Mechanisms of glycine formation in cold interstellar media: a theoretical study

The possibility of the formation of glycine (Gly) from fundamental gas molecules in cold interstellar media was studied using quantum chemical methods, transition state theory and microcanonical molecular dynamics simulations with surface hopping dynamics (NVE-MDSH). This theoretical study emphasized five photochemical pathways in the lowest singlet-excited (S 1) state, thermochemical processes after non-radiative S 1→S 0 relaxations, and photo-to-thermal energy conversion in the NVE ensemble. The optimized reaction pathways suggested that to generate a reactive singlet dihydroxy carbene (HOCOH) intermediate, photochemical pathways involving the H2O…CO van der Waals and H2O−OC hydrogen bond precursors (Ch (1)_Step (1)) possess considerably lower energy barriers than the S 0 state pathways. The Gibbs free energy barriers (∆G ǂ ) calculated after the non-radiative S 1 →S 0 relaxations indicated higher spontaneous temperatures (T s) for the formation of the HOCOH intermediate (Ch (1)_Step (1)) than for Gly formation (Ch (1)_Step (2) and Ch (4)). Although the termolecular reaction in Ch (4) possesses a low energy barrier, and is thermodynamically favourable, the high exothermic S 1 →S 0 relaxation energy leads to the separation of the weakly associated H2O…CH2NH…CO complex into single molecules. The NVE-MDSH results also confirmed that the molecular processes after the S 1 →S 0 relaxations are thermally selective, and because the non-radiative S 1 →S 0 relaxation temperatures are exceedingly higher than T s, the formation of Gly on consecutive reaction pathways is non-synergistic with low yields and several side products. Based on the theoretical results, photo-to-thermal control strategies to promote desirable photochemical products are proposed. They could be used as guidelines for future theoretical and experimental research on photochemical reactions.

The possibility of the formation of glycine (Gly) from fundamental gas molecules in cold interstellar media was studied using quantum chemical methods, transition state theory and microcanonical molecular dynamics simulations with surface hopping dynamics (NVE-MDSH).This theoretical study emphasized five photochemical pathways in the lowest singlet-excited (S 1 ) state, thermochemical processes after non-radiative S 1 →S 0 relaxations, and phototo-thermal energy conversion in the NVE ensemble.The optimized reaction pathways suggested that to generate a reactive singlet dihydroxy carbene (HOCOH) intermediate, photochemical pathways involving the H 2 O…CO van der Waals and H 2 O−OC hydrogen bond precursors (Ch (1)_Step (1)) possess considerably lower energy barriers than the S 0 state pathways.The Gibbs free energy barriers (∆G ǂ ) calculated after the non-radiative S 1 →S 0 relaxations indicated higher spontaneous temperatures (T s ) for the formation of the HOCOH intermediate (Ch (1)_Step (1)) than for Gly formation (Ch (1)_Step (2) and Ch (4)).Although the termolecular reaction in Ch (4) possesses a low energy barrier, and is thermodynamically favourable, the high exothermic S 1 →S 0 relaxation energy leads to the separation of the weakly associated H 2 O…CH 2 NH…CO complex into single molecules.The NVE-MDSH results also confirmed that the molecular processes after the S 1 →S 0 relaxations are thermally selective, and because the non-radiative S 1 →S 0 relaxation temperatures are exceedingly higher than T s , the formation of Gly on consecutive reaction pathways is non-synergistic with low yields and several side products.Based on the theoretical results, photo-to-thermal control strategies to promote desirable photochemical products are proposed.They could be used as guidelines for future theoretical and experimental research on photochemical reactions.

