Arxiv:2108.01790V1 [Astro-Ph.EP] 3 Aug 2021 Hot Jupiters Are the ﬁrst Discovered and Best Characterized Dance Ratio Remains an Ongoing Task

Home , 51 Eridani, 51 Eridani b, 64 Eridani

DRAFTVERSION AUGUST 5, 2021 Typeset using LATEX twocolumn style in AASTeX63

A Comparative Study of Atmospheric Chemistry with VULCAN

SHANG-MIN TSAI,1 MATEJ MALIK,2 DANIEL KITZMANN,3 JAMES R.LYONS,4 ALEXANDER FATEEV,5 ELSPETH LEE,6 AND KEVIN HENG6

1Atmospheric, Ocean, and Planetary Physics, Department of Physics, University of Oxford, UK 2Department of Astronomy, University of Maryland, College Park, MD 20742, USA 3University of Bern, Center for Space and Habitability 4Arizona State University, School of Earth and Space Exploration, Bateman Physical Sciences, Tempe, USA 5Technical University of Denmark, Department of Chemical and Biochemical, Denmark 6University of Bern, Center for Space and Habitability, Switzerland

ABSTRACT We present an update of the open-source photochemical kinetics code VULCAN (Tsai et al.(2017); https://github.com/exoclime/VULCAN) to include C-H-N-O-S networks and photochemistry. Additional new features are advection transport, condensation, various boundary conditions, and temperature-dependent UV cross-sections. First, we validate our photochemical model for hot Jupiter atmospheres by performing an in- tercomparison of HD 189733b models between Moses et al.(2011), Venot et al.(2012), and VULCAN, to diagnose possible sources of discrepancy. Second, we set up a model of Jupiter extending from the deep troposphere to upper stratosphere to verify the kinetics for low temperature. Our model reproduces hydrocarbons consistent with observations, and the condensation scheme successfully predicts the locations of water and ammonia ice clouds. We show that vertical advection can regulate the local ammonia distribution in the deep atmosphere. Third, we validate the model for oxidizing atmospheres by simulating Earth and find agreement with observations. Last, VULCAN is applied to four representative cases of extrasolar giant planets: WASP- 33b, HD 189733b, GJ 436b, and 51 Eridani b. We look into the effects of the C/O ratio and chemistry of titanium/vanadium species for WASP-33b; we revisit HD 189733b for the effects of sulfur and carbon condensation; the effects of internal heating and vertical mixing (Kzz) are explored for GJ 436b; we test updated planetary properties for 51 Eridani b with S8 condensates. We find sulfur can couple to carbon or nitrogen and impact other species such as hydrogen, methane, and ammonia. The observable features of the synthetic spectra and trends in the photochemical haze precursors are discussed for each case. 1. INTRODUCTION Birkby et al. 2013; Brogi et al. 2014; Hoeijmakers et al. Understanding the chemical compositions has been a cen- 2018). tral aspect in atmospheric characterization for planets within The majority of discovered exoplanets have sizes between and beyond the Solar System. Photochemical kinetics mod- Earth and Neptune. Their heavy elemental abundances (i.e. els establish the link between our knowledge of chemical metallicity) can vary considerably, as often inferred by the reactions and various planetary processes (e.g., atmospheric water detection (e.g., Wakeford et al. 2017; Chachan et al. dynamics, radiative transfer, outgassing process, etc.), pro- 2019). While CH4 is expected to be more abundant in cooler viding a theoretical basis for interpreting observations and (Teq . 1000 K) atmospheres, understanding how disequilib- addressing habitability. rium chemistry and other processes alter the CH4/CO abun- arXiv:2108.01790v1 [astro-ph.EP] 3 Aug 2021 Hot Jupiters are the first discovered and best characterized dance ratio remains an ongoing task. class of exoplanets. Transit and eclipse observations have The direct imaging technique provides a complementary made various initial detections of chemical species in their window to resolve young planets at a far orbit (e.g., see the reviews of Crossfield 2015; Pueyo 2018)). The new gener- atmospheres such as Na, K, H2O, CH4, CO, CO2 (e.g. see the review of Kreidberg 2018). An extreme class of exceed- ation of instruments like GPI and SPHERE (Chauvin 2018) ingly irradiated hot Jupiter around bright stars have equilib- have identified a number of interesting young Jupiter ana- rium temperature higher than 2000 K. They are prime tar- log. These young planets are self-luminous from their heat gets for emission observations, and recent high-resolution of formation and receive UV fluxes from the star at the same spectroscopic measurements reveal atomic and ionic features time, giving insights on the planet forming conditions outside that make their atmospheres resemble low-mass stars (e.g., the snow lines and the transition between planets and brown dwarfs. 2

Across the various types of aforementioned planetary at- cles are possibly ubiquitous, with diverse compositions (Gao mospheres, photochemical kinetics and atmospheric trans- et al. 2020) including cloud particles formed from conden- port are the dominant mechanisms that control the ma- sation or produced by photolysis at high altitudes. Micro- jor chemical abundances. Photodissociation occurs when physics models (Helling & Woitke 2006; Lavvas & Koski- molecules are split into reactive radicals by high-energy pho- nen 2017; Kawashima & Ikoma 2018; Gao & Benneke 2018; tons while atmospheric transport shapes the abundance distri- Ohno et al. 2020) have investigated trends and properties of bution. Disequilibrium processes can drive abundances con- aerosols for various environments. One particularly interest- siderably away from the chemical equilibrium state and are ing candidate of aerosols is the sulfur family, such as sulfuric best studied in chemical kinetics models. clouds (Hu et al. 2013; Misra et al. 2015; Loftus et al. 2019) Kinetics models stem from simulating the atmospheric in an oxidizing atmosphere or elemental sulfur in a reducing compositions in Solar System planets (e.g., Kasting et al. atmosphere (Hu et al. 2013; Gao et al. 2017). Photochem- 1979; Yung et al. 1984; Nair et al. 1994; Wilson & Atreya istry generally sets off the initial steps in the gas phase, then 2004; Lavvas et al. 2008; Hu et al. 2012; Krasnopolsky the condensable species can form particles when saturated in 2012), which focus on photochemistry and radical reactions. a broad range of altitudes (Gao et al. 2017). The relatively The low temperature regime makes thermochemistry less rel- simple sulfur particles in H2-dominated atmospheres allow a evant in most cases. Liang et al.(2003) first applied a pho- consistent photochemical-aerosol kinetics modeling, which tochemical kinetics model, Caltech/JPL KINETICS (Allen we will conduct in this work. Although the formation path- et al. 1981), to the hot Jupiter HD 209458b and identified ways of organic haze particles are highly complex, we will the photochemical source of water for producing atomic H. focus on a group of haze precursors and investigate their pho- However, some reaction rates in their study are extrapolated tochemical stability in the hope of providing complementary from measurements at low temperatures and not suitable for insights on the haze-forming conditions. hot Jupiter conditions. Line et al.(2010) adopt the high- The exclusive access to often proprietary chemical mod- temperature rate coefficients for the major molecules and els motivates us to develop an open-source, chemical kinet- use the lower boundary to mimic mixing from the thermo- ics code VULCAN (Tsai et al. 2017). The initial version chemical equilibrium region. A new group of models incor- of VULCAN includes a reduced-size C-H-O thermochemical porating kinetics data valid at high temperatures started to network and treats eddy diffusion. In Tsai et al.(2017), VUL- emerge since then. Zahnle et al.(2009) reverse the reactions CAN is validated by comparing the quench behavior with to ensure kinetics consistent with thermodynamic calcula- ARGO (Rimmer & Helling 2016) and Moses et al.(2011). tions and consider sulfur chemistry on hot Jupiters. Moses Since then, VULCAN has been continuously updated and ap- et al.(2011) implement high-temperature reactions in KI- plied to several studies such as Zilinskas et al.(2020) who NETICS to model hot Jupiters HD 189733b and HD 209458b identify key molecules of hot super-Earths with nitrogen- with detailed pathway analysis. Venot et al.(2012) adopt the dominated atmospheres, and Shulyak et al.(2020) who ex- combustion mechanisms validated for industrial applications plore the effects of XUV for different stellar types. to model the same canonical hot Jupiters but find different In this work, we present the new version of 1-D photo- quenching and photolysis profiles from Moses et al.(2011). chemical model VULCAN, with embedded chemical net- Hobbs et al.(2021) recently extend Zahnle et al.(2009) to works now including hydrogen, oxygen, carbon, nitrogen, include sulfur photochemistry and find the inclusion of sul- and sulfur. The chemical network is customizable and does fur can impact other non-sulfur species on HD 209458b and not require separating fast and slow species. The major up- 51 Eridani b. As the discovery of diverse exoplanets pro- dates of VULCAN from Tsai et al.(2017) are: gresses, more kinetics models have been applied to study a wide range of aspects, such as the compositional diversity • C-H-N-O-S chemical networks with about 100 within an atmospheric-grid framework (Moses et al. 2013; species, including a simplified benzene forming mech- Miguel & Kaltenegger 2014; Molaverdikhani et al. 2019), anism atmospheric evolution with loss and/or outgassing processes (Hu et al. 2015; Wordsworth et al. 2018; Lincowski et al. • Photochemistry with options for temperature- 2018), prebiotic chemistry driven by high-energy radiation dependent UV cross sections input (Rimmer & Helling 2016; Rimmer & Rugheimer 2019), and detectability of habitable planets (Arney 2019; Schwieter- • Condensation and particle settling included man et al. 2018). A number of recent attempts of atmospheric composi- • Advection, eddy diffusion, and molecular diffusion in- tion measurements are hindered by aerosol layers (Kreid- cluded for the transport processes berg et al. 2014; Parviainen et al. 2018). Aerosol parti- • Choice of various boundary conditions 3

In Section2, we describe model details that have been effect compared to eddy diffusion or molecular diffusion, ex- updated since Tsai et al.(2017). In Section3, we validate cept for the light species in the thermosphere with large tem- photochemistry and various new features of VULCAN with perature gradients (Nicolet 1968). The molecular diffusion simulations of HD 189733b, Jupiter, and Earth. A compre- coefficient has the expression of b/N from the gas kinetic, hensive model comparison for HD 189733b between Moses where b is a parameter for binary gas mixtures. The binary et al.(2011), Venot et al.(2012), and VULCAN is given. In parameter b and the thermal diffusion factor αT are ideally Section4, we perform case studies with focus on the effects determined experimentally for each binary mixture. In prac- of sulfur chemistry and haze precursors. We discuss caveats, tice, we simplify the atmosphere to a binary system with the implications and opportunities for future work in Section5 dominant gas as the main constituent and the rest in turn as and summarize the highlights in Section6. the minor constituent. Specifically, we adopt the molecular diffusion coefficient of a binary mixture that is available from 2. KINETICS MODEL the experimental data and scale that of other mixtures based 2.1. Basic Equations and Numerics on the fact that b is proportional to the mean relative speed of two gases, i.e. given D1−2 for the dominant gas 1 and minor The 1D photochemical kinetics model solves a set of Eu- gas 2, the molecular diffusion coefficient for gas 1 and any lerian continuity equations, other minor gas i can be scaled as

∂ni ∂φi p = P − L − , (1) D1−i = D1−2 m2/mi((m1 + mi)/(m1 + m2)). (3) ∂t i i ∂z

−3 The molecular diffusion coefficient and the thermal diffusion where ni is the number density (cm ) of species i and t de- factor for atmospheres dominated by H2,N2, and CO2 are notes the time. P and L are the production and loss rates i i listed in AppendixA. (cm−3 s−1) of species i, from both thermochemical and pho- A second-ordered central difference is used to discretize tochemical reactions. The system of (1) has the same form as the spatial derivative of diffusion flux, as in Tsai et al.(2017), that in Tsai et al.(2017), except only eddy diffusion is con- except a first-order upwind scheme (Brasseur & Jacob 2017) sidered for the transport flux φ in Tsai et al.(2017). The i is applied for advection. The finite difference form for the transport flux including advection, eddy diffusion, molecular derivative of the transport flux of layer j is and thermal diffusion while assuming hydrostatic balance is now written as (e.g., Chamberlain & Hunten 1987) φ − φ i,j+1/2 i,j−1/2 (4) ∆zj ∂Xi ∂ni 1 1 + αT dT φi = ni v−Kzzntot −Di[ +ni( + )], ∂z ∂z Hi T dz , with the upper and lower interfaces of layer j labeled as (2) j + 1/2 and j − 1/2, respectively, in the staggered structure. where v is the vertical wind velocity, Kzz and Di are the The full expression for the transport flux in Equation (2) at eddy diffusion and molecular diffusion coefficient, respec- the upper and lower interfaces is then tively, Hi is the molecular scale height for species i with adv mig molecular mass mi , i.e. Hi = (g: gravity; T : temper- φi,j+1/2 = φi,j+1/2 − (Kzz,j+1/2 + Di,j+1/2)ntot,j+1/2 kB T ature; kB: the Boltzmann constant ), and αT is the thermal Xi,j+1 − Xi,j 1 1 × − Dj+1/2Xi,j+1/2( − + diffusion factor. While advection is commonly ignored in 1- ∆zj+1/2 Hi H0 D models, we keep the advection term and distinguish it from αT Tj+1 − Tj eddy diffusion with respect to their intrinsic differences. For ) Tj+1/2 ∆zj+1/2 example, a plume of smoke transports the initial abundance adv along the direction of wind until diffusion becomes important φi,j−1/2 = φi,j−1/2 − (Kzz,j−1/2 + Di,j−1/2)ntot,j−1/2 and dissipates the smoke to the surrounding air. Xi,j − Xi,j−1 1 1 × − Dj−1/2Xi,j−1/2( − + Physically, the first term of the transport flux (2) describes ∆zj−1/2 Hi H0 advection in the direction of the wind. The second term is αT Tj − Tj−1 eddy diffusion that acts to smear out the compositional gra- ) Tj−1/2 ∆zj−1/2 dient. The molecular diffusion in the third term becomes im-  portant at low pressure and drives each constituent toward v ni,j, for v > 0 φadv = j+1/2 j+1/2 diffusive equilibrium, which is different for each species i,j+1/2 vj+1/2ni,j+1, for vj+1/2 < 0 based on its individual scale height. The direction of thermal  diffusion depends on the sign of the thermal diffusion factor. adv vj−1/2ni,j−1, for vj−1/2 > 0 Positive sign means the component will diffuse toward colder φi,j−1/2 = v n , for v < 0 region and vice versa. Thermal diffusion is often a secondary j−1/2 i,j j−1/2 (5) 4 where H0 is the atmospheric scale height with altitude depen- be assigned to account for escape velocity or for any process dent gravity and we have approximated the physical quan- producing inflow/outflow (Krasnopolsky 2012). The flux and tities at the interface by the average of two adjacent layers velocity can also be assigned together to describe the final ntot,j +ntot,j±1 Xi,j +Xi,j±1 ntot,j±1/2 = 2 , Xi,j±1/2 = 2 , and boundary condition of a single species. Ttot,j +Ttot,j±1 adv Constant mixing ratios are prescribed for the boundary Ttot,j±1/2 = 2 . The advection flux φ in Equation (5) only depends on the property of the upstream condition when the detail exchange is complex but the layer in the upwind scheme. Equation (1) can be reduced to knowledge of precise abundance is available. For example, a system of ordinary differential equations (ODEs) after re- the water vapor at the surface is expected to be set by satura- placing the spatial derivative of transport flux in Equation (1) tion according to relative humidity on an ocean planet with a with (4) and (5) and assigning proper boundary conditions. substantial reservoir of water. Assigning constant mixing ra- The numerical scheme using the Rosenbrock method to inte- tios is also practical for regional models, such as the compo- grate the “stiff” system (1) forward in time until steady state sition around the cloud layers for the Venus model with lower is achieved is described in detail in Tsai et al.(2017). boundary placed at the cloud layer (Krasnopolsky 2012). Since constant mixing ratio does not allow changes of the 2.2. Boundary Conditions composition at the boundary, this boundary condition should The solutions to the system of ODEs derived from Equa- not be used in conjunction with flux or velocity boundary tion (1) need to satisfy the given boundary conditions. The conditions. boundary conditions encompass various planetary processes 2.3. Chemical Networks that are crucial in regulating the atmosphere. There are three basic quantities commonly used to describe the boundary We have extended the previous C-H-O network in (Tsai conditions (e.g. Hu et al. 2012): flux, velocity, and mixing et al. 2017) to include nitrogen and sulfur in a hierarchi- 1 ratio. We will elucidate their corresponding implications for cal manner, e.g., C-H-O , C-H-N-O, C-H-N-O-S networks. the lower and upper boundaries. Each network is provided with a reduced version and a full The flux term in Equation (5) depends on the layers above version, where “reduced” is referred to both oxidation state and below. Hence the fluxes at the top and bottom are un- and network size. The reduced version has species and specified. Assigning constant fluxes is common to repre- mechanisms (e.g., the ozone cycle) that are only important sent surface emission at the lower boundary for rocky plan- in oxidizing conditions stripped off, which are more com- ets and inflow/outflow at the upper boundary. For example, putationally efficient and suited for the general hydrogen- dominated atmospheres. The full version of networks are CO and CH4 surface sources play a key role to Earth’s troposphere; meteoritic inflow or hydrodynamic escape outflow designed for a wide range of main atmospheric constituents, can be prescribed as constant flux at the upper boundary (e.g., from reducing to oxidizing. Hydrocarbon species are trun- Wordsworth et al. 2018). Alternatively, diffusion-limited flux cated at two carbons, while some higher-order hydrocarbons can be assigned at the upper boundary, which assumes the es- are present as necessary sinks for the two-carbon species or cape flux is limited by the diffusion transport into exosphere. hazy precursors. The chemical network files with rate co- The diffusion-limited flux reads efficients for the forward reactions can be found at https: //github.com/exoclime/VULCAN/tree/master/atm. 1 1 The full version of C-H-N-O-S network includes 96 φi,top = −Di,topni( − ) (6) 1 Hi H0 species: H, H2, O, O, O2,O3, OH, H2O, HO2,H2O2, CH, C, CH , 1CH , CH , CH ,C ,C H ,C H, C H ,C H , and can be applied to any set of light species in the code. 2 2 3 4 2 2 2 2 2 3 2 4 C H ,C H ,C H ,C H ,C H ,C H ,C H ,C H ,C H , Without additional constraints, we often simply assume the 2 5 2 6 4 2 3 3 3 2 3 4 6 5 6 6 4 3 C H , CO, CO , CH OH, HCO, H CO, CH O, CH OH, flux to be zero, which means no net material exchange. This 4 5 2 2 2 3 3 CH CO, H CCO, HCCO, CH O , CH OOH, N, N(2D), N , zero-flux boundary condition is generally suited for the lower 3 2 3 2 3 2 NH, CN, HCN, NH , NH , NO, N H ,N H, N H ,N H , boundary conditions while placed at a sufficient depth of 2 3 2 2 2 2 3 2 4 HNO, H CN, HC N, CH CN, CH CN, C H CN, HNCO, most gas giants (Moses et al. 2011; Rimmer & Helling 2016; 2 3 3 2 2 3 NO ,N O, CH NH , CH NH, CH NH , CH CHO, NO , Tsai et al. 2017). While not specifying the boundary condi- 2 2 2 2 2 3 2 3 3 HNO , HNO , NCO, N O , S, S ,S ,S ,S , SH, H S, tion, zero flux is implied as default in VULCAN. 3 2 2 5 2 3 4 8 2 HS , SO, SO , SO , CS, OCS, CS , NS, HCS, HSO, HSO , In addition to the flux, velocity is useful to represent 2 2 3 2 3 H SO , CH S, CH SH, S O and about 570 forward thermo- sources and sinks that scale with the species abundance. 2 4 3 3 2 chemical reactions and 69 photodissociation branches. All For example, (dry/wet) deposition velocity is conventionally used to parametrize removal processes such as gas absorp- 1 tion or uptake into the surface (Hu et al. 2012; Seinfeld & We have updated C-H-O network from (Tsai et al. 2017) by adding HO2 Pandis 2016). At the upper boundary, upward velocity can and H2O2. 5 thermochemical reactions are reversed using the equilibrium Stars are the ultimate energy source of disequilibrium constant derived from the NASA polynomials as described in chemistry. The stellar radiation interacting with the atmo- Tsai et al.(2017) to ensure chemical equilibrium can be ki- sphere can be converted into internal energy or initiates netically achieved2. We also provide an option for customiz- chemical reactions. Photodissociation describes the process ing modular networks. A subgroup of species can be freely in which energetic photons break molecules apart, schemat- picked and only reactions that involve the selected species ically written as an unimolecular reaction with photons (hν) will form a new modular chemical network. Unlike mini- mizing Gibbs free energy for equilibrium chemistry, caution A −−→hν B + C. (7) is required in this process to incorporate trace species that are important intermediates to set up a sensible network. Photodissociation typically produces active free radicals and We have incorporated a simplified benzene mechanism initiate a chain of reactions that are essential to atmospheric into the generally two-carbon based kinetics, with the mo- chemistry (e.g., the ozone cycle on Earth or the organic haze tivation of considering it in the context of haze precursors, as formation on Titan). will be discussed in Section 2.8. The intention is to capture The radiative flux that drives photolysis is conventionally the main formation pathways at minimum cost in terms of the defined by the number of photons from all directions per unit size of the network. We adopt one of the possible benzene time per unit area per unit wavelength and referred as the actinic flux, J(z, λ), with z being altitude λ being wavelength. forming pathways through propargyl (C3H3) recombination M J(z, λ) consists of two components, direct beam and dif- C H + C H −−→ C H (Frenklach 2002), whereas C H 3 3 3 3 6 6 3 3 fuse radiation: is produced by CH3 + C2H −−→ C3H3 + H. We then add hydrocarbons such as C3H2,C3H4, and C6H5 for the hydro- −τ(z,λ)/µ J(z, λ) = J(∞, λ)e + Jdiff(z, λ). (8) gen abstraction reactions of C3H3 and C6H6 to complete the mechanism. where τ is the optical depth and µ = cosθ with θ being The rate coefficients of the reactions are broadly drawn the zenith angle of the incident beam. The first term of 3 4 from the following: (1) NIST database (2) KIDA database Equation(8) describes the attenuated actinic flux reaching the (3) literature sources including Moses et al.(2005); Lavvas plane perpendicular to the direction of beam (there is no co- et al.(2008); Moses et al.(2011); Zahnle et al.(2016). Al- sine pre-factor as for radiative heating since the number of though most rate coefficients are chosen to be validated for a intercepted molecules is randomly oriented and independent wide range as possible (300 - 2500 K), some of the rate coef- of the direction of the stellar beam). ficients are still only measured at limited temperature ranges, The optical depth τ accounts for the extinction from both which has been a long standing issue in kinetics. The ki- absorption and scattering is calculated as netics becomes even more uncertain while sulfur is involved. For example, elemental sulfur in the gas phase exists in many Z τ = [Σ (σ + σ )n ]dz (9) allotropic forms but the chain-forming reactions between the i a,i s,i i allotropes were poorly constrained. The recombination rates where σ and σ are the cross section of absorption and of S that form the first sulfur bond S + S −−→M S from two a,i s,i 2 scattering, respectively. The absorption cross section σ can early measurements Fair & Thrush(1969) and Nicholas et al. a,i be different from the photodissociation cross section because (1979) are differed by four orders of magnitude. A recent absorption is not necessarily followed by dissociation. The calculation by Du et al.(2008) confirms the value by Fair diffusive flux J is the scattered radiation defined by inte- & Thrush(1969) and we adopt the rate coefficient from Du diff grating the diffuse specific intensity over all directions. We et al.(2008) in our network. To address the uncertainties in use the two-stream approximation in Malik et al.(2019) to sulfur kinetics, we perform sensitivity tests for selective key first solve for the diffuse flux and convert it to total intensity reactions in Section4. using the first Eddington coefficient (Heng et al. 2018): 2.4. Computing Photochemistry Jdiff(z, λ) = Fdiff/ (10)

