Draft version June 16, 2021 Typeset using LATEX twocolumn style in AASTeX63

A Large-scale Approach to Modelling Molecular Biosignatures: The Diatomics Thomas M. Cross,1 David M. Benoit,1 Marco Pignatari,1, 2, 3, 4 and Brad K. Gibson1

1E. A. Milne Centre for Astrophysics, Department of Physics and Mathematics, University of Hull, HU6 7RX, United Kingdom 2Konkoly Observatory, Research Centre for Astronomy and Earth Sciences, Hungarian Academy of Sciences, Konkoly Thege Miklos ut 15-17, H-1121 Budapest, Hungary 3NuGrid Collaboration, http:// nugridstars.org 4Joint Institute for Nuclear Astrophysics - Center for the Evolution of the Elements

Submitted to ApJ

ABSTRACT This work presents the first steps to modelling synthetic rovibrational spectra for all of astrophysical interest using the new code Prometheus. The goal is to create a new comprehensive source of first-principles molecular spectra, thus bridging the gap for missing data to help drive future high-resolution studies. Our primary application domain is on molecules identified as signatures of in planetary atmospheres (biosignatures). As a starting point, in this work we evaluate the accuracy of our method by studying the diatomics molecules H2,O2,N2 and CO, all of which have well-known spectra. Prometheus uses the Transition-Optimised Shifted Hermite (TOSH) theory to account for anharmonicity for the fundamental ν = 0 → ν = 1 band, along with thermal profile modeling for the rotational transitions. We present a novel new application of the TOSH theory with regards to rotational constants. Our results show that this method can achieve results that are a better approximation than the ones produced through the basic harmonic method. We discuss the current limitations of our method. In particular, we compare our results with high-resolution HITRAN spectral data. We find that modelling accuracy tends to diminish for rovibrational transition away from the band origin, thus highlighting the need for the theory to be further adapted.

Keywords: Biosignatures –

1. INTRODUCTION (1961), Be´cet al.(2016)), the basic data needed to de- In the past decades the search of the origin of life in tect most biomolecules is incomplete. Existing data was the Universe and the detection of chemical signatures gathered through meticulous experiments and compu- of life and of its main building blocks has been driving tations and built from the bottom up to cover about remarkable scientific achievements in astronomy and as- a hundred of molecules and their ro-vibrational spec- trobiology over decades (see e.g., Des Marais et al. 2008). tra (Tennyson et al. 2020). The molecular data has A fundamental step to answer some of those questions is been carefully curated into well-known databases in the the observation of biologically-relevant molecules in the field such as HITRAN (Gordon et al. 2017), ExoMol Universe. However, apart from extremely well studied (Tennyson et al. 2020), CDMS (Endres et al. 2016), arXiv:2106.07647v1 [physics.chem-ph] 14 Jun 2021 molecules such as (Viti et al.(1997), Polyansky JPL (Pearson et al. 2005), for example. Today those et al.(2018)), (Yurchenko et al.(2009), Coles databases are fundamental resources for molecular line et al.(2019)), (Brown et al.(2013), Yurchenko detection and form the backbone of astrochemistry and & Tennyson(2014)) and (Falk & Whalley research using data from present observa- tion missions like ALMA (Harada et al. 2018), HUB- BLE (Evans et al. 2018; Damiano et al. 2017), TESS Corresponding author: Thomas M. Cross (Col´onet al. 2020) and from future missions like PLATO [email protected] (Katyal et al. 2019), JWST (Barstow et al. 2015) and Ariel (e.g., Tinetti et al. 2018), to name a few. 2 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

Molecular lines are ideal observation targets to study with each additional non-H . Moreover, smaller different astrophysical sources. Fundamental informa- molecules are more likely to be found in gaseous form in tion about the properties of the early Universe can be planetary atmospheres. collected from integrated galactic spectra at different The All Small Molecules (ASM) catalogue described redshifts (e.g., Muller et al. 2006; Costagliola et al. 2011; by Seager et al.(2016) contains over 14,000 biosignature Aladro et al. 2015; Zhang et al. 2018), or from a num- molecules that comply with the criteria discussed above. ber of local sources (e.g., Bacmann For ease throughout this paper we will refer exclusively et al. 2012; Rivilla et al. 2016; Shimajiri et al. 2017), to the biosignature portion of this catalogue as the SBP, from stars (e.g., Yong et al. 2003; Hedrosa et al. 2013) an acronym which is simply made from the first letter and from both planets (e.g., Greaves et al. 2020; Web- from each of its author’s last names, as ASM also details ster et al. 2015; Tran et al. 2006) and exoplanets (e.g., non-biogenic molecules. Swain et al. 2009; Tinetti et al. 2013; Tessenyi et al. In order to be able to detect molecules in the SBP cat- 2013; Guilluy et al. 2019). alogue, there is a need for the laboratory astrophysics The evolution of in the Universe has cap- community to characterise the spectral features of each tured the increase of complexity from a metal-free en- entry. Despite the simple nature of the entries in the vironment (e.g., Galli & Palla 1998) to a metal-, dust- catalogue, and considering only the rovibrational por- and ice-rich environment (e.g., Wakelam & Herbst 2008; tion of the spectrum, there are thousands of biosig- Morales et al. 1998). The stellar production of elements, natures within SBP that have incorrect, incomplete or such as and iron from the first genera- completely unknown spectra (Sousa-Silva et al. 2019). tion of core-collapse supernovae (e.g., Rauscher et al. There is therefore a pressing need for vast quanti- 2002; Nomoto et al. 2013; Sukhbold et al. 2016; Rit- ties of spectral signatures to be characterised, as shown ter et al. 2018) and and from low-mass by the very recent spectroscopic discovery of stars (e.g., Karakas & Lattanzio 2014; Cristallo et al. on Venus by Greaves et al.(2020). Not only is this 2015; Pignatari et al. 2016), formed the building blocks believed to be a biosignature but it was de- of bio-molecules that are observed today. tected in such quantities that it cannot be currently Within this same context, the search for molecular fin- explained abiotically Bains et al.(2020). One could gerprints of life in atmospheres of exoplanets represents argue that the spectroscopic analysis which discovered a fundamental goal of astrobiology. A biosignature phosphine would have not been possible without the ef- is defined as a gas that is produced by life and accu- forts to produce accurate and complete computed line mulates in a planet’s atmosphere. An ideal biosignature lists, ranging from room temperature (Sousa-Silva et al. would be unambiguous with living organisms being its (2013) to up to 1500K (Sousa-Silva et al.(2015)). Es- unique source (Greaves et al. 2020). In reality, many pecially when we consider that suspected inaccuracies biosignatures can be also produced through abiotic pro- found in regions of previous phosphine data may have cesses and can act as false positives. contributed to past misinterpretations of astronomical For a molecule to be classified as a biosignature, un- spectra (Malathy Devi et al.(2014)). der the framework of Seager et al.(2016), the molecule Following this discovery, (Zapata Trujillo et al. 2021) needs to fulfill a series of criteria. Here we will de- noted the importance of being able to detect spectro- tail what we believe to be the three main points, fur- scopically the presence of phosphorous bearing species. ther classifications can be found in the paper referenced. They enumerated a list of phosphorous bearing species, The first main criterion is molecular stability, identify- which could potentially be detected in planetary atmo- ing compounds that are stable on the order of days as a spheres, and compiled all available spectral data. As pure compound at standard conditions and if they are expected the data was scarce, therefore they used a stable to reactions with water. The second main crite- high-throughput computational algorithm, RASCALL rion is volatility: the likelihood a molecule would be in (Sousa-Silva et al. 2019), to produce approximate spec- gaseous form at standard conditions. Volatility is diffi- tra to fill in the gaps. cult to assess so Seager et al.(2016) used boiling points Generating an entire set of SBP spectra for a given instead. A molecule with a boiling below 150◦c was frequency range is a huge task. Indeed, experimentally it considered sufficiently volatile. The third main criterion would be hard to record spectra for the entire catalogue focuses on molecular size. Only molecules of up to six without a concerted worldwide effort. non- were considered during the search. Alternatively, a significant part of the SBP catalogue This criterion was used to limit the number of possi- could be explored computationally. This is our current ble molecules as their number increases exponentially approach for this study, which focuses on the rovibra- Modelling Biosignatures: Diatomics 3

Primarily we are trying to address the gap between ap- proximate, fundamental only models (such as harmonic, SBP 25 HITRAN non-adapted TOSH or RASCALL), and labour inten- ExoMol sive, extremely accurate, line-lists (such as ExoMol and 20 HITRAN). This is done by developing a method which borrows key physics from both approaches to create an 15 intermediate approach. We present here a model that

