© The Authors 2025

1 Introduction

The emergence of chemical complexity in the interstellar medium (ISM) is one of the most fascinating topics in astrochemistry. Although the concept of “complexity” is generally ill-defined, a molecule composed of six or more atoms, with at least one that is carbon, is typically considered a complex organic interstellar molecule (Herbst & Van Dishoeck 2009). Given the harsh conditions in the ISM, the formation mechanisms of these molecules are vastly different from those on Earth. In the ISM, particularly within the ice-covered dust grains that account for approximately 1% of its mass (Bohlin et al. 1978), many molecules are formed through radical chemistry.

Significant effort has been devoted to studying the synthesis of complex organic molecules (COMs) on interstellar ices (Watanabe & Kouchi 2002; Fuchs et al. 2009; Rimola et al. 2014; Garrod 2013; Simons et al. 2020; Jin & Garrod 2020; Garrod et al. 2022; Enrique-Romero et al. 2022; Ferrero et al. 2023; Molpeceres et al. 2024a, to give a few representative examples). Although the processing reactions of COMs, those that convert or destroy already-formed COMs, have also garnered attention, they remain comparatively underexplored.

Examples in the literature include the systematic study of COM hydrogenation (Alvarez-Barcia et al. 2018), the hydrogenation of heterocycles (Miksch et al. 2021), and the processing of formic acid and thioformic acid (Molpeceres et al. 2022a). However, the chemical processing of COMs on interstellar grains has important implications for the evolution of chemical complexity. Among these, hydrogen abstraction (H-abstraction) reactions play a pivotal role. These reactions involve a radical (e.g., H, OH, NH₂) abstracting a hydrogen atom from a COM, producing a closed-shell hydrogenated molecule (e.g., H₂, H₂O, NH₃) and leaving behind a parent radical of the COM. For instance, we recently demonstrated that for hydroxylamine (NH₂OH), H-abstraction reactions, followed by O or N addition, can account for the apparent disappearance of this molecule in the ISM (Molpeceres et al. 2023). Besides, following similar arguments, deuterium enrichment of molecules and COMs (see, e.g., the review of Ceccarelli et al. 2014), is enhanced through H-abstraction reactions, following a cycle of H abstraction and D addition (Molpeceres et al. 2022a; Nguyen et al. 2021a; Lamberts & Kästner 2017b). Furthermore, the same cycle can fuel the process of reactive desorption to return the ice-synthesized molecules to the gas phase. This is due to the increase in reactive events for a fixed probability of chemical desorption per reaction (Garrod et al. 2007), as was evinced by chemical desorption experiments by some of us (Oba et al. 2018; Nguyen et al. 2020) on H₂S or PH₃. Clearly, H-abstraction reactions are critical components of modern astrochemical reaction networks and warrant more detailed investigation.

With advancements in astrochemistry, the pool of molecules requiring detailed study is growing not only in number but also in complexity. For instance, considering the examples of H₂S and PH₃ mentioned above, these molecules have relatively simple reaction networks (Lamberts & Kästner 2017b; Molpeceres & Kästner 2021; Nguyen et al. 2021a). Starting from their atomic precursors (S and P), the fully hydrogenated molecule is the most stable product. Additionally, H-abstraction reactions exhibit increasing activation energies, while H addition remains barrierless. These traits make such molecules easier to detect in experimental setups, as the reaction products are limited to one stable product and undetectable reactive radicals. However, the complexity increases significantly when a carbon atom is introduced, as in carbon monosulfide (CS). Adding carbon leads to reaction networks with approximately 15 reactions en route to forming CH₃SH (Lamberts 2018; Nguyen et al. 2023), including irreversible pathways such as CH₃ + H₂S. This highlights the intricate nature of hydrogenation sequences. Studying these networks for COMs routinely identified in cold environments (Bacmann et al. 2012; Bacmann & Faure 2014; Jiménez-Serra et al. 2021; Jiménez-Serra et al. 2016; Megías et al. 2022; Scibelli et al. 2021) becomes increasingly challenging, whether approached experimentally or theoretically.

The molecule examined in this paper, acetaldehyde (CH₃CHO), is one of the most abundant and extensively studied COMs that can form on ice grains (Jin & Garrod 2020; Lamberts et al. 2019; Ferrero et al. 2023; Molpeceres et al. 2024a), albeit with some challenges (Enrique-Romero et al. 2021; Perrero et al. 2023). Acetaldehyde can also be synthesized in the gas phase (Vazart et al. 2020). Although CH₃CHO qualifies as a COM, its reaction with hydrogen remains experimentally tractable (Bisschop et al. 2007), yet interpreting the experimental results is far from straightforward. Hydrogenation of CH₃CHO reportedly produces a mixture of acetaldehyde (CH₃CHO), ethanol (C₂ H₅ OH), methane (CH₄), carbon monoxide (CO), formaldehyde (H₂CO), and methanol (CH₃OH). However, the latter three products are linked through a sequential hydrogenation pathway (Watanabe & Kouchi 2002). Meanwhile, CH₄ and C₂H₅OH could, in principle, arise from alternative pathways rather than the direct hydrogenation of CH₃CHO. This complexity underscores the need to better understand the precise mechanism of CH₃CHO hydrogenation¹. Such a detailed understanding is essential for proposing potential deuteration pathways and assessing the likelihood of reactive desorption, as was discussed earlier.

A mechanistic understanding of this reaction can only be achieved through a synergistic combination of accurate theoretical calculations and experimental work. This is the approach we have adopted in this study paper, in which we integrate quantum chemical calculations of the hydrogenation of CH₃CHO, incorporating a rigorous treatment of nuclear quantum effects, with tailored experiments on the CH₃CHO + H reaction. Furthermore, our work aims to unify the existing knowledge on this reaction and related reactions, such as CO + H and H₂CCO + H (Watanabe & Kouchi 2002; Fuchs et al. 2009; Mondal et al. 2021; Fedoseev et al. 2022; Ibrahim et al. 2022; Ferrero et al. 2023), by resolving discrepancies where they exist or providing justified support where they align. The structure of this manuscript is as follows. In Sect. 2, we provide a concise overview of our methodology. Section 3 is dedicated to the presentation of our findings. In Sect. 4, we discuss the results, placing them in the context of experimental, modeling, and broader astrochemical perspectives. Finally, we conclude with key remarks in Sect. 5.

2 Methods

2.1 Computational details

Quantum chemical calculations were performed with a twofold objective. First, we used the results of a preliminary exploration to guide our experimental search. Second, when the full experimental procedure was completed, we expanded our quantum chemical calculations to explain the experiments, leading to a self-consistent explanation of CH₃CHO chemistry under ISM conditions. We can categorize our simulations into three different categories. First, we carried out calibration calculations using a small two-water model to test the influence of explicit water molecules in the hydrogenation of CH₃CHO (Appendix A). Second, we derived the instanton rate constants (Rommel et al. 2011; Kästner 2014) for H addition, H abstraction, and D addition on CH₃CHO. Instanton theory is a unification of path integral theory and transition state theory to simulate the quantum nature of the nuclei in the reactant and transition states of a reaction. An instanton represents the most probable tunneling path, which in our formulation is discretized in a number of points or beads, and optimized to a first-order saddle point, like in conventional transition state theory. In this case, an optimization of the tunneling path with a maximum transition probability (or minimum Euclidean action) was carried out. Third and last, we investigated additional reactions with hydrogen in the acetyl radical (CH₃CO), the preferred product of the first reaction with H (see Sect. 3). Our quantum chemical calculations are designed to prioritize accuracy in the electronic structure calculations over cluster representativity. Acetaldehyde is shown to be a COM with a relatively weak binding with the water surface (Molpeceres et al. 2022b; Ferrero et al. 2022), which normally leads to a small influence of the water matrix on the reaction rate constants (see, for example Ferrero et al. 2023; Molpeceres et al. 2024b, for two recent examples). The instanton rate constants were calculated using a sequential cooling scheme starting from a temperature close to the crossover temperature defined as: $$T_{\text{c}}=\frac{ħ\mathrm{\Omega }}{2\pi k_{\text{B}}},$$(1)