Introduction
The photochemistry of small molecules is fundamental to understanding photochemical reactions in the Earth's atmosphere and interstellar media [1].The mechanisms for the formation of amino acids through photochemical processes have received special attention in recent decades because experiments have shown that they can be generated in cold interstellar media through UV irradiation [2,3].The reactions generally involve reactive intermediates, which cannot be detected easily in experiments [4][5][6].As a small proteinogenic molecule, glycine (NH 2 CH 2 COOH, abbreviated Gly) has been frequently chosen as a model molecule for studying the photochemistry of polypeptides [7].As the smallest amino acid found in meteorites [8], attempts have been made to synthesize Gly under extraterrestrial [2] and primitive Earth conditions from fundamental gases such as H 2 , CO, H 2 O, NH 3 and CH 4 .The Strecker synthesis [9] and Miller-Urey experiments [10] are well-known classical examples used to study prebiotic chemistry and the origin of life.In the Miller-Urey prebiotic experiment, more than 20 different amino acids were produced in continuous electrical sparks caused by a pair of electrodes.
To study the multiple Gly synthesis pathways in the Miller-Urey experiment, the so-called 'ab initio nanoreactor' was used [11].The theoretical method is based on ab initio molecular dynamics (AIMD) simulations on reacting molecules and analysis and refinements to suggest an accurate reaction network.The ab initio method was conducted in the electronic ground (S 0 ) state using the Hartree-Fock (HF) method with the 3-21G basis set (abbreviated HF/3-21G).The minimum energy pathways obtained based on AIMD simulations indicated that in the S 0 state, a significant fraction of the elementary reactions takes place with energy barriers (ΔE ǂ ) less than 209 kJ mol −1 , and Gly could be formed via four different pathways, in which singlet dihydroxy carbene (HOCOH), methanimine (CH 2 NH) and methanolamine (HOCH 2 NH 2 ) play an important role, e.g. in channels (1), ( 3) and (6), in figure 1a, hereafter abbreviated Ch (1), Ch (3) and Ch (6), respectively.
Alternative bimolecular and termolecular reactions to generate Gly in the S 0 state were proposed in [12], in which the precursors were suggested to be CO 2 and H 2 , and CO 2 , H 2 and CH 2 NH in Ch (2) and ( 5) in figure 1b, respectively.Based on the S 0 potential energy curves computed using the DFT/ 6-31++G(d,p) method with zero-point vibrational energy corrections (ZPC), formation of the HOCOH intermediate (Ch (2)_Step (1)) possesses a high energy barrier, whereas the reaction between HOCOH and CH 2 NH to generate Gly (Ch (2)_Step ( 2)) is almost barrierless.In contrast, while the concerted termolecular reaction (CO, H 2 O and CH 2 NH) in Ch (4) proceeds on a low energy barrier path [11], the termolecular reaction (CO 2 , H 2 and CH 2 NH) in Ch (5) possesses a considerably higher energy barrier with a possibility for quantum mechanical tunnelling of the hydrogen atom.
In our previous study [13], photodissociations of Gly in the lowest singlet-excited (S 1 ) state were theoretically studied using the DFT, time-dependent DFT (TD-DFT) and complete active space second-order perturbation theory (CASPT2) methods with the aug-cc-pVDZ basis set and microcanonical molecular dynamics simulations with surface hopping dynamics (NVE-MDSH).This theoretical study emphasized unimolecular dissociation and isomerization and non-radiative S 1 →S 0 relaxations that generate molecular products in cold interstellar media.The results showed that the high exothermic S 1 →S 0 relaxation temperature could lead to at least eight molecular product channels, in which the vertically excited Gly was primarily dissociated and isomerized, leading to reactive and stable molecules in different temperature ranges.The thermally selective reactions favoured the lowest energy conformer structure Ip as the precursor.These theoretical results showed in detail for the first time how non-radiative S 1 →S 0 relaxations generate molecules in cold interstellar media without heat transfer from the surroundings.
Because the reported energy barriers for the formation of Gly in the S 0 state are in our opinion too high to occur in cold interstellar media [11], and our previous studies showed that reactive intermediates can be effectively generated through photochemical processes in the S 1 state, from which molecular products can be subsequently formed in the S 0 state, an attempt was made in this work to search for low energy barrier photochemical pathways for Gly formation.The theoretical study began with the S 0 →S 1 vertical excitation of fundamental gas precursors in the five reaction channels (Ch (1)−( 5) in figure 1b), for which non-radiative photochemical pathways to generate intermediates in the S 1 state were primarily studied using the DFT/B3LYP/aug-cc-pVDZ, TD-DFT/B3LYP/aug-cc-pVDZ and nudged elastic band (NEB) methods.Because the DFT and TD-DFT methods are based on a single-reference wavefunction approximation and multiconfigurational characteristics of covalent bond dissociation and formation could be important, the structures at the S 0 /S 1 intersection were refined using the state average-coupled perturbed multiconfigurational self-consistent field (SA-CPMCSCF/ 6-31G(d)) method, from which the non-radiative S 1 →S 0 relaxations of the intermediates and formations of Gly and side products were studied in the S 0 state using the DFT/B3LYP/aug-cc-pVDZ method.The performance of the DFT and TD-DFT methods was also assessed using the CASPT2 method.
Because formations of reactive intermediates in the S 1 state are usually ultrafast [4][5][6]13], the kinetic and thermodynamic properties were studied only after the non-radiative S 1 →S 0 relaxations using the transition state theory (TST) method.The dynamics and mechanisms for the formation of Gly and side products were further studied using NVE-MDSH simulations, from which the role played by the exothermic S 1 →S 0 relaxation energy and the interplay between the photo and thermal energies were analysed and discussed in detail.