10 approximates the fundamental (ν = 0 → ν = 1) band

Number of Molecules with a rotational profile computed at 300K. The inten- 5 sities for the spectra use thermal population equations, and as such have an additional ability of modelling rovi- 0 Diatomic Triatomic Tetratomic Pentatomic Atomic Type brational lines at differing temperatures. The investiga- tion into the effect of differing temperatures on spectra Figure 1. Comparing the availability of data from HITRAN is not considered in this paper. and ExoMol with the molecules needed for the SBP. Only up Additionally our approach is designed to be simple, to pentatomic molecules has been included here. open-source, and computationally cheap, but still more detailed than the aforementioned approximate funda- tional spectra of diatomic molecules in SBP. Nonetheless mental only methods. For example, we do not cur- the envisaged task requires a reliable means of produc- rently model effects such as external electric or magnetic ing good quality spectral data at reasonably low com- fields like a program such as PGOPHER (Western 2017) putational cost so that it can later be easily extended to would, to keep simplicity and low computational cost. produce spectral signatures for the rest of the catalogue. Once the catalogue is constructed it should, and will, The current publicly-available version of HITRAN act as a living document, preferably being continually contains detailed spectra for 49 molecules and also updated with better results. The rationale for this ap- atomic oxygen. A further explanation of HITRAN is proach is to provide a wide coverage of the SBP database provided in section3. From the 49 molecules only 26 are to help with both detection and atmospheric models actually believed to be biosignatures according to SBP. with a reasonable accuracy that could be improved with To break this down even further, Figure1 compares the further releases of the database. number of biogenic molecules for each total number of In this study we have applied our work to the diatomic atoms, up to pentatomics. Rounding the SBP down to biosignature molecules, of which there are three in the exactly 14,000 biosignatures, gives HITRAN a comple- SBP catalogue. We have also additionally included a tion rate of roughly 0.19%. fourth diatomic, (CO), for reasons dis- The current version of ExoMol (see also further back- cussed in section 3.4. ground in section3) contains high-quality spectra at a The paper is organised in the following manner. In large range of temperatures for 81 molecules, however section2 we introduce the methodology and theory re- only 18 of these molecules are biogenic. We report in quired to produce the synthetic data. The following sec- Fig.1 a similar breakdown to the one done for HITRAN. tion, section3, presents and then discusses our results. Here we see that, if the SBP is rounded down, ExoMol We conclude our findings in section4. has a completion rate of approximately 0.13%. Combining both databases gives a maximum total of 2. METHODOLOGY 26 molecules (excluding SBP entries overlaps). Assum- 2.1. Requirements ing a constant rate of growth for these databases, with- When modelling the rovibrational spectrum, we con- out significant change to available technology and re- sidered two key parameters: the band origin and the sources, it will be difficult to produce spectra for even rotational constants. The band origin crucially deter- 1% of SBP, let alone the entire catalogue. mines the position of its corresponding rotational lines The work described here provides a new approach that and thus has a strong influence on the overall shape of is meant to complement and support the high-resolution the rovibrational band. In our model, we have assumed databases mentioned above. We aim to obtain spectra this aspect to be the most important. from first principles, but faster than existing approaches The rotational constants have an effect on the spacing (such as VSCF/VCI shown by Clark & Benoit(2020) or of the transitions on each of the rotational branches. As DVR shown by McKemmish et al.(2019) for example) a secondary effect, anharmonicity causes the rotational possibly at the expense of accuracy. constant to depend on the vibrational state. Indeed, the 4 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson √ bond lengthening in excited vibrational states allows a weighted displacement coordinate, Q = µx, can be Q-branch progression (a set of purely vibrational tran- expressed as: sitions with no rotational transitions, see eq. 28), since the rotational constants of both starting and final vi- 1 ∂2 1 1 1 brational states are different. An identical rotational Hˆ = − + η Q2 + η Q3 + η Q4 (1) 2 ∂Q2 2! ii 3! iii 4! iiii constant for all vibrational level (such as predicted by a simple harmonic oscillator approximation, for exam- ple) incorrectly suggests that vibrations and rotations Now imagine a shift, σ along the coordinate Q. If the are independent. This leads to the Q branch transitions centre of the wave function is shifted by σ, the shape bunching up at a single position, rather than display- will remain the same but the anharmonic correction can ing a typical progression. Therefore, for an accurate be incorporated into the wave function. approximation for the entirety of the spectrum, the ro- The shifted wave functions for TOSH, are now differ- tational constant will need to vary with vibrational level ent to the harmonic wave functions, and are described (Paw lowski et al. 2002). by: The transition-optimised shifted Hermite (TOSH)  1/2 1/2 ω − ω (Q−σ)2 h 1/2i theory is a modern approach to treating the nuclear ψ = e 2 H (Q − σ)ω (2) n π1/22nn! n wave function that does predict correctly most band origins Lin et al.(2008). TOSH is effectively a simpli- where Hn(x) naturally refers to the Hermite polyno- fication of second-order vibrational perturbation theory mial, and ω describes the harmonic frequency (defined (VPT2), and applies limited-order perturbation theory using TOSH constants in equation 13, see below). to the nuclear Schr¨odingerequation. By doing this, it The energy of this ground vibrational state is: avoids any degeneracy issue that may arise from stan- ˆ dard second order vibrational perturbation theory ( Wil- E0 = hψ0|H|ψ0i (3) letts et al.(1990) and V´azquez& Stanton(2006)), for example. 1 1  1  1  3σ  Fundamentally TOSH works by introducing a shifting E = ω + η + σ2 + η + σ3 0 4 2! ii 2ω 3! iii 2ω parameter (σ) which is used to shift the harmonic basis (4)  2  functions from their equilibrium position. This shift is 1 3 3σ 4 + ηiiii 2 + + σ optimised for the vibrational transition energy expan- 4! 4ω ω sion, specifically for the fundamental vibrational transi- The equation for the energy of the first excited vibra- tion, and as such does an excellent job at correcting the tional level is: band origins (this is shown by our results in section3 E1 = hψ1|Hˆ |ψ1i (5) and Table1). An expression for the magnitude of the shift is described in equations2-10 in section 2.2. We     3 1 3 2 1 9σ 3 have discovered however that this shift can be applied E1 = ω + + σ ηii + ηiii + σ to the equilibrium distance to recover the anharmonic 4 2! 2ω 3! 2ω (6)  2  effects of the first vibrationally excited level (see 2.6). 1 15 9σ 4 + ηiiii + + σ Despite its approximate nature, TOSH is roughly a 4! 4ω2 ω factor of two cheaper than VPT2, with only some loss Therefore the energy difference between the first vi- in accuracy as a compromise (Lin et al. 2008). This brational state and the ground state is: approach also has the potential to be of reasonable ac- TOSH ∗ ˆ ∗ ˆ curacy for a large ensemble of molecule (Hanson-Heine ∆E = hψ1 |H|ψ1i − hψ0 |H|ψ0i (7) 2019) and is straightforward to extend to larger systems 2 ηiiii ηiiiσ ηiiiiσ (Lin et al. 2008). = ω + + + (8) 8ω2 2ω 4ω Which can now be compared with the energy from unshifted wave function obtained through second-order 2.2. TOSH Theory perturbation theory, VPT2. We will now briefly discuss the necessary theory 2 2 required to understand the process to produce the ηiiii 5η η ∆EVPT 2 = ω + − iii − iiii (9) Prometheus spectra. 8ω2 24ω4 32ω4 For a , the Hamiltonian with up to Within the TOSH theory derivations, σ is assumed to fourth order terms (quartic potential) and using mass- be small and therefore the σ2 term can be neglected. By Modelling Biosignatures: Diatomics 5 comparing the coefficient of ηiii in both TOSH (8) and The quartic fit applied to the PEC is described by VPT2 (9) expressions, a suitable value for the shift can equation: be obtained: 1 2 1 3 1 4 5 ηiii V = E + ζ X + ζ X + ζ X (11) σ = − (10) 0 2! ii 3! iii 4! iiii 12 ω3 Where X = r − r is a displacement coordinate, with Due to its derivation, this value of the shift parame- e r describing the equilibrium bond length. Note: E ter is only optimal for the 0 → 1 transition. It is also e 0 has been included to help Prometheus with the fitting worth noting that the TOSH paper does not match the by centering around 0, and has little importance beyond η terms for the VPT2 energy expression, but reme- iiii this. died this by stating this term is often neglected anyways. The fit only returns ζ constants, ζ , ζ and ζ . Many intermediate processes for the derivations above ii iii iiii Which need to be massed weighted for them to become have been assumed and therefore omitted. For a full the TOSH constants, as described in the Hamiltonian in derivation, please see the appendix, section A.1. equation1. This was done in the following manner: One lesser considered aspect of having a displaced vi- brational wave function (mimicking what happens for ζii ζiii ζiiii η = , η = , η = (12) the exact wave function) is the usage the shift parameter ii µ iii µ3/2 iiii µ2 in order to include anharmonic effects in the computa- Note that fits of order higher than quartic lead to tion of rotational constants. We explore this possibility sizable errors and inaccurate results. in section 2.6. Additionally, for PEC datasets containing only few points at the lower inclusion ranges, we used spline in- 2.3. Potential Energy Curve, or PEC terpolation to ensure the data spans the entire range The TOSH approach (and our Prometheus implemen- specified. tation, see (2.8)) requires a potential energy curve for the chosen diatomic. Once this is inputted into the code, 2.4. Choice of a PEC for diatomics a quartic fit is applied. We use this methodology as his- In order to avoid issues originating from PEC of sub- torically quartic force fields have been used for defining spectroscopic quality, we choose to use only data that the potential of the inter-nuclear Hamiltonian. In par- has been validated through spectroscopic comparisons. ticular it is used as a tool for analyzing and producing Indeed, while a number of PEC for diatomics are avail- rovibrational spectra for molecules of interest to astro- able in the literature, some of the molecules in our physical observation (Fortenberry & Lee 2019). selected set have a particularly challenging electronic The quality of the PEC is paramount and will af- structure (N2 and O2). All the potentials used within fect the accuracy of the TOSH constants derived from this study are for the ground state of their respective our quartic fit. The code performs a local fit, centered molecule. around the equilibrium bond length, rather than a global This in turn affects the quality of the rovibrational fit which would include the potential wall. We found spectra that can be predicted. As our study focuses on that a global fit is more likely to lead to inaccurate the quality of the ro-vibrational approach, rather than results. The local fit is determined via an “inclusion the quality of the electronic structure approach used for range”. The ranges are determined by first taking the the potential, we choose mostly PECs obtained by RKR equilibrium bond length. From there a series of differing inversion (Rees 1947), apart from H2 where near-exact percentages of the equilibrium bond length is selected. first-principles results already exist (see below). This The ranges tested within the code span from ±20% to latter choice provides an example of using first-principles ±7% of the equilibrium bond length, in intervals of 1%. PEC for our approach. The fit accuracy has been tested for each range us- The potential energy curves used are those from Ko- 1 + ing a simple chi-square test. The range with the lowest los & Wolniewicz(1964) for the Σg state of H2; the chi-square value, and hence the highest correlation with RKR potential energy curve for 1Σ+ of CO from Pardo the reference data, is then selected and used to ulti- (1986), with equilibrium bond length from Herzberg & mately produce a rovibrational spectrum. This method Huber(1950) and the RKR potential energy curves for 3 − 1 + has been selected as a means to ensure a consistent ar- the Σg state of O2 and the Σg state of N2 described gument for the selected inclusion range applied to the in Krupenie(1972) and Lofthus & Krupenie(1977), re- potential energy curve or each molecule. It also removes spectively. any bias we could have introduced into the analysis by The operation of extrapolating PECs from external selecting the ranges ourselves. sources, as opposed to producing our own, is solely for 6 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

T H this initial investigation into modelling diatomics and Where ψn and ψn , represent the TOSH (equation2) will be altered for larger molecules. This is due to this and harmonic wave functions, respectively. current approach limiting our ability to model to only T T H H H H molecules which have pre-existing PECs. hψn |x|ψn i = hψn (u)|u|ψn (u)i + hψn (u)|σ|ψn (u)i (17) 2.5. Spectroscopic Constants The first expectation value is the position expectation The TOSH framework allows us to determine key con- position for a harmonic wave function (i.e. zero) and stants, such as the harmonic frequency, ω, the anhar- thus we can see that the expectation of the position for TOSH monic fundamental transition, ∆E or ν0→1, and TOSH is simply: the coordinate shift, σ. T T The value of ω is calculated using: hψn |x|ψn i = σ (18)