where Ω corresponds to the imaginary frequency of vibration at the transition state. T_c is an estimator of the temperature at which the tunneling effects in the rate constant start to dominate over the purely thermal ones (Gillan 1987). Because the determining quantity to compare with experiments is the reaction rate constant, which we obtained from accurate semiclassical instanton theory calculations, the inclusion of surface effects in our calculations was done via the implicit surface approach (Meisner et al. 2017); in other words, by making the rotational partition faction equal to one in the calculation of the reaction rate constant. Omitting the ice matrix allows us to use a significantly better level of theory than when including an ice matrix. In particular, the level of theory used in this work is the combination of the rev-DSD-PBEP86(D4)/jun-cc-pV(T+d)Z level (i.e., double hybrid functional with a large basis set) (Kozuch & Martin 2011; Santra et al. 2019; Dunning 1989; Papajak et al. 2011; Caldeweyher et al. 2019) for the molecular geometries and molecular Hessian and the (U)CCSD(T)/aug-cc-pVTZ (Bartlett & Purvis 1978; Purvis & Bartlett 1982; Dunning 1989; Knowles et al. 1993; Woon & Dunning 1994) gold standard method for correcting the electronic energies². This level was used for the determination of stationary points in the respective potential energy surface (PES), calculation of the zero-point vibrational energies (ZPVEs), and the calculation of rate constants in the interaction of H with CH₃CHO. To incorporate the effect of a polar environment – that is, H₂O – we also computed single point energies using the conductor-like polarizable continuum (CPCM) (Truong & Stefanovich 1995; Barone & Cossi 1998; Garcia-Ratés & Neese 2020) model, employing a dielectric constant extrapolated for amorphous solid water (ASW) of є = 600, as has been used in recent astrochemical literature (Nguyen et al. 2019). We present models with and without CPCM correction. The hydrogenation of the CH₃CO was investigated with rev-DSD-PBEP86(D4)/jun-cc-pV(T+d)Z; that is, without further coupled cluster refinement, although CPCM single points at the density functional theory (DFT) level are also reported. The last set of calculations was carried out using a broken-symmetry DFT formalism, first converging the triplet electronic state and then rotating the orbitals of the incoming reactant (H) to guarantee the correct spin state for the system. The rate constants for the reactions starting with the CH₃CO as a reactant were obtained from classical transition state theory, including the tunneling correction from an asymmetric Eckart barrier (Eckart 1930). The energies coming from pure broken symmetry DFT calculations were subjected to a higher error than when correcting the electronic energies with a high-level method. However, the lack of a reliable single reference solution in the case of biradical species such as the ones involved in the CH₃CO + H reaction prompted us to continue with this protocol. The agreement between experiments and theory found in this work serves as a proxy benchmark for the validity of our method in this particular case.

All our electronic structure calculations used the ORCA 5.0.4 code (Neese et al. 2020). Geometry optimizations and instanton calculations used a developer version of the DL-FIND code (Kästner et al. 2009) interfaced with CHEMSHELL (Metz et al. 2014)³.

2.2 Experiments

All experiments were performed using an experimental apparatus named Apparatus for Surface Reaction in Astrophysics (ASURA). The details of the ASURA system have been described in previous studies (Watanabe et al. 2006; Nagaoka et al. 2007; Nguyen et al. 2020, 2021b, 2023). In brief, it consists of an ultrahigh vacuum chamber with the basic pressure of 10⁻⁸ Pa; an aluminum (Al) substrate attached to a He cryostat, an atomic source, a quadrupole mass spectrometer (QMS), and a Fourier transform infrared spectrometer (FTIR). The surface temperature was regulated between 5 and 300 K.

The chemical reactions of solid acetaldehyde (CH₃CHO) with H (or D) atoms were studied on both an Al substrate and compact amorphous solid water (c-ASW) at low temperatures (typically 10 K). The c-ASW was made by the vapor deposition of water onto the substrate maintained at 110 K, with an estimated thickness of ~20 monolayers (ML; 1 ML = 1 × 10¹⁵ molecules cm⁻²). The substrate was cooled down to 10 K after the c-ASW layer production. Gaseous CH₃CHO was pre-deposited onto c-ASW with the deposition rate of 1 ML minute⁻¹. The layer thickness of CH₃CHO was adjusted to 1 ML, which was estimated using the peak area of the CO stretching band at 1728 cm⁻¹ and its absorption coefficient of 8.0 × 10⁻¹⁸ cm molecule⁻¹ (Bisschop et al. 2007). The pre-deposited CH₃CHO layer was exposed to the H and D atoms produced via the dissociation of H₂ (or D₂) in a microwave-discharged plasma in the atomic source chamber. The formed H (D) atoms were cooled to 100 K by multiple collisions with the inner wall of the Al pipe at 100 K (Nagaoka et al. 2007). The flux of H and D atoms was estimated as 5.0 × 10¹⁴ and 3.4 × 10¹⁴ cm⁻² s⁻¹, respectively, following the method of Oba et al. (2014). During the process of H (or D) exposure on the sample solid, interactions between CH₃CHO and H (D) atoms were observed in situ using FTIR with a resolution of 2 cm⁻¹. The reactants and products desorbed from the surface were monitored by the QMS via the temperature programmed desorption (TPD) method, with a ramping rate of 4 K minute⁻¹ .

Fig. 1 Acetaldehyde molecule with the positions where the hydrogenation and deuteration reactions were sampled and Newman projection of the molecular model showing the considered CH3CHO syn conformer.

3 Results

3.1 Theoretical exploration

3.1.1 Description of CH₃CHO + H

The study of the H-abstraction and addition reactions necessary to rationalize our experiments begins with the evaluation of the energetics for all the possible reactions reaction channels for CH₃CHO + H. The molecular structure of syn-CH₃CHO and the atomic positions where the reaction can take place are depicted in Fig 1. For each reaction, we assume that energy dissipation occurs via the surface, treating the reactions presented in this study as elementary. The possible reactions are: $${\text{CH}}_3\text{CHO}+\text{H}\to {\text{CH}}_3\text{CO}+{\text{H}}_2$$(R1) $${\text{CH}}_3\text{CHO}+\text{H}\to {\text{CH}}_3{\text{CH}}_2\text{O}$$(R2) $${\text{CH}}_3\text{CHO}+\text{H}\to {\text{CH}}_2\text{CHO}+{\text{H}}_2$$(R3) $${\text{CH}}_3\text{CHO}+\text{H}\to {\text{CH}}_3\text{CHOH}$$(R4) $${\text{CH}}_3\text{CHO}+\text{H}\to {\text{CH}}_4+\text{HCO}.$$(R5)

The energetic parameters for the reactions, reaction energies (∆H^R), activation energies (∆H^‡), and imaginary frequencies of vibration at the TS (Ω) are gathered in Table 1. In the first place, we confirm that all possible reactions are exothermic reactions that proceed with a moderate ∆H^‡, as is expected for closed- shell organic molecules. It is evident from the data in Table 1 that reaction (R5) is, under no circumstances, competitive with Reactions (R1)–(R4), due to the extremely high activation energies. This conclusion allows us to determine that all CH₄ and its isotopologs are formed in the experiment from radical-radical recombinations of secondary products of the reaction (see Sect. 4.1). The other reactions present barriers between ~20 and 50 kJ mol⁻¹ with varying values of Ω. For example, the reactions with the two lowest ∆H^‡ also show the lowest Ω. To be completely certain of the dominant reaction product, a kinetic analysis is required. Despite such an analysis being carried out in Sect. 3.1.2, a qualitative assessment of the influence of quantum tunneling can be made from the intrinsic reaction coordinate (IRC) profiles. Such IRC profiles can be visualized in Fig. 2. From them, we observe that both Reactions (R1) and (R3) have wider barriers than Reactions (R2) and (R4). Besides, both H- addition Reactions, (R2) and (R4), are also more exothermic, which also has an impact on the tunneling probability. Overall, based on the energetics and T_c (Eq. (1)) it is not possible to determine which is the dominant product of the reaction. Kinetic simulations such as the ones in Sect. 3.1.2 are necessary to discriminate them.