diff where Fdiff is the total diffuse flux given by Fdiff ≡ F↑ + 2 We report a significant discrepancy in the new NASA 9-polynomials of diff F↓ and is the first Eddington coefficient with value 0.5 for CH2NH (http://garfield.chem.elte.hu/Burcat/NEWNASA.TXT) compared to the early NASA 7-polynomials and other sources, which can lead to isotropic flux. Although multiple scattering is not explicitly several orders of magnitude errors. We use the fit from the NASA 7- included in the expression in Malik et al.(2019), the process polynomials for CH2NH instead. can be approached through iteration and we find the equilib- 3 https://kinetics.nist.gov rium state of multiple scattering can normally be achieved 4 http://kida.obs.u-bordeaux1.fr/ within 200 iterations for a strongly irradiated hot Jupiter. In 6 the code, we have the option to update the actinic flux peri- 1160 K from EXOMOL), NH3(EXOMOL), O2 (Frederick & odically to save computing time. Mentall 1982; Vattulainen et al. 1997), SH (Gorman et al. Once the actinic flux has been obtained, the photolysis rate 2019), H2S(Gorman et al. 2019), COS (Gorman et al. 2019), coefficient can be determined from integrating the actinic CS2 (Gorman et al. 2019) in the current version of VUL- flux and the absorption cross section over the wavelength CAN. The temperature dependence of the UV cross sections Z of these molecules can be found in Figure 38. It is evi- k = q(λ)σa(λ)J(z, λ)dλ. (11) dent that both the absorption threshold and cross sections of λ CO2 exhibit strong temperature dependence. For H2O, we and the photolysis rate of Reaction (7) is have incorporated the recent measurement for the cross section above 200 nm (Ranjan et al. 2020). We follow Ranjan dn A = −kn (12) et al.(2020) taking a log-linear fit for the noisy data above dt A 216 nm. In addition, we have included measured data from , where q(λ) is the quantum yield (photons−1), describing Schulz et al.(2002) for temperature above 1500 K, . the probability of triggering a photolysis branch for each A layer-by-layer interpolation for the temperature- absorbed photon. In VULCAN, we adopt the cross sec- dependent cross sections is implemented in the model, i.e. tions from the Leiden Observatory database5 (Heays et al. the cross section of one single species is allowed to vary 2017) whenever possible, which provides tabulated data across the atmosphere due to the temperature variation. The of photoabsorption, photodissociation, and photoionisation interpolation is linear in the temperature space and logarith- cross sections with uncertainty ranking. The data has been mic in the cross-section space. With limited data, we find benchmarked against other established databases such as the the linear interpolation in temperature generally underesti- PHIDRATES database6 (Huebner et al. 1992; Huebner & mates the cross sections and therefore our implementation is Mukherjee 2015) which is detailed in Heays et al.(2017). considered as a conservative estimate for how photolysis in- The full list of photolysis reactions and references are listed creases with temperature. in Table6. Condensation and rainout The spectral resolution with respect to the stellar flux and 2.6. cross sections can be important while computing Equation VULCAN handles condensation and evaporation using the (11) numerically. The minimum resolution used in the model growth rate of particles, assuming sufficient activated nuclei. should be capable of resolving the line structures in the stellar For a schematic condensation/evaporation reaction spectra and cross sections. We discuss the errors from under- A(gas) ←−→ A(particle) (13) resolving in AppendixB. the reaction rate is given by the mass balance equation (Sein- 2.5. Temperature-Dependent UV Cross Sections feld & Pandis 2016) Most laboratory measurements of UV cross sections are dnA DAmA sat conducted at room temperature or lower, which might raise = − 2 (nA − nA ) (14) dt ρprp reliability issues with application to high-temperature atmospheres. Heays et al.(2017) suggested that as temperatures where DA and mA are the molecular diffusion coefficient increased by a few hundred K, the excitation of vibrational and molecular mass of gas A, ρp and rp are the density and sat and rotational levels (limited to v ≤ 2) in many cases only radius of the particle, nA and nA are the number density cause minor broadening of the cross sections and does not al- and saturation number density of gas A, respectively. Equa- ter its wavelength integration. However, for molecules with tion (14) describes the growth rate by diffusion for particles prominent transition between excited vibrational states (e.g. with size rp in the continuum regime (particles larger than the mean free path i.e. Knudsen number (K ) smaller than CO2), the temperature dependence on the cross section and n photolysis rate can be important. 1). The negative value of Equation (14) corresponds to con- sat Recent work has started to investigate the high- densation when nA > nA and the positive value corresponds sat temperatures UV cross sections of a few molecules (Venot to evaporation when nA < nA . Our condensation expression et al. 2015, 2018). Given the available data, we have takes the same form as Hu et al.(2012); Rimmer & Helling included temperature-dependent photoabsorption cross sec- (2016), except that the growth rate of particles in the kinetic 7 regime (particles smaller than their mean free path i.e. Knud- tions of H2O (EXOMOL ), CO2 (Venot et al. 2018) (with sen number (Kn) greater than 1) is used in Hu et al.(2012); πµvth Rimmer & Helling(2016). When applying Kn = 4P 5 ∼ http://home.strw.leidenuniv.nl/ ewine/photo where µ is the dynamic viscosity, vth the thermal veloc- 6 http://phidrates.space.swri.edu ity, and P the pressure, a H2 atmosphere enters the kinet- 7 http://www.exomol.com/data/data-types/xsec VUV/ ics regime with Kn > 10 above 1mbar for a temperature of 7

400 K and above 0.1 µbar for a temperature of 1000 K. We While there are a few measurements for the reactions of ti- find that for most of the application, condensation occurs in tanium/vanadium species with laser vaporization at low tem- the lower atmosphere with micron-size or larger particle and perature, the kinetics data at high temperature is nearly non- the continuum regime is more suitable. Since condensation existent. As a first step, we perform simple estimates on the typically operates in a relatively short timescale, we imple- unknown rate constants of titanium/vanadium species. First, ment an option to switch off condensation and fix the abun- we look for kinetics data of analogous transition metals, such dances of condensing species and the condensates after the as Fe. We assume the same rate coefficient as the analo- dynamic equilibrium has reached. The approach is similar gous reaction if it is measured at high temperature. When to the quasi-steady-state assumption (QSSA) method, which high-temperature data are not available, we estimate the tem- decouples the fast and slow reactions to ease the computa- perature dependence based on transition state theory. For tional load. an endothermic reaction, we approximate the activation en- After the gas condenses to particles, they fall following the ergy (the exponential term in the Arrhenius expression) by terminal settling velocity (vs) derived from the Stoke’s law the enthalpy difference between the products and reactants, (Seinfeld & Pandis 2016) as assuming the energy increase of the transition state is small 2 compared to the enthalpy difference for reactions involving 2 ρpr g v = p (15) radicals 8. Once the activation energy is obtained, the pre- s 9 µ exponential factor is adjusted to fit the reference value at low where µ is the atmospheric dynamic viscosity with value temperature. The titanium/vanadium kinetics we adopted is taken from Cloutman(2000) for the corresponding back- listed in Table5. For photolysis, we include photodissocia- ground gas. We have again assumed large particle size to tion of TiO, TiO2, TiH, TiC, and VO. We estimate their UV simplify the slip correction factor (the correction for non- cross sections from FeO (Chestakov et al. 2005) at 252.39 continuum) to unity in Equation (15). In this work, we have nm and scale the photolysis threshold according to their bond implemented and will demonstrate the condensation of H2O, dissociation energy. NH3,S2, and S8 in the following sections. 2.8. Photochemical Hazy Precursors 2.7. Chemistry of Ti and V Compounds Observations have informed us that clouds or photochemi- TiO (titanium oxide) and VO (vanadium oxide) are present cal hazes are ubiquitous in a diverse range of planetary atmo- in the gas phase in cool stars and brown dwarfs where tem- spheres. Microphysics models that include processes such perature exceeds 2000 K. The highly irradiated hot Jupiters as nucleation, coagulation, condensation, and evaporation of have been suggested to manifest inverted temperature struc- particles (e.g., Gao & Benneke 2018; Kawashima & Ikoma tures due to the strong optical absorption of TiO and VO va- 2019; Lavvas & Koskinen 2017) simulate the formation and por (Hubeny et al. 2003) in the stratosphere. The pioneering distribution of various-size aerosol particles. Given the com- work proposing the role of TiO and VO in irradiated atmo- plexity and uncertainty of the polymerizing pathways, one sphere (Fortney et al. 2008) is based on equilibrium chem- common approach is to select precursor species as a proxy istry, where the authors argue that the conversion between and assume they will further grow into complex hydrocar- TiO and TiO2 is fast enough for TiO to remain in chemical bons (Morley et al. 2013; Kawashima & Ikoma 2018). Typi- equilibrium. However, it is not clear for conversion reactions cal choices of haze precursors include C2Hx and HCN, which with Ti or other titanium species. For example, the intercon- is also limited by our kinetics knowledge and computing ca- version of CO ←−→ CO2 is relatively fast but the ultimate pacity. CO abundance is still controlled by the slower CO ←−→ CH4 In this work, we preferentially consider precursors that are interconversion. In addition to TiO, titanium hydride (TiH), more closely related to forming polycyclic aromatic hydro- has been suggested important in brown dwarfs by Burrows carbon (PAH) or nitriles. PAH is a group of complex hy- et al.(2005). As the thermodynamics data of TiH is not avail- drocarbon made of multiple aromatic rings, which has been able in the literature or standard databases, Burrows et al. commonly found in the smog pollution on Earth and ex- (2005) perform ab initio calculations of the Gibbs free en- pected to be associated with the organic haze on Titan (Zhao ergy of TiH (based on the partition function obtained from et al. 2018). In the polar region of Jupiter where charged the spectroscopic constants). To explore the kinetics of titanium and vanadium, we expand the species list to include Ti, 8 TiO, TiO2, TiH, TiC, TiN, V, VO. As only Ti, TiO, and TiO2 To verify our approach, we compared the activation energy estimated from are available for titanium compounds in the NASA polyno- the enthalpy difference to that of well measured reactions. e.g., endothermic reactions H2O + H −−→ OH + H2 and CO2 + H −−→ CO + OH have mials, we adopt the thermodynamics data of TiH from Bur- activation energy 10800 K (Davidson et al. 1989) and 13300 K (Tsang rows et al.(2005), TiC from Woitke et al.(2018), and the rest & Hampson 1986), respectively, whereas our estimate yields 7200 K and from Tsuji(1973). 10300 K, respectively. 8 particles are the main energy source, ionchemistry has also V11 has been developed (Venot et al. 2019), with the mo- been suggested to promote the formation of PAHs and or- tivation to support computationally heavy simulations. In ganic haze (Wong et al. 2003). Once the first aromatic ring, particular, the controversial methanol mechanism, which has benzene, has formed, the thermodynamics state (enthalpy been identified to cause the differences in CH4-CO conver- and entropy) does not vary much with the processes of at- sion (Moses et al. 2011; Moses 2014), is further updated and taching and arranging the rings. From the kinetics point analyzed in (Venot et al. 2020). Therefore, aiming to con- of view, the classic mechanism of making complex hydro- solidate the model discrepancy, we run an additional model carbons, H-Abstraction-Carbon-Addition (HACA), requires with VULCAN but implemented with the updated reduced aromatic hydrocarbon and acetylene in the primary abstrac- network from Venot et al.(2020). The planetary parameters tion and addition steps (e.g. Frenklach & Mebel 2020). It and model setting are listed in Table3. Before diving into the is conceivable that benzene formation is the rate-limiting detailed comparison, we provide an overview of the chemical step in forming complex hydrocarbons as the growth rate profiles and absorption properties for HD 189733b and HD increases downstream from benzene. In practice, while the 209458b in Figure1. fundamental pathways leading to PAH remain elusive (Wang 2011; Zhao et al. 2018), the combustion study can provide a 3.1.1. Disequilibrium Effects good handle on the formation of benzene to a certain degree. The left panels of Figure1 depict how vertical mixing and Therefore, we suggest considering benzene as an important photochemistry drives the compositions out of equilibrium haze precursor. on HD 189733b, by isolating the two effects. The under- One important caveat about modeling benzene is that its lying processes can be understood as a general property of photodissociation branches are poorly quantified across vari- hot Jupiters, as discussed in (Moses et al. 2011; Venot et al. ous branches (see e.g. Lebonnois 2005). The main photolysis 2012; Moses 2014; Hobbs et al. 2019; Molaverdikhani et al. products are possibly phenyl radical (C H ) and benzyne rad- 6 5 2019). Equilibrium chemistry prevails in the deep, hot region ical (C H )(Suto et al. 1992). If they further absorb photons 6 4 whereas energetic photons dissociate molecules and produce again, they could fragment into smaller, linear molecules like reactive radicals in the upper atmosphere. Between the two C H and C H . We adopt the cross sections of C H from 4 3 3 3 6 6 regions, the composition distribution is controlled by vertical Boechat-Roberty et al.(2004) and Capalbo et al.(2016). For transport, viz., species in equilibrium at depth are transported simplicity, we assume the main dissociation of benzene pri- upward and become quenched when vertical mixing predom- marily goes into phenyl radical (C H ) with a small fraction 6 5 inates chemical reactions; photochemical products are also leading to C H (∼ 15% based on (Kislov et al. 2004)). 3 3 mixed downward and initiate a sequence of reactions. Although HCN is the basic molecule for nitrile chemistry, The right panels of Figure1 show the UV photosphere it is unlikely that most of HCN will convert into complex ni- where the optical depth equals one, with decomposition of triles. The nitrile formation is more likely to be limited by the contribution from the main molecules. Our photochemical less abundant H CN, CH NH, or CH CN. Hence we include 2 2 3 model captures several general transmission properties of ir- these species along with HC N to represent the nitrile family 3 radiated H -atmospheres: H provides the dominant absorp- precursor. For sulfur gases, in addition to the condensation 2 2 tion in EUV (10–120 nm) whereas H2O and CO are the of sulfur allotropes (Sx), we also consider CS according to 2 dominant absorbers in FUV (120–200 nm). The window the laboratory experiments by He et al.(2020). Overall, we around 160–200 nm is particularly important for water disso- compose the following species as photochemical haze pre- ciation, which makes a catalytic cycle turning H into atomic cursors: C H ,C H ,C H ,C H , HCN, HC N, CH NH, 2 2 2 2 6 4 2 6 6 3 2 H(Liang et al. 2003; Moses et al. 2011). In the NUV (300– CH CN, CS . 3 2 400nm), radiation can penetrate deep down to ∼ 1 bar until 3. MODEL VALIDATION being scattered. The photospheres in Figure1 descend from about 1 µbar to 10 mbar (from the end of H2-shielding to the 3.1. HD 189733b tail of ammonia absorption) which denote the photochemi- We have benchmarked our thermochemical kinetics results cally active region in the atmosphere. using a C-H-O network with vertical transport against Moses HD 209458b shares qualitatively similar results with HD et al.(2011) for HD 189733b and HD 209458b in Tsai et al. 189733b. Owning to its higher temperature and the inverted (2017). In this work, we compare our results including N-C- thermal structure (see Figure 1. in Moses et al.(2011)), the H-O photochemistry to Moses et al.(2011) and Venot et al. quench level is lifted higher and the photolysis has little in- (2012) (M11 and V12 hereafter). V12 use a chemical ki- fluence, as can be seen in Figure1. The composition distribu- netics scheme derived from combustion application and find tion on HD 209458b can be described by a lower equilibrium different disequilibrium abundances of CH4 and NH3 from region and an upper quenched region. We will now only fo- those in M11. Since then, a size-reduced network based on cus on HD 189733b for the model comparison as disequilib- 9

Table 1. Model Validation Setup

Planet P-T profile Networka stellar UV Gravityb Upper Boundary Lower Boundary (cm2/s) HD 189733b Moses et al.(2011) N-C-H-O Eps Eri c 2140 H escaped zero-flux Moses et al.(2005) N-C-H-O-lowT Gueymard(2018) 2479 H O, CO, CO zero-flux Jupiter 2 2 + dry adiabat inflow

Earth COSPAR S-N-C-H-O-full Gueymard(2018) 980 H, H 2 escape Table2

a files available in supplementary material b at the surface for Earth and defined at 1 bar for gaseous planet c from the StarCat database (https://casa.colorado.edu/∼ayres/StarCAT) (Ayres 2010) and following the same scaling adjustment as Moses et al. (2011) d Assuming diffusion-limited escape rate rium processes contribute more compared to the hotter HD high energy barrier of the reaction. In response, Venot et al. 209458b (see Hobbs et al.(2019) for model comparison of (2019) removed the controversial reaction by Hidaka et al. HD 209458b). (1989) and updated their chemical scheme with a newly validated CH3OH combustion work (Burke et al. 2016), given 3.1.2. Model Comparison with Moses et al.(2011) and Venot the importance of methanol as an intermediate species for et al.(2012) CH4-CO conversion. Intriguingly, Venot et al.(2019) still The HD 189733b model comparisons between VULCAN, find a methane abundance rather close to that in V12. M11, and V12 are showcased in Figure2, where the top Attempting to resolve this mystery, we further run our row highlights the major species and the following rows are model with the Venot et al.(2020) reduced scheme 9 inte- grouped into carbon, oxygen, and nitrogen species. For the grated with new CH3OH mechanism. We did not incorporate major species, VULCAN produces profiles more consistent the same photolysis scheme from V12 but here photolysis has with M11, while there are notable differences with V12 in no effects on the quenching comparison below 1 bar. Con- H, CH4, NH3, and HCN. CH4 and NH3 are quenched from trary to the findings in Venot et al.(2020), the new scheme of below 1 bar level until being attacked by H around 1 mbar. Venot et al.(2020) implemented in our model indeed shows a Hence the differences with V12 in the photospheric region slower CH4-CO conversion and brings the CH4 profile closer (∼ 1 bar – 1 mbar) are due to thermochemical kinetics, to VULCAN and M11 (dotted line in Figure2-(b)). Our rather than photochemical sources. Nitrogen species gener- model implemented with the Venot et al.(2020) scheme pre- ally manifest higher variances, reflecting the kinetics uncer- dicts a quenched methane mixing ratio 1.13 ×10−5, close to tainties. 1.51 ×10−5 in M11 and 1.26 ×10−5 in our nominal model, whereas V12 with the faster methanol decomposition from Quenching of CH and NH — 4 3 Hidaka et al.(1989) predicts 5.20 ×10−6. We conclude that The sharp gradients of the equilibrium distribution of CH 4 the methanol decomposition indeed results in faster CH -CO and NH (Figure1) imply the abundances are sensitive to the 4 3 conversion and subsequently lowers the CH abundance in quench levels, viz. small differences in the quench levels can 4 V12. lead to considerable differences. The key reactions responsi- For nitrogen chemistry, the high-temperature kinetics is ble for the conversion at quench levels deserve a closer look. more uncertain and many reducing reactions relevant for H - The match of quenched CH abundance between VUL- 2 4 atmospheres are not available on the NIST database. We CAN and M11 has been discussed in Tsai et al.(2017), in drew data from the combustion literature (Dean & Bozzelli which we identify a similar pathway of CH destruction as 4 (2000), same as M11) and the KIDA database. In particu- M11. The inclusion of nitrogen does not change the fact lar, there are considerable uncertainties regarding the rates since nitrogen does not participate in the CH -CO conver- 4 for the reactions that control the NH -N conversion, as ex- sion. It can be seen that CH is quenched at a higher level 3 2 4 tensively discussed in Moses(2014). We follow the sugges- with lower mixing ratio in V12, as a result of faster CH -CO 4 tions in Moses(2014) and adopt the rate coefficient of NH + conversion. Moses(2014) identified the faster methanol de- 3 composition H + CH3OH −−→ CH3 + H2O measured by Hi- daka et al.(1989) adopted in V12 as the key reaction that CH 4 9 The reduced scheme captures the key reactions at work from V12 and has exhibits a shorter timescale in V12. Moses(2014) suggested been benchmarked against V12 (Venot et al. 2019). The two schemes are the rate is overestimated by Hidaka et al.(1989) based on the approximately equivalent regarding the quenching of main species. 10

HD 189733b Photosphere 10 8 10 8 =1

H2 10 6 10 6 H2O CO H CH4 10 4 10 4 CO H2 2 NH H2O 3 10 2 CO 10 2 HCN

CO2 Rayleigh Pressure (bar) Pressure (bar) CH4 0 0 10 N2 10

NH3 HCN 102 102 C2H2

10 14 10 12 10 10 10 8 10 6 10 4 10 2 100 50 100 150 200 250 300 350 Mixing Ratio Wavelength (nm)

HD 209458b Photosphere 10 8 10 8 =1

H2 10 6 10 6 H2O CO H CH4 10 4 10 4 CO H2 2 NH H2O 3 10 2 CO 10 2 N2 Rayleigh CO2 Pressure (bar) Pressure (bar) CH4 100 N2 100 NH3 HCN 102 102 C2H2

10 14 10 12 10 10 10 8 10 6 10 4 10 2 100 50 100 150 200 250 300 350 Mixing Ratio Wavelength (nm)

Figure 1. C-H-N-O Photochemical kinetics results (top-left) of HD 189733b (solid), compared with including vertical mixing but no photochemistry (dashed), and thermochemical equilibrium (dotted). The temperature-pressure structure and eddy diffusion (Kzz) proﬁle are taken from the dayside-average proﬁle in Moses et al.(2011) (their Figure 1 and 2). On the top right, we show the pressure level where energetic photons are mostly absorbed, i.e. optical depth τ = 1 (black), and decomposed into the main absorbers. The bottom panels show same as the top panels except for HD 209458 b.