1/2 ω = (ηii) (13) The full derivation, for both required levels, can be found in the appendix section A.2. The value of ∆ETOSH is obtained from eq.8, and Thus the TOSH model leads to r = r + σ. As is the the shift, σ, is given by eq. 10. The associated errors for v e case for the harmonic model, the TOSH approach only each spectroscopic constant is also calculated to allow an produces a single static value for all rotational constants assessment of the error on the positions of transitions in (see section A.2). the spectra produced by Prometheus (see (2.8)). We cannot accurately reproduce rotational constants 2.6. Rotational Constants that exhibit rotation-vibration interaction, using solely either a TOSH or a harmonic model: a hybrid theory is The state-specific rotational constant, B , for a di- v required. The approach we use in Prometheus, approx- atomic molecule is defined as: imates the ground-state rotational constant using the 2 h harmonic value. The first excited vibrational level uses Bv = 2 2 (14) 8π cµrv the modified TOSH bond length derived above, where Where v refers to the vibrational level, µ is the reduced sigma is combined with the equilibrium bond length. The vibrationally-averaged bond distances are thus ex- mass and rv is a bond length that depends on the vi- brational energy level considered. We can see that the pressed as follows: bond constant has a reciprocal squared relationship to r = r (19) the bond length, meaning as length increases the con- 0 e stant will decrease. This is one of the fundamental equa- r1 = re + σ (20) tions used throughout Prometheus to ultimately create where σ is not mass weighted, to harmonise units. the rovibrational spectra. The harmonic approach effectively costs nothing (us- In the harmonic approximation, r is independent of v ing r = r ), so can be exploited to provide the ground the vibrational energy level and thus r = r , the equi- 0 e v e state rotational constant with little effort. Our ap- librium bond length. Consequently, in that approxima- proach does not necessarily provide a strong quantita- tion B = B , a fixed rotational constant obtained from v e tive agreement but is qualitatively correct - bond length r . More generally, the value of the bond length for e gets larger with increased vibration, hence the rotational the rotational constant for each vibrational state can constant changes (gets smaller). be obtained from the expectation value of the position, r = hψ |r|ψ i. Using this expression for the harmonic v n n 2.7. Rovibrational Spectra model, leads to the same conclusions as earlier: rv = re, since the harmonic wave function is symmetric around In order to generate a ro-vibrational linelist and corre- sponding spectrum, we compute the maximum allowed re. If we now consider the TOSH model, the shifted quantum rotational number J for a given input temper- TOSH position, x, can be described using the harmonic ature. In the present study, the temperature was set to position, u, and the shift constant, σ: 300K to match the other data sets. The spectra from each of the experimental databases are all set at 300K x = u + σ (15) also. Rotation symmetry also influences the spectrum the As shown earlier, we can take the expectation value of molecule will produce and therefore conditional argu- the position, but this time implementing TOSH: ments within the intensity calculations have been cre- T T H H hψn |x|ψn i = hψn (u)|u + σ|ψn (u)i (16) ated to account for this. If a diatomic molecule has a Modelling Biosignatures: Diatomics 7 center of symmetry, hence meaning it is homonuclear, Where ν0→1 represents the band origin, B0 and B1 it is Raman active. This occurs instead of IR activity represent the rotational constants for their appropriate due to having no dipole moment, instead stretching and vibrational levels. As previously mentioned, for our code contraction of the bond leads to changes in the polariz- we use the solution calculated by TOSH (equation8) as ability. This type of molecular response implies different the band origin. transition selection rules leading to the O, Q, S branches, Here is a reminder of the labelling for the line series: rather than the typical P and R branches in infrared. ∆J = −2 (O Branch), ∆J = −1 (P Branch), ∆J = 0 Nuclear spin also has an effect on the spectra pro- (Q Branch), ∆J = +1 (R Branch) and ∆J = +2 (S duced. This will be discussed further in section3. The Branch). code is automated to account for additional effects by Finally, the code combines the intensity calculations using the initial inputs of the mass of a molecule. with the transition positions to produce the synthetic The code employs a Boltzmann distribution to obtain spectra. the relative intensities for each transition. This statis- tical law states that, for a system of N total molecules, 2.8. Prometheus NJ only a fraction N will occupy particular energy level Prometheus is the name given to our code, which is EJ a fraction. This can be written as: written in Python. It is maintained by the authors and

 −EJ  is available on the Milne Centre Github ( https://github. N g e kT J = J (21) com/Milne-Centre/Prometheus). N f As previously mentioned it is designed to be an ap- proximate method of simulating spectra, potentially Where, the degeneracy, g , the energy of a given level, i trading off accuracy for simplicity and speed. The E , and the partition function, f are described as: J main idea behind this code is to offer a new com- plementary approach to modelling the vast amount of gJ = (2J + 1) (22) molecules which have astrophysical and astrobiological importance.

EJ = B0J(J + 1) (23) Our analysis may have applications beyond the SBP catalogue, as we have done ourselves within this work

 −EJ  by modelling carbon monoxide. X kT f = gJ e (24) Prometheus primarily implements the TOSH theory Hence, the full equation now becomes: detailed in section 2.2 and makes use of the shift parame- ter, σ to calculate band origins and provide anharmonic −BJ(J+1) N (2J + 1)e( kT ) corrections to spectroscopic constants. To our knowl- J = (25) N f edge this is the first time it has been used to provide anharmonic corrections to rotational constants, rather From here the code calculates the transition positions than purely being used for anharmonically corrected vi- using the appropriate equations depending on the rota- brational frequency calculations. tional branches considered. The full derivations for these Prometheus then produces a stick spectrum (without equations have been included in the appendix, section any line broadening effects) using the constants it has A.3. They are summarised as follows: determined.

0 0 0 0 νO = ν0→1 + B1J (J + 1) − B0(J + 2)(J + 3) (26) 3. RESULTS AND DISCUSSION The spectra Prometheus produces can be compared to ν = ν − (B + B )(J 0 + 1) + (B − B )(J 0 + 1)2 P 0→1 1 0 1 0 different sources to evaluate the accuracy of our method. (27) This doesn’t necessarily mean that if the transitions line up well to another spectrum that Prometheus is “cor- ν = ν − J 0(J 0 + 1)(B − B ) (28) Q 0→1 1 0 rect”, rather that it has a considerable accuracy to one of the mainstream methods. ν = ν + (B + B )(J 00 + 1) + (B − B )(J 00 + 1)2 R 0→1 1 0 1 0 When producing spectra our program can either use (29) purely TOSH, purely harmonic or a mixture of both sets of constants (such as the hybrid approach which we 00 00 00 00 νS = ν0→1 + B1(J + 2)(J + 3) − B0J (J + 2) use within this study, detailed further in section 2.6). (30) Additionally, Prometheus is capable of using any other 8 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson experimental/literature values for the spectroscopic con- et al.(2016) and Chubb et al.(2021). The web- stants. A caveat however is that although the exper- site: http://www.exomol.com/ is the source of the imental/literature results may be more accurate, they ExoMol data used for comparisons. Like the HI- are still bound to our code’s capabilities and don’t in- TRAN we have normalised the intensities of the clude all the effects that ExoMol or HITRAN models, lines, to allow comparisons in intensities. for example. • Harmonic: This is simply using Prometheus con- Within this paper we will be only considering the data stants, calculated from the potential energy curve from HITRAN and ExoMol, and not RASCALL. This is (PEC), without any anharmonic corrections. For because, as previously mentioned, the spectra produced example, the band origin is the harmonic fre- by RASCALL 1.0 focus mainly on approximate band quency, ω, instead of using equation7 for the ori- centres (Sousa-Silva et al.(2019)) rather than rotational gin. This data is meant to highlight how harmonic structure. methods, whilst simple and easy to do, is typically • Literature: To locate high quality spectroscopic the weakest in accuracy and therefore there is a constants, the NIST (National Institute of Stan- need for anharmonic corrections. dards and Technology) Chemistry WebBook was The comparison spectra that are available (ExoMol used (P. J. Linstrom and W. G. Mallard(2003)). or HITRAN or both) have been inverted on the figures Once the diatomic was located, we browsed the below. This is merely an aesthetic choice to prevent the “Constants of diatomic molecules”. Here we then figures from becoming difficult to interpret, by having found the electronic state required (ground) and up to 4 sets of data layered upon each other. the corresponding constants. We then selected the We will briefly outline the colour scheme used appropriate paper referenced, and used the spec- throughout the spectra. Prometheus is designated as troscopic constants calculated by said paper. We red, and the spectra produced by literature constants have included these results to highlight the the- via Prometheus as blue. HITRAN 2016 spectra is de- oretical “best” that the code we are using could noted by the colour black whereas ExoMol spectra is the produce. For the discussion of each diatomic, the colour purple. Finally the harmonic spectra is green. appropriate paper has been stated and annotated The computed error on the line positions for the spec- on the results. tra generate by Prometheus are typically less than 0.1 wavenumbers, except H2 which is less than 4 wavenum- • HITRAN2016: data from HITRAN’s 2016 release. bers. We did not include error bars on each line position HITRAN is an acronym for high-resolution trans- in the figures for clarity. mission molecular absorption database and is a 3.1. Molecular Hydrogen compilation of spectroscopic parameters that a va- riety of computer codes use to predict and simu- The first biosignature molecule is molecular hydrogen. late the transmission and emission of light in the H2 is the most abundant molecule in the Universe by or- atmosphere (Gordon et al. 2017). This data was ders of magnitude (Wakelam et al. 2017). It is difficult obtained from a SpectralCalc URL and read into to observe directly in the interstellar medium due to the code. This data is currently acting as the co- a lack of permanent electric dipole moment and most optimal result along with ExoMol. We normalised transitions are of quadrupolar nature (and extremely the intensities of the lines, to match the relative weak) (Herzberg 1949; Roueff et al. 2019), or in emis- intensities produced by Prometheus. Future work sion for selected environments (see (Habart et al. 2005) will potentially look at the intensities of the rovi- for a review). Indirect techniques, such as the amount brational spectrum - whether that be regarding of dust present using the gas-to-dust ratio (GDR) (Jo opacity functions or using units of atmospheric et al. 2017), also allow some degree of detection. See concentration rather than the current relative in- (Seager et al. 2020) for further discussion of possible H2 tensities. presence and detection on exoplanets. From a biology standpoint, many microorganisms produce hydrogen as • ExoMol: A database which provides high-accurate primary product through metabolism under anaerobic and complete linelists for application in hot as- conditions (Schlegel 1974). trophysical environments (Tennyson et al. 2020). For H2, no data for HITRAN2016 was available in The ExoMol data structure can be used to gener- SpectralCalc but accurate spectroscopic constants were ate lifetimes, cooling functions and partition func- obtained from Foltz et al.(1966). Fortunately, spectra tions. More details can be found in Tennyson was available from ExoMol (Roueff et al. 2019). Modelling Biosignatures: Diatomics 9

Band Origin Rotational Constant Rotational Constant −1 −1 −1 ν0→1 (cm ) B0 (cm ) B1 (cm ) Calculated ∆Lit. Calculated ∆Lit. Calculated ∆Lit.