Fig. 2 IRC profiles for Reactions (R1)–(R4) (We omit reaction (R5) due to the high ∆H‡). IRC profiles are presented at the rev-DSD- PBEP86(D4)/jun-cc-pV(T+d)Z not corrected by ZPVE, double-level CCSD(T), or implicit water ice environment (CPCM).

Fig. 3 Arrhenius plot of the hydrogenation rate constants (kH) for Reactions (R1)–(R4) including (dashed lines) and not including (solid lines) an implicit solvation model. All rate constants have been corrected for the implicit surface. Rate constants are presented until the lowest temperature achievable for each instanton in the sequential cooling scheme.

3.1.2 Kinetic analysis and reactions with deuterium

We began our instanton calculations using the sequential cooling scheme at a temperature of approximately 0.7T_c, starting from a closed instanton path consisting of 36 beads. As is described in Sect. 2, we progressively converged the instanton paths at lower temperatures. To ensure convergence with respect to the number of beads at low temperatures, we increased the number of beads to 70 for temperatures below approximately 140 K in all reactions. The hydrogenation rate constants, k_H, are presented in Fig. 3. We omitted the calculation of rate constants for Reaction (R5) because instanton calculations are computationally demanding, and this reaction has too high an activation energy to be competitive. The rate constants were computed down to different lower temperature limits for each reaction. This variation arises because the convergence of instanton paths depends on the reaction, with the abstraction on the CH₃ moiety being particularly challenging to converge. However, in the cases we studied, the rate constants are converged or close to the asymptotic tunneling limit, so k_H(T_last) ≈ k_H(10). Since the reaction outcome is dominated by Reaction (R1), the lack of rate constants at temperatures below 50 K for Reactions (R3) or (R4) does not affect our conclusions. It is worth noting that the effect of the implicit solvation scheme varies among different reactions, with Reactions (R1) and (R4) showing the most significant changes. Specifically, at low temperatures, these reactions are slower when considering an implicit water matrix due to the increase in $\text{Δ}{H}_{\text{CPCM}}^{‡}$ found for the reactions. Nevertheless, this increase does not alter the overall kinetics of the system regarding competitive reaction pathway’s.

A quick inspection of Fig. 3 reveals that Reaction (R1) is the fastest across all temperatures, consistently dominating all other reaction pathways. Furthermore, aside from Reaction (R1), all other reactions exhibit rate constants that are competitive with hydrogen atom diffusion (Asgeirsson et al. 2017; Senevirathne et al. 2017), with diffusion rates (k_Diff) in typical potential sites ranging from 10³ to 10⁵ s⁻¹ at 10 K (see, for example, Fig. 7, bottom panel of Senevirathne et al. 2017). This competition suggests that even if a pre-reactive complex favoring reactions other than Reaction (R1) is formed – perhaps due to a favorable orientation of the radical attack – diffusion away from the reaction site in a reaction-diffusion competition (Chang et al. 2007) will either completely or partially suppress these alternative reaction pathways. Consequently, we conclude that Reaction (R1) is the sole outcome of the CH₃CHO + H reaction. This conclusion allows us to rationalize the results obtained in our experiments (see Sect. 3.1.3).

We also examined deuteration rate constants, k_D. For these calculations, we focused on the H1 and C2 positions. For example: $${\text{CH}}_3\text{CHO}+\text{D}\to {\text{CH}}_3\text{CO+HD}$$(R6) $${\text{CH}}_3\text{CHO}+\text{D}\to {\text{CH}}_3\text{CHDO,}$$(R7)

because they have the lowest ∆H^‡ and the highest k_H. The rate constants are shown in Fig. 4. In both cases, we observe a reduction of k_D over k_H due to the reduced efficiency of quantum tunneling. The kinetic isotope effects (KIEs; calculated as k_H/k_D) at 50 K⁴ that we observe according to our calculations are KIE_R1/R6 = 34. 7 and KIE_R2/R7 = 734.1, calculated including the implicit solvent approach. The KIE is larger in the case of Reaction (R7), because this reaction is a D addition and quantum tunneling is reduced, compared to Reaction (R6), which is an H abstraction and which has a higher tunneling probability.

Fig. 4 Similar to Fig. 3 but including deuterium as an incoming particle (kD) for Reactions (R6) and (R7)

Fig. 5 Potential energy scan for Reactions (R8) (blue curve), (R9) (red curve), and (R10) (green curve) at the revDSD-PBEP86(D4)/jun-cc- pV(T+d)z level using a broken symmetry formalism. The discontinuities in r(H-H3) and r(H-O4) are caused by the insufficient resolution of the PES scan. The small bump in r(H-O4) is attributed to the constraints in the C-H bonds required to carry out the scan (see text).

3.1.3 Hydrogenation of the CH₃CO radical

The Reaction CH₃CHO + H leads almost exclusively to CH₃CO. Therefore, the CH₃CHO reaction network must be completed by relying on the chemistry of such a radical. In this section, we study the hydrogenation of CH₃CO following a scheme similar to the one shown in Fig. 1 (only this time without the H1 position). The study of the CH₃O + H reaction has the peculiarity that it should be studied in the open-shell singlet channel; that is, using a biradical wavefunction. Radical-radical reactions generally show lower barriers than the radical-closed shell reactions, so we first checked which of the following reactions are barrierless using PES scans: $${\text{CH}}_3\text{CO+H}\to {\text{CH}}_3\text{CHO}$$(R8) $${\text{CH}}_3\text{CO+H}\to {\text{H}}_2{\text{CCO+H}}_2$$(R9) $${\text{CH}}_3\text{CO+H}\to {\text{CH}}_3\text{COH}$$(R10) $${\text{CH}}_3\text{CO+H}\to {\text{CH}}_4\text{+CO}.$$(R11)

The scans for the barrierless reactions are shown in Fig. 5. Of all the above reactions, only Reaction (R11) was not found to be barrierless. All of the other reactions show a barrierless profile. We highlight the difficulty of obtaining clean scans for these, except for Reaction (R8), because in the latter case there are no competing channels in the direction of the scan. The other scans have discontinuities that we attribute to a combination of a low resolution and hindered conformational changes from competitive barrierless pathways during the coordinate scan. For example, the scan for Reaction (R10) shows a small bump at about 1.4 Å. This is due to the fact that in this scan we had to constrain the C-H (CH₃ moiety) bond distance to prevent the optimizer from visiting the barrierless Reaction (R9) at long distances. This results in a small stiffness of the structure leading to this bump. Our calculations still predict the barrierless reaction. We recall that the calculations involving these reactions were carried out using a broken-symmetry formalism, at the revDSD-PBEP86(D4)/jun-cc-pV(T+d)z level, without correction with coupled-cluster methods. This, combined with the approximate nature of the CPCM implicit solvent scheme, makes the second set of calculations less accurate than the ones shown in Sect. 3.1.1. The energetic values for the reactions are collected in Table 2. From the table, we can see that the ∆H^‡ forReaction (R11) (counterpart of (R5)) is significantly lower in this second set than in the previous reactions for the hydrogenation of acetaldehyde. We also observe the effect of the implicit solvation correction on some energetic parameters of the reaction. The abundant presence of CH₄ and its isotopologs in our experiments supports the Reaction (R11), even with a barrier (see next paragraph and Sects. 3.2.1 and (4.1)). Although our results for this reaction are qualitative, they are in agreement with the experiments. However, for a more detailed study of (R11), we suggest using multireference methods. All other reactions are barrierless. Finally, it is also worth noting that our results in this section are not affected by deuteration, since tunneling does not determine the viability of the reaction (for Reactions (R8)–(R10)) or contribute significantly (Reaction (R11), which requires the fragmentation of a C-C bond).