NH2 −−→ N2H3 + H2 from Dean et al.(1984) and that of where the rate-limiting step switches from (16)-(i) to (16)-(ii) NH2 + NH2 −−→ N2H2 + H2 from Klippenstein et al.(2009), with increasing pressure. In the region with pressure between since Konnov & De Ruyck(2001) used in V12 is measured 30 and 1 bar, we ﬁnd two pathways with close contribution: at low temperatures. As NH3 progressively become fully quenched in the region between a few hundreds bar and 1 bar, there are more than a single pathway and rate-limiting step for NH3-N2 conversion that effectively control the NH3 abundance. For pressure greater than ∼ 30 bar, we identify the pathway 2(NH3 + H −−→ NH2 + H2)

NH3 + H −−→ NH2 + H2 NH2 + H −−→ NH + H2

NH3 + NH2 −−→ N2H3 + H2 (i) NH + NH2 −−→ N2H2 + H (iii)

M N2H2 + H −−→ N2H + H2 N2H3 −−→ N2H2 + H (ii) (17) (16) M N2H2 + H −−→ N2H + H2 N2H −−→ N2 + H M M N2H −−→ N2 + H H2 −−→ 2 H

net : 2 NH3 −−→ N2 + 3 H2. net : 2 NH3 −−→ N2 + 3 H2. 11 and abundant carbon-bearing molecule next to CO in the upper atmosphere. We identify the pathway in the HCN-dominated NH + H −−→ NH + H 3 2 2 region between 1 mbar and 1 µbar as NH2 + H −−→ NH + H2 CH4 + H −−→ CH3 + H2 NH + H −−→ N + H2 NH3 + H −−→ NH2 + H2 NH3 + N −−→ N2H + H2 (iv) (18) NH + H −−→ NH + H M 2 2 N2H −−→ N2 + H NH + H −−→ N + H2 M H2 −−→ 2 H CH + N −−→ H CN + H 3 2 (20) H2CN + H −−→ HCN + H2 net : 2 NH3 −−→ N2 + 3 H2. 2(H O −−→hν OH + H) where (17)-(iii) and (18)-(iv) are the rate-limiting steps. 2 Our pathways (16) and (17) are identical to those in M11 2(OH + H2 −−→ H2O + H) ((5) and (6) in Moses et al.(2011)), although we ﬁnd (16)-(i) net : CH4 + NH3 −−→ HCN + 3 H2 still play a role for controlling NH3 quenching, even with the high energy barrier given by Dean et al.(1984). As we adopt , which is identical to (14) of Moses et al.(2011). HCN in the same rates for several key reaction relevant for NH3-N2 V12 naturally follows the more scarce CH4 and NH3 and conversion, our model reproduces NH3 very close to M11, presents in a lower abundance. We note that Pearce et al. whereas V12 with a faster NH3-N2 conversion predicts a (2019) have run simulations and discovered previous un- higher quench level and lower abundance for NH3 (Figure known rate coefﬁcients, e.g., the destruction of HCN by re- 2-(b)). In all, we reiterate that further investigation for the acting with the excited N(2D) could be an important sink of key reactions (e.g., (16)-(i), (16)-(ii), (17)-(iii), (18)-(iv)) at HCN. high temperatures are required to improve our ability to ac- Photolysis Effects — curately model the NH -N system. 3 2 In the upper stratosphere above 1 mbar, the model differ-

Production of CO2 and HCN — ences most likely come from photochemical sources. How- Another unexpected change in Venot et al.(2020) is that ever, it is less straightforward to compare model discrepancy CO2 remains in chemical equilibrium across the atmosphere. originated from photochemistry as each step in converting Our model with the implementation of Venot et al.(2020) photon fluxes into photolysis rates can give rise to deviation, scheme confirmed the same result. This is remarkably dif- including stellar fluxes, cross sections, branching ratios, and fered from all other models, including V12, where CO2 is radiative transfer, etc. For simplicity, we will directly inspect enhanced by photochemically produced OH: the computed photolysis rates from M11, V12, and VUL- CAN. We limit our comparison to water photolysis, owing to CO + OH −−→ CO2 + H (19) its importance of producing H radicals and the frontline role This reaction with the OH radical is expected to rapidly con- of H in reacting with molecules such as CH4 and NH3 (Liang vert CO into CO2 while the reaction rate is well studied own- et al. 2003; Moses et al. 2011). ing to its importance in the terrestrial atmosphere as well as Figure3 compares the photodissociation rates of the main hν combustion kinetics. The rate coefficient of Reaction (19) branch H2O −−→ OH + H computed by three models. The adopted in Venot et al.(2020): 2.589 × 10−16 (T /300)1.5 water photodissociation rate in VULCAN is about twice as exp(251.4 / T ), has a pre-exponential factor about two orders large as that in M11 and around one order of magnitude larger of magnitude smaller than the typical values listed on NIST, than that in V12. The H2O photolysis rates evidently cor- as compared in Figure 41. The slow CO oxidation shuts off relate with the H and OH profiles in Figure2-(a), -(e) and the CO2 production and makes CO2 retain chemical equilib- molecules in V12 (e.g. CH4) generally tend to survive to- rium in Venot et al.(2020). We are not sure if this rate con- ward higher altitude. The disagreement started even from the stant is part of the updated methanol scheme from Burke et al. top of the models, with the same deviation also found across (2016) at this point, as to our knowledge, the base network other photolytic species, such as CH4 and NH3. This im- in Burke et al.(2016) takes the rate coefficient of Reaction plies the model implementation of stellar fluxes is the first- (19) from Joshi & Wang(2006), which is consistent with the order contribution to photochemical differences. Although literature and faster than that in Venot et al.(2020). according to Venot et al.(2012), they found no differences in The dissociation of CH4 and NH3 leads to the formation of switching to the same stellar flux from M11 and suggested HCN, the primary photochemical product that coupled car- that Rayleigh scattering could be the source of disagreement. bon and nitrogen on HD 189733b. HCN becomes the most We have tested switching off Rayleigh scattering and found 12

(a) (b) 10 8 10 8

10 6 10 6

10 4 10 4

10 2 10 2 Pressure (bar) Pressure (bar) CH4 100 100 CO

CO2

NH3 102 102 HCN

10 7 10 6 10 5 10 4 10 3 10 2 10 1 100 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 Mixing Ratio Mixing Ratio

10 6 10 6

10 4 10 4

10 2 10 2

Pressure (bar) C Pressure (bar) 0 0 10 CH3 10

C2H2 C3H3

C2H4 C4H2 102 102 C2H6 C6H6

10 14 10 12 10 10 10 8 10 6 10 4 10 14 10 13 10 12 10 11 10 10 10 9 10 8 10 7 Mixing Ratio Mixing Ratio

(e) (f) 10 8 10 8 O HCO

OH H2CO 10 6 10 6 O2 CH2OH

CH3OH

10 4 10 4

10 2 10 2 Pressure (bar) Pressure (bar)

100 100

102 102

10 20 10 17 10 14 10 11 10 8 10 5 10 18 10 16 10 14 10 12 10 10 10 8 Mixing Ratio Mixing Ratio

Figure 2. Comparison of atmospheric compositions on HD 189733b computed by VULCAN (solid), Moses et al.(2011) (dashed), and Venot et al.(2012) (dashed-dotted), showing volume mixing ratios of main species ((a), (b)), carbon species ((c), (d)), oxygen species ((e), (f)), and nitrogen species ((g),(h)) (some species not included in V12). Additionally, dotted lines for CH4 and CO2 are from running VULCAN with the updated methanol scheme from Venot et al.(2020). negligible changes, since Rayleigh scattering only dominates resolution can contribute to the photolysis rates as well (see in the deep region where photochemistry has ceased (see Fig- AppendixB). Overall, more attention should be paid to cali- ure1). We note that potential errors with insufﬁcient spectral brating the stellar irradiation for future photochemical model 13

(g) (h) 10 8 10 8

10 6 10 6

10 4 10 4

10 2 10 2 Pressure (bar) Pressure (bar)

100 N 100 CN

N2 CH2NH

NH2 HNCO 102 102 NO HC3N

10 16 10 14 10 12 10 10 10 8 10 6 10 4 10 16 10 14 10 12 10 10 10 8 10 6 Mixing Ratio Mixing Ratio

Figure 2. (cont.) benchmarks and we suggest using H2O photolysis as a base- proﬁles predicted by M11, V12, and our model are consis- line. tent with the divergence of parent CH4, except that acetylene (C2H2) is also governed by atomic C in the upper atmosphere.

10 7 C2H2 is the most favoured unsaturated hydrocarbon on HD 189733b. In the CO-photolysis region, atomic C can couple 10 6 with nitrogen into CN and eventually produce C2H2 by dis-

10 5 sociation of HC3N. Yet we ﬁnd CH4 to still be the dominant source for producing C2H2 below 1 µbar via a pathway such 10 4 as

10 3 Pressure (bar)

10 2 VULCAN 10 1 M11 V12 2(CH4 + H −−→ CH3 + H2) 100 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 CH3 + H −−→ CH2 + H2 1 H2O OH + H photolysis rate (s ) CH2 + H −−→ CH + H2

CH + H −−→ C + H2 Figure 3. Comparison of the photodissociation rates (s−1) of the CH3 + C −−→ C2H2 + H (21) main branch of H2O in HD 189733b computed by VULCAN, M11 (Moses et al. 2011), and V12 (Venot et al. 2012). hν 2(H2O −−→ H + OH)

2(OH + H2 −−→ H2O + H)

Carbon Species Comparison — net : 2 CH4 −−→ C2H2 + 3 H2. Panels (c) and (d) in Figure2 show the same comparison for other important carbon-bearing species. Atomic carbon is liberated from CO photodissociation near the top atmosphere. CO photolysis appears to be stronger in M11 and generates more atomic carbon around µbar level. The carbon vapor exceeds saturation and can potentially condense in Our scheme predicts C2H2 with the maximum abundance a the upper atmosphere. We will examine the implication of C few factor smaller than V12 and about an order-of-magnitude condensation in Section 4.2.3. In the lower stratosphere, var- smaller than M11. ious hydrocarbon production is initiated by methane abstrac- Ethylene (C2H4) is the next most abundant hydrocarbon tion, i.e. H being successively stripped from CH4 to form after acetylene and peaks around 10 mbar. C2H4 and other more reactive unsaturated hydrocarbons. The hydrocarbon C2Hx production stems from CH3 association reaction via the 14 pathway possibly from the thermodynamic data difference between

hν JANAF and the NASA polynomial, as pointed out in Tsai H2O −−→ H + OH et al.(2017). All three models exhibit somewhat different OH + H2 −−→ H2O + H quench levels and profiles for CH2OH and CH3OH, which are generally important intermediates for CH -CO intercon- 2(CH4 + H −−→ CH3 + H2) 4 version Moses et al.(2011); Tsai et al.(2018); Venot et al. CH + CH −−→M C H 3 3 2 6 (22) (2020). Nevertheless, this does not reflect in the CH4 abun- C2H6 + H −−→ C2H5 + H2 dance since CH4 has already quenched in the deeper region. M The updated methanol scheme in Venot et al.(2020) also pro- C2H5 −−→ C2H4 + H vides more consistent CH2OH and CH3OH distributions with M11 and VULCAN. Since VULCAN adopted the same rate net : 2 CH4 −−→ C2H4 + 2 H2 coefficients from the ab initio calculation from M11 for the , where forming C2H6 is usually the rate-limiting step. The three methanol reactions, the difference between VULCAN abundances of C2H4 and C2H6 in our model are in agreement and M11 is more likely associated with reactions involving with M11 within an order of magnitude. CH2OH. The kinetics beyond C2 hydrocarbons becomes less constrained (Moses et al. 2011; Venot et al. 2015). As discussed Nitrogen Species Comparison — in Section 2.3, we intended to capture the major pathways of Panel (g) and (h) of Figure2 compare nitrogen-bearing species. N and NH2 follow the same quench level as NH3 producing C6H6 as a proxy for haze precursors without in- voking an exhaustive suite of hydrocarbons. In our model, (panel (b)), since they are part of the NH3-N2 conversion. Considerable amount of atomic N is produced above mbar C6H6 is formed by the pathway level by hydrogen abstraction of ammonia, similar to that of 9(H2O + hν −−→ H + OH) methane. Atomic N is oxidized by OH into NO in the up-

9(OH + H2 −−→ H2O + H) per atmosphere. NO reacts rapidly with atomic C into CN, as C-N bond is stronger than N-O bond. CN is an impor- 6(CH4 + H −−→ CH3 + H2) tant source of nitrile production, e.g., CN reacts with C2H2 4(CH3 + H −−→ CH2 + H2) to form HC3N. Our model shows a slower HC3N produc- 4(CH2 + H −−→ CH + H2) tion and predicts HC3N with a peak value about two orders 4(CH + H −−→ C + H2) of magnitude lower than M11. The carbon-nitrogen-bearing species are grouped in Fig- 2(CH3 + C −−→ C2H2 + H) ure2-(h). Since NH 3 quenched first in the deeper layers 2(C + C2H2 −−→ C3H2) than CH4, the quench levels of general carbon-nitrogen- 2(C3H2 + H + M −−→ C3H3 + M) bearing species also follow NH3. Despite in trace abundance, C3H3 + C3H3 + M −−→ C6H6 + M CH2NH and HNCO participate in HCN forming mechanism net : 6 CH4 −−→ C6H6 + 9 H2. and become important at high pressures. We find HCN formed around 10 mbar via CH NH and CH NH in a path- where the recombination of C H is the rate-limiting step 2 3 2 3 3 way identical to (7) in Moses et al.(2011). (akin to the cooler atmosphere of Jupiter (Moses et al. 2005)). Figure2-(d) shows that C H and C H predicted by our 4 2 6 6 We conclude that we validate our model of HD 189733b reduced scheme have considerably lower abundances than by thoroughly reproducing composition distribution within those in M11. Given the agreement of C H up until 10−5 3 3 the uncertainty range enclosed by M11 and V12. The ki- bar, we suspect that the differences of C H between VUL- 6 6 netics data we employed generally yields quenching behav- CAN and M11 are due to photodissociation effects from ior close to M11, while our model appeared to predict lower C H as well as other species such as CO. Given all the un- 6 6 C H ,C H ,C H , and HC N than M11 in the upper atmo- certainties and complexity as we mentioned in Section 2.8, 2 2 4 2 6 6 3 sphere. Contrary to what have been reported in Venot et al. we do not consider the predicted abundances of C H and 4 2 (2020), we find that the update methanol scheme in fact in- C H to be accurate, but it should rather serve the purpose 6 6 creases the quenched CH abundance and more consistent for accessing the haze precursors. 4 with that in M11 and this work. The photochemical part of Oxygen Species Comparison — the atmosphere is more complex to diagnose but we suggest Panels (e) and (f) of Figure2 compare oxygen-bearing the implementation of stellar fluxes is the main factor in the species. The deviation of O, OH, and O2 again follows the discrepancy between M11, V12, and VULCAN. discrepency in H O photolysis, similar to H. There is a mi- 2 3.2. Jupiter nor shift of the equilibrium abundance of H2CO in V12, 15

The modeling work for Jovian chemistry broadly falls into 10 8 10 8 updraft two categories addressing two main regions: the stratosphere downdraft 0 and the deep troposphere. The stratospheric compositions 10 6 10 6 10 are governed by photochemical kinetics with the main fo- 10 4 10 4 cus on understanding the formation of various hydrocarbons. 101 For stratospheric models, fixed mixing ratios or fluxes at the lower boundary need to be specified (Yung & Strobel 1980; 10 2 10 2 102

Moses et al. 2005). As for the deep tropospheric composi- Pressure (bar) 0 0 tions below the clouds with sparse observational constraints, 10 10 kinetics models attempt to infer the interior water content 3 2 2 10 based on other quenched species (Visscher et al. 2010; Wang 10 10 et al. 2016). Since chemical equilibrium is expected to hold 500 1000 1500 104 106 108 5 0 5 in the deep interior, the elemental ratio essentially controls T(K) Eddy Diffusion Vertical Velocity the reservoir of gases and vertical mixing determines the Coefficient (cm2 s 1) (cm 1) quenched compositions in the upper troposphere. In this validation, our objective is to validate the chemical scheme at low temperatures with observed hydrocar- Figure 4. The temperature, eddy diffusion, and deep vertical velocities used for our Jupiter model. The temperature and eddy diffusion bons and verify the condensation scheme. We take a general in the stratosphere are taken from Moses et al.(2005), while a dry approach by connecting the deep troposphere to the strato- 8 2 adiabat and uniform eddy diffusion with Kzz = 10 (cm /s) are as- sphere and solve the continuity equations consistently. Our sumed for the troposphere. The upward (positive) and downward lower boundary at 5 kbar is far down in the region ruled by (negative) vertical velocities are prescribed by Equation (24) with equilibrium chemistry and zero flux can be applied to the the maximum speed of 5 cm/s at 0.7 bar. lower boundary. In this setup, fixed-abundance lower boundary conditions are not required as in the stratosphere models 3.2.2. Comparing to Stratospheric Observations and Moses et al. (e.g., Moses et al. 2005; Hue et al. 2018). The compositions (2005) at the lower stratosphere are physically determined by condensation and transport from the troposphere in the model. The top panel of Figure5 displays the vertical distribution of key species computed by our model, compared to 3.2.1. Model Setup Moses et al.(2005) and various observations. First, CH 4 is the major carbon-bearing species across the atmosphere. It is The temperature profile in the stratosphere and top of the well-mixed until photolysis and separation by molecular dif- troposphere (above 6.7 bar) is taken from Moses et al.(2005) fusion take in place at low pressure. The CH4 distribution in and extended to 5000 bar following the dry adiabat, with T our model matches well with the observation (Drossart et al. = 427.71 K at 22 bar measured by the Galileo probe as the 1999). We verify that our treatment of molecular diffusion reference point. We use the same eddy diffusion profile for accurately reproduces the decrease of CH4 due to molecular the stratosphere as Model A of Moses et al.(2005), which diffusion above the homopause. is derived from multiple observations. The eddy diffusion is Second, our model successfully predicts the major C2 hy- assumed to be constant with 108 (cm2/s) in the convective drocarbons, which stem from CH4 photolysis in the strato- region below 6.7 bar. The temperature and eddy-diffusion sphere. Our model tends to predict lower abundances for the profiles adopted for our Jupiter model are shown in Figure4. unsaturated hydrocarbons C2H2 and C2H4 than Moses et al. Heavy elements in Jupiter are enhanced compared to so- (2005) in the lower stratosphere, but both profiles are within lar metallicity, except the oxygen abundance is still unclear. the observational constraints. The UV photosphere in Figure We assign the elemental abundances for the Jupiter model 5 indicates that CH4 predominates the absorption from Ly- × −3 as He/H = 0.0785 (Atreya et al. 2020), C/H = 1.19 10 α to about 150 nm. We find the main scheme of converting (Atreya et al. 2020), O/H = 3.03×10−4 (0.5 times solar), and CH4 to C2H6 in the upper atmosphere is N/H = 2.28×10−4 (Li et al. 2017). Sulfur is not included in our Jupiter validation for simplicity. We include condensation of H2O and NH3, assuming a single particle size with hν average radius equal to 0.5 µm for the cloud condensates. 2(CH4 −−→ CH3 + H) Oxygen sources from micrometeoroids are prescribed at the M 2(CH3 + CH3 −−→ C2H6) upper boundary at 10−8 bar following Moses et al.(2005), M (23) −2 −1 4 with influx (molecules cm s ) of H2O = 4 × 10 , CO = 2 H −−→ H2 6 4 4 × 10 , and CO2 = 1 × 10 . net : 2 CH4 −−→ C2H6 + H2 16 and the photodissociation branch of methane is replaced by hν 1 1 10 8 CH4 −−→ CH2 + H2 followed by CH2 + H2 −−→ CH3 + CH4 H at higher pressures. We confirm that hydrogen abstraction CO 10 6 and three-body association reactions are sensitive to the for- C2H2 C2H4 mation of hydrocarbons on Jupiter as discussed in detail in C2H6 10 4 Moses et al.(2005). Particularly in the lower stratosphere C4H2 C H where temperature drops below 200 K, the rate constants 6 6 10 2 fall out of the valid temperature range or are not well con- Drossart et al.(1999) Bézard et al.(2002) strained. We find it is particularly important to adopt the low- Pressure (bar) 100 Moses et al.(2005) temperature rate constants for CH4 and C2Hx recombination Gladstone et al.(1999) M M Romani et al.(2008) reactions, i.e. CH3 + H −−→ CH4, H + C2H2 −−→ C2H3, 102 Bézard et al.(2002) M M Gladstone et al.(1999) H + C2H3 −−→ C2H4, H + C2H4 −−→ C2H5, and H + Fouchet et al.(2000) M C2H5 −−→ C2H6. We also adopt the limit of rate constants 10 12 10 10 10 8 10 6 10 4 10 2 below certain threshold temperatures derived by Moses et al. Mixing Ratio (2005). Third, our condensation scheme predicts the location of 10 8 H water-ice clouds starts at 3.6 bar and ammonia clouds at 0.7 H2O bar as shown in Figure5, consistent with the thermodynam- 10 6 NH3 H2O ice ics prediction with 0.5 solar O/H (Atreya et al. 2005; Wei- NH3 ice 10 4 H2O denschilling & Lewis 1973). The ammonium hydrosulfide saturation

NH3 (NH4SH) is not considered since sulfur is not included. Last, saturation 10 2 our model produces lower abundances of C4H2 and C6H6 is

produced at higher altitude compared to those in Moses et al. Pressure (bar) 0 (2005), which reflects the uncertainties in high-order hydro- 10 carbons and the photolysis branches of C6H6. 102 3.2.3. Spatial Variation of Ammonia Due to Vertical Advection During the Juno spacecraft’s first flyby in 2016, the mi- 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 Mixing Ratio crowave radiometer (MWR) on Juno measured the thermal emission below the clouds, which was inverted to global dis- Photosphere tribution of ammonia from the cloud level down to a few hun- =1 10 7 dreds bar level. A plume-like feature was curiously seen as- H2 sociated with latitudinal variation of ammonia (Bolton et al. CH4 2017). To explore the local impact of advection, we test how 10 5 C2H2 C2H6 the upward and downward motion in a plume can shape the CO 3 deep ammonia distribution in Jupiter. 10 Rayleigh Although Galileo probe has provided constraints on the 10 1 structure of Jupiter’s deep zonal wind (Atkinson et al. 1997) Pressure (bar) and Juno also sheds light on the vertical extension of the 101 zonal wind (Stevenson 2020), we do not have observational constraints on the deep vertical wind. Hence we consider a 103 simple but physically motivated (mass-conserving) vertical 50 100 150 200 250 300 wind structure without tuning to fit the data. We assume an wavelength (nm) updraft and downdraft plumes starting from the bottom of NH -ice clouds at 0.7 bar, in addition to eddy diffusion, as 3 Figure 5. The top panel shows the vertical mixing profiles of impor- depicted in the right panel of Figure4. For the non-divergent tant chemical species in our Jupiter model (solid), compared with advection to conserve mass in a 1-D column, the vertical ve- various observations (data points) of hydrocarbons and the strato- locity at layer j with number density nj follows spheric distributions from Model A of Moses et al.(2005) (dashed). We follow Rimmer & Helling(2016) placing a factor of two error vjnj = vtopntop = constant (24) bars in pressure when they are not given in the observational data. The vapor mixing ratios and cloud densities (g/cm3) of the conden- such that the net flux remains zero at each layer. For this sible H2O and NH3 are displayed in the middle panel. The bottom test, we arbitrarily choose the maximum wind velocity at panel illustrates the UV photosphere where τ = 1 with decomposition of main absorbers. 17

3.3. Present-Day Earth

1 10 Our chemical network has only been applied to hydrogen- dominated, reducing atmospheres so far. In this section, we 100 validate our full S-N-C-H-O network with the oxidizing atmosphere of present-day Earth. The interaction with the surface is particularly crucial in regulating the composition for 101 the terrestrial atmosphere. Surface emission and deposition

Equilibrium via biological and geological activities have to be taken into 2

Pressure (bar) 10 Kzz account. Our implementation of the top boundary ﬂuxes and updraft + Kzz condensation scheme has been validated for Jupiter in the downdraft + Kzz 103 Juno lat. = 2 N previous section. We will proceed to verify the lower bound- Juno lat. = 12 N ary with surface emission and deposition in the Earth model.