H2 Foltz et al.(1966) 4161.18 — 59.3362 — 56.3712 — Harmonic 4407.03 ± 0.02 +245.85 60.8643 +1.5281 60.8643 +4.4931 TOSH 4179.39 ±2.84 +18.21 54.5241 ± 0.0006 −4.8121 54.5241 ±0.0006 −1.8471

O2 Fletcher & Rayside(1974) 1556.38 ± 0.01 — 1.4376 ± 0.0002 — 1.4218 ± 0.0002 — Harmonic 1580.13 +23.75 1.4456 +0.0080 1.4456 +0.0238 TOSH 1556.67 ±0.06 +0.29 1.4257 −0.0119 1.4257 +0.0039

N2 Bendtsen(1974) 2329.92 ± 0.02 — 1.9896 — 1.9722 — Harmonic 2360.83 ± 0.01 +30.91 1.9982 +0.0086 1.9982 +0.0260 TOSH 2329.62 ±0.10 −0.30 1.9749 −0.0147 1.9749 +0.0027 CO Mantz et al.(1975) 2143.24 — 1.9225 — 1.9050 — Harmonic 2170.54 ± 0.01 +27.30 1.9313 +0.0088 1.9313 +0.0263 TOSH 2143.22 ±0.13 −0.02 1.9080 −0.0145 1.9080 +0.0030 Table 1. A comparison of the spectroscopic constants between Prometheus (TOSH), the literature and the harmonic method. ∆Lit. represents the difference between the calculated value and the appropriate literature. The highlighted values represent the values which Prometheus uses to create spectra, with the band origin and upper rotational coming from TOSH theory and the ground level rotational constant coming from the harmonic approximation. The band origins are calculated to 2 decimal places. The rotational constants have been calculated to 4 decimal places. Errors have been included where available (some literature sources did not provide them) and other errors have been omitted if they were of a lower order than the rounding criterion of the data.

In Table1, Foltz et al.(1966) determined the band correct region by considering symmetry of the diatomic origin to be 4161.18 cm−1. The harmonic approach using the initial mass inputs. gives 4407.03 cm−1 which deviates from the literature As also mentioned by Banwell(1972), a molecule com- −1 −1 by 245.85 cm . TOSH, with a result of 4179.39 cm , posed of nuclei with non-zero nuclear spin (such as H2) manages to produce a band origin with a difference of will exhibit spectral lines that show an alternation in only 18.21 cm−1 — an order of magnitude less than the intensity. Unfortunately Prometheus does not currently harmonic approach. model this. Instead the intensities have the typical The experimental ground state rotational constant, Boltzmann distribution regardless of the nuclear spin B0, was determined by Foltz et al.(1966) to be of the molecule’s nuclei components. −1 59.3362 cm . The harmonic method approximates B0 better, with a value of 60.8643 cm−1, than TOSH, which O Branch Q Branch S Branch appears to drastically over-correct, leading to a value of 54.5241 cm−1. j f

, y t i

However, since the predicted rotational constants has s n e t n I

no vibrational dependence for TOSH or the harmonic d e z i l a m model, the reverse is true for B1. Foltz et al.(1966) de- r o termined 56.3712 cm−1, with TOSH and harmonic now N −1 −1 ExoMol differing by 1.8471 cm and 4.4931 cm , respectively. Prometheus Foltz et al. (1966) Figures2 and3, show the resulting simulated spectra 2500 3000 3500 4000 4500 5000 5500 6000 Wavenumber [cm 1] from Prometheus compared to ExoMol, data by Foltz et al.(1966) and the harmonic method. Figure 2. A comparison of the H2 spectra produced by Two major effects need to be considered when the Prometheus, ExoMol (Roueff et al. 2019) and Foltz et al. (1966) via Prometheus. The intensities for all sets of data rovibrational spectrum of H2 is analyzed. First, this homonuclear molecule exhibits a center of symmetry. As have been normalised and therefore are relative values. a consequence of this, it is Raman active and IR inactive with O, Q and S branches present in its spectra (Ban- A key aspect of Figure2, is how well all three band well(1972)). Prometheus will model a molecule in the origins align. Using spectroscopic constants of Foltz et al.(1966), Prometheus can quite accurately reproduce the ExoMol positions for the transitions. Prometheus 10 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

TOSH approch on its own however appears to slightly O Branch Q Branch S Branch overestimate the band origin by about 20 cm−1. This is likely due to the strong quantum nature of the hydro- j f

, y t i gen nuclei and thus this molecule displays strong anhar- s n e t n I

monic effects that are less well modeled by the TOSH ap- d e z i l a m proach. Regardless, Figure2 shows all three approaches r o N are in relatively good agreement with one another. Harm. Band Origin ExoMol Prometheus Regarding the Q-branch, Prometheus displays a larger Harmonic

2500 3000 3500 4000 4500 5000 5500 6000 spread of transitions than ExoMol and Foltz et al. Wavenumber [cm 1] (1966). The Prometheus central transitions potentially would roughly line up provided the band origin was cor- Figure 3. A comparison of the H2 spectra produced by rected. Once again Foltz et al.(1966) adequately models Prometheus, ExoMol (Roueff et al. 2019) and the Harmonic the ExoMol data. It does falter slightly with the distri- method. The intensities for all sets of data have been nor- bution of the branch, as the Foltz et al.(1966) Q branch malised and therefore are relative values. appears to model a greater spread in the transitions than ExoMol. Finally let’s review the Prometheus spectrum with re- The O branch, indicated by the labels on Figures3 gards to to the Harmonic as shown by Figure3. As dis- and2, shows an interesting result. Foltz et al.(1966) cussed earlier, it can be seen that Prometheus slightly appears to decrease slightly in accuracy the higher the overestimate the origin, whereas the harmonic method transition, whereas Prometheus does the same but the substantially overestimates it. Due to this, it is difficult decrease in accuracy is larger. This deterioration is to to comment on much of the harmonic spectrum, as it such a point it becomes difficult to match the ExoMol’s barely matches ExoMol. transitions to Prometheus. In this regard, we can see that despite Prometheus For the S branch (also indicated by the label on also having some difficulties with modelling H2, par- Figures3 and2), Prometheus seems to initially fare ticularly at higher transitions and on the O branch, it quite well. The Prometheus transition positions ap- still manages to do a better job than the Harmonic ap- pear to slightly increase in difference to the ExoMol data proach. Certainly, this shows anharmonic corrections with each higher transition. Possibly if the origins had are required to produce a qualitatively correct result. aligned better Prometheus would be able to continue to 3.2. Molecular Oxygen quite effectively model this branch at higher rotational The second biosignature is molecular oxygen. Oxygen transitions. Like before, Foltz et al.(1966) performs at is the third most abundant element in the universe, yet a satisfactory level of comparison to ExoMol, till the molecular oxygen is one of the most elusive molecules higher transitions are reached at which point it begins (Wang et al. 2020). This is often cited to be due to to falter. Some difficulty occurs with the comparisons O ’s high chemical reactivity and lack of electric dipole at higher transitions as the relative intensities for the 2 moment (Luspay-Kuti et al. 2018). Even now, a com- ExoMol data relegate very low intensities to them. The prehensive picture of oxygen chemistry in interstellar en- ExoMol S branch seems to have a far greater weight- vironments is still missing (Wang et al. 2020; Wakelam ing than the O, something which is not modelled in any et al. 2010; Agundez & Cernicharo 2006). other spectra. Oxygen is crucial to our understanding of most life on Earth with oxygenic photosynthesis being the dom- inant producing metabolism on our planet (Domagal- Goldman et al. 2014). Concerning exoplanets observa- tions, oxygen is a potential biosignature since it can be produced through photosynthesis processes. However, it is also a possible false positive since it can be formed in larger quantities through abiotic processes such as runaway greenhouse effects (Meadows 2017). Oxygen’s potential to act as a false positive biosigna- ture is rooted in the diversity of exoplanets. For exam- ple, on Earth there are no abiotic processes that would produce it in large abundance Meadows et al.(2018). It has been shown for a very different star and planetary Modelling Biosignatures: Diatomics 11

system, O2 could be generated in a planetary environ- O Branch Q Branch S Branch ment without life (Wordsworth & Pierrehumbert 2014). The literature spectroscopic constants for O2 were ob- j f

, y t i tained from Fletcher & Rayside(1974), and our values s n e t n I

are compared to data from the HITRAN 2016 release d e z i l a m as no ExoMol data was available in a format currently r o readable by Prometheus. N HITRAN 2016 Like H ,O has a centre of symmetry which causes Prometheus 2 2 Harmonic Harm. Band Origin 1350 1400 1450 1500 1550 1600 1650 1700 1750 it to be Raman active but IR inactive. This once again Wavenumber [cm 1] means that the O, Q and S branches are present in the spectra. In addition the symmetry also means the effects Figure 5. A comparison of the O2 spectra produced by of nuclear spin will be observed, but these effects differ Prometheus, harmonic method and HITRAN 2016 data. to that of the other homonuclear diatomic molecules. The intensities for all sets of data have been normalised and therefore are relative values.

When considering the band origins (see Table1), TOSH produces a better approximation to the litera- ture than the harmonic method, despite the points men- tioned previously. Banwell(1972) reports the band ori- −1 O Branch Q Branch S Branch gin to be at 1556.38 cm , and this is extremely well es- timated by TOSH, with a result of 1556.67 cm−1. The j f

harmonic approximation on the other hand displays a , y t i s −1 n e

t significant displacement, roughly 24 cm . We also note n I

d e z i

l again that the harmonic approximation provides a bet- a m r o N ter approximation of B0, with TOSH doing the same for HITRAN 2016 B1. Prometheus Fletcher and Rayside (1974) In Figure5, we can quite clearly see the shift in band 1350 1400 1450 1500 1550 1600 1650 1700 1750 Wavenumber [cm 1] origin has a significant effect on the harmonic approxi- mations ability to mimic the literature. Figure 4. A comparison of the O spectra produced by 2 Prometheus, by using the hybrid methodology, models Prometheus, Fletcher & Rayside(1974) via Prometheus and HITRAN 2016 data. The intensities for all sets of data have the Q branch well (in good agreement with Fletcher & been normalised and therefore are relative values. Rayside 1974). This is something the purely harmonic method (and naturally a purely TOSH approach) re- peatedly fails at due to the fixed rotational constants between levels. 3.3. Molecular Nitrogen The third diatomic biosignature is molecular nitrogen. Unlike the other diatomic molecules in this study, O2 This is a key species in cosmology and has been observed has a zero nuclear spin which has a different effect on in various galactic environments (Vangioni et al. 2018). the spectra than the alternating intensities effect. In- Although direct observation of N2 is often difficult, as it stead, every rotational level with an even value is absent lacks strong pure rotational or vibrational lines (Li et al. from the spectra. This presents itself as a spectra with 2013). seemingly large gaps between the rotational transitions Within our own solar system we find N2 present in var- Banwell(1972) (See Fig.4). ious planetary and satellites’ atmospheres; it accounts O2 exhibits a triplet structure for each of these remain- for 3.5% (Oyama et al. 1980) of the Venusian atmo- ing transition lines which arises from the splitting of the sphere, 78.1% (Cox 2002) of the terrestrial atmosphere, rotational levels. This is another effect that Prometheus 2.8% (Franz et al. 2017) of the Martian atmosphere, and does not currently model and needs to be taken into perhaps most significantly, 98.4% (Strobel & Shemansky consideration when evaluating spectra. However, the 1982) of Titan’s atmosphere. weighting of the relative intensities due to triplet split- From a biogenic perspective, N2 is an essential ingre- ting have been included into Prometheus, to allow for dient for the building blocks for life as we know it(Sproß ease of comparison. et al. 2018) since it is required, along with carbon and 12 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

, for the formation of nucleic acids and pro- O Branch Q Branch S Branch teins (Lammer et al. 2019). The spectroscopic constants were obtained from j f