The possibility of finding products beyond the ones obtained in barrierless channels can be easily rationalized if we consider the typical reaction-diffusion competition scheme (Chang et al. 2007). Assuming that the desorption of both reactants and the diffusion of CH₃CHO are negligible, the rate constant for the reaction under this formalism is (Chang et al. 2007): $$k=\frac{k_{\text{R}}}{k_{\text{R}}+k_{\text{Diff}}},$$(2)

where “R” and “Diff” represent reaction and diffusion, respectively. In the anisotropic potential created by an amorphous polar environment such as ASW, the orientation from which H atoms approach is random. Assuming that all directions are equally possible for Reactions (R8)–(R11), Eq. (2) shows that if each reaction starts from different positions, as is the case, then moving out of the reaction center requires at least one diffusion step. If k_R is much larger than k_Diff, then the reaction can start from that orientation, even if a barrierless channel is available. This is essentially different from gas phase reactivity. A scheme for this explanation is shown in Figure 6. The arguments based on reaction-diffusion competition are therefore able to rationalize the experimental results in the light of the theoretical calculations, as we show in Sect. 4.1. However, not all reaction orientations are equally weighted, resulting in different branching ratios for reactions. Determining the exact branching ratios for these barrierless reactions from theoretical calculations is challenging. The same arguments presented in this section apply to the reactions studied in Sect. 3.1.1 (Reactions (R1)- (R5)), but in this case reactions other than Reaction (R1) are competitive with diffusion, k_R ~ k_Diff, and therefore Reaction (R1) is dominant.

Our theoretical results for the hydrogenation of CH₃CO are in reasonable agreement with the ones presented in Ibrahim et al. (2022), although several differences appear. In their study, Ibrahim et al. (2022) indicate that the Reaction (R11) is barrierless, while the Reaction (R9) has an activation energy of ~18 kJ mol⁻¹. This tendency contradicts our calculations and experiments. However, this discrepancy does not change the picture of both works, except for the absence of H₂CCO by Reaction (R9), which we found to bevery fast, and which is supported by experiments, since all other reactions remain very fast in both our works. We believe that the reason for the discrepancy could be a non-biradical wavefunction in the case of Ibrahim et al. (2022). Starting our calculations from an ionic wavefunction, we can reproduce the results of Ibrahim et al. (2022) for the Reaction (R9) with even higher activation energies (~30 kJ mol⁻¹).

Fig. 6 Schematic view of the CH3CO + H reaction. Depending on the initial configuration of the H atoms, the different kR must compete with diffusion, kDiff. If kR ≫ kDiff, then the reaction would dominate even in the presence of a barrierless channel, as is shown in the figure. BL means barrierless.

Fig. 7 Fourier-transform infrared profile of CH3CHO on c-ASW at 10 K.

3.2 Experiments

3.2.1 Hydrogenation experiments

Figure 7 shows an FTIR spectrum of the solid CH₃CHO on c- ASW at 10 K. The most intense absorption peak at 1728 cm⁻¹ was attributed to the C-O stretching band (ν_sC=O) in CH₃CHO. Additionally, other infrared bands were observed at 1432, 1349, and 1124 cm⁻¹ can be assigned to the CH₂ scissors, CH₃ symmetric deformation, and C-C stretch of the solid CH₃CHO, respectively (Bennett et al. 2005; Bisschop et al. 2007).

Figure 8a shows the difference spectrum of solid CH₃CHO after exposure to H atoms for up to 2 hours compared to the initial spectrum of unexposed CH₃CHO. The intensity of the C-O stretching band at 1728 cm⁻¹ decreases with atom exposure time, indicating the loss of CH₃CHO. Figure 8b shows variations in the relative abundance of CH₃CHO after exposure to H atoms at 10 K with relevance to atom exposure times, with a blank experiment for H₂ exposure. The consumption of CH₃CHO was estimated to be only 10% relative to the initial amount of CH₃CHO after exposure to H atoms for up to 2 hours, while no decrease was observed after exposure to H₂ molecules. We expect the formed CH₃CO radicals from Reaction (R1) to further react with H atoms on the surface according to the Reactions ((R8) – (R11)). Unfortunately, we could not observe any distinct infrared features for products from the pre-deposition of CH₃CHO and H atoms on c-ASW at 10 K, probably because of the small amount of products. The loss of CH₃CHO may be partially due to reactive desorption. We cannot quantitatively evaluate this in any of our experiments, even considering the low desorption energy of CH₃CHO of ~3500 K (Molpeceres et al. 2022b; Ferrero et al. 2022), something we discuss in more detail in Sect. 4.2.

The TPD experiments demonstrate the formation of CH₄ and CO based on the desorption peak observed at m/z = 16 and m/z = 28 in the temperature range of 30–60 K and 25–35 K, respectively (Fig. 9, top and middle panels). These results are in agreement with the computational predictions for the Reaction ((R11); see Sect. 3.1.3) and Ibrahim et al. (2022) experiments. In addition, we observe a desorption peak at m/z = 46 in the desorption temperature range 150–165 K (Fig. 9, bottom panel), suggesting the formation of ethanol (C₂H₅OH). This ethanol should be formed by the H addition reaction of CH₃COH, which was obtained from the Reaction (R10), followed by: $${\text{CH}}_3\text{COH+H}\to {\text{CH}}_3\text{CHOH}$$(R12) $${\text{CH}}_3\text{CHOH+H}\to {\text{C}}_2{\text{H}}_5\text{OH}.$$(R13)

No distinct IR features of the products of the reaction between CH₃CHO and H atoms are observed in the predeposition experiments, probably due to the small amount of solid CH₃CHO (1 ML). In order to identify the products of the reaction between CH₃CHO and H atoms in more detail by FTIR, we performed an additional experiment in which 30 ML of CH₃CHO was co-deposited with H atoms on the Al substrate at 10 K, a technique suitable for identifying reaction products. Figure 10 shows an FTIR spectrum obtained after co-deposition of CH₃CHO with H atoms on Al substrate at 10 K. In contrast to the behavior of the pre-deposited CH₃CHO with H atoms (Fig. 8a), various peaks appear after the co-deposition of CH₃CHO with H atoms. The appearance of a sharp peak at 2133 cm⁻¹ was assigned to CO (Gerakines et al. 1995). After its formation on the surface, CO should be further hydrogenated to give formaldehyde (H₂CO) and methanol (CH₃OH), with peaks at 1724 and 1500 cm⁻¹ for H₂CO (Hidaka et al. 2004) and 1044 cm⁻¹ for CH₃OH (Nagaoka et al. 2007), as is observed in the spectrum (Fig. 10). A strong peak observed at 1305 cm⁻¹ is attributed to CH₄, which is in good agreement with the CH stretching band of CH₄ (Qasim et al. 2020). In addition, two small peaks observed at 1081 and 1050 cm⁻¹ are attributed to the CCO asymmetric bending and C–O stretching bands of CH₃CH₂OH, respectively (Hudson 2017).

Finally, the presence of ketene (H₂CCO) is also observed in our co-deposition experiments. We could not detect it in the IR because H₂CCO is easily hydrogenated (Ferrero et al. 2023), and its characteristic vibrational frequency (at 2133 cm⁻¹) overlaps with the absorption of CO. However, clear evidence for the presence of H₂CCO in the reaction mixture is observed in our TPD experiments. The desorption peaks corresponding to m/z = 42 and 14 at about 75–90 K and 150–165 K (Fig. 11) can be derived from H₂CCO itself and its fragment (i.e., CH₂), respectively. The detection of these m/z = 14 (CH₂) and 42 (H₂CCO) features in the TPD experiments, consistent with previous studies (Maity et al. 2015; Chuang et al. 2020; Fedoseev et al. 2022), provides strong evidence that H₂CCO was formed under the current experimental conditions, as is supported by our theoretical models.

Fig. 8 (a) Variation in the difference spectra of the solid CH3CHO after exposure to H atoms for 3, 60, and 120 min at 10 K. (b) Relative abundance of CH3CHO following a 2-hour exposure to H atoms (blue circles) compared to that with H2 molecules (red squares) on c-ASW at 10 K.

Fig. 9 TPD-QMS profile of pre-deposited CH3CHO with H atoms (black line) comparable with H2 molecules (blue line) on c-ASW at 10 K: m/z = 16 (top panel), m/z = 28 (middle panel), and m/z = 46 (bottom panel).

Fig. 10 FTIR spectrum of codeposition of CH3CHO and H atoms on Al at 10 K.