0 50 100 150 200 250 300 350 400 450 NH3 Mixing Ratio (ppm) Table 2. Lower Boundary Conditions for the Earth Validation

a b Species Surface Emission Vdep Figure 6. The deep ammonia distribution in parts per million (molecules cm−2 s−1) (cm s−1) (ppm) computed by our Jupiter model (black), while assuming COc 3.7 × 1011 0.03 chemical equilibrium, with eddy diffusion only, and including up- d 11 ward/downward advection for the updraft/downdraft branch (Figure CH4 1.6 × 10 0 4), respectively. The red and green profiles show the inferred am- NOd 1.3 × 1010 0.001 ◦ ◦ d 9 monia distribution at 2 N latitude and 12 N latitude based on Juno N2O 2.3 × 10 0.0001 microwave measurement by Li et al.(2017), where the shaded areas d 9 NH3 1.5 × 10 1 enclose the 16th and 84th percentile of the samples in their MCMC NO 0 0.01 runs. 2 NO3 0 0.1 SO d 9 × 109 1 the top to be 5 cm/s. This choice has advection timescale 2 d × 8 shorter than diffusion timescale in the lower pressure region, H2S 2 10 0.015 d 7 i.e. v K /H, which allows us to see the influence of COS 5.4 × 10 0.003 j & zz d 8 advection. Figure6 compares the computed distribution of H2SO4 7 × 10 1 e 8 ammonia to that retrieved from Juno measurements (Li et al. HCN 1.7 × 10 0.13 e 8 (2017); also see updates in Li et al.(2020)) at two different CH3CN 1.3 × 10 0.13 latitudes. First, the ammonia distribution predicted by chem- HNO3 0 4 ical equilibrium is rather uniform with depth, only slightly H2SO4 0 1 increasing from 350 ppm to 400 ppm. Next, vertical mixing by eddy diffusion alone makes ammonia quenched from a Global emission typically measured in mass budget (Tg/yr), which is the deep interior below 1000 bar and thus brings ammonia converted to molar flux with the surface area of the Earth = 5.1 × 1018 to a slightly lower but uniform concentration of 300 ppm. cm2 for our 1-D photochemical model. There is almost no visible difference while including the up- b Adopted from Hauglustaine et al.(1994) c ward advection since ammonia has already been quenched by Smithson(2001) d eddy diffusion from the deep region. Last, the uniform dis- Seinfeld & Pandis(2016) e Li et al.(2003) tribution of ammonia is altered in the downdraft, where the downward motion transports the lower concentration of NH3 from the condensing region. Our NH3 distribution is qualita- ◦ 3.3.1. Model Setup tively consistent with the NH3-depleted branch at 12 N from Li et al.(2017), where NH 3 reaches a local minimum around We follow Hu et al.(2012), taking the monthly mean tem- 7 bar. We emphisize that this shape cannot be reproduced by perature at the equator in January 1986 (CIRA-86) from the eddy diffusion alone. empirical model COSPAR International Reference Atmo- Although eddy diffusion is probably still essential in prac- sphere10 (Rees 1988; Rees et al. 1990) as the background tice for parametrizing a range of mixing processes, we temperature profile and the eddy diffusion coefficients from demonstrate that including vertical advection can be useful. Massie & Hunten(1981), as shown in Figure7. The winter The advection processes can especially play a bigger part in 2-D systems (Zhang et al. 2013; Hue et al. 2015). 10 https://ccmc.gsfc.nasa.gov/pub/modelweb/atmospheric/cira/cira86ascii 18

Eddy Diffusion Coefficient (cm2 s 1) 103 104 105 106

100 120 =1

100 O3 80 N2

H2O 80 CO2 60 CH4

60 N2O

Height (km) Rayleigh 40 Altitude (km) 40

20 20

0 0 160 180 200 220 240 260 280 300 50 100 150 200 250 300 350 Temperature (K) wavelength (nm)

Figure 7. The temperature (at the equator in January from CIRA- 86 with references described in the text) and eddy diffusion (Kzz) Figure 8. The UV photosphere i.e. optical depth τ = 1 (black) in profiles (Massie & Hunten 1981) for the Earth model. our Earth model, overlaid with the composition-decomposed photosphere for several key molecules. atmosphere has a colder and hence drier tropopause and bet- urally accounted for. We have also tried including the ab- ter represents the global averaged water vapor content (see sorption from atomic oxygen and nitrogen and found no dif- Chiou et al.(1997) and the discussion in Hu et al.(2012)). ferences regarding the neutral chemistry in the lower atmo- Unlike gas giants, terrestrial atmospheres typically do not sphere, since N2 and O2 have already screened out the bulk extend to a thermochemical equilibrium region. Instead, bio- EUV. The chlorine chemistry and lightning sources for odd chemical (e.g., plants and anthropogenic pollution) and ge- nitrogen are not included in this validation. ological (e.g. volcanic outgassing) fluxes provide surface sources and sinks that are key to regulate the atmosphere. For the lower boundary condition, global emission budget provides estimates for the surface fluxes, which is conventionally recorded in the unit of mass rate (Tg yr−1) and needed 3.3.2. Results to convert to flux (molecules cm−2 s−1) in our 1-D model. For Earth and any ocean worlds with large bodies of surface water reservoir, the standard setup is to fix the surface- Molecular oxygen (O2) and ozone (O3) are the main play- water mixing ratio (Kasting & Donahue 1980; Hu et al. 2012; ers in Earth’s photochemistry. O2 absorbs VUV below 200 Lincowski et al. 2018). We set the surface mixing ratio of wa- nm and O3 takes up the radiation longward than about 200 ter to 0.00894, corresponding to 25% relative humidity. Sur- nm, which blocks the harmful UV from life on the surface. face CO2 is also fixed at 400 ppm for simplicity, since we did The penetration level of solar UV flux shown in Figure8 in- not consider several major sources and sinks of CO2 at the dicates that ozone absorbs predominately between 20 km and surface, such as respiration, photosynthesis, ocean uptake, 50 km. The basics of oxygen–ozone cycle are described by weathering, etc. The specific emission fluxes and deposition the Chapman mechanism (e.g., Yung & DeMore 1999; Jacob velocities for the lower boundary are listed in Table2, while 2011). Our full chemical network encompasses the catalytic zero-flux boundary is assumed for all remaining species. We cycles involving hydrogen oxide and nitrogen oxide radicals initialize the atmospheric composition with well-mixed (con- that are responsible for the ozone sinks in the stratosphere. stant with altitude) 78% N2, 20% O2, 400 ppm CO2, 934 ppm Although the catalytic cycle of chlorine which accounts for Ar, and 0.2 ppb SO2. additional ozone loss is not included, we are able to repro- For the solar flux, we adopt a recently revised high- duce the observed global average ozone distribution in Figure resolution spectrum (Gueymard 2018), which is derived from 9. various observations and models (see Table 1. of Gueymard Our condensation scheme captures the cold trap of water in (2018)). The solar radiation was cut from 100 nm in Hu et al. the troposphere, i.e. the water vapor entering the stratosphere (2012) for the missing absorption from the thermosphere. We is set by the tropopause temperature. Above the tropopause, do not find it necessary as we set the top layer to the lower water is supplied by diffusion transport from the troposphere thermosphere around 110 km and the EUV absorption is nat- and oxidation of CH4. We find the conversion in the strato- 19

sphere go through the steps

100 US Standard H2O CH + OH −−→ H O + CH Atmosphere 1976 4 2 3 H2O saturation hν 1 2(O3 −−→ O2 + O( D)) 80 O3 1 NO 2(O( D) + N2 −−→ O + N2) NO2 60 N2O CH3 + O −−→ H2CO + H

HNO3 O + H2CO −−→ HCO + OH 40 Altitude (km) (25) HCO + O2 −−→ CO + HO2

CO + OH −−→ CO2 + H 20 HO2 + OH −−→ H2O + O2 M 0 10 14 10 12 10 10 10 8 10 6 10 4 10 2 100 2(H + O2 −−→ HO2) Mixing Ratio net : CH4 + 2 OH + 2 O3 −−→ CO2 + 2 H2O + 2 HO2

100 , effectively turning one CH4 molecule into two H2O CH4 CO molecules (Noel¨ et al. 2018). H2O eventually photodisso- CO2 ciated in the mesosphere and produced H2, as indicated by 80 NH3 the proﬁles in Figure9. Overall, our model produces wa- H2 OH ter distribution consistent with observations considering the 60 diurnal and spatial variations. Funke et al.(2009) Höpfner et al.(2016) The two oxides of nitrogen, NO and NO2, cycle rapidly in 40 Altitude (km) Ehhalt et al.(1973) the presence of ozone:

20 NO + O3 −−→ NO2 + O2

NO2 + O −−→ NO + O2 0 net :O3 + O −−→ 2 O2 10 14 10 12 10 10 10 8 10 6 10 4 Mixing Ratio Thus NO and NO2 are conventionally grouped as NOx. The burning of fossil fuel accounts for about half of the global tropospheric emission (e.g. Table 2.6 of Seinfeld & Pandis OCS 50 SO2 (2016)). NOx is mainly lost by oxidation into nitric acid H SO M 2 4 (HNO3): NO2 + OH −−→ HNO3. Our model reproduces 40 distribution of NOx, whereas our higher HNO3 in the upper stratosphere is seemingly attributed to missing the hydration 30 removal in the actual atmosphere. Nitrous oxide (N2O) is mainly emitted by soil bacteria, Altitude (km) 20 prescribed by the surface emission at the lower boundary. Barkley et al.(2008) Inn et al.(1979) There is no efﬁcient N2O removal reactions in the tropo- 10 Georgii et al.(1980) sphere and N2O remains well-mixed as one of the impor- Jaeschike et al.(1976) Viggiano and Arnold(1981) tant greenhouse gases. N2O is predominantly removed by 0 10 14 10 13 10 12 10 11 10 10 10 9 photodissociation in the stratosphere. Our calculated N2O is Mixing Ratio in agreement with the observations for the troposphere and stratosphere. Although similar to Hu et al.(2012), our model Figure 9. The global average vertical distribution of key composi- slightly overpredicts its abundance above 50 km, which is tions in present-day Earth’s atmosphere compared to observations. likely due to missing the photolysis branch of N2O that pro- * 1 The H2O mixing ratio is from the US Standard Atmosphere 1976 . duces excited oxygen O( S). ◦ ◦ Satellite observations of CO in the tropics and NH3 within 30 – 40 CH4 is the most abundant hydrocarbon in Earth’s atmo- ◦ ◦ N and 70 – 80 E in 2003 are measured by the Michelson Interfer- sphere, with the surface emission largely comes from human ometer for Passive Atmospheric Sounding (MIPAS) (Fischer et al. activities (e.g. agriculture) as well as natural sources (e.g. 2008). The rest unlabelled observational data are from Massie & wetlands). CH is oxidized into CO and eventually CO by Hunten(1981); Hudson & Reed(1979). When errors are not in- 4 2 cluded in the published observations, we follow Hu et al.(2012) OH through multiple steps similar to (25) in the stratosphere. placing one order of magnitude error bars for the diurnal and spatial variations. *e.g. https://www.digitaldutch.com/atmoscalc/help.htm 20

CO is produced by combustion activities with about 0.1 ppm concentration near the surface Seinfeld & Pandis(2016), as 100 a result of the balance among the emission ﬂux, OH oxdiza- dyn CH4 tion, and dry deposition. CO is continuously removed by CH4 + h 80 OH through the troposphere and generated by photodissocia- NH3 NH3 + h tion of CO2 in the thermosphere and mesosphere, as depicted CO by their distributions in Figure9. As the major oxidizing 60 N2O agent, OH is an important diagnostic species for Earth’s photochemical model. It is mainly produced in the stratosphere Altitude (km) 40 during daytime initiated by ozone photolysis and regenerated in the troposphere by NOx (see e.g., Jacob 2011). The OH 20 distribution in our model is consistent with that in Massie

& Hunten(1981). We will further discuss using calculated 0 104 105 106 107 108 109 1010 OH concentration to estimate the chemical timescale of long- Timescale (s) lived species against oxidation in the next section. Carbonyl sulfide (OCS) is the main sulfur species in the troposphere, emitted by direct outgassing or oxidation of car- Figure 10. Calculated chemical timescales of some environmen- tally important gases compared to the dynamical timescale of eddy bon disulfide (CS2) and dimethyl sulfide (DMS) released by diffusion in the Earth validation model. The thick lines indicate the the ocean (Seinfeld & Pandis 2016; Barkley et al. 2008). region where the oxidation is dominated by OH (i.e. τOH ' τchem). OCS is rather stable in the troposphere until entering the stratosphere where it is photodissociated or oxidized by OH where k is the rate coefficient of the oxidizing reac- and ultimately turned into sulfuric acid. Sulfur dioxide (SO ) A−OH 2 tion of A + OH. In the upper atmosphere where molec- is another important sulfur containing pollutant from fossil ular collision is less frequent, the excited O(1D) produced fuel combustion. SO oxidation begins from the troposphere 2 by ozone photolysis is not immediately stablized and be- with comes the main oxidant. We consider the two major oxi- SO + OH −−→M HSO 2 3 dizing paths across the atmosphere and write the chemical HSO3 radical rapidly reacts with oxygen to form SO3 timescale against oxidation as

HSO3 + O2 −−→ SO3 + HO2 A 1 τchem = 1 (28) followed by sulfuric acid formation kA−OH[OH] + kA−O(1D)[O( D)]

SO3 + H2O −−→ H2SO4 Figure 10 illustrates the chemical timescales (τchem) along with photolysis timescales (1/k ) for several trace gases, The sulfur-containing gases in our model generally agree photo where τOH (thick lines) inversely correlates with tempera- with the global distribution, while the mismatch of H SO 2 4 ture in general. We can gain some insights by comparing is expected as our model does not include H SO photodis- 2 4 (τ ) to the dynamical timescale of vertical mixing (τ sociation and heterogeneous reactions that efﬁciently remove chem dyn = H2/K ): In the troposphere, CH and N O display rather H SO from the gas phase. zz 4 2 2 4 well-mixed abundances due to their longer chemical lifetime. 3.3.3. Chemical Lifetime CO and NH3 have comparable τchem with τdyn and and ex- The oxidizing capacity of Earth’s atmosphere is important hibit negative gradient with altitude from oxidation removal. for decontaminating toxic and greenhouse gases, such as CO, In the stratosphere, NH3 is rapidly photodissociated while CH4 is transported from the troposphere and oxidized into CH4, and various volatile organic compounds . The oxidizing power is not only essential for regulating habitable conditions CO. In the thermosphere above ∼ 80 km, the oxidation by 1 but also key to address the stability of biosignature gases for O( D) takes over for most species, but mixing processes with other terrestrial planets. Here we present a brief overview of a shorter timescale here controls the chemical distribution. the key timescales for some important trace gases from our For example, CO abundance starts to increase with altitude Earth model. from about 60 km as a result of downward transport of CO OH radical is the primary daytime oxidizing agent in our produced by CO2 photodissociation in the upper atmosphere. biosphere. The chemical timescale of species A against ox- In summary, we validate our photochemical model with A HD 189733b, Jupiter, and Earth, for a wide range of tem- idization (τOH) can be estimated by the computed OH concentration as peratures and oxidizing states. The inclusion of nitrogen and sulfur chemistry, along with the implementation of advec- A [A] 1 τOH = = (27) tion, condensation, and boundary conditions are veriﬁed by kA−OH[A][OH] kA−OH[OH] 21

Table 3. Parameters of the planetary systems. from Moses et al.(2011)), we assume Kzz to be constant Parameter WASP-33b HD 189733b GJ 436b 51 Eri b in the convective region and increasing roughly with inverse aa (AU) 0.02558 0.03142 0.02887 11.1 square root of pressure in the stratosphere (Lindzen 1981; Parmentier et al. 2013). The expression takes a similar form Tint (K) 200 — 100/400 760 as Charnay et al.(2015) or Moses et al.(2016): Rs (R ) 1.51 0.805 0.464 1.45

Rp (RJ) 1.603 1.138 0.38 1.11 Ptran 0.4 gb (cm2/s) 2700 2140 1156 18197 Kzz = Kdeep( ) , (29) Pbar θc 58 48 58 67 stellar type A5 K1-K2 M2.5 F0 where Ptran is the transition pressure level informed by the radiative transfer calculation. The more irradiated atmospheres a orbital distance have deeper radiative-convective transition levels and greater b gravity at 1 bar level Ptran. c mean stellar zenith angle We run nominal models for all planets in this section with the S-N-C-H-O chemical network11. We recognize there are comparing with models and/or observations. The discrep- considerable uncertainties in sulfur kinetics, as discussed in ancies in previous models of HD 189733b are identified for Section 2.3. In order to gauge the uncertainty effects of our future investigation. sulfur scheme, we explore the sensitivity to sulfur chain- forming reactions for GJ 436b and OCS recombination for 4. CASE STUDY 51 Eridani b. After chemical abundances are obtained, we In this section, we select WASP-33b (ultra-hot Jupiter), use the open-source tool PLATON (Zhang et al. 2019, 2020) HD 189733b (hot Jupiter), GJ 436 b (warm Neptune), and to generate transmission spectra and HELIOS for the emis- 51 Eridani b (young Jupiter) to perform case studies. Each sion spectra. case represents a distinctive class among gas giants with H2- dominated atmospheres. The effective temperatures of these 4.1. WASP-33b objects span across 700–3000 K while having host stars of WASP-33b is among the hottest gas giants discovered various stellar types. We investigate how disequilibrium pro- with dayside temperature around 3000 K (von Essen et al. cesses play a part for these cases with additional attention on 2020). To date it remains the only case showing evidence the effects of sulfur chemistry and photochemical haze pre- of both temperature inversion and TiO features (Serindag cursors. et al. 2021), which makes WASP-33b an interesting target for All the P-T profiles in this section are generated using testing the stability of TiO/VO along with other molecules. the open-source radiative-transfer model, HELIOS, except Previous work on ultra-hot Jupiters are limited by the as- we keep the same P-T profile of HD 189733b as in Sec- sumption of chemical equilibrium chemistry (Kitzmann et al. tion 3.1 for comparative purposes. HELIOS employs two- 2018; Parmentier et al. 2018; Zhang et al. 2018). Here, we stream approximation and correlated-k method to solve for will verify the equilibrium assumption by exploring how dis- the radiative-convective equilibrium temperature consistent equilibrium processes affect the titanium and vanadium com- with thermochemical equilibrium abundances. The gaseous pounds with different C/O ratios. opacities include H2O, CH4, CO, CO2, NH3, HCN, C2H2, 4.1.1. Stellar UV-flux and Eddy Diffusion NO, SH, H2S SO2, SO3, SiH, CaH, MgH, NaH, AlH, CrH,

AlO, SiO, CaO, CIAH2−H2 , CIAH2−He, and additionally The host star WASP-33 is an A5 type star with effective TiO, VO, Na, K, H- for WASP-33b. The P-T profiles are temperature about 7400 K. We use the UV spectrum of HD fixed without taking into account of the radiative feedback 40136 (F0 type) merged with a 7000 K atlas spectrum from from disequilibrium chemistry (but see Drummond et al. Rugheimer et al.(2013) as an analogue. The star is fast ro- (2016) for the effects on HD 189733b). The astronomical tating and exhibits pulsations, which might add more uncer- parameters used are listed in Table3. The dayside-average tainties to the UV flux. Nevertheless, as we will see in Sec- stellar zenith angle is used for WASP-33b and GJ436b and tion 4.1.3, photodissociation solely converts more molecules the global-average stellar zenith angle is used for 51 Eri b to atoms at this high temperature and the results should be (see AppendixC), except that we keep the same value for qualitatively robust. HD 189733b to compare with the results in Section 3.1. The Vertical wind generally correlates with the planet’s effec- stellar UV fluxes adopted for each system are compared in tive temperature (Tan & Komacek 2019; Komacek et al. Figure 11, with detailed description in each section. For the eddy diffusion (Kzz) profiles in our case studies 11 (except that we again retain the same profile for HD 189733b included in the supplementary material 22

Eddy Diffusion Coefficient (cm2 s 1) 106 107 108 109 1010 1011 1012 1013 10 8

103 10 6 ) 2

m 1

c 10 4 1 10 m n 1 1 10 2 s 10

g r Pressure (bar) e ( 10 3 x

u 0 l WASP-33 10 F HD 189733 10 5 GJ 436 51 Eri 102

50 100 150 200 250 300 350 400 2000 2500 3000 3500 4000 4500 Wavelength (nm) Temperature (K)

Figure 12. The temperature-pressure and eddy diffusion (Kzz) pro- 106 ﬁles for WASP-33b. Solar elemental abundance (solid) and C/O = 1.1 (dashed) are assumed for calculating the temperature structure.

) 4 2 10 m c

1 due to the scarcity of oxygen, as oxygen preferably combines 102

m with the excess carbon to form CO (Madhusudhan 2012). n

1 0 Atomic titanium is the major species across this temperature s 10

g r range and TiC, TiH, and TiO have close abundances. e ( 2

x 10

u The effects of thermal dissociation on WASP-33b are

l WASP-33 F HD 189733 clearly visible in the equilibrium profiles in Figure 14. The 4 10 GJ 436 blistering heat of WASP-33b makes all elements predomi- 51 Eri nantly exist in the atomic form above 0.1 bar, where tem- 50 100 150 200 250 300 350 400 Wavelength (nm) perature starts to increase with altitude and exceeds 3000 K, while CO with the strong C–O bond is the only molecule that survives the high temperature. For solar C/O ratio, as Figure 11. Stellar UV fluxes normalized at 1 AU (top) and at the the majority of C is locked in CO, atomic C tracks the tem- top of the planet’s atmosphere (bottom) adopted in our case study perature structure whereas oxides such as H O, TiO, VO, and models. 2 TiO2 display inverse trends with temperature. For C/O = 1.1, atomic O swaps place with C and TiO and VO are signifi- 2019; Baxter et al. 2021). We assume the value of K based zz cantly depleted. on the simulations in Tan & Komacek(2019), where the global RMS vertical wind increases with decreasing pressure 4.1.3. The effects of Disequilibrium Chemistry and reaches about 100 m/s at 1 mbar (personal communica- For a typical hot Jupiter (e.g. HD 189733b), vertical mix- tion). The vertical wind translates to K ∼ 1011 cm2s−1 zz ing plays a major role controlling the chemical distribution in around 1 mbar. The temperature and eddy diffusion profiles the photosphere. However, it is not the case for WASP-33b as for WASP-33b are shown in Figure 12. we compare the equilibrium and disequilibrium mixing ratio profiles in Figure 14. Although the strength of eddy diffu- 4.1.2. Chemical Equilibrium sion also increases with temperature, faster thermochemical We first look at the trends associated with thermal dissocia- reactions still prevail upon vertical mixing. The deviation tion governed by thermochemical equilibrium under carbon- of disequilibrium profiles above the temperature-inverted re- poor and carbon-rich conditions, for which we assume a solar gion ( ∼ 10−4 bar) is due to photodissociation, which reduces C/O and C/O = 1.1, respectively. Figure 13 illustrates how ti- molecular species and produces more atoms. In the absence tanium compounds vary with temperature in equilibrium at of vertical quenching, the depleted TiO in a carbon rich con- 1 mbar. For solar C/O, titanium mainly exists in the form dition is unable to be replenished by vertical transport from of Ti and TiO. As temperature exceeds about 2500 K, TiO the deep region, as seen in Figure 14. In the photodissocia- becomes unstable against thermal dissociation and its abun- tion region, in principle, stronger vertical mixing can trans- dance falls with temperature. For C/O = 1.1, TiO is depleted port more molecules upward against photodissociation. We 23

ferences in both transmission and emission spectra. The pho- 10 6 tochemical region above the temperature-inverted region is Ti too optically thin, even when molecules like H O and TiO TiO 2 10 8 TiH are strongly photodisscoiated here. High-resolution spec- TiC troscopy might be more sensitive to probe the atomic species TiN in this region. 10 10 TiO2 Alternatively, the equilibrium abundances of TiO/VO are sensitive to the change of elemental abundance. Figure 15 12 10 demonstrates that the opacity in the optical is most sensitive to the change of TiO/VO as C/O is close to unity, which Mixing Ratio at 1 mbar 14 10 shows even greater variation than H2O absorption between 1.2 and 2 µm.

10 16 In conclusion, we ﬁnd photodissociation only impacts the 2000 2500 3000 3500 4000 4500 Temperature (K) upper atmosphere of WASP-33b where P < 0.1 mbar, chemical equilibrium is generally a valid assumption, as has been found for KELT9-b (Kitzmann et al. 2018) and ultra hot 10 6 Ti Jupiters with dayside temperatures above 3000 K. Atmo- TiO spheric mixing might still play an important role in an at- 10 8 TiH mosphere with temperature lower than WASP-33b. Our ﬁrst TiC TiN attempt to solve the kinetics of titanium species can provide 10 10 TiO2 an interesting avenue for investigating other transition metals such as Fe and Ca for future study of ultra hot Jupiters.