Quantum chemical calculations
To study the structures and energetics of the precursors in the S 0 and S 1 states, DFT and TD-DFT methods with the Becke, 3-parameter, Lee-Yang-Parr (B3LYP) hybrid functional and aug-cc-pVDZ basis set were used.For the TD-DFT method, the Tamm-Dancoff approximation was applied to avoid singlet instabilities in the S 1 state calculations [14].Although the DFT/B3LYP/aug-cc-pVDZ and TD-DFT/B3LYP/aug-cc-pVDZ methods had been systematically tested and were applied successfully in our previous work on Gly [13], they were assessed again in this study using the CASPT2/aug-cc-pVDZ method.The active spaces used in the CASPT2 calculations are summarized in electronic supplementary material, table S1.Both DFT and TD-DFT calculations were made using the TURBO-MOLE 7.50 software package [15].The state-averaged CASPT2 calculations were performed in the S 0 and S 1 states with equal weights using the MOLPRO software package [16,17], for which the Werner-Mayer-Knowles nonlinear method was used in orbital/state optimizations [18][19][20].

Geometry optimizations
Because our previous study [13] showed that photochemical products depend strongly on the structures of the vertically excited precursors in the S 0 state and the structures at the S 0 /S 1 intersection and because fundamental gases such as H 2 , CO, CO 2 and H 2 O were suggested to be involved in Gly formation (Ch (1)−( 5)) [11], their equilibrium structures in the S 0 state and S 0 →S 1 vertical excitation energies (ΔE Ex ) were calculated using the DFT/B3LYP/aug-cc-pVDZ and TD-DFT/B3LYP/aug-cc-pVDZ methods, respectively.Based on the proposed mechanisms in [11,12] (figure 1b), the equilibrium structures of the van der Waals (VdW) and hydrogen bond (H-bond) precursors and intermediates in Ch (1)−( 5) were also optimized in the S 0 state.To assess the DFT method, the optimized structures were reoptimized using the CASPT2 method.

Reaction pathway optimizations
Because the photochemical pathways that generate Gly in cold interstellar media were hypothesized in this work to involve the S 0 →S 1 vertical excitation of the precursors (binary or ternary complexes formed from fundamental gases), non-radiative S 1 →S 0 relaxations of the excited precursors, and generation of the intermediates/products in the S 0 state, reaction pathway optimizations were performed for Ch (1)−(5) in both S 1 and S 0 states using the NEB method [21].The NEB potential energy curves were computed using limited-memory Broyden-Fletcher-Goldfarb-Shanno (LBFGS) optimizers included in the ChemShell software package [22].
In this work, to verify the results obtained from the TD-DFT and DFT methods, the S 1 and S 0 potential energy curves were compared with CASPT2 calculations with the geometries obtained from the TD-DFT and NEB methods.The structures at the S 0 /S 1 intersection were confirmed using the SA-CPMCSCF/6-31G(d) conical intersection optimization method (electronic supplementary material, table S2).A schematic diagram showing step-by-step non-radiative photochemical pathway optimizations and steps to optimize the reaction pathways for the formation of the HOCOH intermediate from the H 2 O…CO VdW precursor in Ch (1)_Step (1) are explained in detail in the electronic supplementary material.In this work, to discuss the reaction pathways, […] eq and […] denote the equilibrium structures and structures on the S 0 potential energy curves, respectively, whereas […] * are the structures on the S 1 potential energy curves.The transition structures and structures at the S 0 /S 1 intersection are […] ǂ and […] § , respectively.

Transition state theory calculations
Because kinetic and thermodynamic properties were not reported in all previous quantum chemical studies, the rate constants, enthalpies and Gibbs free energies for the formation of Gly and side products were calculated in this work using the TST method [21,23,24].The calculations of the kinetic and thermodynamic properties using the TST method are discussed in [25] and in the electronic supplementary material (tables S3−S5).

Kinetics of reaction pathways
The classical (k Class ) and quantized-vibrational (k Q-vib ) rate constants [26] were primarily computed over the temperature range of 100-4500 K. k Q-vib was computed using the ZPC barrier obtained by including the zero-point correction energy (ΔE ZPE ) to the energy barrier obtained from the NEB method (ΔE ǂ ).To study the effects of quantum mechanical tunnelling on the reaction rates, the crossover temperatures (T c ) were computed; for the pathways that involve proton/H-atom transfer, quantum mechanical tunnelling could dominate below T c [27,28].In addition, the simple Wigner-corrected rate constants (k S-Wig ) were calculated, for which k S-Wig = k Q-vib at the classical limit (ħ = 0) [29].

Thermodynamics of reaction pathways
The activation free energies (ΔG ǂ ), enthalpies (ΔH ǂ ) and entropies (ΔS ǂ ) of the elementary reactions were computed from the rate constants (k Q-vib ) obtained from the TST method.All the kinetic and thermodynamic calculations were performed using the DL-FIND program [30] included in the ChemShell package [22].