Fig. 11 TPD-QMS profile for the m/z = 14 and 42. Black arrows correspond to features attributable to H2CCO, obtained after the reaction of the pre-deposited CH3CHO with H atoms on c-ASW at 10K. The desorption peaks at 100–130 K and 140–150 K are derived from the remaining CH3CHO on c-ASW after exposure to H atoms. The peaks at 75–90 K and 150–165 K represent the desorption of H2CCO (Maity et al. 2015; Chuang et al. 2020; Fedoseev et al. 2022).

3.2.2 Isotopic labeling

Figure 12 shows the evolution of the difference spectra of CH₃ CHO after exposure to D atoms for up to 2 hours as well as that of CH₃ CHO without atomic exposure. Similar to the behavior after exposure to H atoms, the C-O stretching band of CH₃CHO at 1728 cm⁻¹ decreases with exposure times, indicating the loss of CH₃CHO after interaction with D atoms on the Al substrate. Unfortunately, we cannot identify clear FTIR features for the products of the pre-deposited CH₃CHO with D atoms.

On the other hand, we observe various features for the formation of isotopologs of the products based on the TPD experiments. Figure 13 shows the TPD spectrum after the reaction of CH₃CHO with D atoms on A1 substrate at 10 K. Since CH₃CHO desorbs at the temperature between 90 and 145 K and shows a CHO fragment (m/z = 29) upon ionization, the desorption peak for m/z = 30 (the CDO fragment) is attributed to the deuterated counterpart of acetaldehyde, CH₃CDO. Thus, the desorption peak (m/z = 30 – top panel) observed at the same temperature after exposure to D atoms identifies the formation of CH₃CDO. The formation of CH₃CDO provides compelling evidence for the reaction between CH₃CO and D atoms: CH₃CO +D → CH₃CDO. This process, in which the CH₃CO radical is produced by the Reaction (R6), is consistent with the calculated results.

Furthermore, the desorption peaks at m/z = 17 and 18 (Fig. 13, middle and bottom panels) within the temperature range of 35–65 K are indicative of the formation of deuterated methane isotopologs, CH₃D and CH₂D₂. These species could be produced by the reactions in (R16), as discussed in Sect. 4.1, although we cannot be certain due to the high barriers associated with H abstraction in CH₃D (Lamberts et al. 2022).

A tiny peak corresponding to deuterated methanol (m/z = 36; CD₃OD) is observed to desorb at about 135 K (Fig. 14). The formation of CD₃ OD has been proposed to result from the deuteration of CO (Watanabe & Kouchi 2002), implying that the CO molecule is likely produced by the reaction of CH₃ CO (Reaction (R11)) with H or D atoms under the current experimental conditions, a finding consistent with our calculated results. In addition, we also observe the desorption peaks at m/z = 49 and 50 in the temperature range of about 130–200 K. These peaks can be attributed to the formation of ethanol isotopologs corresponding to CH₃ CD₂OD and/or CH₂DCD₂OD (see Figure 15), which were obtained by reactions of CH₃CO radicals with D atoms, followed by: $${\text{CH}}_3\text{CO}\stackrel{\text{+D}}{\to }{\text{H}}_2\text{CCO}\stackrel{\text{+D}}{\to }{\text{CH}}_2\text{COD}\stackrel{\text{+3D}}{\to }{\text{CH}}_2{\text{DCD}}_2\text{OD,}$$(R14)

and $${\text{CH}}_3\text{CO}\stackrel{\text{+D}}{\to }{\text{CH}}_3\text{COD}\stackrel{\text{+D}}{\to }{\text{CH}}_3\text{CDOD}\stackrel{\text{+D}}{\to }{\text{CH}}_3{\text{CD}}_2\text{OD}.$$(R15)

The formation of ethanol isotopologs is consistent with the computational results obtained from Reactions (R9) and (R10) (see Sect. 3.1.3).

The appearance of a desorption peak observed at m/z = 48 (see Fig. 16) likely indicates the formation of CH₃CHDOD, resulting from the addition reaction of D to CH₃CHO under the current experimental conditions. Nevertheless, based on the quantitative analysis of desorption areas for the reaction products, CH₃CHDOD is found to be a minor species in the reaction between CH₃CHO and D atoms. The predominant reaction involves CH₃CO radicals with D atoms, leading a higher yield of different products.

Fig. 12 Variation in the difference spectra of the solid CH3 CHO on the Al substrate at 10 K after exposure to D atoms for 10, 60, and 120 min comparable with the initial CH3CHO.

Fig. 13 TPD-QMS profile of pre-deposition of CH3 CHO with D atoms (black line) comparable with D2 (blue line) on Al substrate at 10 K, m/z = 30 (top panel); m/z = 17 (middle panel), and m/z =18 (bottom panel). The m/z = 30 (top panel) in the blue line represents the fragmentation of CH3CHO.

Fig. 14 TPD-QMS profile of m/z = 36 observed for the pre-deposition of CH3 CHO and D atoms (black line) is compared with that of D2 molecules (blue line) on the A1 substrate.

Fig. 15 TPD-QMS profiles of m/z = 49 and 50 (black lines) desorbed at around 130–200 K, likely corresponding to ethanol isotopologs CH3CD2OD and/or CH2DCD2OD. These species were yielded through the reaction of CH3CHO and D atoms, with behavior comparable to that of D2 molecules (blue line) on Al substrate at 10 K.

Fig. 16 TPD-QMS profile for the desorption peak observed at m/z = 48 (black line) may be attributed to the formation of CH3CHDOD through the addition reaction of D to CH3CHO. This is in the comparison to the desorption behavior for D2 (blue line) on the Al substrate at 10 K.

4 Discussion

4.1 Merging theory and experiments: Comparison with previous works

Once the theoretical and experimental results have been discussed, we are in a privileged position to discuss the chemical mechanism behind the surface hydrogenation of acetaldehyde. The main experimental finding comes after the analysis of the 1432, 1349, and 1124 cm⁻¹ IR decay bands that, after comparison with the amount of deposited CH₃CHO, reveals a limited destruction of around of a 10%, with the rest of the CH₃CHO either unaffected or reformed after a H-abstraction and H- addition cycle. This finding is the easiest to rationalize based on the quantum chemical calculations. The limited destruction or conversion of CH₃CHO is due to the dominance of the (R1) channel in the CH₃CHO + H/D reaction. Previous literature studies on the hydrogenation of aldehydes (Mondal et al. 2021) have not considered the H-abstraction reactions, perhaps analogously to the limited importance of this type of reaction in the parent molecule formaldehyde, H₂CO (Song & Kästner 2017). However, for CH₃CHO we found otherwise, with H-abstraction dominating the reactivity. Our deuteration experiments show a tiny fraction of CH₃CHDOD, indicating the possibility also of some hydrogen or deuterium addition in the hydrogenation or deuteration of CH₃CHO. This is somewhat coherent with the rate constants for H addition found in Reactions (R2) and (R7), when CH₃CHO and H are in deep binding sites with slow diffusion rates (see Fig. 6 and associated discussion). However, H addition is a minor channel of the reaction, whose quantification to the total branching ratio is difficult. Therefore, in Appendix B we recommend considering a 1.0 branching ratio toward H abstraction (Reaction (R1)).

