Vacuum ultraviolet photodissociation of sulfur dioxide and its implications for oxygen production in the early Earth's atmosphere

The emergence of molecular oxygen (O2) in the Earth's primitive atmosphere is an issue of major interest. Although the biological processes leading to its accumulation in the Earth's atmosphere are well understood, its abiotic source is still not fully established. Here, we report a new direct dissociation channel yielding S(1D) + O2(a1Δg/X3Σg−) products from vacuum ultraviolet (VUV) photodissociation of SO2 in the wavelength range between 120 and 160 nm. Experimental results show O2 production to be an important channel from SO2 VUV photodissociation, with a branching ratio of 30 ± 5% at the H Lyman-α wavelength (121.6 nm). The relatively large amounts of SO2 emitted from volcanic eruptions in the Earth's late Archaean eon imply that VUV photodissociation of SO2 could have provided a crucial additional source term in the O2 budget in the Earth's primitive atmosphere. The results could also have implications for abiotic oxygen formation on other planets with atmospheres rich in volcanically outgassed SO2.


S1. Experimental Methods
The present experimental apparatus is equipped with two independently tunable vacuum ultraviolet (VUV) laser systems coupled with the time-sliced velocity map ion imaging (TS-VMI) detection scheme, as shown schematically in Fig. S1. The VUV-VUV TS-VMI apparatus has been described previously in detail. [1][2][3] Briefly, two independently tunable VUV laser beams have been generated by, respectively, the VUV free electron laser (FEL) and a sum-/difference-frequency four-wave mixing (FWM) scheme by a table-top laser system. Henceforth, these sources are denoted as VUV-FEL and VUV-FWM, respectively. The VUV-FEL, which runs in the high gain harmonic generation (HGHG) mode, was used to photoexcite SO 2 molecules to higher electronic states of interest. The typical pulse energy of the VUV FEL is >200 μJ (attenuated to ~ 10 μJ in the present experiments), with the duration of ~1.5 ps in the wavelength range 50 ≤ λ ≤ 150 nm. The polarization of the VUV FEL was fixed in the horizontal plane and thus parallel to the front face of the MCP detector. The VUV-FWM beam at λ=130.092 nm was used to ionize the S( 1 D) photofragment via the autoionization Rydberg state S*[3s 2 3p 3 ( 2 D)5s ( 1 D 2 )]. The 130.092 nm light was generated as the difference-frequency (2ω 1 -ω 2 ) output in Kr gas, in which ω 1 was set at the frequency corresponding to a wavelength  =212.556 nm and the ω 2 was scanned around  ~580.654 nm so as to cover the Doppler profile. The ω 1 ( =212.556 nm) radiation was produced by frequency doubling the output of a 355 nm (Nd:YAG laser, Spectra Physics Pro-290) pumped dye laser (Sirah, PESC-G-24, 3000 l/mm) operating at λ ~425 nm. Half of the 532 nm output of the same Nd:YAG laser was used to pump another dye laser (Sirah, PESC-G-24) which operated around λ ~580.654 nm (~8-10 mJ pulse energy). To avoid any unwanted secondary dissociations, an off-axis biconvex LiF lens was used to disperse the 212.556 nm and 580.654 nm outputs from the direction of 130.092 nm beam, and the 130.092 nm radiation remained defocused. The polarization vector of the VUV-FWM output was determined by the polarization of the 580.654 nm radiation, which was fixed in the horizontal direction.
The pulsed supersonic beam was generated by expanding a mixture of 1% SO 2 3 and Ar into the source chamber where it was skimmed before entering (through a 2 mm hole in the first electrode) and propagating along the centre axis of the ion optics assembly (IOA, 23-plate ion optics) mounted in the reaction chamber. The molecular beam was intersected at right angles by the photolysis and probe laser beams between the second and the third plates of the IOA. The SO 2 molecules were photodissociated by the VUV-FEL pulse, and the S( 1 D) photoproducts were subsequently probed by single photon excitation at λ = 130.092 nm. The resulting S + ions were accelerated through the remaining ion optics and detected by a dual microchannel plate (MCP) detector coupled with a P43 phosphor screen at the end of the 740 mm ion flight tube.
In the current experiments, the detector was time gated to select ions with m/z 32 (15 ns gain pulse width). To confirm that the signal was from the intended two-color VUV pump-probe scheme, three images were recorded specifically under different experimental conditions: (i) with both pump (VUV-FEL) and probe (VUV-FWM) beams present in the interaction region; (ii) with the pump beam present but the probe beam being blocked, and (iii) with the pump beam blocked and the probe beam being present. The one-color photon induced background images recorded under conditions (ii) and (iii) were very weak and were subtracted from the two-color image recorded under condition (i). The speed of the S + product was calibrated by imaging the O + ions from the well-characterized multiphoton excitation and ionization of O 2 at 225.000 nm. 4,5