Surface hopping molecular dynamics simulations
To study the time evolutions of the photochemical pathways suggested based on the NEB and TST methods, NVE-MDSH simulations were performed using the equilibrium structures of the VdW and H-bond precursors.The initial configurations for the S 0 →S 1 vertical excitations were generated using the Wigner distribution (n = 307), from which NVE-MDSH simulations were performed in the S 1 and S 0 states using the TD-DFT and DFT methods, respectively.
In this work, the NEWTON-X software package [31,32] interfaced with TURBOMOLE 7.50 was used to study the surface hopping dynamics over a time span of ~4 ps.The integration of Newton's equations of motion was conducted with a timestep of 0.5 fs.The improved version of the fewest switches algorithm introduced by Hammes-Shiffer and Tully [33,34] was used in the calculations of the non-adiabatic transition probabilities between the S 1 and S 0 states; the Tully type fewest switches surface hopping algorithm allows switches between the S 1 and S 0 states [32].
For NEWTON-X/TURBOMOLE 7.50 calculations with the TD-DFT method, NVE-MDSH simulations were based on the overlap of the wavefunctions [31,32].The populations of states, structures of the VdW and H-bond complexes, and dynamics before and after the non-radiative S 1 →S 0 relaxation (surface hopping) were analysed in detail using the tools provided with NEWTON-X.The focus was on the reaction conditions (e.g. the exothermic S 1 →S 0 relaxation temperatures and vertical excitation energies) and probabilities for formation of Gly and side products in the S 0 state and interplay between the photo and thermal energies.

Equilibrium structures
The equilibrium structures of molecules, VdW and H-bond precursors obtained from the DFT and CASPT2 methods are summarized in electronic supplementary material, table S1.It appears that all the equilibrium structures obtained from the DFT method are in excellent agreement with the CASPT2 optimized geometries.For example, two equilibrium structures are obtained for HOCOH, namely, the s-and m-forms, [HOCOH] s,eq and [HOCOH] m,eq , respectively.The VdW and H-bond complexes formed from H 2 O and CO are suggested by the DFT method to be the global and local minimum energy geometries, [H 2 O…CO] eq and [H 2 O…OC] eq , respectively.The equilibrium structures in electronic supplementary material, table S1, were used in the NEB reaction pathway optimizations.

The S 1 and S 0 potential energy curves
All the S 0 and S 1 potential energy curves and profiles obtained from the DFT, TD-DFT, CASPT2 and NEB methods are included in the electronic supplementary material.Examples of the potential energy profiles used in the discussion are illustrated in figure 3.
It appears that all the S 1 potential energy curves obtained from the TD-DFT and NEB methods reveal barrierless energies for non-radiative S 1 →S 0 relaxations of the vertically excited VdW and H-bond precursors and SA-CPMCSCF/6-31G(d) calculations confirm all the structures at the S 0 /S 1 intersections (electronic supplementary material, table S2 3a,b, respectively. It should be noted that because the CASPT2 potential energy curves were constructed using the geometries obtained from the TD-DFT and NEB methods, low energy barriers were observed on the S 1 potential energy curves for [H 2 O…OC] * →[COH…OH] § and [H 2 O…CO] * →[OCH…OH] § .To confirm barrierless potentials in the S 1 state, all the degrees of freedom in the precursors, except the O-H…O H-bond and O…C VdW distances, were allowed to relax in CASPT2 calculations.The barrierless potentials obtained from the CASPT2 relax-scan calculations (electronic supplementary material, figures S2a and S3a) lead to the conclusion that the S 0 →S 1 vertically excited H-bond and VdW precursors are instantly transformed into structures at the S 0 /S 1 intersections, and the kinetic and thermodynamic properties of the photochemical pathways should be studied after non-radiative S 1 →S 0 relaxations.

Bimolecular reactions 3.3.1. Ch (1)_Step (1)
For the bimolecular reaction in Ch (1)_Step (1) (figure 1b), both [H 2 O…CO] eq and [H 2 O…OC] eq were used as the precursors in NEB calculations.The S 0 potential energy profiles after the non-radiative S 1 →S 0 relaxation show that the S 0 →S 1 vertically excited [H 2 O…CO] eq VdW precursor involves a higher energy barrier for the formation of [HOCOH] s,eq than the [H 2 O…OC] eq H-bond precursor; for [OCH…OH] § →[HOCOH] s,eq , ΔE ǂ = 167 kJ mol −1 (figure 3a), whereas ΔE ǂ = 123 kJ mol −1 for [COH… OH] § →[HOCOH] s,eq (electronic supplementary material, figure S2b).The higher energy barrier could be attributed to the formation of [HCOOH] on a barrierless potential prior to the formation of [HOCOH] s,eq .Similar results were obtained for the formation of [HOCOH] m,eq , for which the S 0 →S 1 vertically excited [H 2 O…CO] eq involves a high energy barrier due to the formation of HCOOH; for [OCH…OH] § →[HOCOH] m,eq , ΔE ǂ = 148 kJ mol −1 (electronic supplementary material, figure S3c).It should be noted that although [COH…OH] § →[HOCOH] m,eq possesses a lower energy barrier (ΔE ǂ = 124 kJ mol −1 in electronic supplementary material, figure S2c), the formation of [HOCOH] m,eq on this reaction pathway has to pass through [HOCOH] s and the energy barrier for the interconversion between the s-and m-forms on the S 0 state pathway is moderate, ΔE ǂ = 68 kJ mol −1 (electronic supplementary material, figure S3d).Therefore, only [HOCOH] s,eq was chosen in further study on Ch (1).Comparison of the above-discussed photochemical pathways with the S 0 state pathways shows considerably lower energy barriers; for [H 2 O…CO] eq →[HOCOH] s,eq and [H 2 O…OC] eq →[HOCOH] s,eq on the S 0 state pathways, ΔE ǂ = 236 (figure 3a) and 242 kJ mol −1 (electronic supplementary material, figure S2c), respectively.These high energy barriers result mainly from the O−H→O unimolecular isomerization on the S 0 state pathways.