After Reaction (R1), any subsequent hydrogenation on the new system involves the very reactive CH₃CO radical. From a theoretical standpoint, we determined that all reaction positions are, in principle, possible. Beginning with Reaction (R8), the H (or D) addition on this position reforms CH₃CHO and, based on the experiment, is the most likely outcome, with 90% of the reactive events in this direction, determined from the total fraction of CH₃CHO remaining after concluding our experiments. However, neither our calculations nor our experiments can rule out the contribution of the reforming of CH₃CHO from secondary reactions; that is, hydrogenation of H₂CCO (see below and Ibrahim et al. 2022; Ferrero et al. 2023; Ibrahim et al. 2024). For Reactions (R9), (R10), and (R11), the products are H₂CCO (and H₂), CH₃ COH (methyl-methoxy carbene; Schreiner et al. 2011), and CH₄ + CO. Beginning with H₂CCO, as we mention in Sect. 3.2, we cannot identify it in the IR spectrum because its main transition falls at 2133 cm⁻¹, overlapping with the CO band. However, the appearance of additional desorption features for m/z = 14 (CH₂) and 42 (H₂CCO) between 75–90 and 150– 165 K in the TPD experiments (Fig. 11) is in line with previous studies (Maity et al. 2015; Chuang et al. 2020; Fedoseev et al. 2022), which confirmed the desorption of H₂CCO. Nevertheless, Ferrero et al. (2023), using instanton calculations such as the ones shown in this work, show that the H₂CCO hydrogenation is very efficient with final products CH₃CHO and perhaps C₂H₅OH (ethanol). Bisschop et al. (2007); Ibrahim et al. (2022) also report CO + CH₄ and H₂CO as products of the H₂CCO + H reaction, which is also coherent with our results. Combining experiments and theory, we determined that H₂CCO must be a product of the CH₃CHO hydrogenation under interstellar conditions, but in small amounts. Continuing with CH₃COH, even though it is formed in our calculations, we are unable to comment much on this species from an experimental point of view. This is due to the limited experimental data, but also to its transient nature. As a reactive carbene, we expect that the CH₃COH+ 2H→ C₂H₅OH reaction will dominate in our experimental setup. Indeed, C₂H₅OH is seen in our co-deposition experiments, and the experimental and theoretical synergy carried out in this work allowed us to conclude that it is a secondary product of the hydrogenation of CH₃CHO rather than a direct one. The presence of C₂ H₅ OH is puzzling when checking the previous literature. In the experiments of Fedoseev et al. (2022) (for H₂CCO + H) C₂H₅OH is seen, as in our experiments. However, Ibrahim et al. (2022) and Ibrahim et al. (2024) do not report its presence after the same reaction. The reason for this disagreement is likely due to the experimental conditions, and our results provide a reconciling scenario. We could not detect C₂H₅OH using the IR spectra in our pre-deposition experiment, which would be the one directly comparable to the experiments of Ibrahim et al. (2022) needing to rely on TPD-QMS measurements (see Fig. 9) for that. However, the presence of C₂H₅OH is seen in our co-deposition experiments starting from CH₃ CHO mimicking Fedoseev et al. (2022). Although other experimental conditions differ between our setups, such as the H flux or irradiation time, we think the reason for the disagreement stems from the deposition strategy.

Finally, the last experimental product that we observe is CH₄ (and its isotopologs, see below) and CO. Given the enormous ∆H^‡ found for Reaction (R5), the formation of CH₄ must come from the hydrogenation of CH₃CO, or perhaps of further radicals in the hydrogenation sequence of CH₃CO, out of the scope of the present paper. However, we find it likely that CH₄ comes directly from CH₃CO via Reaction (R11). CO is seen in our experiments, either directly as CO or as its hydrogenated products H₂CO and CH₃OH as discussed above.

Our deuteration experiments do show a variety of isotopologs of the products discussed in the previous paragraphs. The presence or absence of certain isotopologs not only aids our experiments but also allows us to confirm our theoretical observations. For CH₃ CHO + D, our experiments reveal the majoritarian formation of CH₃CO that later reforms CH₃CDO. Likewise, we clearly see the formation of CH₃D and CH₂D₂, which might arise from the sequence of reactions (omitting HD when applicable): $${\text{CH}}_3\text{CHO}\stackrel{\text{D}}{\to }{\text{CH}}_3\text{CO}\stackrel{\text{D}}{\to }{\text{CH}}_3\text{D+CO}\stackrel{\text{D}}{\to }{\text{CH}}_2\text{D}\stackrel{\text{D}}{\to }{\text{CH}}_2{\text{D}}_2\text{.}$$(R16)

These reactions serve as a proxy to suggest that the CH₃D + H/D→ CH₂D + H₂ / HD could be a viable reaction, although we cannot confidently ensure it, because some reaction steps in (R16) have high barriers. A preliminary search at the rev-DSD- PBEP86(D4)/jun-cc-pV(T+d)Z level for the CH₃D + D → CH₂D + HD (The third reaction in the scheme of reactions shown in (R16)) shows that the reaction is slightly exothermic, unlike the full hydrogenated counterpart, by ∆H^R = −3.3 kJ mol⁻¹ (−404 K), making the reaction scheme (R16) in principle viable albeit probably very slow. Overall, while the presence of CH₃D is easy to explain from experiments and theory, the presence of CH₂D₂ is not that easily explained, with secondary and tertiary hydrogenation branches being important, and requiring a careful validation out of the scope of this work. The CO formed in the previous reaction can also be hydrogenated to fully deuterated methanol (CD₃OD) (Watanabe & Kouchi 2002), a reaction that could not happen without the formation of CO, further increasing our confidence in the viability of Reaction (R11). We also observe clear signals of the formation of ethanol isotopologs, especially of CH₃CD₂OD or CH₂DCD2OD (Reaction scheme in (R14) and (R15) and Fig. 15), once again merging experimental and theoretical results and highlighting the importance of Reaction (R1) as the initiator of the complex chemistry observed in the experiments. We reiterate that a small fraction of CH₃CHDOD is also found (Fig. 16).

We briefly comment on the differences between our new set of experiments and the classical ones of Bisschop et al. (2007) that inspired this work. In both our experiments, we obtain the same products of the reaction, although in different proportions. While we only find a conversion of CH₃CHO equal to 10% for the duration of our experiment, Bisschop et al. (2007) report variable conversion rates, varying from between 30% to almost total conversion (see their Table 5). These quantitative differences should come from different factors, the most likely reasons being the temperature of the H atoms and the selection of the substrate. In our setup, the H atoms are cooled down before hitting the surface, guaranteeing immediate thermalization on the surface. In Bisschop et al. (2007), the H-atoms were introduced in the chamber at 300 K, which can promote a plethora of reactions through different reaction mechanisms, such as the Eley-Rideal or Harris-Kasemo ones. Moreover, in our setup, we carried out our experiments on ASW, where the interactions between CH₃CHO molecules with the surface will be different than in pure CH₃CHO or than in a metallic substrate.

Lastly, in light of what has been presented in this section, we have prepared a series of recommendations to include this reaction in astrochemical rate equation models. We have gathered our conclusions in Appendix B.

4.2 Astrophysical implications

Our results carry a series of implications for the astrochemical community, in particular regarding the deuteration and reactive desorption of CH₃CHO. Acetaldehyde is a molecule routinely detected in a variety of interstellar environments (Cazaux et al. 2003; Occhiogrosso et al. 2014; Walsh et al. 2014; Imai et al. 2016; Codella et al. 2015; Holdship et al. 2019; Bonfand et al. 2019), including several detections in prestellar cores (Bacmann et al. 2012; Cernicharo et al. 2012; Scibelli et al. 2021; Jiménez-Serra et al. 2016; Megías et al. 2022), where there is little energy for bringing the molecules to the gas phase, where they are detected. The formation of CH₃CHO is debated, including the gas-phase route C₂H₅ + O → CH₃CHO + H (Vazart et al. 2020) or ice surface routes. These routes involve the radical-radical recombination of HCO + CH₃, on ice (Garrod 2013; Molpeceres et al. 2024a; Lamberts et al. 2019), although on water ice this route is sometimes considered less effective (Enrique-Romero et al. 2021). Alternative routes include the radiolytic (Shingledecker et al. 2018) or nonthermal formation (Jin & Garrod 2020). Regardless of the formation route, CH₃CHO will form or deplete on the grain, and it is at this stage when the title reaction can take place.