10 12 4.2. HD189733b

Mixing Ratio at 1 mbar We have benchmarked our model of HD 189733b in Sec- 10 14 tion 3.1, where we attempt to keep the astronomical and chemical setup as close to Moses et al.(2011); Venot et al. 10 16 2000 2500 3000 3500 4000 4500 (2012) as possible for comparison. In this section, we in- Temperature (K) clude the following updates and aspects that have not been considered in previous work: Figure 13. The equilibrium mixing ratios of several gas phase tita- • Recently observed stellar UV-ﬂux of HD 189733 (Bourrier nium species at 1 mbar as a function of temperature for solar ele- et al. 2020) mental abundance (top) and C/O = 1.1 (bottom). • Sulfur chemistry

• Condensation of carbon vapor have perform additional runs with eddy diffusion profile varied by a factor of 10. Yet we found the change is minor and 4.2.1. Stellar UV-flux our results are not too sensitive to Kzz. For sulfur species, sulfur atom S is also the favored form Bourrier et al.(2020) combine HST and XMM-Newton followed by hydrogen sulfide (SH). The formation of S2 and observations and derive semi-synthetic UV spectra up to 160 other polysulfur (Sx) is entirely shut down at this extremely nm. For our model benchmark in Section 3.1, solar flux is high temperature. Sulfur does not couple to oxygen, carbon, used for wavelengths below 115 nm and the observed spec- or nitrogen since it mostly remains in the atomic form. Over- trum of epsilon Eridani (a K2-type analogue) is adopted for all, the composition distribution of WASP-33b resembles that 115 - 283 nm. The previously adopted and newly observed of a hot Jupiter except without a vertical quench region. The stellar fluxes are compared in Figure 16. The EUV flux of atmosphere of WASP-33b can be divided into a photochem- HD 189733 is modestly higher than that of the Sun, but the ically influenced region and a thermochemical equilibrium photochemically important FUV (λ > 122 nm) flux appears region, with the transition at the top of temperature-inverted to be weaker. Nevertheless, the change in the UV flux turns layer around 10−4 bar. out to only slightly decrease the atomic H (by about 20%). The overall impact of the updated EUV flux on neutral chem- 4.1.4. Transmission Spectra istry is in fact insignificant. We have computed the synthetic spectra from equilibrium 4.2.2. Sulfur Chemistry and disequilibrium abundances and found no observable dif- 24

Solar C/O = 1.1 10 8 10 8 H H O O 10 6 C 10 6 C N N S S 10 4 4 OH 10 OH

H2O H2O

2 CO CO 10 10 2 SH C2H2

Pressure (bar) H2S Pressure (bar) SH 100 H2S 100

102 102

10 12 10 10 10 8 10 6 10 4 10 2 100 10 12 10 10 10 8 10 6 10 4 10 2 100 Mixing Ratio Mixing Ratio

10 8 10 8 Ti Ti V V 10 6 TiO 10 6 TiO VO VO

TiO2 TiO2 10 4 4 TiC 10 TiC TiH TiH

2 10 10 2 Pressure (bar) Pressure (bar)

100 100

102 102

10 16 10 14 10 12 10 10 10 8 10 6 10 16 10 14 10 12 10 10 10 8 10 6 Mixing Ratio Mixing Ratio

Figure 14. The composition proﬁles for the main species of interest for WASP-33b, assuming solar C/O (left) and C/O = 1.1 (right). The equilibrium abundances are plotted in dotted lines. We next run the same model except including sulfur ki- 0.0115 netics. The sulfur species from our photochemical calcula- Solar 0.0114 C/O = 0.75 tion are illustrated in Figure 17 and are broadly consistent C/O = 1.1 with previous work (Zahnle et al. 2009). Hydrogen sulﬁde 0.0113 (H2S) is the thermodynamically stable form of sulfur in a 0.0112 hydrogen-dominated atmosphere. H2S is mostly destroyed

0.0111 by hydrogen abstraction

0.0110 H2S + H −−→ SH + H2 (30) Transit Depth 0.0109 and restored by the reverse reaction of (30). The forward

0.0108 and backward reactions of (30) essentially dictates the level where H2S starts to loss its stability. On HD 189733b, H2S 0.0107 TiO H2O is dissociated above 1 mbar and predominantly turned into 0.4 0.5 0.6 0.7 0.8 0.91.0 1.2 1.4 1.6 1.82.0 3.0 4.0 Wavelength ( m) S. SH and S2 also reach maximum values at the level where H2S dissociation kicks off. The implication is both SH and S2 absorb shortwave radiation and could potentially provide Figure 15. Synthetic transmission spectra for WASP-33b computed stratospheric heating, especially with super-solar metallicity from modeled compositions assuming solar elemental abundance, condition as discussed in Zahnle et al.(2009). C/O = 0.75, and C/O = 1.1. The absorption features of TiO and We ﬁnd SO accumulated in the upper atmosphere from the H2O are indicated by the color bands. oxidation of S + OH → SO + H. The highly reactive SO is 25

such as 102 Sun + Eps Eri CH4 + H −−→ CH3 + H2 HD 189733 (Bourrier et al. 2020) CH3 + H −−→ CH2 + H2 101 CH2 + S −−→ HCS + H

100 HCS + H −−→ CS + H2

SH + H −−→ S + H2 10 1 S + OH −−→ SO + H CS + SO −−→ OCS + S (31) 10 2 Flux (erg s-1 nm-1 cm-2) OCS + H −−→ CO + SH

3 hν 10 H2O −−→ OH + H 20 40 60 80 100 120 140 160 180 hν Wavelength (nm) SH −−→ S + H

S + H2 −−→ SH + H Figure 16. The UV flux of HD 189733 received at 1 A.U. derived from recent observations Bourrier et al.(2020) compared to the pre- net : CH4 + H2O −−→ CO + 3 H2. viously adopted spectrum, which consists of solar EUV and epsilon Note there is no net change of sulfur species in the cycle. The Eridani following the same approach as Moses et al.(2011) and rate-limiting reaction in pathway (33) is the carbon-sulfur used in Section 3.1. The spectra are binned for clarity. step CH2 + S −−→ HCS + H, which is about three orders of magnitude faster than the pathway without sulfur around known to self-react into SO dimer ((SO) ) and may facilitate 2 1mbar. Interestingly, we find SH play an analogous role to formation of S O and S (Pinto et al. 2021) or back into S 2 2 H O in catalytically converting H to H, causing the minor H and SO . What actually happened in our model is SO either 2 2 2 increase in Figure 18. photodissociated or reacted with atomic H back to S in the The presence of sulfur species enhances the destruction low pressure region. The elemental S might be subject to of methane and might partly contribute to the scarcity of photoionization, as we will discuss in Section5. methane detection on hot Jupiters (e.g., Baxter et al. 2021, One notable effect of photochemistry with sulfur is sev- and references within). H S has also been reported to eral sulfur species absorb in the MUV/NUV. As illustrated in 2 speed up the oxidation of methane in combustion experiment Figure 17, sulfur species raised the UV photosphere above (Gersen et al. 2017), in the oxidizing and high-pressure con- ∼ 230 nm, compared to that without sulfur where no effi- ditions of gas engines. The decreasing of CH naturally re- cient absorption beyond the ammonia bands. We find H S 4 2 duces its offspring products to a great extent. The column responsible for the dominant absorbption in the NUV (300– density shown in Figure 37 reflects the reduction of haze pre- 400 nm), rather than SH as reported in Zahnle et al.(2009), cursors with the participation of sulfur. Based on our fiducial which might be caused by the isothermal atmosphere at 1400 analysis on HD 189733 b, we suggest that organic haze for- K used in Zahnle et al.(2009). The absorption of S between 2 mation is likely to be partly suppressed by sulfur kinetics on 250 and 300 nm and the SH peaks around 325 nm can make a hot Jupiter, as opposed to enhanced by sulfur kinetics in prospective observable features. a CO -rich condition suggested by experimental simulations Figure 18 highlights the compositional differences when 2 (He et al. 2020). sulfur is present. Sulfur species can play an interesting role in catalyzing conversion schemes that take multiple steps. In 4.2.3. Condensation of Carbon Vapor particular, CH4 is more diminished down to about 1 mbar. Atomic carbon vapor (C) is produced by CO dissociation We find sulfur provide a catalytic pathway for CH4-CO con- (including both photodissociation and thermal dissociation in version. As CH and H S react with atomic H to liberate 4 2 the thermosphere) or the reaction N + CN −−→ C + N2 above carbon and sulfur, they couple to form carbon monosulfide ∼ 0.1 mbar and also by H abstraction with CHx species in the (CS). Carbon in CS is further oxidized into OCS and even- lower region. The saturation vapor of C falls off rapidly with tually ends up in CO through H-abstraction, via a pathway decreasing temperature in the upper atmosphere, as shown in Figure 19. In fact, the disequilibrium abundance of C starts to exceed the saturation concentration above 10 mbar. The realistic timescale for graphite growth by condensation in- volves detailed microphysics and is beyond the scope of this study. As a simple test, we explore the kinetic effects after 26

1 10 8 =1 (no S) =1 6 1 10 SH

H2S

S2 1 10 4 Rayleigh

1 10 2 Pressure (bar) 1 100 1 2 10

11 111 11 1 1 1 1 1 1 50 100 150 200 250 300 350 400 Wavelength (nm)

Figure 17. Left: Mixing ratios of the major sulfur species computed in the model of HD 189733b. The photochemical kinetics results are in solid lines and equilibrium abundances in dotted lines. Right: The pressure level of optical depth τ = 1 as a function of wavelength while including (black) and excluding (grey) sulfur chemistry, along with the main individual contribution from sulfur species. to the model without sulfur). In the end, we ﬁnd that the con- 10 8 densation of C has limited effects to other gas compositions in the upper atmosphere.

10 6

10 4 10 8

10 2 10 6 H Pressure (bar) C 0 4 10 CH4 10

NH3

C2H2 102 2 HCN 10 C

10 8 6 4 2 0 Pressure (bar) 10 10 10 10 10 10 CH4 100 Mixing Ratio C2H2 HCN CS 102 Figure 18. Mixing ratio profiles of main species on HD 189733b C saturation that exhibit differences from models including sulfur kinetics (solid) 10 30 10 26 10 22 10 18 10 14 10 10 10 6 10 2 and without sulfur kinetics (dashed). Mixing Ratio carbon vapor is fully condensed. We run our HD 189733b Figure 19. Several carbon-containing species from the nominal model including sulfur chemistry and do not allow C to be- model (dashed) compared to those from the model with limited C come supersaturated but simply fix the abundance of C in the due to instantaneous condensation. The saturation mixing ratio of gas phase to its saturation mixing ratio. This is physically C is plotted in dotted curve. equivalent to assuming instantaneous condensation and un- limited condensation nuclei. Figure 19 demonstrates the consequences when C is in- 4.2.4. Transmission Spectra stantaneously condensed, which mainly impacts the region Here we first take a look at the observational consequences above 0.1 mbar. Without the condensation of C, CH4 can due to model uncertainties among Moses et al.(2011), Venot be replenished by the hydrogenation sequence of C (i.e. C et al.(2012), and VULCAN which we examined in Sec- → CH → CH2 → CH3 → CH4). This channel is closed as C tion 3.1. Figure 20 showcases the transmission spectra of condensed out and CH4 is further depleted in the upper atmo- HD 189733b generated from the compositions computed by sphere. CS is reduced in the same way but C2H2 and HCN re- VULCAN, Moses et al.(2011), and Venot et al.(2012). The main almost unaffected (they are already reduced compared lower quenched abundances of CH4 and NH3 in Venot et al. 27

(2012) are responsible for the primary spectral differences while the spectra from VULCAN and Moses et al.(2011) are With S fairly close. The ammonia absorption around 8–12 µm could Without S be a useful diagnosis for the quenching mechanism of nitro- 0.0244 Equilibrium McCulloughet al.(2014) gen chemistry. Overall, we ﬁnd the model uncertainties lead Pont et al.(2013) to about half of the spectral deviation caused by disequilib- 0.0242 rium chemistry.

We then examine the effects of including sulfur chemistry 0.0240

to the transmission spectra in Figure 21. While the features Transit Depth from sulfur-containing species are almost obscured by other 0.0238 SO molecules such as H2O and CH4 in the near-IR, there are 2 NH3 still visible differences due to sulfur’s impact on CH4 and CH4 0.0236 H2O NH3. Since the coupling to sulfur reduces the abundances of 1 2 3 4 5 6 7 8 9 10 12 CH4 and NH3, the transit depth is smaller in the presence of Wavelength ( m) sulfur. The differences caused by sulfur chemistry is smaller than that between equilibrium and disequilibrium CH4 and NH abundances but not trivial. Current observations are not Figure 21. Same as Figure 20 but with abundances computed from 3 our model while including or excluding sulfur species. capable of placing conclusive constraints and we need to rely on future facilities with higher resolving power (e.g., JWST, ARIEL). well as transmission spectroscopic study with HST WFC3 (Demory et al. 2007; Knutson et al. 2014). Spitzer 3.6 µm and 4.5 µm emission indicates the atmosphere is rich in CO/CO2 and poor in CH4. Yet precise constraint and the VULCAN mechanism on CO/CH4 ratio still remain inconclusive. For- M11 0.0244 V12 ward models have suggested high metallicity (Moses et al. Equilibrium 2013; Morley et al. 2017) and hot interior from tidal heating McCulloughet al.(2014) (Agundez´ et al. 2014) can explain the observed CO/CH but 0.0242 Pont et al.(2013) 4 inconsistent with the low water content (less than 10−4) obtained by the retrieval model Madhusudhan & Seager(2011). 0.0240 Hu et al.(2015) propose that a remnant helium-dominated Transit Depth atmosphere as a result of hydrogen escape can naturally de- 0.0238 plete CH4 and H2O. However, the Ly-alpha absorption still NH3 appears to indicate a hydrogen-dominated atmosphere for GJ CH4 0.0236 H2O 436b (Khodachenko et al. 2019). For this work, we restrict

1 2 3 4 5 6 7 8 9 10 12 ourself to 100 times solar metallicity (Neptune-like) and ex- Wavelength ( m) plore the effects of vertical mixing and internal heat with the presence of sulfur. Figure 20. Synthetic transmission spectra for HD 189733b gen- 4.3.1. Model Input erated from chemical abundances computed by VULCAN, Moses et al.(2011), Venot et al.(2012), and when assuming chemical equi- Following the best-fit parameters in Morley et al.(2017), librium. Transit observations from Pont et al.(2013) and McCul- we set up a low and a high internal heating scenario by run- lough et al.(2014) are shown as data points with error bars. The ab- ning HELIOS assuming Tint = 100 K and 400 K, respec- sorption features for the main molecules are indicated by the color tively. The stellar UV flux of GJ 436 is adopted from the bands. MUSCLES survey (version 2.2; France et al.(2016); Young- blood et al.(2016); Loyd et al.(2016)). The eddy diffusion profile is prescribed by (29) with Ptran = 1 bar, as where 4.3. GJ 436 b the radiative-convective transition is located in our radiative GJ 436b is a Neptune-sized planet in a close orbit around transfer calculation. We also explore the weak and strong a M dwarf star. This warm Neptune has received tremendous vertical mixing scenarios, based on the GCM simulation by attention since its first discovery (Butler et al. 2004), includ- Lewis et al.(2010). The average vertical wind from Lewis ing multiple primary transit and secondary eclipse observed et al.(2010) translates to effective eddy diffusion coefficient with Spitzer (Stevenson et al.(2010); Madhusudhan & Sea- from 108 cm2 s−1 at 100 bar to 1011 cm2 s−1 at 0.1 mbar, as- ger(2011); Morley et al.(2017) and references within), as suming the mixing length to be the atmospheric scale height. 28

2 1 Kzz (cm s ) abundance for low internal heating and conversely produces 106 107 108 109 1010 1011 8 10 lower quenched CH4 abundance for high internal heating. For low internal heating, CH4 remains in considerable

10 6 amounts in both weak and strong mixing cases, with photospheric CO/CH4 ratio about 25 and 4, respectively. The amount of methane efﬁciently converts to other hydrocar- 10 4 bons (e.g., C2H2 and C6H6) and HCN, especially in the pho- CO tochemically active region above 1 mbar. For high internal 10 2 CH4

Pressure (bar) heating, CH4 abundance is significantly reduced compared to that with low internal heating, regardless of vertical mixing. 0 10 The photospheric CO/CH4 ratio for weak and strong mixing Tint = 100 K is about 2000 and 7000. C2H2 and HCN are also diminished Tint = 400 K 102 with CH4, except CH4 can still be transported to the upper 500 1000 1500 2000 2500 3000 Temperature (K) atmosphere in the strong mixing scenario. For nitrogen species, N2 predominants under high metallicity and quenched NH3 exhibits greater abundances than Figure 22. The temperature-pressure and eddy diffusion (Kzz) pro- equilibrium in all cases. The NH3–N2 conversion is slower files for GJ 436b, showing low (Tint = 100 K) and high (Tint = 400 K) internal heating and weak (dashed) and strong (solid) vertical than CH4–CO and hence NH3 is quenched deeper than CH4. Photochemically produced HCN can take over NH and CH mixing. The [CH4]/[CO] = 1 equilibrium transition curve for 100 3 4 times solar metallicity is shown by the dotted curve. in the upper atmosphere, but at higher altitude compared to HD 189733b due to the weaker UV flux of GJ 436. Inter- Since this choice of mixing length generally overestimates estingly, the quench level of NH3 appears to vary little with the strength of eddy diffusion (Smith 1998; Parmentier et al. vertical mixing and mainly responds to the change of inter- 2013; Bordwell et al. 2018), we consider it as the upper limit nal heating (see the pressure and temperature dependence of and set it for the strong vertical mixing scenario. We corre- NH3–N2 conversion in Tsai et al.(2018)). The insensitiv- 8 2 −1 ity of NH3 to vertical mixing could provide additional con- spondingly have Kdeep = 10 cm s for the strong mixing 6 2 −1 straints to the deep thermal property. scenario and assume Kdeep = 10 cm s for the weak mixing scenario. 4.3.3. Effects of Sulfur Species

4.3.2. Effects of Vertical Mixing and Internal Heating Figure 24 shows the distribution of main sulfur species for each scenario. H2S still remains the major sulfur-bearing Resolving the CO/CH4 abundance ratio is the leading molecule. The region where H2S is stable extends to a lower question for the atmospheric compositions of GJ 436b. Since pressure of about 0.1 mbar compared to HD 189733b be- we did not vary the elemental abundance, the photospheric cause of less photochemically produced atomic H on GJ abundance of CH4 primarily depends on the quench level, 436b. Above the H2S-stable region, the sulfur species is more which is controlled physically by the strength of vertical mix- diverse than the S/H2S dichotomy in a hot Jupiter. The strato- ing and thermal structures. Figure 22 shows that the 100- spheric temperature of GJ 436b is too warm for elemental times solar metallicity constrains both temperature profiles sulfur to grow into large allotropes but allows rich interaction within the CO dominated region. As illustrated by the equi- of sulfur and oxygen species in the upper stratosphere. The librium profiles in Figure 23, for low internal heating (Tint = distribution is sensitive to mixing processes: SO2 takes over 100 K), the temperature is close to the CH4/CO transition and H2S for weak vertical mixing while S, S2, and SO are in turn −4 the equilibrium CH4 abundance oscillates below 10 bar, the leading sulfur species for strong vertical mixing. Since whereas the equilibrium CH4 abundance decreases mono- sulfur species do not quench in the deep region like CH4 and −4 tonically with increasing pressure from about 10 bar for NH3, they are not affected by internal heating. Instead, sulfur high internal heating (Tint = 400 K). The twisting variation species are more sensitive to photochemical products trans- of CH4 with depth was pointed out by Molaverdikhani et al. ported from the upper atmosphere. (2019), suggesting it can lead to a non-monotonic correla- Sulfur also impacts other non-sulfur species. Figure 25 tion with increasing Kzz for low internal heating. However, compares our models of GJ 436b that include and exclude in the physically-motivated range of Kzz we explored, CH4 sulfur chemistry. H is considerably reduced between 0.1 and is always quenched in the confined region below 1 bar where 10−4 bar in the presence of sulfur, opposite to what is seen CH4 increases with depth, as shown in Figure 23. There- on HD 189733b. This is because the photolysis of SH which fore, stronger vertical mixing produces higher quenched CH4 provides the catalytic H production on HD 189733b is ab- 29

Tint = 100 K Tint = 400 K 10 8 10 8

10 6 10 6

10 4 10 4

H2O

CO H2O 10 2 10 2 Weak Mixing CH Pressure (bar) 4 Pressure (bar) CO

CO2 CH4

NH3 CO2 100 100 HCN NH3

C2H2 HCN

C6H6 C2H2 102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 Mixing Ratio Mixing Ratio

10 8 10 8

10 6 10 6

10 4 10 4

H2O

CO H2O 10 2 10 2

Strong Mixing CH Pressure (bar) 4 Pressure (bar) CO

CO2 CH4

NH3 CO2 100 100 HCN NH3

C2H2 HCN

C6H6 C2H2 102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 Mixing Ratio Mixing Ratio

Figure 23. The mixing ratio profiles (solid) along with equilibrium profiles (dotted) of several main species on GJ 436b for different assumption of internal temperature and vertical mixing. The left/right columns correspond to low/high (Tint = 100 K/Tint = 400 K) internal heating and the top/bottom rows correspond to weak/strong vertical mixing. sent as SH is less favored on GJ 436b. Instead, hydrogen is Similar to HD 189733b, sulfur species raise the UV pho- recycled faster by H + H2S −−→ H2 + SH in the H2S-stable tosphere longward of 200 nm, as shown in Figure 26. The region of GJ 436b, as indicated in Figure 27. The reduction absorption feature around 200–300 nm reflects the aforemen- of H subsequently slows down the production of C2H2 and tioned sensitivity to vertical mixing, with SO2 predominating HCN, even when CH4 abundance is almost unchanged. In in the weak mixing scenario and SO and S2 in the strong mix- the photochemically active region above ∼ 0.1 mbar, atomic ing scenario. C preferably combines with sulfur into OCS or CS, which further lowers C2H2 and HCN in the upper atmosphere. As 4.3.4. Sensitivity to Sx Polymerization NH3 being oxidized by atomic O into NO in this region, nitrogen sulfide (NS) accelerates NH3 oxidization while cou- Since the growth from S2 to larger sulfur allotropes is sup- pling to sulfur, analogous to the role of HCS for destroying pressed in our GJ 436b model, we perform a sensitivity test CH4 on HD 189733b. The catalytic pathway for oxidizing to see if Sx beyond S2 can be produced with faster polysul- NH3 is fur recombination rates. The three-body recombination reactions that interconvert S2–S4–S8 are the main polymerization steps. We follow Zahnle et al.(2016) and raise the rate of S 4 NH3 + H −−→ NH2 + H2 recombination by 10 and that of S8 recombination by 100 for NH2 + S −−→ NS + H2 a faster polysulfur forming test. We find the abundances of S (32) 4 NS + O −−→ NO + S and S8 remain low and almost unchanged. We confirm that the stratosphere of GJ 436b is too warm for elemental sulfur net : NH3 + H + O −−→ NO + 2 H2. 30

Tint = 100 K Tint = 400 K 10 8 10 8

10 6 10 6

10 4 10 4 O O S S

S2 S2 10 2 10 2 Weak Mixing

Pressure (bar) SH SH

H2S H2S OCS OCS 0 0 10 SO 10 SO

SO2 SO2 NS NS 102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 Mixing Ratio Mixing Ratio

10 8 10 8

10 6 10 6

10 4 10 4 O O S S

S2 S2 10 2 10 2 Strong Mixing

Pressure (bar) SH SH

H2S H2S OCS OCS 0 0 10 SO 10 SO

SO2 SO2 NS NS 102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 Mixing Ratio Mixing Ratio