, y t i Bendtsen(1974). Our N 2 results are compared to the s n e t n I

harmonic approximation, literature optimal results and d e z i l a m the data from the HITRAN 2016 release. The ExoMol r o data was not in a usable format for Prometheus to ex- N HITRAN 2016 Prometheus trapolate. Bendtsen (1974)

2100 2200 2300 2400 2500 Finally N2, like O2 and H2, has a center of symmetry Wavenumber [cm 1] and thus the spectrum is Raman active but IR inactive. As N2 is composed of nuclei of non-zero nuclear spin, Figure 6. A comparison of the N2 spectra produced by the HITRAN spectrum shows the characteristic alter- Prometheus, Bendtsen(1974) via Prometheus and HITRAN nation in intensities (again, not modelled here by our 2016 data. The intensities for all sets of data have been Prometheus approach). normalised and therefore are relative values. By looking at the results of Table1, we see that N 2 rotational constants follow the typical pattern of the pre- vious molecules within this study. Bendtsen(1974) de- A comparison with the harmonic spectrum is shown termined the first excited level rotational constant to be in Figure7. The harmonic band origin is displaced from 1.9722 cm−1 and the ground rotational constants to be the literature by roughly 30 wavenumbers (see Table1). 1.9896 cm−1. The TOSH approach, with a fixed value of This effectively renders any comparisons of the transi- 1.9749 cm−1, over-estimates the ground state but pro- tion positions to the other spectra pointless. vides a good approximation for the first excited level. As was the case for H2 the relative intensities of N2 al- Whereas the harmonic approximation, with a constant ternate in weighting, an effect that Prometheus does not value of 1.9982 cm−1, achieves the opposite. currently model, once again causing the difficulties in drawing comparisons. The general intensities of the har- We can see in Table1 that for N 2, TOSH (2329.62 cm−1) once again provides a superior approxi- monic and the Bendtsen(1974) spectra have the same mation of the band origin than the harmonic approxima- problem, which is to expected, as it is produced via our tion (2360.83 cm−1), when comparing to the literature Prometheus methodology. value (2329.92 cm−1). In Figure7, we can draw compar- Comparing the relative intensities of Prometheus and isons between HITRAN, Prometheus and the Bendtsen harmonic to HITRAN does raise interesting points. The (1974) constants used in a Prometheus spectrum. As ex- “peak” for the O branch for HITRAN appears to oc- pected from using the TOSH band origin, Prometheus cur 10-20 wavenumbers away from Prometheus. Poten- provides an excellent approximation of both Bendtsen tially the alternating intensities is artificially causing the (1974) and HITRAN. “peak” to occur at a higher wavenumbers. The S branch For the lower rotational transitions of both the O and does not appear to be as exaggerated. As discussed ear- S branches, Prometheus does a satisfactory job of the lier, this discrepancy is not a major concern as intensities modelling the positions. In most cases it is lining up calculations are the not the focus of our present study. incredibly well with not only Bendtsen(1974) but also HITRAN data. As was the case for the other molecules this accuracy is lower for higher J transitions. O Branch Q Branch S Branch Prometheus does an excellent job at approximating j the Q branch for N and the relative spread of the transi- f 2 , y t i s n tions of this branch is comparable to the HITRAN data. e t n I

d e z i

For each increasing J transition the position is in- l a m r o crementally over-estimated, but not to the point where N the Prometheus spectrum is incorrectly matching lower HITRAN 2016 Prometheus number transitions to higher HITRAN ones. With this Harmonic Harm. Band Origin 2100 2200 2300 2400 2500 is mind, Prometheus is a very good approximation of Wavenumber [cm 1] both Bendtsen(1974) and HITRAN. Figure 7. A comparison of the N2 spectra produced by Prometheus, the harmonic method and HITRAN 2016 data. The intensities for all sets of data have been normalised and therefore are relative values. Modelling Biosignatures: Diatomics 13

We may conclude that for N2, Prometheus is pro- P Branch R Branch ducing a better approximate spectrum than the har- monic method. Like the other molecules seen in pre- j f

, y t i vious sections, Prometheus’ predictive power is reduced s n e t n I

at higher rotational transitions. Presumably the un- d e z i l a m corrected harmonic value for ground rotational constant r o N and the TOSH approximate values for the rotational HITRAN 2016 ExoMol Prometheus constant of the first excited level is having an effect on Harmonic Harm. P Branch Harm. R Branch 2050 2100 2150 2200 2250 the later transitions positions. Wavenumber [cm 1]

3.4. Carbon Monoxide Figure 9. A comparison of the CO spectra produced by Prometheus, the harmonic method, ExoMol (Li et al. 2015) The last molecule considered in this work is carbon and HITRAN 2016 data. The intensities for all sets of data monoxide. Unlike the previous molecules, CO has a per- have been normalised and therefore are relative values. manent dipole moment and therefore readily absorbs, even at the low temperatures found in molecular clouds has a difference of 27.30 cm−1 to Mantz et al.(1975). (Whitworth & Jaffa 2018). More relevantly, CO has al- It is also displaced from ExoMol/HITRAN by approx- ready been detected in an exoplanet’s atmosphere (HD imately 20-30 wavenumbers. Such a displacement nat- 189733, de Kok et al. 2013). urally renders the harmonic CO spectrum considerably Within a biogenic context, CO is often classified as an inaccurate when considering the positions of individual anti-biosignature. For example, for an inhabited planet, J transitions. it could be difficult for CO to accumulate in the atmo- sphere as it acts as an energy source for some microbes P Branch R Branch on Earth (Wang et al. 2016). Anthropologically, it can be formed instead when any organic substance is com- j f

, y t busted incompletely and therefore can be an indication i s n e t n I of not just life but intelligent life (Horner 2000). d e z i l a

The literature spectroscopic constants were obtained m r o N from Mantz et al.(1975). Our results for CO are com- HITRAN 2016 ExoMol pared to the harmonic approximation, optimal results Prometheus Mantz et al. (1975) and data from both HITRAN 2016 release and ExoMol’s 2050 2100 2150 2200 2250 Wavenumber [cm 1] Li2015 line list (Li et al. 2015). Carbon monoxide does not have a center symmetry Figure 8. A comparison of the CO spectra produced by therefore it is infra-red (IR) active and Raman inactive. Prometheus, Mantz et al.(1975) via Prometheus, ExoMol It is easily modelled by Prometheus to an accuracy akin (Li et al. 2015) and HITRAN 2016 data. The intensities to the HITRAN data at the lower transitions. Unlike for all sets of data have been normalised and therefore are the other biogenic diatomic molecules in this analysis, relative values. carbon monoxide solely demonstrates P and R branches, with no Q branch transitions. In general the distributions of the intensities between Mantz et al.(1975), as shown in Table1, provided HITRAN and Prometheus are similar. On closer in- −1 −1 1.9225 cm for B0 and 1.9050 cm for B1. As spection, only the fundamental transitions at lower ro- before, the harmonic approximation for B0 is better tational energy levels, J, are modelled with high accu- (1.9313 cm−1) than TOSH’s 1.9080 cm−1. For the up- racy. per rotational constant, B1, TOSH provides a rotational Indeed, most of the lower target transitions lie within constant closer to literature. a couple of wave numbers of Prometheus’ transitions, The band origin of the Prometheus spectrum (using but this starts to falter for the higher values of J. The TOSH), is 2143.22 cm−1, and easily lies within a single error for the higher transitions is then exaggerated due wavenumber of Mantz et al.(1975) (at 2143 .24 cm−1). to the slight position differences of the band origin. As shown by Figure9, even when compared to the HI- The relative intensities appear in a good agreement, TRAN/ExoMol spectra, Prometheus’ band origin still although it is worth mentioning that ExoMol is exhibit- lies within a few wavenumbers. ing some increased intensities for certain higher level The spectrum produced via the harmonic method transitions. This is not shown by either HITRAN2016 (Figure9) gives a band origin of 2170 .54 cm−1, which nor Prometheus data. 14 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

3.5. Comparisons of Constants As shown and discussed in section3, the TOSH As shown in Table1, the harmonic approximation con- method produces full fundamental rovibrational spectra sistently describes the ground level rotational constant at higher accuracy than the harmonic approximation for a slight increase in computational complexity. (B0) well but poorly estimates the upper rotational con- Provided the potential energy curve is of spectroscopic stant (B1). This makes realistically modelling of any Q branch impossible since it relies on a difference between quality, Prometheus can seemingly replicate a spectrum the ground and excited state rotational constant (see that is comparable to HITRAN at the lower transitions. Eq. 28 and Sec.2.7). Despite the breakdown in accuracy at higher transitions, The TOSH corrected rotational constants, rather than Prometheus still produces a satisfactory approximation offering the predicted “vibrationally averaged” value be- at extremely low computational cost (essentially only tween to the first and ground states, actually model the the cost of determining the necessary quartic constants: first excited state well (see Table1). Unfortunately, ηii, ηiii and ηiiii). the TOSH rotational constant over estimate the ground A major caveat is that the current theory for the shift state value, with the harmonic approach typically of- sigma does not accommodate for two differing vibra- fering a better approximation. Other studies have also tional constants for the upper and lower states. In- shown (e.g. Table 1 in Endres 1967) that the uncor- stead the lower value must be calculated via a harmonic rected harmonic approximation is a good approximation methodology, which in turn leads to similar inaccuracies for the ground state, whereas for the higher levels this in the transition positions. is not the case. A second issue is the Prometheus spectra do not cur- Both the ground and first excited level rotational re- rently model specific phenomena such as alteration in sults have been included in Table1 and discussed within line intensity due to nuclear spin and the splitting of ro- the previous results sections. This has been done to tational levels. This arguably is not a crucial element, highlight that using the current modified theory, where as the code is currently focussing on line position rather we amalgamate the TOSH and harmonic results, is the than intensities. In the case of effects due to nuclear best option to match reference spectra. The results used spin, the intensities variation is more of a cosmetic is- for Prometheus are highlighted in the table for each case. sue. The absent lines due to zero spin is arguably a Typically Prometheus is able to replicate the band ori- more important effect, and as such has been modelled gin within a single wavenumber using TOSH, with the by Prometheus. A recurring note for improvement throughout this pa- exception of H2. Note that H2 was the only molecule for which we did not use a spectroscopic RKR curve per has pertained to the shift variable sigma. It is a but a first-principle PEC instead. On the other hand, useful tool to use to account for anharmonic corrections Prometheus has shown it is capable of estimating the of the rotational constant, however in its current form band origins to a greater accuracy than that of the har- it is only suitable for the fundamental transition. To monic method, often with order of magnitudes improve- improve the spectra, we need to modify the theory to ments. obtain a shift that varies with vibrational level instead Whilst TOSH can approximate the upper rotational of being optimised for the fundamental, 0 → 1, vibra- constant better than the harmonic with the lower, both tional transition. still only do an adequate job at approximating the lit- Another aspect of further work would revolve around erature. Hence Prometheus, even by using a mixed se- creating a more robust means of evaluating the addi- lection of spectroscopic constants, struggle to achieve a tional effects arising from symmetry of a molecule. This high degree of accuracy. would include considering the phenomena arising due to This has the further effect of causing incorrect spacing properties such a spin and triplet structure. of transitions in the branches, which has been shown Finally, the next stage is to adapt the theory to tri- at the higher transitions in all spectra produced. We atomic molecules. Once triatomics have been success- can verify this conjecture by reviewing the equations 26- fully modelled we can then progress to modelling larger 30. Each of these position equations have a quadratic polyatomic molecules. dependence on the quantum number J, therefore any variation between the rotational constants (such as the ACKNOWLEDGEMENTS differences between the literature and Prometheus), will We would like to thank our colleagues in the Milne become more apparent at the higher transitions. Centre for all their input and contributions. We ac- knowledge the support of JINA-CEE (NSF Grant PHY- 4. CONCLUSIONS 1430152) and STFC (through the University of Hull’s Modelling Biosignatures: Diatomics 15

Consolidated Grant ST/R000840/1), and ongoing ac- “Lend¨ulet-2014”Programme of the Hungarian Academy cess to viper, the University of Hull High Performance of Sciences (Hungary), and the ERC Consolidator Grant Computing Facility. MP acknowledges support from the (Hungary) programme (RADIOSTAR, G.A. n. 724560).

