Introduction

For the origin of life on earth it has long been presumed that prebiotic molecules were delivered from interstellar space through meteorites or comets or asteroids1. The simplest amino acid, glycine NH2CH2C(O)OH, is a key building block of proteins. Both glycine and methylamine, CH3NH2, were detected in comets Wild 2 and 67P/Churyumov-Gerasimenko (67P/C-G); these observations provided strong evidence for a cosmic origin of amino acids on Earth2,3, but the nature of the formation of glycine and other prebiotic molecules during the star-forming process remains unclear.

Several paths have been proposed for the formation of glycine according to theoretical calculations. Among them, the radical–radical reactions

$$\bullet {{{{{{\rm{CH}}}}}}}_{2}{{{{{{\rm{NH}}}}}}}_{2}+\bullet {{{{{\rm{HOCO}}}}}}\to {{{{{{\rm{NH}}}}}}}_{2}{{{{{{\rm{CH}}}}}}}_{2}{{{{{\rm{C}}}}}}({{{{{\rm{O}}}}}}){{{{{\rm{OH}}}}}}$$
(1)
$$\bullet {{{{{{\rm{NH}}}}}}}_{2}+\bullet {{{{{{\rm{CH}}}}}}}_{2}{{{{{\rm{C}}}}}}({{{{{\rm{O}}}}}}){{{{{\rm{OH}}}}}}\to {{{{{{\rm{NH}}}}}}}_{2}{{{{{{\rm{CH}}}}}}}_{2}{{{{{\rm{C}}}}}}({{{{{\rm{O}}}}}}){{{{{\rm{OH}}}}}}$$
(2)

appear to be more important than the ionic channels4. Garrod5 and Suzuki et al6. employed a model consisting of processes in the gaseous phase, on grain surface, and in bulk ice in hot cores to indicate the key roles of reactions (1) and (2) in the formation of glycine on the grains. Sato et al. employed state-of-the-art DFT calculations and reported that reaction (1) is the most feasible route7; these authors proposed that the aminomethyl radical, •CH2NH2, might be produced from successive hydrogenation of HCN or H abstraction from CH3NH2 by •OH or •NH25,6,7,8,9. A similar theoretical chemical model involving NH3 and •HOCO was also proposed to understand the main routes for the formation and decomposition of CH3NH210.

Laboratory investigations to produce glycine from smaller precursors mimicking interstellar conditions have been extensive. UV photolysis and electron bombardment of various interstellar ice analogues such as CO/NH3/H2O or H2O/CH3NH2/CO2 at low temperatures were demonstrated to yield glycine11,12,13,14,15,16,17,18. Recently, Ioppolo et al.19 observed glycine formation from ices containing CH3NH2, CO, O2, and atomic H under conditions similar to dark interstellar clouds; this result indicates that glycine can be formed with no need for energetic irradiation such as UV photons or cosmic rays, that is, during a much earlier star-formation stage than previously assumed. These authors proposed that glycine was produced via a barrierless radical-radical surface reaction (1), in which •CH2NH2 was produced through H abstraction from CH3NH2 by •OH (produced from H + O2) or H atom and •HOCO was produced via reaction between •OH and CO.

Even though the radical •CH2NH2 plays a key role in the formation of glycine, its spectrum and mechanism of formation have yet to be directly identified. Bossa et al. reported that the isolation of this radical is difficult because of the reformation of CH3NH2 after the recombination of •CH2NH2 with H atom; they consequently employed CO as an H-atom scavenger to diminish the recombination, but observed only formamide and N-methylformamide, not •CH2NH2, after VUV (vacuum ultraviolet) irradiation of a CH3NH2/CO binary ice mixture20.

Here, we present direct experimental evidence, via IR spectra, of the formation of •CH2NH2 and CH2NH from the reaction of H atom with CH3NH2 via a H-abstraction tunneling reaction at low temperature, even in darkness. Furthermore, our experimental results showed a tight chemical connection between CH3NH2 and CH2NH through dual H-abstraction and H-addition cycles.

Results and discussion

To perform H-atom reactions in the laboratory, we co-deposited Cl2, CH3NH2, and para-hydrogen (p-H2) at 3.2 K and irradiated the matrix with light at 365 nm from a light-emitting diode, followed by IR irradiation. The UV photolysis of Cl2 at 365 nm generated Cl atom, which is stable toward H2 because the reaction Cl + H2 → HCl + H is endothermic and has a large barrier21. The subsequent IR irradiation excites H2 from ν = 0 to ν = 1 to overcome the energetic limitations so that the Cl atom reacts with H2 (ν = 1) to form HCl + H. The H atom thus produced reacted with CH3NH2 during IR irradiation, but the reaction continued even when the matrix was maintained in darkness for a long period because the H atom could migrate slowly in the matrix via tunneling reactions to break and form neighboring H−H bonds (so-called quantum diffusion) to approach CH3NH2 and react via tunneling reactions. The efficient production of Cl from photolysis of Cl2 in a low-temperature matrix requires a diminished cage effect, the production of H requires H2 (ν = 1), and the migration of H in darkness requires quantum tunneling reaction; all these become possible only in quantum solid p-H2 having associated unique properties22,23.