S2. The Absorption Spectrum of SO 2
As shown in Fig. S2, the room temperature absorption spectrum of SO 2 displays extensive absorption at all UV wavelengths below ~400 nm. 6,7 The strong absorption band between 185 and 235 nm is associated with the transition from ground (X̃1A 1 ) state to the C̃1B 2 state (C̃1B 2 X̃1A 1 ). A weaker absorption band between 240-350 nm corresponds to the transition to the B̃1B 1 state, which couples with the Ã 1 A 2 state via a conical intersection. 8,9 The direct Ã 1 A 2 X̃1A 1 transition is electric dipole forbidden at C 2v geometries, and the Ã state is expected to be a dark state (i.e., not observed directly by optical spectroscopy). Another weak absorption band at  >350 nm is ascribed to 4 the spin-forbidden transition to a triplet state (ã 3 B 1 X̃1A 1 ), not shown in Fig. S2.
Photons of wavelength  <219 nm have enough energy to break one of the SO bonds, leading to fragmentation, 10 whereas radiation at all longer wavelengths can only excite SO 2 to one of the excited states (the coupled B̃1B 1 /Ã 1 A 2 states and/or the ã 3 B 1 state) which then decay radiatively. 11 Photodissociation of SO 2 via the C̃1B 2 state has been investigated by several groups 12,13 due to its strong relation with the sulfur massindependent fractionation (S-MIF). Since the C̃1B 2 state correlates adiabatically to SO(a 1 Δ) + O( 1 D) products, the observed production of the lower energy SO(X 3 Σ  ) + O( 3 P) products must arise via non-adiabatic coupling. One possible pathway is internal conversion to the X̃1A 1 state, which then dissociates to SO(X 3 Σ  ) and O( 3 P) products. 14 Alternatively, a conical intersection formed with the higher lying D̃1A 1 state, which is repulsive and correlates with the SO(X 3 Σ  ) + O( 3 P) asymptote, facilitates predissociation of SO 2 via the C̃1B 2 state. 15 Another potential non-adiabatic pathway would involve intersystem crossing to the repulsive 2 3 A 1 state, which correlates to the same triplet asymptote. 15 The diffuse absorption bands at shorter wavelengths are assigned to transitions to Rydberg states. The Ẽ state was observed by Vuskovic and Trajmar 16 by electron impact excitation and assigned to the absorption band centred at ~154 nm. Features identified at higher energy include the F̃ Rydberg state, with an origin at 134 nm, 17 and the and Rydberg states which can be reached at wavelengths ~120 nm. 18  The well-resolved concentric rings evident in the images can be directly attributed to population of different vibrational states of the O 2 co-product in its ground (X 3 Σ g  ) or first excited (a 1 Δ g ) electronic state.

S4. The Branching Ratio Measurements
Any quantitative assessment of the importance of SO 2 photolysis in the VUV region and its possible role in abiotic oxygen formation in the early Earth's atmosphere requires determination of the branching ratios of all dissociation channels. Since Lyman-α (121.6 nm) photons dominate the stellar VUV radiation field, we take this as the representative wavelength at which to measure the branching ratios. The energetically allowed channels from photodissociation of SO 2 at 121.6 nm are listed below,  Table S1. Thus, we estimate a branching ratio of 30 ± 5 % for the S( 1 D) + O 2 (X 3 Σ  g /a 1 Δ g ) channel in SO 2 photolysis at 121.6 nm. The overall error associated with these experimental results was estimated to be ~15%.

S5. Computational Details
The present work employed the explicitly correlated version (F12) of the internally contracted multi-reference configuration interaction (ic-MRCI-F12) method, together with Dunning's augmented correlation-consistent triple-zeta (aug-cc-pVTZ) basis set, to calculate potential energy curves relevant to SO 2 photodissociation. In some cases, the Davidson correction (+Q) was also employed for the MRCI-F12, to better account for contributions from higher excitations; these results are denoted as ic-MRCI-F12+Q.
The electronic structure calculations were performed using the MOLPRO 2018 suite of programs. 22 The reference wave functions were based on state averaged complete active space self-consistent field (SA-CASSCF) wavefunctions and the full-valence active space employed consisted of 18 electrons in 12 active orbitals (i.e. an 18 (12) active space, with the 1s orbitals of O and S and 2s and 2p orbitals of S frozen). Potential 9 energies were calculated in C s symmetry, and the SA-CASSCF wavefunctions were averaged over the six lowest energy singlet and six lowest energy triplet states of both Aꞌ and Aꞌꞌ symmetry, i.e. a total of twelve Aꞌ and twelve Aꞌꞌ states were finally obtained. Table S2 compares the equilibrium geometry of the global minimum of the ground state of SO 2 calculated in the present work with available experimental data [23][24][25] and with previous theoretical results. 8,14 The current theoretical results are in very good agreement with the recent experimental results and confirm the C 2v equilibrium geometry of the ground state. 23,24 To obtain the correct asymptotic energies for different S + O 2 product pairs (i.e. dissociation energies for the various electronic states), the MRCI-F12+Q energies of S and O 2 were calculated separately, using 12(8) and 6 (5) active spaces for O 2 and S, respectively, in both the CASSCF/AVTZ and MRCI-F12+Q/AVTZ calculations. Table S3 shows that the dissociation energies derived here accord well with the experimental values. the many A' and A" states that correlate with these various low energy dissociation limits (and we note that two excited states with quintet spin multiplicity will also correlate with the ground state products). The density of electronic states in the vertical Franck-Condon region in the energy range of current interest (8-10 eV) will be very high, which can be expected to facilitate easy transfer of population between states (via internal conversions and/or intersystem crossings) once SO 2 is excited to these states.