Our results show that acetaldehyde is mostly processed and reformed through Reactions (R1) and (R8), and, although other products and routes exist, as is extensively discussed in this manuscript, the astrophysical implications of our work are mainly carried by this limited destruction. We shall focus on those in this section. First, we shall talk about reactive desorption. When an H atom is abstracted from the HCO moiety in CH₃CHO, the hydrogenation of the resulting CH₃CO radical could trigger a chemical desorption event, considering that the binding energy of CH₃CHO is among the lowest for COMs. As we indicated in Sect. 3.2, we cannot confirm this possibility, as the different hydrogenation branches opened from our reaction make it very difficult to confirm or deny the possibility of chemical desorption. Optimistic estimates of the probability of chemical desorption for COMs place the probability per reaction event at around 1% (Garrod et al. 2007). Because this amount should be included in the 10% of nonreformed CH₃ CHO, it is experimentally extremely hard to prove. Yet, CH₃ CHO constitutes the best-case scenario with an effective H-abstraction/H-addition cycle, the high reaction energy for the CH₃ CO + H reaction, and the relatively low binding energy on water ice (Molpeceres et al. 2022b; Ferrero et al. 2022).

Another important implication of our work pertains to the formation of ethanol, C₂H₅OH, which is predicted to be formed in very low amounts from Reaction (R2), but actually from the evolution of the CH₃ CO + H reaction. This is in contrast with the hydrogenation of formaldehyde (H₂CO), the prototypical interstellar aldehyde, which leads to methanol CH₃OH (Watanabe & Kouchi 2002; Hidaka et al. 2004). In Mondal et al. (2021), the authors indicate that the connection between CH₃CHO and C₂H₅OH through direct hydrogenation is not enough to account for the abundances of C₂H₅OH in the hot core G10.47+0.03, and that their models reveal that CH₃ + CH₂OH → C₂H₅OH is needed to explain the observations. Our results fully adhere to this picture and strongly suggest that the link between CH₃CHO and C₂H₅OH is not straightforward.

Finally, we reserve some words for the possibility of finding deuterated acetaldehyde. A relatively recent publication Jørgensen et al. (2018) shows that CH₃CDO is the isotopolog of a COM with the highest ratio to the non-deuterated molecule, with a percentage of 8% in IRAS16293–2422. This result reinforces the results obtained in this work and suggests as a possible explanation for this observation the H-abstraction or H-addition route discussed in this work, in addition to the explicit detection of CH₃CDO in our experiments. Nevertheless, it is important to remark that other CH₃CHO isotopologs, CH₂DCHO and CD₂HCHO, have also been detected in the same source (Coudert et al. 2019; Ferrer Asensio et al. 2023). These last findings do not rule out the importance of our reaction routes but show the complexity associated with deuterium fractionation in the ISM, where gas and surface chemistry are in constant interplay.

5 Conclusions

We conclude by providing a series of bullet points that summarize our work.

1. Our calculations and experiments show that the CH₃CHO + H does not lead to significant conversion. The reason for that limited conversion is found in the dominance of hydrogen abstraction in the HCO moiety of CH₃CHO, which in turn leads to facile reforming of CH₃CHO. Quantifying this conversion leads us to a maximum of 10% of CH₃CHO under our experimental conditions.

2. Other minor products of the reaction of CH₃CHO with H come from the hydrogenation of the CH₃ CO reactive radical and are H₂CO, CH₃OH, C₂H₅OH, H₂CCO, CO, or CH₄, revealing a very complex chemistry. In general, our calculations and previous knowledge of the chemistry of the different intermediates found in our calculations can satisfactorily explain the different products that we obtain. However, and because the study of the CH₃ CO + H is more qualitative due to theoretical constraints, more sophisticated approaches are required to derive accurate branching ratios for the reaction.

3. Our experiments reveal the effective formation of CH₃CDO, supporting the idea of a H/D-abstraction and H/D-addition cycle as the major route for the CH₃CHO + H reaction.

4. Although we can explain the vast majority of our experimental observations with our calculations, the presence of doubly deuterated methane CH₂ D₂ and the CH₃CHDOD isotopolog in our deuteration experiments remains difficult to explain. In general, several hydrogenation and deuteration branches of the minor products could lead to different products, making explaining the complete reaction network a very difficult task.

5. Our results align with the results found in the modeling literature (Mondal et al. 2021) suggesting that the link between acetaldehyde and ethanol is not direct from the former to the latter via hydrogenation.

6. We are unable to determine whether reactive desorption happens for CH₃CHO, although in light of the energetic parameters found by our theoretical calculations, summed to the H-abstraction and addition cycle present in the reaction, we consider it a likely outcome.

7. Our experiments and calculations nicely explain the excess of CH₃CDO found in the IRAS 16293-2422 hot core (Jørgensen et al. 2018). However, we remark that a single chemical reaction is not enough to explain complex deuterium enrichment processes in space.

Future avenues of the present work could address the presence of certain non-easy-to-explain isotopologs in our experiments, or the study of the impact of the quantities derived here in realistic astrochemical models. Likewise, we are currently tackling the complex problem of CH₃ CHO chemical desorption using more sophisticated molecular dynamics approaches. We shall continue combining experiments and calculations to fully unravel complex interstellar reaction networks.

Data availability

The Cartesian coordinates for the structures shown in this paper, as well as the numerical values for the rate constants, can be retrieved at https://zenodo.org/records/14278652. Any further data can be requested from the corresponding authors.

Acknowledgements

The authors thank the anonymous reviewer for the thorough revision of the manuscript. We thank Professor Johannes Kästner at Stuttgart University for providing a development version of CHEMSHELL. We also thank Dr. Masashi Tsuge and Dr. Hiroshi Hidaka for fruitful discussions on the experimental results. G.M acknowledges the support of the grant RYC2022-035442-I funded by MICIU/AEI/10.13039/501100011033 and ESF+. G.M. also received support from project 20245AT016 (Proyectos Intramurales CSIC). We acknowledge the computational resources provided by bwHPC and the German Research Foundation (DFG) through grant no INST 40/575-1 FUGG (JUSTUS 2 cluster), the DRAGO computer cluster managed by SGAI-CSIC, and the Galician Supercomputing Center (CESGA). The supercomputer FinisTerrae III and its permanent data storage system have been funded by the Spanish Ministry of Science and Innovation, the Galician Government and the European Regional Development Fund (ERDF). Y.O and N.W. acknowledge the funding support from JSPS KAKENHI grant nos. JP23H03980, JP21H04501, and JP22H00159.

Appendix A Determination of reaction barriers using a small 2 H₂O cluster model. Validity of the implicit surface approximation

Fig. A.1 Same as Fig. 2 but including reactions using an explicit 2 H2O model (dashed line).

In the main text, we obtain the reaction descriptors and subsequent kinetic rate constants using two implicit approximations. In addition to an implicit solvation approach, we assume that the ASW surface does not play a major role in the reaction, introducing surface effects via rotational partition function fixing. These approximations are required for including nuclear quantum effects in the calculation of our rate constant while keeping an accurate potential (CCSD(T)/aug-cc-pVTZ//rev- DSD-PBEP86(D4)/jun-cc-pv(T+d)Z). The applicability of these approximations is ensured by a single condition, that the reaction descriptors are not affected by the water matrix. Such a condition is fulfilled normally in adsorbates bound to ASW via weak physisorption forces, although it has been applied to molecules as tightly bound as OH (Meisner et al. 2017; Lamberts & Kästner 2017a,b; Molpeceres & Kästner 2021; Molpeceres et al. 2022a; Molpeceres & Rivilla 2022; Ferrero et al. 2023). Acetaldehyde has one of the lowest binding energies found for any COM (Ferrero et al. 2022; Molpeceres et al. 2022b) so we do not expect a different behavior to the cases cited above. Nevertheless, we explicitly tested whether the introduction of explicit water affects the reaction descriptors and reaction profiles in our studied reactions. The tests we performed include comparing the activation energies, with and without water molecules, $\text{Δ}{H}_{\text{Gas}}^{‡}$, and $\text{Δ}{H}_{2{\text{H}}_2\text{O}}^{‡}$. For the 2 H₂O coupled cluster calculations, we used the domain local pair natural orbital (DLPNO) (Guo et al. 2018) version of CCSD(T), using a TightPNO localization scheme.