REFERENCES

Agundez, M., & Cernicharo, J. 2006, The Astrophysical de Kok, R., Brogi, M., Snellen, I., et al. 2013, Astronomy & Journal, 650, 374, doi: 10.1086/506313 Astrophysics, 554, A82, Aladro, R., Mart´ın,S., Riquelme, D., et al. 2015, A&A, doi: 10.1051/0004-6361/201321381 579, A101, doi: 10.1051/0004-6361/201424918 Des Marais, D. J., Nuth, Joseph A., I., Allamandola, L. J., Bacmann, A., Taquet, V., Faure, A., Kahane, C., & et al. 2008, Astrobiology, 8, 715, Ceccarelli, C. 2012, A&A, 541, L12, doi: 10.1089/ast.2008.0819 doi: 10.1051/0004-6361/201219207 Domagal-Goldman, S. D., Segura, A., Claire, M. W., Bains, W., Petkowski, J. J., Seager, S., et al. 2020, arXiv Robinson, T. D., & Meadows, V. S. 2014, Astrophysical e-prints. https://arxiv.org/abs/2009.06499 Journal, 792, doi: 10.1088/0004-637X/792/2/90 Banwell, C. N. 1972, Fundamentals of molecular Endres, C. P., Schlemmer, S., Schilke, P., Stutzki, J., & [by] C. N. Banwell, 2nd edn. (McGraw-Hill M¨uller,H. S. P. 2016, Journal of Molecular Spectroscopy, London, New York), xii, 348 p. 327, 95, doi: 10.1016/j.jms.2016.03.005 Barstow, J. K., Aigrain, S., Irwin, P. G. J., Kendrew, S., & Endres, P. F. 1967, The Journal of Chemical Physics, 47, Fletcher, L. N. 2015, Monthly Notices of the Royal 798, doi: 10.1063/1.2140500 Astronomical Society, 448, 2546, Evans, T. M., Sing, D. K., Goyal, J. M., et al. 2018, AJ, doi: 10.1093/mnras/stv186 156, 283, doi: 10.3847/1538-3881/aaebff Be´c,K., Futami, Y., W´ojcik,M., & Ozaki, Y. 2016, Phys. Falk, M., & Whalley, E. 1961, The Journal of Chemical Chem. Chem. Phys., 18, doi: 10.1039/C6CP00924G Physics, 34, 1554, doi: 10.1063/1.1701044 Bendtsen, J. 1974, Journal of Raman Spectroscopy, 2, 133, Fletcher, W. H., & Rayside, J. S. 1974, Journal of Raman doi: 10.1002/jrs.1250020204 Spectroscopy, 2, 3, doi: 10.1002/jrs.1250020102 Brown, L., Sung, K., Benner, D., et al. 2013, \jqsrt, 130, Foltz, J., Rank, D., & Wiggins, T. 1966, Journal of 201, doi: 10.1016/j.jqsrt.2013.06.020 Molecular Spectroscopy, 21, 203, Chubb, K. L., Rocchetto, M., Yurchenko, S. N., et al. 2021, doi: 10.1016/0022-2852(66)90138-X Astronomy & Astrophysics, 646, A21, Fortenberry, R. C., & Lee, T. J. 2019, Computational doi: 10.1051/0004-6361/202038350 vibrational spectroscopy for the detection of molecules in Clark, V. H., & Benoit, D. M. 2020, Proceedings of the space, 1st edn., Vol. 15 (Elsevier B.V.), 173–202, International Astronomical Union, 468, doi: 10.1016/bs.arcc.2019.08.006 doi: 10.1017/S174392131900944X Franz, H. B., Trainer, M. G., Malespin, C. A., et al. 2017, Coles, P. A., Yurchenko, S. N., & Tennyson, J. 2019, Planetary and Space Science, 138, 44, Monthly Notices of the Royal Astronomical Society, 490, doi: 10.1016/j.pss.2017.01.014 4638, doi: 10.1093/mnras/stz2778 Galli, D., & Palla, F. 1998, A&A, 335, 403. Col´on,K. D., Kreidberg, L., Welbanks, L., et al. 2020, AJ, https://arxiv.org/abs/astro-ph/9803315 160, 280, doi: 10.3847/1538-3881/abc1e9 Gordon, I. E., Rothman, L. S., Hill, C., et al. 2017, Journal Costagliola, F., Aalto, S., Rodriguez, M. I., et al. 2011, of Quantitative Spectroscopy and Radiative Transfer, A&A, 528, A30, doi: 10.1051/0004-6361/201015628 203, 3, doi: 10.1016/j.jqsrt.2017.06.038 Cox, A. N., ed. 2002, Allen’s Astrophysical Quantities (New Greaves, J. S., Richards, A. M. S., Bains, W., et al. 2020, York, NY: Springer New York), Nature Astronomy, doi: 10.1038/s41550-020-1174-4 doi: 10.1007/978-1-4612-1186-0 Greaves, J. S., Richards, A. M., Bains, W., et al. 2020, Cristallo, S., Straniero, O., Piersanti, L., & Gobrecht, D. Nature Astronomy, doi: 10.1038/s41550-020-1174-4 2015, ApJS, 219, 40, doi: 10.1088/0067-0049/219/2/40 Guilluy, G., Sozzetti, A., Brogi, M., et al. 2019, arXiv, 107, Damiano, M., Morello, G., Tsiaras, A., Zingales, T., & 1 Tinetti, G. 2017, The Astronomical Journal, 154, 39, Habart, E., Walmsley, M., Verstraete, L., et al. 2005, Space doi: 10.3847/1538-3881/aa738b Science Reviews, 119, 71, doi: 10.1007/s11214-005-8062-1 16 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

Hanson-Heine, M. W. D. 2019, The Journal of Physical Meadows, V. S. 2017, Astrobiology, 17, 1022, Chemistry A, 123, 9800, doi: 10.1021/acs.jpca.9b07886 doi: 10.1089/ast.2016.1578 Harada, N., Sakamoto, K., Mart´ın,S., et al. 2018, The Meadows, V. S., Reinhard, C. T., Arney, G. N., et al. 2018, Astrophysical Journal, 855, 49, Astrobiology, 18, 630, doi: 10.1089/ast.2017.1727 doi: 10.3847/1538-4357/aaaa70 Morales, R., Espinosa, G., Borrell, Y. J., et al. 1998, Hedrosa, R. P., Abia, C., Busso, M., et al. 2013, ApJL, 768, Biotecnologia Aplicada, 15, 149 L11, doi: 10.1088/2041-8205/768/1/L11 Muller, S., Gu´elin,M., Dumke, M., Lucas, R., & Combes, Herzberg, G. 1949, Nature, 163, 170, doi: 10.1038/163170a0 F. 2006, A&A, 458, 417, doi: 10.1051/0004-6361:20065187 Herzberg, G., & Huber, K.-P. 1950, Molecular spectra and Nomoto, K., Kobayashi, C., & Tominaga, N. 2013, molecular structure, (New York: Van Nostrand), ARA&A, 51, 457, doi: https://doi.org/10.1007/978-1-4757-0961-2 doi: 10.1146/annurev-astro-082812-140956 Horner, J. M. 2000, Reviews on environmental health, Oyama, V. I., Carle, G. C., Woeller, F., et al. 1980, Journal 15(3), 289–298, doi: 10.1515/reveh.2000.15.3.289 of Geophysical Research, 85, 7891, Jo, Y. S., Seon, K. I., Min, K. W., Edelstein, J., & Han, W. doi: 10.1029/JA085iA13p07891 2017, arXiv, 231, 21, doi: 10.3847/1538-4365/aa8091 P. J. Linstrom and W. G. Mallard. 2003, Nist standard Karakas, A. I., & Lattanzio, J. C. 2014, PASA, 31, e030, reference database number 69 (National Institute of Standards and Technology), 20899, doi: 10.1017/pasa.2014.21 doi: https://doi.org/10.18434/T4D303 Katyal, N., Nikolaou, A., Godolt, M., et al. 2019, The Pardo, A. 1986, Journal of Molecular Structure, 141, 137, Astrophysical Journal, 875, 31, doi: 10.1016/0022-2860(86)80317-9 doi: 10.3847/1538-4357/ab0d85 Pawlowski, F., Jørgensen, P., Olsen, J., et al. 2002, The Kolos, W., & Wolniewicz, L. 1964, The Journal of Chemical Journal of Chemical Physics, 116, 6482, Physics, 41, 3663, doi: 10.1063/1.1725796 doi: 10.1063/1.1459782 Krupenie, P. H. 1972, Journal of Physical and Chemical Pearson, J. C., Drouin, B. J., & Pickett, H. M. 2005, in Reference Data, 1, 423, doi: 10.1063/1.3253101 Astrochemistry: Recent Successes and Current Lammer, H., Sproß, L., Grenfell, J. L., et al. 2019, Challenges, ed. D. C. Lis, G. A. Blake, & E. Herbst, Vol. Astrobiology, 19, 927, doi: 10.1089/ast.2018.1914 231, 270 Li, G., Gordon, I. E., Rothman, L. S., et al. 2015, The Pignatari, M., Herwig, F., Hirschi, R., et al. 2016, ApJS, Astrophysical Journal Supplement Series, 216, 15, 225, 24, doi: 10.3847/0067-0049/225/2/24 doi: 10.1088/0067-0049/216/1/15 Polyansky, O. L., Kyuberis, A. A., Zobov, N. F., et al. Li, X., Heays, A. N., Visser, R., et al. 2013, A&A, 555, 2018, Monthly Notices of the Royal Astronomical A14, doi: 10.1051/0004-6361/201220625 Society, 480, 2597, doi: 10.1093/mnras/sty1877 Lin, C. Y., Gilbert, A. T. B., & Gill, P. M. W. 2008, Rauscher, T., Heger, A., Hoffman, R. D., & Woosley, S. E. Theoretical Chemistry Accounts, 120, 23, 2002, ApJ, 576, 323, doi: 10.1086/341728 doi: 10.1007/s00214-007-0292-8 Rees, A. L. 1947, Proceedings of the Physical Society, 59, Lofthus, A., & Krupenie, P. H. 1977, Journal of Physical 998, doi: 10.1088/0959-5309/59/6/310 and Chemical Reference Data, 6, 113, Ritter, C., Herwig, F., Jones, S., et al. 2018, MNRAS, 480, doi: 10.1063/1.555546 538, doi: 10.1093/mnras/sty1729 Luspay-Kuti, A., Mousis, O., Lunine, J. I., et al. 2018, Rivilla, V. M., Fontani, F., Beltr´an,M. T., et al. 2016, Space Science Reviews, 214, 1, ApJ, 826, 161, doi: 10.3847/0004-637X/826/2/161 doi: 10.1007/s11214-018-0541-2 Roueff, E., Abgrall, H., Czachorowski, P., et al. 2019, Malathy Devi, V., Kleiner, I., Sams, R. L., et al. 2014, Astronomy & Astrophysics, 630, A58, Journal of Molecular Spectroscopy, 298, 11, doi: 10.1051/0004-6361/201936249 doi: 10.1016/j.jms.2014.01.013 Schlegel, H. G. 1974, Tellus, 26, 11, Mantz, A. W., Maillard, J. P., Roh, W. B., & Narahari doi: 10.3402/tellusa.v26i1-2.9732 Rao, K. 1975, Journal of Molecular Spectroscopy, 57, Seager, S., Bains, W., & Petkowski, J. J. 2016, 155, doi: 10.1016/0022-2852(75)90049-1 Astrobiology, 16, 465, doi: 10.1089/ast.2015.1404 McKemmish, L. K., Masseron, T., Hoeijmakers, H. J., et al. Seager, S., Huang, J., Petkowski, J. J., & Pajusalu, M. 2019, Monthly Notices of the Royal Astronomical 2020, Nature Astronomy, 4, 802, Society, 488, 2836, doi: 10.1093/mnras/stz1818 doi: 10.1038/s41550-020-1069-4 Modelling Biosignatures: Diatomics 17