Ch (2)_Step (1)
Because H 2 and CO 2 are stable molecules, the formation of [HOCOH] eq on both photochemical and S 0 state pathways involves considerably high energy barriers, ΔE ǂ > 230 kJ mol −1 .Therefore, irradiation seems not to help promote the formation of [HOCOH] eq in Ch (2)_Step (1).It should be mentioned that the potential energy profile reveals the formation of [HCOOH] eq from [HCOO…H] § on a barrierless potential (electronic supplementary material, figure S4b).Because [HCOOH] eq could be generated as a side product from [H 2 O…CO] eq and [H 2 …CO 2 ] eq , and [HCOOH] eq is also a precursor in Ch (3)_Step (1), an attempt was made to study the possibility for the formation of [HOCOH] eq from [HCOOH] eq .The results (electronic supplementary material, figure S5a) show that the S 0 →S 1 vertically excited [HCOOH] eq could non-radiatively relax to the S 0 state, and after the non-radiative S 1 →S 0 relaxation, [OCH…OH] § →[HOCOH] eq involves a moderate energy barrier (ΔE ǂ = 123 kJ mol −1 ).Therefore, [HCOOH] eq could be an effective precursor for the formation of [HOCOH] eq .

Ch (1)_Step (2)
The potential energy profiles for the formation of Gly ([Ip] eq ) from [HOCOH−NHCH 2 ] eq are shown in figure 3b.It appears that the energy barriers for the formation of [Ip] eq on the photochemical and S 0 state pathways are not significantly different.The S 0 →S 1 vertically excited H-bond precursor barrierlessly relaxes to the S 0 /S 1 intersection, and [HOCO…NH 2 CH 2 ] § →[Ip] eq in the S 0 state has a low energy barrier, ΔE ǂ = 41 kJ mol −1 ; the energy barrier for [HOCOH−NHCH 2 ] eq →[Ip] eq on the S 0 state pathway is slightly lower, ΔE ǂ = 34 kJ mol −1 .These results lead to the conclusion that the formation of Gly in Ch (1)_Step (2) could occur on both photochemical and S 0 state pathways.

Ch (3)_Step (1) and Step (2)
For the bimolecular reactions in Ch (3), the potential energy profiles reveal that although the formation of [Ip] eq from [HOCO…NH 2 CH 2 ] § (Ch (3)_Step (2) in figure 3c) involves a barrierless energy, the formation of [NH 2 CH 2 OCOH] eq on the photochemical pathway (Ch (3)_Step (1) in electronic supplementary material, figure S6a) possesses a moderate energy barrier in the S 0 state (ΔE ǂ = 117 kJ mol −1 ), whereas the energy barrier for the formation of [Ip] eq directly from [HCOOH…NHCH 2 ] eq on the S 0 state pathway has a comparable energy barrier (ΔE ǂ = 119 kJ mol −1 ; electronic supplementary material, figure S6c).These results lead to the conclusion that the photochemical process does not help promote the formation of [Ip] eq in Ch (3) and [HCOOH…NHCH 2 ] eq →[Ip] eq could occur on the S 0 state pathway.
Based on the above discussion, one can conclude that for bimolecular reactions, the photochemical pathways in Ch ( 1) and ( 3) are more favourable than those in Ch (2).For the photochemical pathways in Ch (1), the non-radiative S 1 →S 0 relaxation of the vertically excited H 2 O…CO VdW precursor could generate HCOOH as a side product, from which HOCOH could be formed.In addition, the formation of Gly from the HCOOH−CH 2 NH H-bond complex in Ch (1)_Step ( 2) is energetically feasible on both photochemical and S 0 state pathways.
It should be noted, for example, that although the structure after the S 1 →S 0 relaxation ([H 2 O… NHCH 2 …CO] in figure 3d) and reaction precursor are similar, they are not exactly the same.Based on different NEB gradients, the photochemical pathway (right-hand side) and S 0 state pathway (left-hand side) proceed through different transition states; on the photochemical pathway, the S 0 /S 1 structure ([HO…NH 2 CH 2 …CO] § ) is the starting geometry in the NEB calculations, whereas on the S 0 state pathway, the equilibrium structure ([H 2 O…NHCH 2 …CO] eq ) is the starting geometry.Although the photochemical pathway possesses lower ΔE ǂ , without the effect of thermal energy, it is difficult to decide on which pathway the reaction will proceed.Therefore, to complete the mechanistic study, TST calculations and NVE-MDSH simulations must be performed to explore favourable thermodynamic conditions, time evolutions and statistics of the formation of the intermediates and products on the candidate reaction pathways.
For the termolecular reaction in Ch ( 5), while the reaction on the S 0 state pathway is similar to that in Ch (4) (ΔE ǂ = 139 kJ mol −1 ), the vertically excited [H 2 …NHCH 2 …CO 2 ] eq precursor readily transforms into [HOCO…NH 2 CH 2 ] § , and after the non-radiative S 1 →S 0 relaxation, the breaking of the C−O covalent bond and N−H→Ο proton/H-atom transfer lead to [H 2 O…NHCH 2 …CO] eq (electronic supplementary material, figure S7b), which is the precursor in Ch (4).These findings suggest that the photochemical process in Ch (5) does not help promote the formation of [Ip] eq .The photochemical pathways obtained based on the DFT, TD-DFT and NEB methods will be regarded as static reaction pathways.