The results of the tests are shown in Fig. A.1 and Table A.1. Starting with the IRC profiles shown in Fig. A.1 it is evident that in all cases the barrier shapes are almost identical, with subtle differences. These are, for example, that in the 2 H₂O model the IRC path extends beyond the gas phase one, a consequence of the interaction of the reacting molecules with the 2 H₂O. The most important deviation appears for Reaction (R4) that, in the 2 H₂O model, experiences a restructuring of the water dimer to better accommodate the CH₂ CHO radical. Nevertheless, the change in the profile is still very small. Similarly, the deviations in ∆H^‡ are minimal and cannot be ruled out if the deviations are a consequence of the 2 H₂O model or by the DLPNO scheme used in the computation of the molecular energies. Overall, we find that including explicit water molecules does not affect our results. While the inclusion of further water molecules can also have an impact, this impact will not be larger than the one exerted by the closest water molecules. We remember that polar solvent effects are introduced in the main text utilizing an implicit solvation approach. Overall, we confidently conclude that the approximations in this work are justified in light of these results.

Appendix B Modeling recommendations

The picture left by the combination of our calculations and experiments is a complex one, despite the relative simplicity of the CH₃CHO molecule. We think this is a trait common to all COMs (with more than one carbon atom) hydrogenations. As chemical complexity increases, the possibility of reaction branches starts to emerge, and hydrogenation experiments are harder to reconcile with theory because the former are suited to detect final products whereas the latter are more suited to investigate elementary processes. We feel that approaches like ours, merging experiments and theory are fundamental to securely identify representative interstellar reaction routes.

However, the here studied reactions (or reaction schemes) seek to provide reliable data to feed astrochemical models, and in the particular case of this study, COMs and deuterium fractionation reaction networks, aiming to a better description of interstellar environments. Once the complexity of a reaction scheme starts to increase, as the one shown in this work, several assumptions and recommendations for astrochemical models need to be given. In this section we provide our guidelines for the treatment of CH₃CHO hydrogenation, justifying our decisions.⁵ In the first place, and although the most exact approach is to include the exact rate constants for (R2)-(R4) and an accurate modeling of H atom diffusion, we consider safe deactivating the reactions in reaction networks. We base our recommendation on the limited conversion that we observe for CH₃CHO, which remains mostly unaltered in our experiments, an observation supported by the CH₃CHO ⇋ CH₃CO + H cycle that we derive theoretically. The minor presence of CH₃CHDOD in our experiments, the only product that unambiguously can come from Reaction (R7) (as a deuterated proxy of Reaction (R2)) also encourages us to consider the branching ratio of reactions other than Reaction (R1) very low, and in the absence of further evidence, negligible. The second interpretation of our results concerns the branching ratios (α values) of Reactions (R9)- (R11). Rationalizing this choice is easy. If we consider that Reaction (R8) is the only reaction from the CH₃CHO + H then, >the 90% of remaining CH₃CHO in our hydrogenation experiment (Sect. 3.2.1) must come from Reaction (R8) alone. The remaining 10% should be distributed along the different reaction channels (R9)-(R11). It is extremely difficult to determine α for these reactions experimentally or theoretically. We consider that the best compromise is to equally weigh the remaining 10% along the three different reaction channels. Therefore we recommend setting α=0.90 for (R8) and 0.03 for (R9)-(R11).

We would like to indicate that a certain degree of chemical intuition and data interpretation is needed to provide these modeling recommendations. Because our calculations show a significant resistance of CH₃CHO to hydrogenation and deuteration, we interpret our data based on this observation and the evidence from the quantum chemical calculations. Should minor channels along the hydrogenation sequence, for example, H₂CCO hydrogenation reforming CH₃CHO as reported in the literature (Ferrero et al. 2023; Fedoseev et al. 2022) have a larger contribution, several of the α values mentioned above can vary. Nevertheless, we expect these changes to be small.

References

All Tables

All Figures

	Fig. 1 Acetaldehyde molecule with the positions where the hydrogenation and deuteration reactions were sampled and Newman projection of the molecular model showing the considered CH3CHO syn conformer.
In the text

	Fig. 2 IRC profiles for Reactions (R1)–(R4) (We omit reaction (R5) due to the high ∆H‡). IRC profiles are presented at the rev-DSD- PBEP86(D4)/jun-cc-pV(T+d)Z not corrected by ZPVE, double-level CCSD(T), or implicit water ice environment (CPCM).
In the text

	Fig. 3 Arrhenius plot of the hydrogenation rate constants (kH) for Reactions (R1)–(R4) including (dashed lines) and not including (solid lines) an implicit solvation model. All rate constants have been corrected for the implicit surface. Rate constants are presented until the lowest temperature achievable for each instanton in the sequential cooling scheme.
In the text

	Fig. 4 Similar to Fig. 3 but including deuterium as an incoming particle (kD) for Reactions (R6) and (R7)
In the text

Fig. 5 Potential energy scan for Reactions (R8) (blue curve), (R9) (red curve), and (R10) (green curve) at the revDSD-PBEP86(D4)/jun-cc- pV(T+d)z level using a broken symmetry formalism. The discontinuities in r(H-H3) and r(H-O4) are caused by the insufficient resolution of the PES scan. The small bump in r(H-O4) is attributed to the constraints in the C-H bonds required to carry out the scan (see text).

In the text

	Fig. 6 Schematic view of the CH3CO + H reaction. Depending on the initial configuration of the H atoms, the different kR must compete with diffusion, kDiff. If kR ≫ kDiff, then the reaction would dominate even in the presence of a barrierless channel, as is shown in the figure. BL means barrierless.
In the text

	Fig. 7 Fourier-transform infrared profile of CH3CHO on c-ASW at 10 K.
In the text

	Fig. 8 (a) Variation in the difference spectra of the solid CH3CHO after exposure to H atoms for 3, 60, and 120 min at 10 K. (b) Relative abundance of CH3CHO following a 2-hour exposure to H atoms (blue circles) compared to that with H2 molecules (red squares) on c-ASW at 10 K.
In the text

	Fig. 9 TPD-QMS profile of pre-deposited CH3CHO with H atoms (black line) comparable with H2 molecules (blue line) on c-ASW at 10 K: m/z = 16 (top panel), m/z = 28 (middle panel), and m/z = 46 (bottom panel).
In the text

	Fig. 10 FTIR spectrum of codeposition of CH3CHO and H atoms on Al at 10 K.
In the text

Fig. 11 TPD-QMS profile for the m/z = 14 and 42. Black arrows correspond to features attributable to H2CCO, obtained after the reaction of the pre-deposited CH3CHO with H atoms on c-ASW at 10K. The desorption peaks at 100–130 K and 140–150 K are derived from the remaining CH3CHO on c-ASW after exposure to H atoms. The peaks at 75–90 K and 150–165 K represent the desorption of H2CCO (Maity et al. 2015; Chuang et al. 2020; Fedoseev et al. 2022).

In the text

	Fig. 12 Variation in the difference spectra of the solid CH3 CHO on the Al substrate at 10 K after exposure to D atoms for 10, 60, and 120 min comparable with the initial CH3CHO.
In the text

	Fig. 13 TPD-QMS profile of pre-deposition of CH3 CHO with D atoms (black line) comparable with D2 (blue line) on Al substrate at 10 K, m/z = 30 (top panel); m/z = 17 (middle panel), and m/z =18 (bottom panel). The m/z = 30 (top panel) in the blue line represents the fragmentation of CH3CHO.
In the text

	Fig. 14 TPD-QMS profile of m/z = 36 observed for the pre-deposition of CH3 CHO and D atoms (black line) is compared with that of D2 molecules (blue line) on the A1 substrate.
In the text

	Fig. 15 TPD-QMS profiles of m/z = 49 and 50 (black lines) desorbed at around 130–200 K, likely corresponding to ethanol isotopologs CH3CD2OD and/or CH2DCD2OD. These species were yielded through the reaction of CH3CHO and D atoms, with behavior comparable to that of D2 molecules (blue line) on Al substrate at 10 K.
In the text

	Fig. 16 TPD-QMS profile for the desorption peak observed at m/z = 48 (black line) may be attributed to the formation of CH3CHDOD through the addition reaction of D to CH3CHO. This is in the comparison to the desorption behavior for D2 (blue line) on the Al substrate at 10 K.
In the text

	Fig. A.1 Same as Fig. 2 but including reactions using an explicit 2 H2O model (dashed line).
In the text