, Hu et al. (2015) estimates that paleopressures were no higher than ∼1 bar. Grott et al. (2011) argued for a 1 bar atmosphere on ancient Mars; however, that was published before the CO 2 -H 2 hypothesis of Ramirez et al. (2014) and so they were unaware of the latter's implications. The meteoritic analysis there suggested that the mantle fugacity for early Mars was near the iron-wustite buffer (if not lower), which is why the CO 2 -H 2 mechanism of Ramirez et al. (2014) and Ramirez (2017) works in principle. Jakosky et al. (2018) modeled the atmospheric loss rate of Mars and found a lower estimate of 0.8 bar of CO 2 has been lost, suggesting that the ancient Martian atmosphere was greater than 0.8 bar. The variability in Mars paleopressure estimates is well summarized by Kite (2019); most estimates are closer to a pressure of 1 bar, but pressures near 2 bar remain possible. Another issue is that for CH 4 concentrations above a CH 4 /CO 2 ratio of 0.1, photochemical hazes form that cool the planet, canceling the greenhouse effect (Haqq-Misra et al., 2008). Ultimately, CO 2 -CH 4 absorption is not as promising as initially argued in Wordsworth et al. (2017). CO 2 -H 2 , however, still seems much more promising, as liquid water is possible with 5% hydrogen and less than 2 bar of surface pressure. These results agree with recent modeling work by Turbet et al. (2020), which also found that ∼6% of H 2 in a 2 bar atmosphere could achieve liquid water on the surface. Additionally, this work also supports the assertion by Turbet et al. (2020) that CH 4 is insufficient to warm the surface on its own.
The planetary albedo plots are calculated here using the updated CIA and compared to those found in Ramirez (2017); as in that study, the planetary albedo decreases with increasing CH 4 or H 2 concentrations since this causes the overall atmospheric scattering to decrease. Comparing the effect of scaled versus unscaled CIA on planetary albedo, the albedo does not change significantly with the updated CIA for CO-CH 4 as shown in Figures 3c and 3d; this is primarily because water vapor amounts are still small at these temperatures. However, with updated CO 2 -H 2 CIA, there is a slight change in planetary albedo. This is because H 2 has a larger impact on the surface temperature (Figures 3a and 3b) compared to CH 4 , which in turn increases the amount of water vapor in the atmosphere, resulting in the increased sensitivity of the planetary albedo to changes in CO 2 -H 2 CIA. Additionally, atmospheres with CH 4 will have a lower planetary albedo than those with H 2 , partly due to the increased absorption of CH 4 at solar wavelengths, which reduces the planetary albedo compared to H 2 .
Following the analysis in Ramirez (2017), a comparison of temperature-altitude profiles for a fully saturated 3 bar CO 2 early Mars atmosphere containing 1% CH 4 or 5% H 2 for a fully saturated 2 bar CO 2 atmosphere were also performed using the updated CIA from this work. As seen in Figures 3e and 3f, there is little change in the temperature profile in the upper atmosphere when using Wordsworth et al. (2017) CIA or scaled CIA from this work; however, at the surface there is ∼10 K difference in temperature.

Conclusions
This report details the first temperature-dependent experimental measurements of CO 2 -H 2 and CO 2 -CH 4 CIA. It was found that below 600 cm −1 , the experimentally derived CIA cross sections agree with the spectral range and temperature dependence of the calculations by Wordsworth et al. (2017), however, the amplitude is half of what was predicted. Furthermore, the CIA cross sections reported here agree within combined uncertainty with those measured by Turbet et al. (2020), strengthening our confidence in these results. Additionally, the temperature-dependent measurements agree with the new temperature-dependent theoretical calculations performed by Turbet et al. (2020), that the CIA strength is weaker than predicted at other temperatures as well.
GODIN ET AL. 8 of 10 10.1029/2019JE006357 First row: modeled surface temperature as a function of CO 2 partial pressure (log scale), at 1%, 5%, and 10% CH 4 (a) or H 2 (b). Second row: Mars' planetary albedo as a function of CO 2 partial pressure, at 1%, 5%, and 10% CH 4 (c) or H 2 (d). Third row: temperaturealtitude profiles for a fully saturated CO 2 early Mars atmosphere from Ramirez (2017)  With improved spectra, radiative transfer calculations of the early Mars atmosphere were performed and showed that CO 2 -CH 4 CIA is not as promising as initially argued by Wordsworth et al. (2017) for producing a warm and wet early Mars. CO 2 -H 2 , however, seems more promising, as liquid water is possible with 5% hydrogen and less than 2 bar of surface pressure. These results agree with similar climate calculations preformed by Turbet et al. (2020).

Data Availability Statement
Experimentally derived CIA cross sections and analysis codes are available from the York University Dataverse repository (Godin et al., 2020).