Kinetic and thermodynamic properties
To further study the candidate photochemical pathways (Ch (1), ( 3) and ( 4)), kinetic and thermodynamic properties were computed using the TST method, from which all the results are summarized in electronic supplementary material, tables S3−S5.Outstanding results used in the discussion are included in electronic supplementary material, tables S6−S8.Because NVE-MDSH simulations (to be discussed in the forthcoming subsections) confirm no stable intermediate in the S 1 state, the discussions are focused only on the scenarios after the non-radiative S 1 →S 0 relaxations.In addition, because the NVE-MDSH results revealed that the average exothermic S 1 →S 0 relaxation temperatures for all the photochemical pathways are high, the discussion will emphasize the thermodynamic spontaneity for the formation of Gly and side products.

Quantum mechanical effect
Because the average exothermic S 1 →S 0 relaxation temperatures obtained from NVE-MDSH simulations are higher than T c , the quantum mechanical effect is concluded not to be important in the present systems; the highest quantum mechanical effect is for [HOCOH−NHCH 2 ]→[HOCO−CH 2 NH 2 ] ǂ , T c = 216 K (Ch (1)_Step ( 2) in electronic supplementary material, table S6b), below which the rate constants for the O−H→N proton/H-atom transfer are significantly different, e.g. in electronic supplementary material, table S3e, at T = 100 K, k f Class = 2.64 × 10 −10 , k f Q-vib = 2.97 × 10 −6 and k f S-Wig = 2.59 × 10 −5 s −1 , whereas at high temperatures, e.g. at T = 1500 K, k f Q-vib and k f S-Wig are comparable, 4.20 × 10 10 and 4.34 × 10 10 s −1 , respectively.Therefore, it is reasonable to use only k Q-vib in further analysis on the exothermic S 1 →S 0 relaxation processes.
The spontaneous temperature for the formation of the [NH 2 CH 2 COOH] ǂ transition structure is T s = 3763 K.It should be noted that although the termolecular reaction in Ch (4) possesses a low energy barrier, and is thermodynamically favourable, the high exothermic S 1 →S 0 relaxation energy could lead to separation of the weakly associated H 2 O…CH 2 NH…CO ternary complex into single molecules.This issue will be discussed based on the NVE-MDSH results.

Dynamic reaction pathways
The dynamics of the photochemical pathways in Ch ( 1), ( 3) and ( 4) are discussed in detail using the NVE-MDSH results.The emphases are on the statistics and effect of the exothermic S 1 →S 0 relaxation temperature on the thermodynamic spontaneity, as well as the thermal and photo selectivity.Because the reaction pathways after the non-radiative S 1 →S 0 relaxations are thermally selective [13], they can be classified into two categories, as in the case of the Gibbs free energy profiles (figure 2).For the single-step reaction pathway, intermediate(s) or product(s) could be formed directly from the structure at the S 0 /S 1 intersection ((I) § →(II)), characterized by an average exothermic S 1 →S 0 relaxation temperature.However, more than one average exothermic S 1 →S 0 relaxation temperature corresponding to the formation of transient intermediate(s) and product(s) is considered characteristic of consecutive reaction pathways, (I) § →(II) followed by (II)⇌(III) ǂ →(IV).