Figure 24. Same as Figure 23 but for atomic O and sulfur species. to grow beyond S2 into fair quantities, even after taken into and is favored in our weak mixing scenario, which can po- account the uncertainties in the sulfur polymerization rates. tentially be detectable by JWST/MIRI. In addition, S2 is more favored with strong vertical mixing and provides strong ab- 4.3.5. Transmission and Emission Spectra sorption features in the UV. The observational indication in transmission spectroscopy Figure 29 shows the synthetic emission spectra for GJ 436b of varying vertical mixing and internal heating for GJ 436b is compared to Spitzer observations. While the 3.6 µm data shown in Figure 28. The early analysis of Spitzer data (Knut- prefers the Tint = 400 K models, Tint = 100 K models are son et al. 2011) has shown inter-epoch variability, which is favored by the 8 µm data. On the other hand, the already reduced in the investigation of Lanotte et al.(2014); Morello large column abundance of CO makes the thermal emission et al.(2015). Methane absorption at 2.1-2.5 and 3-4 µm and at 4.5 µm insensitive to internal heating or vertical mixing. sulfur dioxide absorption at 7-10 µm are the most promising The models somewhat overpredict the flux at 4.5 µm, as in diagnostic features. For Tint = 100 K, vertical mixing leads the previous study (Morley et al. 2017) to higher CH4 abundance and the strong mixing scenario can We conclude that our models demonstrate and confirm be marginally ruled out by the Spitzer 3.6-µm data. For T = int that the combination of moderately high (& 100 times) solar 400 K, vertical mixing conversely reduces CH abundances, 4 metallicity and internal heating can explain the low CH4/CO confirmed with previous work by Agundez´ et al.(2014) and ratio loosely constrained by the Spitzer 3.6 and 4.5 µm obser- Morley et al.(2017). The 3.6 and 4.5 µm Spitzer data are vations, regardless of the strength of mixing. Sulfur species consistent with our models under weak/strong vertical mix- do not quench in the deep region like CH4 or NH3 but closely ing or in chemical equilibrium. While CH4 is not sensitive to associate with photolysis and mixing processes in the up- the strength of vertical mixing at high internal heating sce- per stratosphere. The independence from the thermal prop- nario, SO2 shows strong dependence on mixing processes 31

Tint = 100 K Tint = 400 K 10 8 10 8

10 6 10 6

10 4 10 4

H H O O 2 2 10 C 10 C Weak Mixing Pressure (bar) CH4 CH4

C2H2 C2H2 100 HCN 100 HCN

NH3 NH3 NO NO 102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 Mixing Ratio Mixing Ratio

10 8 10 8

10 6 10 6

10 4 10 4

H H O O 10 2 10 2 Strong Mixing

Pressure (bar) C C

CH4 CH4

C2H2 C2H2 0 0 10 HCN 10 HCN

NH3 NH3 NO NO 102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 Mixing Ratio Mixing Ratio

Figure 25. The abundances of several main species that show differences from models including sulfur kinetics (solid) and without sulfur kinetics (dashed). Each panel corresponds to the same internal heating and vertical mixing as Figure 23. erty of the interior makes sulfur chemistry a complementary file of 51 Eri b (although we find little difference between avenue for characterizing GJ 436b, in conjunction with the setting Teff = 760 K and Teff = 700 K as assumed in previous long-standing quest for constraining CH4/CO. work (Moses et al. 2016; Zahnle et al. 2016)). Samland et al. (2017) also derive a 10 times super-solar metallicity based 4.4. 51 Eridani b on the K-band emission. To explore the effects of metallic- 51 Eridani b (51 Eri b) is a young Jupiter-mass giant ity, we construct one temperature profile with solar metallic- around an F-type star at a wide orbit. Unlike irradiated hot ity and one with 10 times solar metallicity. The resulting P-T Jupiters, the residual heat from the formation predominates profiles are shown in Figure 30. We did not include a thermo- over the stellar flux. In the discovery work, Macintosh et al. sphere as Moses et al.(2016) have added an arbitrary 1000 (2015) suggest 51 Eri b has an effective temperature 550 – K inversion layer above 1 µbar but found little effects on the 750 K with vertically quenched CH4 and CO. Water vapor chemistry. The eddy diffusion takes the same form by (29), should not condense owing to its heat at depth, in contrast to with the radiative-convective transition Ptran set to 0.1 bar. our colder Jupiter. The combination of the hot-interior and The stellar UV flux of 51 Eridani is assembled from vari- photochemically-active stratosphere makes 51 Eri b a unique ous sources. The observations from the International Ultravi- testbed for atmospheric chemistry, as explored by Zahnle olet Explorer (IUE)12 covers the wavelength range between et al.(2016); Moses et al.(2016). 115 and 198 nm. The EUV flux (λ < 115 nm) is adopted from the synthetic spectrum of HR 8799 (Sanz-Forcada et al. 4.4.1. Model Setup We adopt T = 760 K as suggested by the retrieval work eff 12 of Samland et al.(2017) for calculating the temperature pro- https://archive.stsci.edu/iue/ 32

Tint = 400 K, weak mixing 10 8 10 8 H + H2O H2 + OH

H + H2S H2 + SH 10 6 10 6

4 10 4 10

=1 (no S) 2 10 2 =1 10 Pressure (bar) Pressure (bar) H2S

S2 100 SO 100

SO2 Rayleigh 102 102 50 100 150 200 250 300 350 400 100 104 108 1012 1016 1020 1024 Wavelength (nm) Reaction Rates (cm 3s 1)

Tint = 400 K, strong mixing 10 8 Figure 27. The rates of reactions that are key to recycle H back to H2, in the GJ 436b model including sulfur kinetics with Tint = 400 K and weak vertical mixing (corresponding to the bottom right panel 10 6 of Figure 25).

10 4 N-C-H-O chemical network (∼ 1650 reactions), which in- =1 (no S) clude more complex hydrocarbons, to study the quenching 10 2 =1 and photochemical effects. The mixing ratios of the main Pressure (bar) H2S species in our 51 Eri b model for solar and 10 times solar S2 100 SO metallicity are displayed in the top left panel of Figure 31. SO2 The main C, H, N, O chemical species in our model are over- Rayleigh 102 all consistent with both Zahnle et al.(2016) and Moses et al. 50 100 150 200 250 300 350 400 (2016), which we summarize in the following paragraph. Wavelength (nm) The top row of Figure 31 shows how disequilibrium processes control the composition distribution. First, CH4-CO Figure 26. The pressure level of optical depth τ = 1 for GJ 436b conversion is quenched at about 1 bar thus CO predominates with high internal heating (Tint = 400 K) while including (black) and over CH4. Likewise, N2 is the predominant nitrogen bear- excluding (grey) sulfur chemistry, along with the main contribution ing species over NH . Second, without the fast recycling from sulfur species. The top and bottom panels are for weak and 3 strong vertical mixing. mechanism like that on a hot Jupiter, strong photolysis of water makes the upper atmosphere oxidizing and produces considerable CO and O . Third, C H and HCN are pho- 2011)13, following Moses et al.(2016). For wavelengths 2 2 2 2 tochemically generated in the upper atmosphere, similar to greater than 198 nm, we use a theoretical stellar spectrum hot Jupiters. While C H is produced to about 10 ppb level with a close temperature from ATLAS9 Model Atmosphere 6 6 in Moses et al.(2016), C H is greatly reduced in our model Database14 by BOSZ stellar atmosphere model (Bohlin et al. 6 6 including sulfur, as seen for GJ 436b. 2017), assuming T = 7250 K, log(g) = 4, log[Fe/H] = 0, ra- eff The most outstanding difference between Zahnle et al. dius = 1.45 M . The stellar irradiation received by the planet (2016) and our model in terms of sulfur chemistry is that in our 51 Eridani model is stronger by about 50 % than pre- S reaches condensable level in our atmosphere. Although vious work, since the orbit has been updated from 13-14 AU 8 produced at about the same level, S is close to saturation to 11.1 AU according to De Rosa et al.(2020). 8 but does not condense in the nominal model of Zahnle et al. 4.4.2. Disequilibrium Chemistry Compared with Previous Work (2016). Since we adopt the same saturation vapor pressure of Zahnle et al.(2016) investigate sulfur hazes in the atmo- sulfur allotropes (Lyons 2008) as Zahnle et al.(2016), the dif- sphere of 51 Eri b. Moses et al.(2016) use an extensive ferent condensation behavior should be due to a warmer upper stratosphere in Zahnle et al.(2016). Zahnle et al.(2016) indeed noted that S8 would condense if the temperature were 13 http://sdc.cab.inta-csic.es/xexoplanets just a few degrees lower. 14 https://archive.stsci.edu/prepds/bosz/ 33

T = 100 K, weak mixing T = 100 K int int 0.0010 Tint = 100 K, strong mixing weak mixing 0.0071 Tint = 400 K, weak mixing strong mixing Tint = 400 K, strong mixing Equilibrium 0.0008 Knutson et al. 2011 0.0070 Knutson et al.(2014) Morley et al. 2017 Morello et al.(2015) r

Lanotte et al.(2018) a t 0.0069 s F 0.0006

t e n a l

0.0068 p F 0.0004 Transit Depth 0.0067 SO2 NH3 0.0002 0.0066 CO2 CO CH4 H2O 0.0065 0.0000 1 2 3 4 5 6 7 8 9 10 12 2 4 6 8 10 12 Wavelength ( m) Wavelength ( m)

T = 400 K int Figure 29. Synthetic emission spectra computed for our GJ 436b weak mixing 0.0071 strong mixing models including sulfur chemistry, in comparison with Spitzer Equilibrium secondary-eclipse data. weak mixing w/o sulfur 0.0070 Knutson et al.(2014) Morello et al.(2015) Lanotte et al.(2018) 0.0069 2 1 Kzz (cm s ) 106 107 108 10 8 0.0068 Transit Depth 0.0067 10 6 SO2 NH3 0.0066 CO2 CO CH4 10 4 H2O 0.0065 1 2 3 4 5 6 7 8 9 10 12 Wavelength ( m) 10 2 Pressure (bar)

Figure 28. Synthetic transmission spectra computed for our GJ 100 436b model assuming Tint = 100 K (top) and 400 K (bottom) with solar weak and strong vertical mixing. The model without sulfur chem- 10 × solar istry (for Tint = 400 K and weak vertical mixing) is also plotted 500 1000 1500 2000 2500 3000 for comparison. The HST/WFC3 points from Knutson et al.(2014) Temperature (K) have been shifted down by 200 ppm, following Lothringer et al. (2018). The wavebands of main molecular absorption are indicated. Figure 30. The temperature-pressure and eddy diffusion (Kzz) pro- ﬁles for 51 Eri b, assuming solar and 10 times solar metallicity. The 4.4.3. Effects of Super-Solar Metallicity [CH4]/[CO] = 1 equilibrium transition curves corresponding to two metallicities are shown by the dotted curves. The left and right columns of Figure 31 compare the results of solar and 10 times solar metallicity. The model with 10 4.4.4. Effects of Sulfur times solar metallicity has slightly hotter troposphere (Figure 30) which favored CO over CH4. Although the equilibrium Compared to HD 189733b and GJ 436b, H2S can only re- abundance of CH4 in the stratosphere is increased in the 10 × main stable against hydrogen abstraction (30) at higher pres- solar metallicity model, CH4 is in fact decreased with a lower sure about 0.05 bar. The reverse rate of (30) signiﬁcantly CH4/CO ratio at the quench level. In the end, the 10 × solar drops in the cooler stratosphere of 51 Eri b and prohibits the metallicity model shows higher CO, CO2,H2O and lower reformation of H2S. The active SH radical produced by H2S CH4 abundances, which in turn reduces other hydrocarbons leads to a rich variety of sulfur species, as illustrated in the as well. The mixing ratio of CO2 exceeds CH4 for 10 × solar middle row of Figure 31. Compared to Zahnle et al.(2016), metallicity and can reach ∼ 0.1% in the upper atmosphere. our model exhibits a more oxidized upper stratosphere due The production of O2 also raises with metallicity following to stronger UV irradiaion from the closer orbit in our setting the increase of water. (11.1 AU compared to 13.2 AU). Nonetheless, both models 34

Solar Metallicity 10× Solar Metallicity 10 8 10 8

H2O H2O

O2 O2

10 6 CH4 10 6 CH4 CO CO

CO2 CO2

10 4 C2H2 10 4 C2H2 N2 N2

NH3 NH3 10 2 HCN 10 2 HCN Pressure (bar) Pressure (bar)

100 100

102 102 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 1 Mixing Ratio Mixing Ratio

10 8 10 8 S S

H2S H2S 10 6 SO 10 6 SO

SO2 SO2 OCS OCS

10 4 S4 10 4 S4 NS NS

10 2 10 2 Pressure (bar) Pressure (bar)

100 100

102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 Mixing Ratio Mixing Ratio

10 8 10 8

S8 sat. S8 sat.

S2 sat. S2 sat. 6 6 10 S8 10 S8

S2 S2 10 10 S8(s)×10 S8(s)×10 10 4 10 10 4 10 S2(s)×10 S2(s)×10

10 2 10 2 Pressure (bar) Pressure (bar)

100 100

102 102 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 Mixing Ratio Mixing Ratio

Figure 31. The computed abundance profiles of 51 Eri b, assuming solar (left panels) and 10 times solar (right panels) metallicity. The top row presents the main species, with equilibrium profiles shown in dotted lines. The middle row shows the main sulfur species and the bottom has S2/S8 vapor (solid), S2/S8 condensate particles (dashed-dotted), and the saturation mixing ratios of S2/S8 (dotted). The particles are plotted in the ratio of the number density of particles to the total number density of gas molecules and multiplied by 1010. predict efficient polymerization forming great abundance of Elemental S is still the leading sulfur species above the S8 S8. Since S8 is the end pool of sulfur chain reactions, we find condensing layers until being oxidized into SO and SO2 in the condensation of S8 does not affect other sulfur species. the upper stratosphere. 35

Solar Metallicity 10× Solar Metallicity 10 8 10 8

10 6 10 6

10 4 10 4

10 2 H 10 2 H Pressure (bar) O Pressure (bar) O

O2 O2 0 0 10 NH3 10 NH3 HCN HCN NO NO 102 102 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 10 11 10 10 10 9 10 8 10 7 10 6 10 5 10 4 10 3 10 2 Mixing Ratio Mixing Ratio

Figure 32. The abundances of several main species that show differences from models including sulfur kinetics (solid) and without sulfur kinetics (dashed) for 51 Eri b.

The equilibrium abundance of H2S scales with metallicity, The excess atomic O readily reacts with OH to form O2. This which leads to more production of S8 vapor as metallicity enhanced oxidized region along with NS accelerates the ox- increased. The 10 times solar metallicity model has slightly idization of NH3, via the same pathway (32) but more pro- warmer temperature which allows higher saturation pressure nounced than that on GJ 436b. On the other hand, CH4 is un- of sulfur as well. In the end, both the gas and condensates of affected because the intermediates HCS and CS are deﬁcient S2 and S8 increase with metallicity. in the colder stratosphere of 51 Eri b. Lastly, the coupling to The effects of coupling to sulfur on other species are high- H2S also helps atomic H recycle back to H2 faster, as seen on lighted in Figure 32. The most remarkable feature is the en- GJ 436b. hanced oxygen abundances in the upper atmosphere with sulfur. In the absence of sulfur, atomic O can be released from H2O with the aid of CO2: 1 hν H2O −−→ OH + H

CO + OH −−→ CO2 + H 1 1 s

hν (33) CO2 −−→ CO + O 1

2 hν net :H2O −−→ 2 H + O. 1 While sulfur is present, SO and SO2 dissociate more than 1 CO2 around Ly-α and provide a faster channel to liberate O 1 1 from H2O: 1

hν 1 H2O −−→ OH + H 1

SO + OH −−→ SO2 + H hν (34) M SO2 −−→ SO + O Figure 33. The low-pressure limit rate coefﬁcient of S + CO −−→ OCS estimated in this work (37), compared to those in the literature. 2 hν net :H2O −−→ 2 H + O. or 4.4.5. sensitivity to OCS recombination hν H2O −−→ OH + H The fate of elemental S after being released from H2S is S + OH −−→ SO + H critical in sulfur kinetics. Several reactions potentially con- hν (35) trol whether S proceeds to chain formation into larger poly- SO −−→ S + O sulfur (Sx), forming OCS, or being oxidized to SO, SO2. 2 hν To address the effects of kinetics uncertainties, Zahnle et al. net :H2O −−→ 2 H + O. 36

We compare the rate coefﬁcient (37) with those assumed in 10 8 Zahnle et al.(2016) and Venus literature (Zhang et al. 2012; OCS Krasnopolsky 2013) in Figure 33. The reaction rates show S

10 6 S2 diverse values especially toward lower temperatures, the rel- S8 evant temperature range for the stratosphere of 51 Eri b. Al- S8 sat. beit the rate discrepancy in each model, the rate constants in 10 4 Zhang et al.(2012), Krasnopolsky(2013) and this work exhibit consistent temperature dependence from 1000 K to 200 10 2

Pressure (bar) K, whereas that in Zahnle et al.(2016) has suprisingly almost no temperature dependence. Since rate constant of (36) from 100 Zahnle et al.(2016) is the most different from the literature and yields fastest OCS forming rate, we will use the rate from

102 Zahnle et al.(2016) as the upper limit to test the sensitivity. 10 14 10 13 10 12 10 11 10 10 10 9 10 8 10 7 10 6 10 5 Mixing Ratio We run our nominal model with solar metallicity but adopt the rate constant of (36) from Zahnle et al.(2016). The ef- Figure 34. Main sulfur species from our nominal model with solar fects of faster OCS recombination are illustrated in Figure metallicity (solid) compared to those adopting the faster rate of (36) 34. With the OCS recombination rate from Zahnle et al. from Zahnle et al.(2016) (dashed). (2016), OCS mixing ratio is significantly increased above 0.1 bar. S8 is slightly reduced but remains the major sulfur car- (2016) explore the sensitivity to H2S recombination and Sx rier between 10−2 and 10−4 bar, consistent with the model polymerization. The authors found a faster H S recombina- 2 results in Zahnle et al.(2016). The abundance of S and S 2 tion counteracts the destruction of H2S and reduces the pro- are subsequently affected by more ample OCS photodissoci- duction of S , while their results are not sensitive to the poly- 8 ation but that of S8 remains the same as set by condensation. merizing rates of forming S4 and S8 within the tested ranges. Given these differences, we reiterate further investigation to We have tested the polymerization rates for GJ 436b and con- pin down the reaction rate of OCS recombination. firmed its general insensitivity to S8 formation. For 51 Eri b, we recognize that the recombination of OCS 4.4.6. Emission Spectra Figure 35 demonstrates the effects of disequilibrium chem- M S + CO −−→ OCS (36) istry and S8 clouds on the planetary emission spectra. For both metallicities, quenched CH4 and H2O have lower abun- could be important in determining the oxidizing rate of sul- dances than equilibrium, leading to higher emission in the H fur. The rate coefficient of Reaction (36) has in fact not been and J bands from the deeper region. The 10 times solar metal- measured. Only the reverse step of (36), the dissociation of licity further reduces CH4 and prompts the flux at 1.6 - 1.8 OCS, has available data at high temperatures. Recently, Ran- µm. We assume 1 µm particle size for S8 condensates, which jan et al.(2020) has also identified this reaction to modestly scatter strongly and reduce the emission in this wavelength alter the CO abundance in a CO2-N2 atmosphere and advo- range. However, using the higher effective temperature and cate laboratory investigation. Here, we will explain how the metallicity from Samland et al.(2017), our models generate rate coefficient of Reaction (36) is estimated in our nominal emission that are too high in the H and J bands and fail to model and then explore the sensitivity to the uncertainty for reproduce the observed spectra. We conclude that either Teff 51 Eri b. is lower than that determined by Samland et al.(2017) and/or Reaction (36) is a spin-forbidden reaction and usually additional cloud layers (e.g. Moses et al. 2016) is required to many orders of magnitude slower than a typical three-body match the lower observed emission. reaction. Since the measured high-temperature dissociation reaction has a high activation energy, extrapolating the dis- 4.5. Sulfur Mechanism sociation reaction (the reversal of (36)) to low temperatures Figure 36 summerizes the important pathways for sulfur will result in unrealistically rates. Instead, we estimate the species in the irradiated H2-dominated atmospheres we ex- activation energy from the well-studied analogous reaction, plored in this section. H2S is the dominant molecule, which M O + CO −−→ CO2. The pre-exponential factor is then de- is thermochemically stable for a wide range of tempera- termined by matching the reverse of dissociation reaction at tures in the lower atmosphere followed by OCS. The pho- 2000 K from Oya et al.(1994). The low-pressure limit rate tochemistry of sulfur is initiated from SH and S produced of (36) we estimate is by H2S dissociation, leading to multiple channels including chain formation and oxidization depending on the atmo- −34 k0 = 4.47 × 10 exp(−1510/T ). (37) spheric condition. Sulfur chain formation is highly tempera- 37

ture sensitive where S2 is favored about 600 - 800 K and S8 Solar Metallicity can only form below ∼ 500 K (e.g. the stratosphere of 51 Eri Cloud-free Equilibrium GPI b). When OH is sufﬁciently produced by H O photolysis, S ) 8 Cloud-free Disequilibrium SPHERE 2

1 10 Disequilibrium with S8 clouds SPHERE/IFS-YJ will most likely be oxidized into SO and SO in the upper

m 2

c SPHERE/IFS-YH

2 atmosphere. S also participates in accelerating CH4 and NH3

m 10 9 c

destruction via the coupling to CH2 or NH2, as seen in our 1 s

HD 189733b and GJ 436b models. g r e

( 10

10 x

u 4.6. Trends of Photochemical Hazy Precursors l F

n o

i Figure 37 summarizes the column densities of haze pre- s 10 11 s i cursors above 1 mbar for the simulated atmospheres in Sec- m E tion4. Across the various irradiated H 2-atmospheres we ex- 10 12 plored, we ﬁnd HCN consistently to be the most prevailing 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 Wavelength ( m) precursor. This is not surprising as HCN is a robust photochemical product of CH4 and NH3 and also recently been 10× Solar Metallicity detected on HD 209458b (Giacobbe et al. 2021). Nonethe- Cloud-free Equilibrium GPI less, it does not necessarily imply HCN will lead to com-

) 8 Cloud-free Disequilibrium SPHERE 1 10 plex nitriles formation, since HCN is not the limiting fac- Disequilibrium with S8 clouds SPHERE/IFS-YJ m

c SPHERE/IFS-YH

tor as we discussed in Section 2.8. A more careful assess- 2

m 10 9 ment at high temperatures is required before extrapolating c

1 the haze-forming mechanism below 200 K on Titan. We ob- s

r serve a general increasing trend with decreasing temperature e

( 10

x for the more indicative nitrile precursors HC3N but opposite u l F for CH3CN. The same trend is seen for the hydrocarbon pre- n o i

s 10 11 cursors C4H2 and C6H6. s i

m Only HCN and C H can reach appreciable levels on

E 2 2 WASP-33b, even as photochemical hazes are not expected on 10 12 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 WASP-33b. For GJ 436b, most of the precursors are not too Wavelength ( m) sensitive to eddy diffusion. For 51 Eri b, almost all precur- Figure 35. Synthetic emission spectra of 51 Eri b produced from sors are reduced with increasing metallicity, except for CS2 equilibrium abundances, disequilibrium abundances, and with S8 since it contains no H. In fact, CS2 is most favored on HD condensate layer. Data points show GPI observations from Macin- 189733b, which suggests sulfur-containing hazes in the hot tosh et al.(2015) and SPHERE observations from Samland et al. Jupiter condition as carbon and sulfur can couple closely. In (2017). addition to sulfur condensates, 51 Eri b might also be covered by nitrile-type hazes according to the precursor distribution.