Shimajiri, Y., Andr´e,P., Braine, J., et al. 2017, A&A, 604, Viti, S., Tennyson, J., & Polyansky, O. L. 1997, \mnras, A74, doi: 10.1051/0004-6361/201730633 287, 79, doi: 10.1093/mnras/287.1.79 Sousa-Silva, C., Al-Refaie, A. F., Tennyson, J., & Wakelam, V., & Herbst, E. 2008, ApJ, 680, 371, Yurchenko, S. N. 2015, Monthly Notices of the Royal doi: 10.1086/587734 Astronomical Society, 446, 2337, Wakelam, V., Smith, I. W. M., Herbst, E., et al. 2010, doi: 10.1093/mnras/stu2246 Space Science Reviews, 156, 13, Sousa-Silva, C., Petkowski, J. J., & Seager, S. 2019, doi: 10.1007/s11214-010-9712-5 Physical Chemistry Chemical Physics, 21, 18970, doi: 10.1039/c8cp07057a Wakelam, V., Bron, E., Cazaux, S., et al. 2017, Molecular Sousa-Silva, C., Yurchenko, S. N., & Tennyson, J. 2013, Astrophysics, 9, 1, doi: 10.1016/j.molap.2017.11.001 Journal of Molecular Spectroscopy, 288, 28, Wang, J., Li, D., Goldsmith, P. F., et al. 2020, ApJ, 889, doi: 10.1016/j.jms.2013.04.002 129, doi: 10.3847/1538-4357/ab612d Sproß, L., Lammer, H., Lee Grenfell, J., et al. 2018, Wang, Y., Tian, F., Li, T., & Hu, Y. 2016, Icarus, 266, 15, European Planetary Science Congress, 12, 2018 doi: 10.1016/j.icarus.2015.11.010 Strobel, D. F., & Shemansky, D. E. 1982, Journal of Webster, C. R., Mahaffy, P. R., Atreya, S. K., et al. 2015, Geophysical Research: Space Physics, 87, 1361, Science, 347, 415, doi: 10.1126/science.1261713 doi: 10.1029/ja087ia03p01361 Western, C. M. 2017, Journal of Quantitative Spectroscopy Sukhbold, T., Ertl, T., Woosley, S. E., Brown, J. M., & and Radiative Transfer, 186, 221, Janka, H. T. 2016, ApJ, 821, 38, doi: 10.1016/j.jqsrt.2016.04.010 doi: 10.3847/0004-637X/821/1/38 Swain, M. R., Tinetti, G., Vasisht, G., et al. 2009, ApJ, Whitworth, A. P., & Jaffa, S. E. 2018, A&A, 611, A20, 704, 1616, doi: 10.1088/0004-637X/704/2/1616 doi: 10.1051/0004-6361/201731871 Tennyson, J., Yurchenko, S. N., Al-Refaie, A. F., et al. Willetts, A., Handy, N. C., Green, W. H., & Jayatilaka, D. 2016, Journal of Molecular Spectroscopy, 327, 73, 1990, Journal of Physical Chemistry, 94, 5608, doi: 10.1016/j.jms.2016.05.002 doi: 10.1021/j100377a038 —. 2020, Journal of Quantitative Spectroscopy and Wordsworth, R., & Pierrehumbert, R. 2014, Astrophysical Radiative Transfer, 255, 107228, Journal Letters, 785, 2, doi: 10.1016/j.jqsrt.2020.107228 doi: 10.1088/2041-8205/785/2/L20 Tessenyi, M., Tinetti, G., Savini, G., & Pascale, E. 2013, Yong, D., Lambert, D. L., & Ivans, I. I. 2003, ApJ, 599, Icarus, 226, 1654, doi: 10.1016/j.icarus.2013.08.022 1357, doi: 10.1086/379369 Tinetti, G., Encrenaz, T., & Coustenis, A. 2013, A&A Rv, Yurchenko, S. N., Barber, R. J., Yachmenev, A., et al. 21, 63, doi: 10.1007/s00159-013-0063-6 2009, Journal of Physical Chemistry A, 113, 11845, Tinetti, G., Drossart, P., Eccleston, P., et al. 2018, doi: 10.1021/jp9029425 Experimental Astronomy, 46, 135, doi: 10.1007/s10686-018-9598-x Yurchenko, S. N., & Tennyson, J. 2014, Monthly Notices of Tran, H., Flaud, P. M., Fouchet, T., Gabard, T., & the Royal Astronomical Society, 440, 1649, Hartmann, J. M. 2006, Journal of Quantitative doi: 10.1093/mnras/stu326 Spectroscopy and Radiative Transfer, 101, 306, Zapata Trujillo, J. C., Syme, A.-M., Rowell, K. N., et al. doi: 10.1016/j.jqsrt.2005.11.033 2021, Frontiers in Astronomy and Space Sciences, 8, 1, Vangioni, E., Dvorkin, I., Olive, K. A., et al. 2018, Monthly doi: 10.3389/fspas.2021.639068 Notices of the Royal Astronomical Society, 477, 56, Zhang, Z.-Y., Romano, D., Ivison, R. J., Papadopoulos, doi: 10.1093/mnras/sty559 P. P., & Matteucci, F. 2018, Nature, 558, 260, V´azquez,J., & Stanton, J. F. 2006, Molecular Physics, 104, doi: 10.1038/s41586-018-0196-x 377, doi: 10.1080/00268970500290367 18 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

APPENDIX

A. FURTHER DERIVATIONS A.1. Deriving the Energy We will go through how to derive the corresponding energy for each vibrational level as referenced by equations3 and5. We will use the following notation from section 2.6 within the following appendix sections. As a reminder, the relation between the anharmonic position, x, and the harmonic position, u, can be described as:

x − σ = u (A1)

We will now demonstrate deriving the ground state energy. To begin we must use perturbation theory. As the anharmonic Hamiltonian (equation1) and wavefunctions are known (equation2) we can express it as:

∗ ˆ E0 = hψ0 |H|ψ0i (A2)

Z ∞  2  ∗ 1 ∂ 1 2 1 3 1 4 = ψ0 − 2 + ηiix + ηiiix + ηiiiix ψ0 dx (A3) −∞ 2 ∂x 2! 3! 4! For ease of comprehension we will break the following derivations down into four components and solve individual, then sum back together. We will now describe the separated parts and label for ease:

Z ∞  2  ∗ 1 δ A = ψ0 − 2 ψ0 dx (A4) −∞ 2 δx Z ∞   1 2 B = ψ0 ηiix ψ0 dx (A5) −∞ 2! Z ∞   1 3 C = ψ0 ηiiix ψ0 dx (A6) −∞ 3! Z ∞   1 4 D = ψ0 ηiiiix ψ0 dx (A7) −∞ 4!

Lets begin with A first, and we need to remember that when an operator involves a differentiation, it does not commute. This means we now need to first differentiate ψ0 twice, but leave ψ0 unaffected:

Z ∞  2  ∗ 1 ∂ ψ0 A = ψ0 − 2 dx (A8) −∞ 2 ∂Q Now to insert the equations for the wavefunctions in:

∞ 2 1 ω 1/2 Z ω 2 ∂ ω 2 − 2 (x−σ) − 2 (x−σ) = − e 2 e dx (A9) 2 π −∞ ∂x Recalling u = x − σ therefore, dx = du. Substitute this back in:

∞ 2 1 ω 1/2 Z ω 2 ∂ ω 2 − 2 (u) − 2 (u) = − e 2 e du (A10) 2 π −∞ ∂x Doubly differentiating within the equation we get:

∂ − ω (u)2 − ω (u)2 e 2 = −ωue 2 (A11) ∂x ∂ − ω (u)2 2 2 − ω (u)2 − ω (u)2 − ωue 2 = u ω e 2 − ωe 2 (A12) ∂x Modelling Biosignatures: Diatomics 19

1/2 Z ∞ 1 ω  − ω (u)2  2 2 − ω (u)2 − ω (u)2  ∴ = − e 2 u ω e 2 − ωe 2 du (A13) 2 π −∞ ∞ 1 ω 1/2 Z  2 2  = − u2ω2eω(u) − ωe−ω(u) du (A14) 2 π −∞ 1 ω 1/2  1  π 1/2  π 1/2 = − ω2 − 2ω (A15) 2 π 2ω ω 4ω 1 = ω (A16) 4

Now onto part B, same as before we will introduce the wavefunctions then use substitution to solve:

Z ∞   ∗ 1 2 B = ψ0 ηiix ψ0 dx (A17) −∞ 2! 1/2 Z ∞ 1 ω  2 −ω(x−σ)2 = ηii x e dx (A18) 2! π −∞ 1/2 Z ∞ 1 ω  2 −ωu2 = ηii (u + σ) e du (A19) 2! π −∞ 1/2 Z ∞ 1 ω  2 2 −ωu2 = ηii (u + 2uσ + σ )e du (A20) 2! π −∞

Once again we can break this down into pieces:

Z ∞ 2 −ωu2 1  π  (u )e du = 3 (A21) −∞ 2 ω ∞ Z 2 2σ (u)e−ωu du = 0 (A22) −∞ ∞ Z 2  π  σ2 e−ωu du = σ2 (A23) −∞ ω

Putting these components back together:

1 ω 1/2  1  π 1/2  π 1/2 = η + σ2 (A24) 2! ii π 2ω ω ω 1  1  = η + σ2 (A25) 2! ii 2ω

Part C, using same method as before:

Z ∞   ∗ 1 3 C = ψ0 ηiiix ψ0 dx (A26) −∞ 3! 1/2 Z ∞ 1 ω  3 −ω(Q−σ)2 = ηiii x e dx (A27) 3! π −∞ 1/2 Z ∞ 1 ω  3 −ω(u)2 = ηiii (u + σ) e du (A28) 3! π −∞ 1/2 Z ∞ 1 ω  3 2 2 3 3 −ω(u)2 = ηiii (u + 3σu + σ u + σ ) e du (A29) 3! π −∞ 20 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

Breaking it down into parts again for ease of solving: ∞ Z 2 u3e−ω(u) du = 0 (A30) −∞ Z ∞ 1/2 2 −ω(u)2  π  3σ u e du = 3σ 3 (A31) −∞ 4ω ∞ Z 2 σ2 ue−ω(u) du = 0 (A32) −∞ ∞ Z 2  π 1/2 σ3 e−ω(u) du = σ3 (A33) −∞ ω Substituting these components back into the main equation: 1 ω 1/2  3σ  π 1/2  π 1/2 = η + σ3 (A34) 3! iii π 2ω ω ω 1  3σ  = η + σ3 (A35) 3! iii 2ω Finally onto section D of the original equation: Z ∞   ∗ 1 4 D = ψ0 ηiiiix ψ0 dx (A36) −∞ 4! 1/2 Z ∞ 1 ω  4 −ω(Q−σ)2 = ηiiii x e dx (A37) 4! π −∞ 1/2 Z ∞ 1 ω  4 −ω(u)2 = ηiiii (u + σ) e du (A38) 4! π −∞ 1/2 Z ∞ 1 ω  4 3 2 2 3 4 −ωu2 = ηiiii u + 4σu + 6σ u + 4σ u + u e du (A39) 4! π −∞ Break down into parts: Z ∞ 1/2 4 −ωu2 6  π  u e du = 2 (A40) −∞ 8ω ω ∞ Z 2 4σ u3e−ωu du = 0 (A41) −∞ ∞ 2 Z 2 6σ  π 1/2 6σ2 u2e−ωu du = (A42) −∞ 2ω ω Z ∞ 4σ3 u du = 0 (A43) −∞ ∞ Z 2  π 1/2 σ4 e−ωu du = σ4 (A44) −∞ ω Summing the piece back together: 1 ω 1/2  6  π 1/2 6σ2  π 1/2  π 1/2 = η + + σ4 (A45) 4! iiii π 8ω2 ω 2ω ω ω 1  3 3σ2  = η + + σ4 (A46) 4! iiii 4ω2 ω

If we now take all the parts, A, B, C and D we can now have the full equation for the energy of the ground state, which matches the expression in equation6: 1 1  1  1  3σ  1  3 3σ2  E = ω + η + σ2 + η + σ3 + η + + σ4 (A47) 0 4 2! ii 2ω 3! iii 2ω 4! iiii 4ω2 ω The method described here can be applied to the first excited state to achieve the answer shown in equation6. Modelling Biosignatures: Diatomics 21

A.2. Position Expectation Values Let’s first start with expectation value for the harmonic position, u: Z ∞ hui = u|ψ|2du (A48) −∞ We can use a single simple identity to calculate the expectation value, as we know the wavefunction can be normalised:

Z t o(u)du = 0 ; |ψ|2= 1 (A49) −t

∴ hui = 0 (A50) Now to do the anharmonic position expectation values. Although the anharmonic wavefunctions are also normalised we cannot use only the identities described previously, as the equations are more complex. Therefore, in this case we need to define the wavefunctions. Using2, we see that the for the ground state is:

1/4 ω  − ω (x−σ)2 ψ = e 2 (A51) 0 π Additionally for the first excited vibrational state:

 3 1/4 4ω − ω (x−σ)2 ψ = (x − σ)e 2 (A52) 1 π

Now to derive the expectation value for the shifted anharmonic position, hxi, with respect to the anharmonic wavefunctions. We can now derive the expectation value of the position, x, for the anharmonic ground and first vibrational level. Z ∞ 2 hψ0|x|ψ0i = |ψ0| x dx (A53) −∞

∞ ω 1/2 Z 2 = xe−ω(x−σ) dx (A54) π −∞ Substituting x = u + σ: ∞ ω 1/2 Z 2 = (u + σ)e−ω(u) du (A55) π −∞ Separating into two components (A and B) for easier integration:

∞ Z 2 A = (u)e−ω(u) du (A56) −∞

Using the identity for an odd function (where o(u) = −o(−u) for all u), and knowing the wavefunction can be normalised: Z t o(u)du = 0 (A57) −t Z ∞ −ω(u)2 ∴ A = (u)e du = 0 (A58) −∞ Now for part B: ∞ Z 2 B = σ e−ω(u) du (A59) −∞ An even function e(u) satisfies e(u) = e(−u) for all u. There, for any t,

Z t Z t e(u)du = 2 e(u)du (A60) −t 0 22 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

Z ∞ Z ∞ −ω(u)2 −ω(u)2 ∴ B = σ e du = 2σ e du (A61) −∞ 0 Then using a secondary integration identity: ∞ Z 2  π 1/2 e−ωu du = (A62) 0 4ω  π 1/2 B = 2σ (A63) ∴ 4ω Put A and B back together into the original expression: ω 1/2   π 1/2 = 0 + 2σ (A64) π 4ω

= σ (A65) Now onto the first vibrational level: Z ∞ 2 hψ1|x|ψ1i = |ψ1| x dx (A66) −∞

 3 1/2 ∞ 4ω Z 2 = (x − σ)2xe−ω(x−σ) dx (A67) π −∞ Substituting x = u + σ:  3 1/2 ∞ 4ω Z 2 = u2(u + σ)e−ω(u) du (A68) π −∞

 3 1/2 ∞ 4ω Z 2 = (u3 + σu2)e−ω(u) du (A69) π −∞ Once again separating into two components (A and B) for easier integration: ∞ Z 2 A = (u3)e−ω(u) du (A70) −∞ Using the identity for an odd function (where o(u) = −o(−u) for all u): Z t o(u)du = 0 (A71) −t Z ∞ 3 −ω(u)2 ∴ A = (u )e du = 0 (A72) −∞ ∞ Z 2 B = σ (u2)e−α(u) du (A73) −∞ Using the integration identity: Z ∞ 1/2 2 −ωu2 1  π  u e du = 3 (A74) −∞ 2 ω σ  π 1/2 B = (A75) ∴ 2 ω3 Put A and B back together into the original expression:

4ω3 1/2  σ  π 1/2 = 0 + (A76) π 2 ω3

= σ (A77) Clearly the outcome is independent of the vibrational level. Modelling Biosignatures: Diatomics 23

A.3. Branches This next section will detail the working out required to produce equations 26, 27, 28, 29 and 30. The general analytical expression for a spectrum of a diatomic molecule is the following:

ε(J,v) = εJ + εv (A78) 2 2 =ω ¯v(v + 1/2) + BvJ(J + 1) − DeJ (J + 1) (A79) Where:

ω¯v =ω ¯e − ω¯exe(v + 1/2) (A80) 2 2 2 ∴ ε(J,v) =ω ¯e(v + 1/2) − ω¯exe(v + 1/2) + BvJ(J + 1) − DeJ (J + 1) + ... (A81)

The frequency ν0→1 is usually called the band origin or band centre. Which is represented by the following equation:

ν0→1 =ω ¯e − 2¯ωexe (A82)

De represents the centrifugal distortion constant. For current theory we can ignore the small centrifugal distortions from De etc, and the equation can be rewritten simply as:

2 ε(J,v) =ω ¯e(v + 1/2) − ω¯exe(v + 1/2) + BvJ(J + 1) (A83) For the next part we will be restricting the discussion to solely fundamental vibrational transitions, v = 0 → v = 1. We also take the respective B values as B0 and B1 with B0 > B1. This transition can generally be described as (in cm−1):

∆ε = εJ 0,v=1 − εJ 00,v=0 (A84) 0 0 00 00 = ν0→1 + B1J (J + 1) − B0J (J + 1) (A85) The transitions for the P Branch (∆J = −1, J 00 = J 0 + 1) is given by:

∆ε =ν ¯P (A86) 0 0 2 = ν0→1 − (B1 + B0)(J + 1) + (B1 − B0)(J + 1) (A87) Where; J 0 = 0, 1, 2, 3, ... The transitions for the Q Branch (∆J = 0, J 00 = J 0) is given by:

∆ε =ν ¯Q (A88) 0 0 = ν0→1 − J (J + 1)(B1 − B0) (A89) Where; J 0 = 0, 1, 2, 3, ... Therefore for the R Branch (∆J = +1, J 0 = J 00 + 1) the transition is given as:

∆ε =ν ¯R (A90) 00 00 2 = ν0→1 + (B1 + B0)(J + 1) + (B1 − B0)(J + 1) (A91) Where; J 00 = 0, 1, 2, 3, ... . Taking the general rovibrational equation:

2 ε(J, v) =ω ¯e(v + 1/2) − ωx¯ e(v + 1/2) + BvJ(J + 1) (A92) Lets now apply this for the ground state with quantum number J” and the upper vibrational state with quantum number J’ as per standard notation.

1 1 00 00 ε 00 = ω¯ − ωx¯ + B J (J + 1) (A93) (J ,v=0) 2 e 4 e 0 3 9 0 0 ε 0 = ω¯ − ωx¯ + B J (J + 1) (A94) (J ,v=1) 2 e 4 e 1 0 0 00 00 ∴ ∆ε = ν0→1 + B1J (J + 1) − B0J (J + 1) (A95) 24 Thomas M. Cross, David M. Benoit, Marco Pignatari and Brad K. Gibson

Now consider the S branch where ∆J = +2,J 0 = J 00 + 2. By plugging in these formulas:

∆ε = νS (A96) 00 00 00 00 = ν0→1 + B1(J + 2)(J + 3) − B0J (J + 2) (A97) 00 2 00 00 2 00 = ν0→1 + B1((J ) + 5J + 6) − B0((J ) + J ) (A98)

In the cases where B0 = B1 = B, the expression simplifies to:

00 νS = ν0→1 + B(4J + 6) (A99)

Finally onto the O branch where ∆J = −2,J 00 = J 0 + 2. By plugging in these formulas:

∆ε = νO (A100) 0 0 0 0 = ν0→1 + B1J (J + 1) − B0(J + 2)(J + 3) (A101) 0 2 0 0 2 0 = ν0→1 + B1((J ) + J ) − B0((J ) + 5J + 6) (A102)

In the cases where B0 = B1 = B, the expression simplifies to:

00 νO = ν0→1 − B(4J + 6) (A103)