Pathway statistical analysis
Starting from 307 Wigner-sampled VdW and H-bond precursors in Ch (1), ( 3) and ( 4), 62 NVE-MDSH simulations are without S 1 →S 0 surface hopping.Therefore, 245 NVE-MDSH simulations are used in the dynamic analysis (100%).It appears that in cold interstellar media (NVE ensemble), the formation of Gly is feasible in photochemical pathways with several side products, depending upon the ΔE Ex of the precursors, exothermic S 1 →S 0 relaxation and spontaneous temperatures.For example, the HOCOH intermediate in Ch (1) (figure 3a) could be formed from the HCOOH precursor (1.63% in electronic supplementary material, table S9a), and Gly could be generated in Ch ( 1)_Step (2) (0.82% in electronic supplementary material, table S9b).It also appears that 11 side products, such as OH, HCN, HNC, HCOOH and H 2 (electronic supplementary material, table S10), are observed in the NVE-MDSH simulations.
To explain these findings, the NVE-MDSH results on the VdW and H-bond precursors are systematically analysed.Because the total average exothermic S 1 →S 0 relaxation temperatures ( T S0 Ch(n) , Because NVE-MDSH simulations on Ch (3)_Step_(1) (18.4% in figure 5b) do not generate the NH 2 CH 2 OCOH intermediate because of the high exothermic S 1 →S 0 relaxation temperature (MDSH_ (11)), the photochemical pathway in Ch (3) might not be feasible in cold interstellar media and therefore was ruled out from the discussion.
The time evolutions of the structures, energetics and exothermic S 1 →S 0 relaxation temperatures for the formation of Gly from the bimolecular reaction in Ch (1)_Step (2) (MDSH_( 8)) and side products from the termolecular reaction in Ch (4) (MDSH_( 14)) are included in figure 6 as an example.
C−C covalent bond formation, T S0 Av, MDSH_(8) = 3990 ± 776 K.The low percentage for Gly formation could be attributed again to high T S0 Av, MDSH_( 8) compared with T s (electronic supplementary material, table S6b); the Gibbs free energy plots in electronic supplementary material, figure S8d, suggest T s = 974 K, whereas the formation of Gly in MDSH_(8) begins at a higher temperature, T = 1219 K at 546 fs (figure 6a).

Photoselectivity of precursors
The correlations between the average exothermic S 1 →S 0 relaxation temperatures and S 0 →S 1 vertical excitation energies for the representative single-step and consecutive reaction pathways (T S0 Av, MDSH_(n) , n = 1−9 and 12−14, and ΔE Ex in figure 5) are shown in figure 7, together with the s.d.The statistics for the formation of the intermediate(s)/molecular product(s) through covalent bond breaking and/or formation, proton/H-atom exchange, and isomerization are summarized in electronic supplementary material, table S10.These pieces of information are used to discuss the thermal selectivity for the formation of molecular products.The correlation diagram in figure 7a   CO and eventually separated precursor molecules as the products (electronic supplementary material, figures S9a and S10a).

Thermal selectivity and molecular processes
To acquire fundamental information to control thermochemical processes, molecular mechanisms (processes) underlying the formation of molecular products after the exothermic S 1 →S 0 relaxation in figure 7a are studied in detail.In this study, the ΔE Ex of the Wigner-sampled precursors and the characteristic temperatures for covalent bond dissociation and inter-and intramolecular proton/Hatom exchanges are analysed, classified and included in the photothermal correlation matrix in figure 7b.The analysis of figure 7b suggests that starting from the same type of precursor, the exothermic S 1 →S 0 relaxation temperatures of consecutive reaction pathways are higher than those of single-step reaction pathways due to formation and/or dissociation of the transient intermediates.
It appears that for strong N−H…O H-bonds, the inter-and intramolecular proton/H-atom exchanges dominate in the low average exothermic S 1 →S 0 relaxation temperature range and low ΔE Ex

Photo-to-thermal control strategies
Because the theoretical results in this work suggested that the formation of intermediates or products requires heat to overcome energy barriers in the consecutive reaction pathways, but the average exothermic S 1 →S 0 relaxation temperatures are exceedingly higher than T s (figure 7b), a photo-to-thermal control strategy is required in the experiment; the non-synergistic effect of the photochemical and thermochemical pathways could be one of the reasons for the lack of product or low product yields.
To enhance the product yields, a heat transfer or removal mechanism through inert gas molecules could be incorporated in the photochemical reaction chamber to carry away excess heat and cool down the temperature to below T s .For example, for the formation of the HOCOH intermediate and Gly in Ch (1), the reaction temperatures after the exothermic S 1 →S 0 relaxation should be controlled below T s = 6321 and 974 K, respectively.In this case, non-reactive heat transfer gases (diluent gases), such as He, Ar, N 2 , CO 2 or H 2 , could be used as 'heat absorbing gases' [35], among which CO 2 could be a reasonable choice due to its high heat transfer capability and high ΔE Ex compared with the other VdW and H-bond precursors (figure 7c).
Alternatively, because photochemical reactions could be considered thermal reactions of electronically excited molecules, instead of photoexcitation directly at the VdW and H-bond precursors, photo-to-thermal converters (self-heating molecules or photothermal catalysts) could be applied to generate tuneable thermal energy in the photochemical reaction chamber.For example, through the plasmonic localized heating process, photothermal nanomaterials with high electron mobility, such as Au, Ag, Cu and Al, could transform the absorbed photon energy into highly concentrated thermal energy within 0.1−1.0ps [36].