Oxidization HCS NS Chain-formation CH 2 5. DISCUSSION H 2 NH

OH OH SH S2 S4 SO2 SO S S2 S4 S8 5.1. High-Temperature UV Cross Sections hν hν hν hν H H

2 S We have implemented layer-by-layer UV cross sections ac- H H H condensate CS C S cording to the temperature at each atmospheric level in VUL-

OH SH HS2 Low P CO CO CAN. Due to the sparsely available data, we did not perform H systematic study in this work. Nonetheless, we have gained H

OCS 2 some insights through the case studies in Section4. H High P We found the effects of temperature dependence for H2O H2S is mostly negligible in a H2-dominated atmosphere. How- ever, this is solely based on the limited wavelength-range Figure 36. A Schematic diagram illustrating the main pathways measurements we assembled. For the high-temperature (T > for sulfur kinetics in an H2-dominated atmosphere. The dashed line represents the transition from the lower region where sulfur is pre- 1000 K) cross sections, only wavelengths longer than about dominantly locked in H2S to the upper region where H2S is subject 190 nm are included (Figure 38). The high-temperature cross to dissociation. Rectangles indicate stable species whereas ellipses sections in the FUV could have larger effects. indicate active radical or intermediate species. We conﬁrm the analysis in Venot et al.(2013) that although the CO2 abundance is not directly inﬂuenced by the tempera- 38

WASP33-b (solar) GJ 436b (weak Kzz) 51 Eri b (solar) WASP33-b (C/O=1.1) GJ 436b (strong Kzz) 51 Eri b (10 × solar) HD 189733b

1018 1018 ) ) 2 1016 2 1016 m m c c

s 1014 s 1014 e e l l u u c c e e l 1012 l 1012 o o m m ( (

y 10 y 10 t 10 t 10 i i s s n n e e 8 8 D D

10 10 n n m m u u l 106 l 106 o o C C

104 104

C2H2 C2H4 C4H2 C6H6 HCN HC3N CH2NH CH3CN CS2 C2H2 C2H4 C4H2 C6H6 HCN HC3N CH2NH CH3CN

Figure 37. The column number densities (molecules cm−2) above 1 mbar of haze precursors for the simulated atmospheres in Section4, including sulfur (left) and without sulfur (right). Some molecule abundances are negligible and not shown for WASP-33b. For GJ 436b, the models with Tint = 400 K and weak/strong vertical mixing are used. ture dependence of CO2 photolysis, the shielding effects can be photoionized and ramify into various organic molecules impact other species. As CO2 absorb more strongly with in- through ion-exchange reactions. creasing temperature, the UV photosphere is lifted to lower 5.3. More Intriguing Questions about Sulfur pressure. The production of radicals, such as H and OH, is reduced and subsequently alters other species. However, In Section4, we find the coupling to sulfur chemistry impacts the core C-H-N-O kinetics in several ways for H - we also find that the shielding effects of CO2 are completely 2 shadowed when sulfur species are included (e.g. see the right dominated atmospheres. The coupling effects essentially depend on if the sulfur-containing intermediates are active, panel of Figure 18). The temperature dependence of CO2 which is not well-understood in general as it can vary with photolysis should be more amplified in CO2-dominated atmospheres. atmospheric conditions such as temperature and bulk compositions. Gersen et al.(2017) find that CH 3S and CH3SH provide more efficient pathways for methane oxidization in the combustion (oxidizing) environment. He et al.(2020) also 5.2. Implication of Ionization observe the photochemical formation of CH3S and CH3SH Ions are not included in this work. We are working on in- in a CO2-rich gas mixture in the experiments. Although we cluding ionchemistry in the next update of VULCAN. Pho- have included reactions involving CH3S and CH3SH in our toionization is known to be critical in initiating the haze for- sulfur mechanism, they are not identified to be important in mation (Wong et al. 2003; Krasnopolsky 2009; Plessis et al. the pathway analysis for all of the H2-atmospheres we inves- 2012; Lavvas et al. 2013). Even thermoionization can be im- tigated. The chemical role of CH3SH is worth further study portant for ultra hot Jupiters. In our study about WASP-33b, in the broad context of biologically produced sulfur. atomic Ti and V in the upper atmosphere are expected to be The temperature profiles are fixed without considering the partly ionized and contribute to free electrons. Since Ti has radiative feedback in this whole work. The radiative effects an ionization threshold of 180 nm, compared to about 240 nm might be more prominent in the presence of sulfur, such as for Na, the effects of photoionization on Ti and V should be the absorption of SH and S2 in the optical and NUV. 51 Eri b similar and probably smaller than those on the alkali atoms, or other directly imaged planets with a relatively cold strato- as investigated in Lavvas & Koskinen(2017). An important sphere (. 500 K) and under sufficient UV irradiation sit in outcome of photoionization is that the increased electrons the sweet spot for testing the radiative feedback on sulfur can lead to more hydrogen anions (H-) than predicted by condensates. thermal equilibrium, which are found to be important opacity sources in some hot Jupiters atmospheres (Lewis et al. 6. SUMMARY 2020). In terms of sulfur chemistry, several sulfur species In this paper, we present a thorough theoretical frame- have relatively lower-energy threshold of ionization and can work of the updated photochemical code VULCAN. We val- be subject to photoionization. For example, atomic S starts to idate our models for the atmospheres of hot Jupiters, Jupiter, ionize from Lyman-α. Since S is likely the dominant sulfur and modern Earth and carry out comparative study on repre- species in the hot Jupiter’s stratosphere (Section 4.2.2), S can sentative cases of extrasolar giant planets: WASP-33b, HD 39

189733b, GJ 436b, and 51 Eridani b. The highlights of our ation of SO and SO2 also make the upper atmosphere results are: of 51 Eri b more oxidizing.

• We suggest including several photochemical haze pre- • We have carefully validated the model of HD 189733b. cursors such as C6H6 and HC3N, which are more The updated methanol scheme in Venot et al.(2020) is indicative than the commonly considered HCN and found to bring the quenching behavor of methane close C2H2. We observe a general increasing trend with de- to Moses et al.(2011) and VULCAN. We pointed out creasing temperature for C4H2,C6H6, and HC3N but the photochemical source plays a non trivial part in the opposite for CH3CN. model differences between Moses et al.(2011), Venot et al.(2012), and VULCAN. MODEL AVAILABILITY In addition to the public code, the conﬁguration ﬁles used • We demonstrate advection transport in the downdraft for the models in Section3 are available on https://github. can qualitatively explain the deep ammonia distribu- com/exoclime/VULCAN and the main model output in Sec- tion in Jupiter, which can not be explained by eddy tion3 and Section4 can be found in the supplementary ma- diffusion alone. terial. Software: Python; Numpy (van der Walt et al. 2011); Scipy

• The implementation of surface boundary conditions (Oliphant 2007); Matplotlib (Hunter 2007) and condensation in an oxygen-rich atmosphere is validated in the present-day Earth model. A general oxidation timescale analysis is provided for assessing the ACKNOWLEDGMENTS chemical lifetime of biosignature gases. S.-M.T gratefully thanks O. Venot and J. Moses for shar- ing the output of HD189733b for model comparison, C. Li for providing the retrieved ammonia results from Juno • The atmosphere of WASP-33b is not affected by vertical quenching but consisted of an upper photolytic re- measurements, L.M. Lara for fruitful discussions about set- gion and a thermochemical equilibrium region below. ting up photochemistry, P. Rimmer for the compiled observational data of Jupiter, and M. Zhang for customiz- For GJ 436b, we find NH3 insensitive to vertical mixing and the sulfur species governed by photolysis and ing PLATON to read non-equilibrium compositions. S.- mixing in the upper stratosphere are independent of the M.T acknowledges support from PlanetS National Center of deep thermal structure, which can be complementary Competence in Research (NCCR) and University of Oxford. M.M. acknowledges support from NASA under the XRP to the CH4/CO metric for breaking degeneracies. The grant No. 18-2XRP18 2-0076. E.K.L. acknowledges sup- quenched CO always predominates over CH4 on 51 Eri port from the University of Oxford and CSH Bern through b and sulfur aerosols (chiefly S8) condense out in the stratosphere. the Bernoulli fellowship. K.H. acknowledges support from the PlanetS National Center of Competence in Research (NCCR) of the Swiss National Science Foundation and the • We find the coupling to sulfur chemistry impact C-H- European Research Council Consolidator Grant EXOKLEIN N-O kinetics. Sulfur can provide catalytic paths to de- (No. 771620). This work was supported by the European stroy CH4 and NH3 and generally lower the hydrocar- Research Council Advanced Grant EXOCONDENSE (No. bon abundances. H2S makes H recycled back to H2 740963). faster on the cooler GJ 436b and 51 Eri b. The dissoci-

APPENDIX

A. MOLECULAR DIFFUSION AND THERMAL DIFFUSION FACTOR

For H2-based atmospheres, we take 17 0.765 DH2−CH4 = 2.2965 × 10 T /N (A1) from Marrero & Mason(1972) as the reference and scale the molecular diffusion coefﬁcient of H 2 with other species according to Equation (3). The thermal diffusion factor for H and He are approximately αH ≈ -0.1 and αHe ≈ 0.145 (Moses et al. 2000). We assume αi = -0.25 for the rest of species based on rigid sphere approximation (Banks & Kockarts 1973). 40

For N2-based atmospheres, we take 16 0.75 DN2−CH4 = 7.34 × 10 T /N (A2) from Banks & Kockarts(1973) as the reference and scale the molecular diffusion coefﬁcient of N 2 with other species according to Equation (3). The thermal diffusion factor of Ar is αAr ≈ 0.17 and αi = -0.25 for the rest species. For CO2-based atmospheres, we take 16 0.759 DCO2−H2 = 7.51 × 10 T /N (A3) from Banks & Kockarts(1973) as the reference and scale the molecular diffusion coefﬁcient of CO 2 with other species according to Equation (3). The thermal diffusion factor of Ar is αAr ≈ 0.17 and αi = -0.25 for the rest species.

B. RESOLUTION ERRORS OF PHOTOLYSIS RATES

Table 4. Errors from Low Spectral Resolution

a Bin (nm) Stellar Flux (in %) ∆ JH2 ∆ JH2O 0.2 0 0.0004 0.003 0.5 0.08 0.18 1.13 1 0.08 0.44 0.87 10 2.65 0.22 0.87

a ∆ J = |J-J0| /J0 where J0 is the reference photolysis rate calculated with constant 0.1 nm resolution

Equation (11) is numerically computed in the form of finite sum. The wavelength grid in the code needs to properly resolve the line structures in the stellar flux and cross sections. This is especially important in the XUV where there are more fine structures from the band transition (Rimmer & Helling 2016). We demonstrate the errors of computing Equation (11) with the GJ 436b model in Section 4.3. The effects are emphasized with its host M-star showing more emission lines. Figure 39 shows the stellar flux overplotted with the UV cross sections of H2 and H2O. The stellar flux is adopted from the MUSCLES survey with a constant 0.1 nm resolution and the UV cross sections from the Lieden database have the same resolution of 0.1 nm. Therefore, we consider constant resolution of 0.1 nm as the reference for this test. The resolution for computing Equation (11) is varied from 0.2 nm to 10 nm, where the trapezoidal rule is applied for the integral. The errors with respect to summing-up the total stellar flux and the resultant photolysis rate of H2 and H2O at the top of atmosphere (10−8 bar) are summerized in Table4. Starting from the bin size of 0.5 nm, which is five times the native resolution of the flux and cross section data, the errors become comparable to the absolute value of the photolysis rate. The errors do not behave linearly with the bin size since the overestimate from the peak can offset the underestimate from the trough. The test shows the importance of using matching spectral resolution to attain accurate photolysis rates. 41

H2O CO2 10 16

10 16 10 17 10 17 10 18

) ) 18 2 2 10 Ref. m 19 m c 10 c ( (

19 150 K

n n 10

o Ref. o 170 K i i t 10 20 t c 423 K c 20 195 K

e e 10 S S

573 K 230 K

s 21 s

s 10 1540 K s 21 420 K

o o 10 r 1630 K r 500 K C C 10 22 1773 K 10 22 585 K 1820 K 700 K 23 10 2010 K 10 23 800 K 2360 K 1160 K 10 24 10 24 50 100 150 200 25 50 75 100 125 150 175 200 225 Wavelength (nm) Wavelength (nm)

NH3 O2 10 16

10 17 10 17

10 18 ) ) 2 2

m m 19 c 10 18 c 10 ( (

n n o o i i t t 10 20 c c e e

S 19 S 10

s s 21

s s 10 o o r r Ref. C C 10 22 200 K 10 20 Ref. 250 K 175 K 10 23 873 K 562 K 1073 K 10 21 10 24 0 50 100 150 200 50 100 150 200 250 Wavelength (nm) Wavelength (nm)

SH H2S

16 10 10 16

18 10 10 17 ) ) 2 2 10 20 18 m m 10 c c ( (

n n 22 o o 19 i 10 i 10 t t c c e e S S

24 Ref. 10 20 s s 10 s 500 K s o o r 1000 K r C 26 C 10 21 1500 K 10 Ref. 2000 K 423 K 28 10 2500 K 10 22 573 K 3000 K 773 K 10 30 100 200 300 400 500 50 100 150 200 250 300 350 Wavelength (nm) Wavelength (nm)

OCS CS2 10 15

10 16 10 16

10 17 10 17 ) ) 2 10 18 2 m m

c c 10 18 ( (

19 n 10 n o o i i t t 10 19 c c 20 e 10 e S S

s s 20

s s 10 21 o 10 o r r C C 10 21 10 22 Ref. Ref. 423 K 423 K 22 10 23 573 K 10 573 K 773 K 773 K 10 24 10 23 50 100 150 200 250 300 50 100 150 200 250 300 350 Wavelength (nm) Wavelength (nm)

Figure 38. Photoabsortion cross sections of H2O, CO2, and NH3 across various temperatures, with Ref. denoting the cross sections measured at room temperature. The measured cross sections of H2O at 423 K (dot) and 573 K (cross) are noisy beyond 216 nm and we use linear ﬁt for conservative estimate. The references of the cross sections are described in Section 2.5. 42

10 15 104 H2 H2O 10 16 ) 3 2 10 m )

c 17 2 102 10 1 m c A (

1 1 10 10 18 s n s

o i g t

r 0 10 c e

19 e (

10 S

x s

u 1 s l 10 F o

r r 10 20 C a l l 10 2 e t S 10 21 10 3

10 4 10 22 50 75 100 125 150 175 200 225 250 Wavelength (nm)

Figure 39. The stellar UV ﬂux of GJ 436 and cross sections of H2 and H2O, showing the native resolution 0.1 nm adopted in the model. 43

Table 5. List of forward reactions (backward reactions with even indexes are reversed numerically with thermodynamic data) and rate coefﬁ- 3 −1 −1 cients (cm s for bimolecular reactions and s for k0) of titanium and vanadium species. Ea est. means the activation energy is estimated by the enthalpy difference as described in Section 2.7.

Reaction Rate Coefﬁcient Reference −13 a TiO + O −−→ Ti + O2 5.42 ×10 exp(−20794/T ) Ea est. with A from matching NIST 1994LIA/MIT (at room temperature) −12 TiO + N −−→ Ti + NO 4.76 ×10 exp(−4772/T ) Ea est. with A from matching NIST1993CLE/HON (at room temperature) −12 TiO + N2 −−→ Ti + N2O 5.9 ×10 exp(−62193/T ) Ea est. with A from matching NIST 1994LIA/MIT (at room temperature) −10 TiO + H −−→ Ti + OH 2.0 ×10 exp(−28418/T ) Ea est. with A based on FeO + H −−→ Fe + OH from Rumminger et al.(1999) −11 TiO + H2 −−→ Ti + H2O 2.0 ×10 exp(−21270/T ) Ea est. with A based on FeO + H2 −−→ Fe + OH from Decin et al.(2018) −11 Ti + CO2 −−→ TiO + CO 7.01 ×10 exp(−1790/T ) NIST 1993CAM/MCC7942-7946 −10 TiO2 + H −−→ TiO + OH 1.0 × 10 exp(−19500/T ) Ea est. with A based on FeO2 + H −−→ FeO + OH from Rumminger et al.(1999) −10 TiO2 + O −−→ TiO + O2 1.0 × 10 exp(−11877/T ) Ea est. −10 TiO2 + CO −−→ TiO + CO2 1.0 × 10 exp(−9160/T ) Ea est. with A based on FeO2 + O −−→ FeO + O2 from Rumminger et al.(1999) −11 TiH + H −−→ Ti + H2 8.3 ×10 Est. from FeH + H −−→ Fe + H2 (Rumminger et al. 1999) −10 TiH + O −−→ Ti + OH 1.66 ×10 Est. from FeH + O −−→ Fe + OH (Rumminger et al. 1999) −10 TiH + CH3 −−→ Ti + CH4 1.0 ×10 Est. from FeH + CH3 −−→ Fe + CH4 (Rumminger et al. 1999) −10 TiC + H −−→ Ti + CH 1.0 × 10 exp(−20109/T ) Ea est. −10 Ti + CO −−→ TiC + O 8.0 × 10 exp(−68842/T ) Ea est. −10 Ti + CN −−→ TiC + N 5.0 × 10 exp(−29543/T ) Ea est. −11 Ti + NO −−→ TiN + O 5.0 × 10 exp(−19706/T ) Ea est. with A from matching Campbell & McClean(1993) −10 TiN + H −−→ Ti + NH 10 exp(−15655/T ) Ea est. −12 VO + O −−→ V + O2 6.19 × 10 exp(−14763/T ) Ea est. with A from matching NIST 1990RIT/WEI (at room temperature) −11 V + N2O −−→ VO + N2 4.7 × 10 exp(−1299/T ) NIST 2000CAM/KOL −10 VO + H −−→ V + OH 1.66 × 10 exp(−22386/T ) Ea est. with A based on FeO + H −−→ Fe + OH from Rumminger et al.(1999) −11 VO + H2 −−→ V + H2O 1.0 × 10 exp(−15239/T ) Ea est. with A based on FeO + H2 −−→ Fe + H2O from Decin et al.(2018) −1.8 M k0 = 1.38 × T exp(−94079/T ) M TiO2 −−→ Ti + O2 17 −0.91 Ea est. A based on FeO2 −−→ Fe + O2 from Rumminger et al.(1999) k∞ = 2 × 10 T exp(−94079/T ) −1.8 M k0 = 1.38 × T exp(−82202/T ) M TiO −−→ Ti + O 17 −0.91 Ea est. A based on FeO −−→ Fe + O from Rumminger et al.(1999) k∞ = 2 × 10 T exp(−76171/T ) −1.8 M k0 = 1.38 × T exp(−76171/T ) M VO −−→ V + O 17 −0.91 Ea est. A based on FeO −−→ Fe + O from Rumminger et al.(1999) k∞ = 2 × 10 T exp(−76171/T )

a The pre-exponential factor. 44

Table 6. Full List of Photolysis Reactions in VULCAN

Photolysis Reaction Threshold Temp. Dependence (K) Reference (nm) Cross Sections / Quantum Yields (λ in nm)

H2O −−→ H + OH 207 300, 423, 573, 1230, a, b, e / d, Stief et al.(1975); Slanger 1540, 1630, 1820, 2010, & Black(1982) 2360 1 −−→ H2 + O( D) −−→ O + H + H

CH4 −−→ CH3 + H 145 — a / d (λ < 97.7), Gans et al.(2011) (λ ≥ 118.2) 1 −−→ CH2 + H2 1 −−→ CH2 + H + H

−−→ CH + H2 + H

CH3 −−→ CH + H2 220 — a / Lavvas et al.(2008); Kassner & Stuhl(1994)

−−→ CH2 + H

CH2 −−→ CH + H 275 — a / a CO −−→ C + O 166 — a, b/ a

H2 −−→ H + H 120 — a / a

C2H2 −−→ C2H + H 217 — a / a, Okabe(1983)

CO2 −−→ CO + O 202 150, 170, 195, 230, 300, a, Venot et al.(2018) , e / b 420, 500, 585, 700, 800, 1160 −−→ CO + O(1D)

C2H4 −−→ C2H2 + H2 195 — a / Lavvas et al.(2008) and the references in

−−→ C2H2 + H + H

−−→ C2H3 + H

C2H6 −−→ C2H4 + H2 165 — a / b, Lias et al.(1970) (104 < λ < 105)

−−→ C2H4 + H + H

−−→ C2H2 + H2 + H2 1 −−→ CH4 + CH2

−−→ CH3 + CH3

C4H2 −−→ C2H2 + C2 197 — a / Lavvas et al.(2008)

C6H6 −−→ C6H5 + H 206 — a / Est. from Kislov et al.(2004)

−−→ C3H3 + C3H3 OH −−→ O + H 265 — a / a, b (λ < 91) HCO −−→ H + CO 656 — a / a

H2CO −−→ H2 + CO 360 — a / b (λ < 250), d −−→ H + HCO

CH3OH −−→ CH3 + OH 220 — a / b, Hagege et al.(1968)

−−→ H2CO + H2

−−→ CH3O + H

CH3CHO −−→ CH4 + CO 350 — a / b

−−→ CH3 + HCO

N2 −−→ N + N 150 — a (λ < 100), c (λ > 120) / a

NH3 −−→ NH2 + H 226 175, 300, 562 a / b −−→ NH + H + H HCN −−→ H + CN 179 — a / Nuth & Glicker(1982) NO −−→ N + O 202 a / a 45

NO2 −−→ NO + O 398 b, Voigt et al.(2002) / b, d

NO3 −−→ NO2 + O 703 — b / b, Sander et al.(2015)

−−→ NO + O2 1 N2O −−→ NO + O( D) 340 300, 714, 833, 1000, a,c (Zuev & Starikovskii 1990)) / b 1250, 1667, 2000

HNO2 −−→ NO + OH 591 — b / b

HNO3 −−→ NO2 + OH 598 — b / b

N2O5 −−→ NO3 + NO2 500 — b / b (from here)

−−→ NO3 + NO + O

N2H4 −−→ N2H3 + H 291 — c (BiehlStuhl(1991) and Vaghjiani(1993) 296K 191-291n) / Lavvas et al.(2008)

O2 −−→ O + O 240 200, 250, 300, 873, 1073 a / b,d −−→ O + O(1D)

O3 −−→ O2 + O 900 — a / Matsumi(2002) 1 −−→ O2 + O( D)

HO2 −−→ O + OH 275 — a / a

H2O2 −−→ OH + OH 350 — a / a HNCO −−→ NH + CO 354 — b / b −−→ H + NCO SH −−→ S + H 345 300, 500, 750, 1000, a, Gorman et al.(2019) for λ ≥ 314 1250, 1500, 1750, 2000, nm / a 2250, 2500, 3000

H2S −−→ SH + H 238 423, 573, 773 a, e / a SO −−→ S + O 235 — a / a

SO2 −−→ SO + O 220 — a / a

−−→ S + O2 — a / a

S2 −−→ S + S 283 — a / a

S4 −−→ S + S 575 — a / a OCS −−→ CO + S 280 — a / a CS −−→ C + S 160 — a / a

CS2 −−→ CS + S 278 300, 423, 573, 773 a / a

CH3SH −−→ CH3S + H 310 — a / b

−−→ CH3 + SH

a: Leiden Observatory database15 (Heays et al. 2017) b: PHIDRATES database16 (Huebner et al. 1992) c: MPI-Mainz UV/VIS Spectral Atlas 17 (Keller-Rudek et al. 2013)) d: Sander et al.(2015) e: ExoMol database 18

C. THE CHOICE OF ZENITH ANGLE A global or hemispheric average 1-D photochemical model requires specifying an effective zenith angle (θ) of the stellar beam to represent the planetary-mean actinic ﬂux. The zenith angle of 48◦ – 60◦ has been used for the hemispheric average in various photochemical models (e.g. Moses et al.(2011); Hu et al.(2012)). For instance, a common choice in radiative transfer calculation is to take the ﬂux-weighted cosine of the zenith angle (Cronin 2014):

R 2π R 1 µF0µdµdφ µ = 0 0 (C4) I R 2π R 1 0 0 F0µdµdφ

15 http://home.strw.leidenuniv.nl/∼ewine/photo 16 http://phidrates.space.swri.edu 17 http://satellite.mpic.de/spectral atlas/index.html 18 http://www.exomol.com/data/data-types/xsec VUV 46

◦ where φ is the azimuth angle and F0 is the stellar ﬂux at the top of atmosphere. Equation (C4) yields µI = 2/3 or θI ≈ 48 . For photochemistry calculation, actinic-ﬂux-weighted cosine should be considered and Equation (C4) becomes

R 1 µJ0dµ µ = 0 (C5) II R 1 0 J0dµ where J0 is the actinic flux at the top of atmosphere (i.e. total intensity) and the azimuth term is dropped. Equation (C5) now ◦ yields µII = 1/2 or θII = 60 . Hu et al.(2012) discuss this choice of mean zenith angle by further considering the optical depth of the level where the hemispheric average of the attenuated actinic flux is evaluated, e.g., the authors find a mean zenith angle of 57◦ and 48◦ correspond to optical depth 0.1 and 1, respectively. We will revisit the discussion of Hu et al.(2012) but with a different approach here. Instead of taking the average of µ, we consider an effective zenith angle such that the resulting mean actinic flux matches the hemispheric-mean actinic flux, viz.