Conclusions
The possibility of the formation of Gly from fundamental gas molecules in cold interstellar media was studied using quantum chemical methods, TST and NVE-MDSH simulations.While previous theoretical studies focused on the multiple pathways to generate Gly in the S 0 state, this study emphasized photochemical pathways in the S 1 state, thermal selectivity and thermodynamic spontaneity after non-radiative S 1 →S 0 relaxations, and photo-to-thermal energy conversion in the NVE ensemble.
The optimized bimolecular and termolecular reaction pathways revealed barrierless energies for non-radiative S 1 →S 0 relaxations, and the photochemical pathways to generate the HOCOH intermediate from the H 2 O…CO VdW and H 2 O−OC H-bond precursors (Ch (1)_Step ( 1)) involve considerably lower energy barriers than on the S 0 state pathways and the pathways using the H 2 …CO 2 VdW precursor (Ch (2)_Step ( 1)).The optimized reaction pathways also suggested that HCOOH could be formed as an intermediate in Ch (1)_Step (1), and HOCOH could be generated from the S 0 →S 1 vertical excitation of HCOOH.For the termolecular reactions, the photochemical pathway that involves the H 2 …CH 2 NH…CO 2 precursor (Ch (5)) possesses a significantly higher energy barrier than H 2 O… CH 2 NH…CO (Ch (4)) due to the presence of stable molecules (H 2 and CO 2 ).
The TST results suggested that the T s for the photochemical pathway using the H 2 O…CO VdW precursor (Ch (1)_Step ( 1)) is the highest, whereas T s for formation of Gly from the HOCOH−CH 2 NH intermediate (Ch (1)_Step ( 2)) is significantly lower; T s is the temperature below which the transition structure is spontaneously formed after the exothermic S 1 →S 0 relaxation.The total Gibbs free energy changes (∆G˚, Tot ) confirmed that Ch (1)_Step ( 2) is thermodynamically more favourable than Ch (1)_Step (1).
To study the photo and thermal selectivity after the exothermic S 1 →S 0 relaxation, the NVE-MDSH results on the photochemical pathways in Ch (1), ( 3) and ( 4) were subdivided into 15 dynamic reaction pathways.Because all the exothermic S 1 →S 0 relaxation temperatures are higher than T s , almost all the S 0 →S 1 vertically excited Wigner-sampled VdW and H-bond precursors in Ch (1)_Step (1) take single-step reaction pathways and return to the precursors, whereas HOCOH and HCOOH could be formed as transient intermediates on consecutive reaction pathways.NVE-MDSH simulations anticipated that in cold interstellar media, without heat exchange between the system and surrounding, formation of Gly is feasible on the photochemical pathways through Ch (1)_Step (1) and Ch (1)_Step (2).Analysis of the representative dynamic reaction pathways revealed the thermal selectivity of molecular processes and products, and the formation of Gly is not high because of the non-synergistic effect of the photochemical and thermochemical pathways.
Because the main drawbacks of photochemical reactions are multiple pathways with unwanted side products, to generate desirable products with high yields, photochemical and thermochemical controlled conditions were suggested as an example.For Gly formation, to alleviate the non-synergistic problem, CO 2 could be included in the reaction chamber to carry away excess heat and reduce the temperature to below T s .Alternatively, instead of photoexcitation directly at precursors, a photothermal catalyst could be applied to generate tuneable thermal energy within the reaction chamber; photothermal nanomaterials could transform the absorbed photon energy into highly concentrated thermal energy.
The theoretical results reported in this study confirmed that although quantum chemical and TST methods could give valuable information on the kinetics and thermodynamics of the photochemical and thermochemical pathways, inclusion of the effect of thermal energy through NVE-MDSH simulations is essential if a complete mechanistic study of photochemical processes is to be established.

Figure 6 .
Figure 6.Examples of the results on the representative reaction pathways after the S 1 →S 0 relaxation obtained from NVE-MDSH simulations.(a) MDSH_(8) obtained using [HOCOH… CH 2 NH] as the precursor.(b) MDSH_(14) obtained using [H 2 O…NHCH 2 …CO] as the precursor.E Pot/S0 , E Pot/S1 and E Pot/Cur = potential energies in the S 0 , S 1 and current states.