R 1 τ τ 0 J0 exp(− µ )dµ J0 exp(− ) = µ R 1 dµ 0 (C6) Z 1 τ = J0 exp(− )dµ. 0 µ Equation (C6) can be evaluated numerically at the given optical depth, as illustrated in Figure 40. We find that optical depth unity (i.e. where UV photons are mostly utilized for photochemistry) corresponds to θ ≈ 58◦. Note that the evaluation in Hu et al.(2012) is similar to Equation (C6) but weighted by µ within the integral, which is effectively the mean stellar flux instead of the actinic flux. Evaluating Equation (C6) at τ = 1, we find θ ≈ 58◦ for a dayside-average model. For a terminator-average model, the denominator in Equation (C6) is integrated from -1 to 1 and we have θ ≈ 67◦.

70 ) ( 60 e l g n A 50 h t i n e Z

40 n a e

M 30

10 2 10 1 100 101 102 Optical Depth

Figure 40. Dayside mean zenith angle for photochemical calculation as a function of optical depth from Equation (C6).

REFERENCES

Agundez,´ M., Venot, O., Selsis, F., & Iro, N. 2014, ApJ, 781 Atreya, S. K., Hofstadter, M. H., In, J. H., et al. 2020, SSRv, 216 Allen, M., Yung, Y. L., & Waters, J. W. 1981, Journal of Atreya, S. K., Wong, A. S., Baines, K. H., Wong, M. H., & Owen, Geophysical Research: Space Physics, 86 T. C. 2005, Planet. Space Sci., 53 Arney, G. N. 2019, The Astrophysical Journal, 873 Ayres, T. R. 2010, ApJS, 187 Atkinson, D. H., Ingersoll, A. P., & Seiff, A. 1997, Nature, 388 Banks, P. M., & Kockarts, G. 1973, Aeronomy. 47

0 00 0 0 00 0

0 0 0 0 0 000

Figure 41. Comparisons of selective rate constants of Reaction (19): CO + OH −−→ CO2 + H with wide temperature ranges available on NIST. Barkley, M. P., Palmer, P. I., Boone, C. D., Bernath, P. F., & Chauvin, G. 2018, arXiv e-prints, arXiv:1810.02031. Suntharalingam, P. 2008, Geophysical Research Letters, 35 https://arxiv.org/abs/1810.02031 Baxter, C., Desert,´ J.-M., Tsai, S.-M., et al. 2021, A&A, 648 Chestakov, D. A., Parker, D. H., & Baklanov, A. V. 2005, The Birkby, J. L., de Kok, R. J., Brogi, M., et al. 2013, Monthly Notices Journal of Chemical Physics, 122 of the Royal Astronomical Society: Letters, 436 Chiou, E. W., McCormick, M. P., & Chu, W. P. 1997, Journal of Boechat-Roberty, H. M., Rocco, M. L. M., Lucas, C. A., & Geophysical Research: Atmospheres, 102 de Souza, G. G. B. 2004, Journal of Physics B: Atomic, Cloutman, L. 2000, Lawrence Livermore National Laboratory Molecular and Optical Physics, 37 report, UCRL-ID-139893 Bohlin, R. C., Mesz´ aros,´ S., Fleming, S. W., et al. 2017, AJ, 153 Cronin, T. W. 2014, J. Atmos. Sci., 71 Bolton, S. J., Adriani, A., Adumitroaie, V., et al. 2017, Science Crossﬁeld, I. J. M. 2015, Publications of the Astronomical Society (80-. )., 356 of the Paciﬁc, 127 Bordwell, B., Brown, B. P., & Oishi, J. S. 2018, The Astrophysical Davidson, D. F., Chang, A. Y., & Hanson, R. K. 1989, Symposium Journal, 854 (International) on Combustion, 22 Bourrier, V., Wheatley, P. J., Lecavelier des Etangs, A., et al. 2020, De Rosa, R. J., Nielsen, E. L., Wang, J. J., et al. 2020, AJ, 159 MNRAS, 493 Dean, A. M., & Bozzelli, J. W. 2000, Combustion chemistry of Brasseur, G. P., & Jacob, D. J. 2017, Numerical Methods for nitrogen (Springer), 125–341 Advection (Cambridge University Press), , p275–341 Dean, A. M., Chou, M.-S., & Stern, D. 1984, International Journal Brogi, M., de Kok, R. J., Birkby, J. L., Schwarz, H., & Snellen, of Chemical Kinetics, 16 I. A. G. 2014, A&A, 565 Decin, L., Danilovich, T., Gobrecht, D., et al. 2018, Astrophys. J., Burke, U., Metcalfe, W. K., Burke, S. M., et al. 2016, Combustion 855 and Flame, 165 Demory, B. O., Gillon, M., Barman, T., et al. 2007, A&A, 475 Burrows, A., Dulick, M., Bauschlicher, C. W., et al. 2005, Drossart, P., Fouchet, T., Crovisier, J., et al. 1999, in ESA Special Astrophys. J., 624 Publication, Vol. 427, The Universe as Seen by ISO, ed. P. Cox Butler, R. P., Vogt, S. S., Marcy, G. W., et al. 2004, The & M. Kessler, 169 Astrophysical Journal, 617 Drummond, B., Tremblin, P., Baraffe, I., et al. 2016, A&A, 594 Campbell, M. L., & McClean, R. E. 1993, J. Phys. Chem., 97 Du, S., Francisco, J. S., Shepler, B. C., & Peterson, K. A. 2008, Capalbo, F. J., Benilan,´ Y., Fray, N., et al. 2016, Icarus, 265 JChPh, 128 Chachan, Y., Knutson, H. A., Gao, P., et al. 2019, AJ, 158 Ehhalt, D. H., & Heidt, L. E. 1973, Journal of Geophysical Chamberlain, J. W., & Hunten, D. M. 1987, in International Research (1896-1977), 78 Geophysics, Vol. 36, Theory of Planetary Atmospheres, ed. J. W. Fair, R. W., & Thrush, B. A. 1969, Trans. Faraday Soc., 65 Chamberlain & D. M. Hunten (Academic Press), 330 – 415 Fischer, H., Birk, M., Blom, C., et al. 2008, Atmospheric Charnay, B., Meadows, V., & Leconte, J. 2015, ApJ, 813 Chemistry and Physics, 8 48

Fortney, J., Lodders, K., Marley, M., & Freedman, R. 2008, ApJ, Hubeny, I., Burrows, A., & Sudarsky, D. 2003, ApJ, 594 678 Hudson, R. D., & Reed, E. I. 1979 France, K., Loyd, R. O. P., Youngblood, A., et al. 2016, ApJ, 820 Hue, V., Cavalie,´ T., Dobrijevic, M., Hersant, F., & Greathouse, T. Frederick, J. E., & Mentall, J. E. 1982, Geophysical Research 2015, Icarus, 257 Letters, 9 Hue, V., Hersant, F., Cavalie,´ T., Dobrijevic, M., & Sinclair, J. A. Frenklach, M. 2002, Phys. Chem. Chem. Phys., 4 2018, Icarus, 307 Frenklach, M., & Mebel, A. M. 2020, Phys. Chem. Chem. Phys., Huebner, W., & Mukherjee, J. 2015, Planetary and Space Science, 22 106 Funke, B., Lopez-Puertas,´ M., Garc´ıa-Comas, M., et al. 2009, Huebner, W. F., Keady, J. J., & Lyon, S. P. 1992, Ap&SS, 195 Atmospheric Chemistry and Physics, 9 Hunter, J. D. 2007, Computing in Science Engineering, 9 Gans, B., Boye-P´ eronne,´ S., Broquier, M., et al. 2011, Physical Inn, E. C. Y., Vedder, J. F., Tyson, B. J., & O’Hara, D. 1979, Chemistry Chemical Physics (Incorporating Faraday Geophysical Research Letters, 6 Transactions), 13 Jacob, D. J. 2011, Introduction to Atmospheric Chemistry Gao, P., & Benneke, B. 2018, The Astrophysical Journal, 863 (Princeton University Press) Gao, P., Marley, M. S., Zahnle, K., Robinson, T. D., & Lewis, Jaeschke, W., Schmitt, R., & Georgii, H.-W. 1976, Geophysical N. K. 2017, Astron. J., 153 Research Letters, 3 Gao, P., Thorngren, D. P., Lee, G. K. H., et al. 2020, Nature Joshi, A. V., & Wang, H. 2006, International Journal of Chemical Astronomy, 4 Kinetics, 38 Georgii, H.-W., & Meixner, F. X. 1980, Journal of Geophysical Kassner, C., & Stuhl, F. 1994, Chemical Physics Letters, 222 Research: Oceans, 85 Kasting, J. F., & Donahue, T. M. 1980, Journal of Geophysical Gersen, S., van Essen, M., Darmeveil, H., et al. 2017, Energy and Research: Oceans, 85 Fuels, 31 Kasting, J. F., Liu, S. C., & Donahue, T. M. 1979, Journal of Giacobbe, P., Brogi, M., Gandhi, S., et al. 2021, Nature, 592 Geophysical Research: Oceans, 84 Gorman, M. N., Yurchenko, S. N., & Tennyson, J. 2019, Monthly Kawashima, Y., & Ikoma, M. 2018, The Astrophysical Journal, 853 Notices of the Royal Astronomical Society, 490 —. 2019, The Astrophysical Journal, 877 Gueymard, C. A. 2018, Solar Energy, 169 Keller-Rudek, H., Moortgat, G. K., Sander, R., & Sorensen,¨ R. Hagege, J., Roberge, P. C., & Vermeil, C. 1968, Trans. Faraday 2013, Earth System Science Data, 5 Soc., 64 Khodachenko, M. L., Shaikhislamov, I. F., Lammer, H., et al. Hauglustaine, D. A., Granier, C., Brasseur, G. P., & MeGie,´ G. 2019, ApJ, 885 1994, J. Geophys. Res., 99 Kislov, V. V., Nguyen, T. L., Mebel, A. M., Lin, S. H., & Smith, He, C., Horst,¨ S. M., Lewis, N. K., et al. 2020, The Planetary S. C. 2004, JChPh, 120 Science Journal, 1 Kitzmann, D., Heng, K., Rimmer, P. B., et al. 2018, The He, C., Horst,¨ S. M., Lewis, N. K., et al. 2020, Nature Astronomy, Astrophysical Journal, 863, 183 4 Klippenstein, S. J., Harding, L. B., Ruscic, B., et al. 2009, Journal Heays, A. N., Bosman, A. D., & van Dishoeck, E. F. 2017, 105 of Physical Chemistry A, 113 Helling, C., & Woitke, P. 2006, A&A, 455 Knutson, H. A., Benneke, B., Deming, D., & Homeier, D. 2014, Heng, K., Malik, M., & Kitzmann, D. 2018, ApJS, 237 Nature, 505 Hidaka, Y., Oki, T., Kawano, H., & Higashihara, T. 1989, Journal Knutson, H. A., Madhusudhan, N., Cowan, N. B., et al. 2011, The of Physical Chemistry; (USA), 93:20 Astrophysical Journal, 735 Hobbs, R., Rimmer, P. B., Shorttle, O., & Madhusudhan, N. 2021, Komacek, T. D., Showman, A. P., & Parmentier, V. 2019, ApJ, 881 MNRAS Konnov, A., & De Ruyck, J. 2001, Combustion and Flame, 124 Hobbs, R., Shorttle, O., Madhusudhan, N., & Rimmer, P. 2019, Krasnopolsky, V. A. 2009, Icarus, 201 MNRAS, 487 —. 2012, Icarus, 218 Hoeijmakers, H. J., Ehrenreich, D., Heng, K., et al. 2018, Nature, —. 2013, Icarus, 225 560 Kreidberg, L. 2018, Exoplanet Atmosphere Measurements from Hopfner,¨ M., Volkamer, R., Grabowski, U., et al. 2016, Transmission Spectroscopy and Other Planet Star Combined Atmospheric Chemistry and Physics, 16 Light Observations, ed. H. J. Deeg & J. A. Belmonte, 100 Hu, R., Seager, S., & Bains, W. 2012, Astrophys. J., 761 Kreidberg, L., Bean, J. L., Desert,´ J.-M., et al. 2014, Nature, 505 —. 2013, Astrophys. J., 769 Lanotte, A. A., Gillon, M., Demory, B. O., et al. 2014, A&A, 572 Hu, R., Seager, S., & Yung, Y. L. 2015, Astrophys. J., 807 Lavvas, P., & Koskinen, T. 2017, The Astrophysical Journal, 847 49

Lavvas, P., Yelle, R. V., Koskinen, T., et al. 2013, Proceedings of Moses, J. I., Fouchet, T., Bezard,´ B., et al. 2005, Journal of the National Academy of Science, 110 Geophysical Research (Planets), 110 Lavvas, P. P., Coustenis, A., & Vardavas, I. M. 2008, Moses, J. I., Visscher, C., Fortney, J. J., et al. 2011, Astrophys. J., Planet. Space Sci., 56 737 Lebonnois, S. 2005, Planet. Space Sci., 53 Moses, J. I., Line, M. R., Visscher, C., et al. 2013, The Lewis, N. K., Showman, A. P., Fortney, J. J., et al. 2010, ApJ, 720 Astrophysical Journal, 777 Lewis, N. K., Wakeford, H. R., MacDonald, R. J., et al. 2020, Moses, J. I., Marley, M. S., Zahnle, K., et al. 2016, The ApJL, 902 Astrophysical Journal, 829 Li, C., Ingersoll, A., Janssen, M., et al. 2017, Geophys. Res. Lett., Nair, H., Allen, M., Anbar, A., Yung, Y., & Clancy, R. 1994, 44 Icarus, 111 Li, C., Ingersoll, A., Bolton, S., et al. 2020, Nature Astronomy, 4 Nicholas, J. E., Amodio, C. A., & Baker, M. J. 1979, J. Chem. Li, Q., Jacob, D. J., Yantosca, R. M., et al. 2003, Journal of Soc., Faraday Trans. 1, 75 Geophysical Research: Atmospheres, 108 Nicolet, M. 1968, Geophysical Journal International, 15 Liang, M.-C., Parkinson, C. D., Lee, A. Y.-T., Yung, Y. L., & Noel,¨ S., Weigel, K., Bramstedt, K., et al. 2018, Atmospheric Seager, S. 2003, The Astrophysical Journal, 596 Chemistry and Physics, 18 Lias, S. G., Collin, G. J., Rebbert, R. E., & Ausloos, P. 1970, The Nuth, J. A., & Glicker, S. 1982, JQSRT, 28 Journal of Chemical Physics, 52 Ohno, K., Okuzumi, S., & Tazaki, R. 2020, ApJ, 891 Lincowski, A. P., Meadows, V. S., Crisp, D., et al. 2018, The Okabe, H. 1983, JChPh, 78 Astrophysical Journal, 867 Oliphant, T. E. 2007, Computing in Science Engineering, 9 Lindzen, R. S. 1981, Journal of Geophysical Research: Oceans, 86 Oya, M., Shiina, H., Tsuchiya, K., & Matsui, H. 1994, Bulletin of Line, M. R., Liang, M. C., & Yung, Y. L. 2010, The Astrophysical the Chemical Society of Japan, 67 Journal, 717, 496 Parmentier, V., Showman, A. P., & Lian, Y. 2013, A&A, 558 Loftus, K., Wordsworth, R. D., & Morley, C. V. 2019, The Parmentier, V., Line, M. R., Bean, J. L., et al. 2018, 1 Astrophysical Journal, 887 Parviainen, H., Palle,´ E., Chen, G., et al. 2018, A&A, 609 Lothringer, J. D., Benneke, B., Crossﬁeld, I. J. M., et al. 2018, AJ, Pearce, B. K. D., Ayers, P. W., & Pudritz, R. E. 2019, Journal of 155 Physical Chemistry A, 123 Loyd, R. O. P., France, K., Youngblood, A., et al. 2016, ApJ, 824 Pinto, J. P., Li, J., Mills, F. P., et al. 2021, Nature Communications, Lyons, J. R. 2008, Journal of Sulfur Chemistry, 29 12 Macintosh, B., Graham, J. R., Barman, T., et al. 2015, Science, 350 Plessis, S., Carrasco, N., Dobrijevic, M., & Pernot, P. 2012, Icarus, Madhusudhan, N. 2012, ApJ., 758 219 Madhusudhan, N., & Seager, S. 2011, ApJ, 729 Pont, F., Sing, D. K., Gibson, N. P., et al. 2013, MNRAS, 432 Malik, M., Kitzmann, D., Mendonc¸a, J. M., et al. 2019, The Pueyo, L. 2018, Direct Imaging as a Detection Technique for Astronomical Journal, 157 Exoplanets, ed. H. J. Deeg & J. A. Belmonte (Cham: Springer Marrero, T. R., & Mason, E. A. 1972, Journal of Physical and International Publishing), 1–61 Chemical Reference Data, 1 Ranjan, S., Schwieterman, E. W., Harman, C., et al. 2020, ApJ, 896 Massie, S. T., & Hunten, D. M. 1981, J. Geophys. Res., 86 Rees, D. 1988, Advances in Space Research, 8 McCullough, P. R., Crouzet, N., Deming, D., & Madhusudhan, N. Rees, D., Barnett, J. J., & Labitzke, K. 1990, Advances in Space 2014, ApJ, 791 Research, 10 Miguel, Y., & Kaltenegger, L. 2014, Astrophys. J., 780 Rimmer, P. B., & Helling, C. 2016, ApJS, 224 Misra, A., Krissansen-Totton, J., Koehler, M. C., & Sholes, S. Rimmer, P. B., & Rugheimer, S. 2019, Icarus, 329 2015, Astrobiology, 15 Rugheimer, S., Kaltenegger, L., Zsom, A., Segura, A., & Sasselov, Molaverdikhani, K., Henning, T., & Molliere,` P. 2019, The D. 2013, Astrobiology, 13 Astrophysical Journal, 883 Rumminger, M. D., Reinelt, D., Babushok, V., & Linteris, G. T. Morello, G., Waldmann, I. P., Tinetti, G., et al. 2015, ApJ, 802 1999, Combust. Flame, 116 Morley, C. V., Fortney, J. J., Kempton, E. M.-R., et al. 2013, The Samland, M., Molliere,` P., Bonnefoy, M., et al. 2017, A&A, 603 Astrophysical Journal, 775 Sander, S. P., Friedl, R. R., Abbatt, D., J. P., & et al. 2015, JPL Morley, C. V., Knutson, H., Line, M., et al. 2017, The Publication 15-10, 106 Astronomical Journal, 153 Sanz-Forcada, J., Micela, G., Ribas, I., et al. 2011, A&A, 532 Moses, J. 2014 Schulz, C., Koch, J., Davidson, D., Jeffries, J., & Hanson, R. 2002, Moses, J. I., Bruno, B., Lellouch, E., & Allen, M. 2000, Icarus, 143 Chemical Physics Letters, 355 50

Schwieterman, E. W., Kiang, N. Y., Parenteau, M. N., et al. 2018, Voigt, S., Orphal, J., & Burrows, J. 2002, Journal of Astrobiology, 18 Photochemistry and Photobiology A: Chemistry, 149 Seinfeld, J. H., & Pandis, S. N. 2016, Atmospheric chemistry and von Essen, C., Mallonn, M., Borre, C. C., et al. 2020, A&A, 639 physics: from air pollution to climate change (John Wiley Wakeford, H. R., Sing, D. K., Kataria, T., et al. 2017, Science, 356 & Sons, Inc.) Wang, D., Lunine, J. I., & Mousis, O. 2016, Icarus, 276 Serindag, D. B., Nugroho, S. K., Molliere,` P., et al. 2021, A&A, Wang, H. 2011, Proceedings of the Combustion Institute, 33 645 Weidenschilling, S. J., & Lewis, J. S. 1973, Icarus, 20 Shulyak, D., Lara, L. M., Rengel, M., & Nemec,` N. E. 2020, A&A, Wilson, E. H., & Atreya, S. K. 2004, Journal of Geophysical 639 Research: Planets, 109 Slanger, T. G., & Black, G. 1982, JChPh, 77 Woitke, P., Helling, C., Hunter, G. H., et al. 2018, A&A, 614 Smith, M. D. 1998, Icarus, 132 Wong, A. S., Yung, Y. L., & Friedson, A. J. 2003, Geophys. Res. Smithson, P. A. 2001, International Journal of Climatology, 22 Lett., 30 Stevenson, D. J. 2020, Annual Review of Earth and Planetary Wordsworth, R. D., Schaefer, L. K., & Fischer, R. A. 2018, The Sciences, 48 Astronomical Journal, 155 Stevenson, K. B., Harrington, J., Nymeyer, S., et al. 2010, Nature, Youngblood, A., France, K., Loyd, R. O. P., et al. 2016, ApJ, 824 464 Yung, Y. L., Allen, M., & Pinto, J. P. 1984, ApJS, 55 Stief, L. J., Payne, W. A., & Klemm, R. B. 1975, JChPh, 62 Yung, Y. L., & DeMore, W. B. 1999, Photochemistry of planetary Suto, M., Wang, X., Shan, J., & Lee, L. 1992, Journal of atmospheres (Oxford University Press) Quantitative Spectroscopy and Radiative Transfer, 48 Yung, Y. L., & Strobel, D. F. 1980, ApJ, 239 Tan, X., & Komacek, T. D. 2019, ApJ, 886 Zahnle, K., Marley, M. S., Freedman, R. S., Lodders, K., & Tsai, S.-M., Kitzmann, D., Lyons, J. R., et al. 2018, ApJ, 862 Fortney, J. J. 2009, The Astrophysical Journal, 701 Tsai, S.-M., Lyons, J. R., Grosheintz, L., et al. 2017, Astrophys. J. Zahnle, K., Marley, M. S., Morley, C. V., & Moses, J. I. 2016, The Suppl. Ser., 228 Astrophysical Journal, 824 Tsang, W., & Hampson, R. F. 1986, Journal of Physical and Zhang, M., Chachan, Y., Kempton, E. M. R., & Knutson, H. A. Chemical Reference Data, 15 2019, PASP, 131 Tsuji, T. 1973, A&aA, 23, 411 Zhang, M., Chachan, Y., Kempton, E. M. R., Knutson, H. A., & van der Walt, S., Colbert, S. C., & Varoquaux, G. 2011, Computing Chang, W. H. 2020, ApJ, 899 in Science Engineering, 13 Zhang, M., Knutson, H. A., Kataria, T., et al. 2018, The Astronomical Journal, 155 Vattulainen, J., Wallenius, L., Stenberg, J., Hernberg, R., & Linna, Zhang, X., Liang, M. C., Mills, F. P., Belyaev, D. A., & Yung, Y. L. V. 1997, Applied Spectroscopy, 51 2012, Icarus, 217 Venot, Bounaceur, R., Dobrijevic, M., et al. 2019, A&A, 624 Zhang, X., Shia, R.-L., & Yung, Y. L. 2013, The Astrophysical Venot, Hebrard,´ Eric, Agundez,´ Marcelino, Decin, Leen, & Journal, 767 Bounaceur, Roda. 2015, A&A, 577 Zhao, L., Kaiser, R. I., Xu, B., et al. 2018, Nature Astronomy, 2 Venot, O., Cavalie,´ T., Bounaceur, R., et al. 2020, A&A, 634 Zilinskas, M., Miguel, Y., Molliere,` P., & Tsai, S.-M. 2020, Venot, O., Hebrard,´ E., Agundez,´ M., et al. 2012, A& A, 546 MNRAS, 494 Venot, O., Fray, N., Benilan,´ Y., et al. 2013, A&A, 551 Zuev, A. P., & Starikovskii, A. Y. 1990, Journal of Applied Venot, O., Benilan,´ Y., Fray, N., et al. 2018, A&A, 609 Spectroscopy, 52 Visscher, C., Moses, J. I., & Saslow, S. A. 2010, Icarus, 209