Observation of methylamine radical (•CH2NH2) and methylene imine (CH2NH)

The IR spectrum of a CH3NH2/Cl2/p-H2 (1/10/10000) matrix is shown in Supplementary Fig. 1a. The difference spectra after photolysis of the matrix at 365 nm for 30 min, subsequent IR irradiation for 90 min, maintenance of the matrix in darkness for 10 h, and secondary photolysis at 460 nm for 30 min are shown in Supplementary Fig. 1b–e, respectively. New features that appeared after IR irradiation, decreased by ~6% after being in darkness, and decreased by ~30% after secondary photolysis are indicated as group A, whereas those that appeared after IR irradiation, remained nearly constant after being in darkness, and increased by ~90% after secondary photolysis are indicated as group B. The difference spectrum after secondary photolysis at 460 nm (Supplementary Fig. 1e) in representative spectral regions is reproduced in Fig. 1b; lines in group A are pointing downward and those in group B are pointing upward, as indicated with color-coded arrows and labels.

Fig. 1: Comparison of observed lines in groups A (•CH2NH2) and B (CH2NH) with theoretical calculations.
figure 1

a IR stick spectrum of aminomethyl radical •CH2NH2. b IR difference spectrum after secondary photolysis at 460 nm of a UV/IR- irradiated CH3NH2/Cl2/p-H2 matrix after being maintained in darkness for 10 h; c IR stick spectrum of methylene imine CH2NH. Both IR stick spectra in a and c were simulated according to scaled harmonic vibrational wavenumbers and IR intensities calculated with the B3LYP/aug-cc-pVTZ method.

Full size image

The lines in group A at 3500.5, 3403.6, 3143.3, 3042.6, 1609.9, 1213.6, and 685.5 cm−1 agree well, in terms of wavenumbers and relative intensities, with the IR stick spectrum simulated for •CH2NH2 (Fig. 1a) according to the scaled harmonic vibrational wavenumbers and IR intensities predicted with the B3LYP/aug-cc-pVTZ method; the scaling method is discussed in the Methods section. To understand the perturbations of H2 on the IR spectrum of •CH2NH2, we performed calculations also on •CH2NH2 surrounded by eighteen H2 molecules, either in a hexagonal-closed pack (hcp) lattice or randomly (free optimization). The resultant vibrational wavenumbers and IR intensities are compared in Supplementary Table 1 and the simulated IR stick spectra of •CH2NH2 are presented in Supplementary Fig. 2 to compare with calculations for gaseous •CH2NH2 and experiments. The perturbation by H2 is small (with average absolute deviations 8.8 ± 6.0 and 14.8 ± 8.5 cm−1 from the gaseous phase; listed errors represent one standard deviation in fitting) and within calculation errors. This is in line with the fact that observed IR spectra of matrix-isolated species typically showed <1% matrix shifts so that comparison of observed vibrational wavenumbers with predictions of gaseous species was generally performed. The observed lines in group A agree poorly with stick IR spectra of other possible products, as shown in Supplementary Fig. 3.

The symmetric and antisymmetric NH2-stretching modes of •CH2NH2 predicted near 3388 and 3477 cm−1 were observed at 3403.6 and 3500.5 cm−1, respectively. The symmetric and antisymmetric CH2-stretching modes predicted near 3037 and 3134 cm−1 were observed at 3042.6 and 3143.3 cm−1, respectively. The NH2-scissoring mode predicted at 1616 cm−1 agrees with the observed feature at 1609.9 cm−1. The CN-stretching mode coupled with CH2-scissoring, predicted near 1194 cm−1, was observed at 1213.6 cm−1. The most intense line predicted for the NH2-wagging mode at 634 cm−1 was observed at 685.5 cm−1. Experiments and calculations are compared in Table 1. The observed features in group A can hence be clearly assigned to •CH2NH2. The average absolute deviation between experiment and prediction is 18.7 ± 15.2 cm−1 (1.06 ± 0.8%) for •CH2NH2. The large deviation for ν6 (CH2 wag) is typical for this mode because of the inadequacy in describing the double-well potential experienced by N atom, similar to NH3. All lines of •CH2NH2 located in our detection spectral range with predicted IR intensity greater than 6 km mol−1 were observed; predicted lines near 1448 and 1292 cm−1 have intensity ~2 km mol−1 too small to be observed.

Table 1 Comparison of observed wavenumbers and relative IR intensities of •CH2NH2 in solid p-H2 with their scaled harmonic vibrational wavenumbers and IR intensities predicted with the B3LYP/aug-cc-pVTZ method.
Full size table

The lines in group B at 3260.2, 3022.4, 2912.1, 1637.4, 1450.2, 1343.3, 1125.1, 1060.3, and 1056.9 cm−1 are readily assigned to methylene imine CH2NH, of which the IR spectrum in solid p-H2 was recorded by Ruzi and Anderson on photodissociation at 193 nm of N–methylformamide in solid p-H224. The observed vibrational wavenumbers and relative intensities also agree with those predicted for CH2NH, as shown in Fig. 1c. A comparison of experiments with theoretical calculations is presented in Supplementary Table 2.

To explore the possible products of reaction H + CH3NH2, we performed quantum-chemical calculations with the CCSD(T)/aug-cc-pVTZ//B3LYP/aug-cc-pVTZ method. The potential-energy scheme for H-abstraction (left side) and H-addition (right side) is shown in Fig. 2; to be self-consistent in terms of energy and species involved, we included all H atoms and H2 involved in this reaction network. The first H-abstraction from either moiety CH3 or NH2 of CH3NH2 (light blue background) results in the formation of •CH2NH2 (pink background) or CH3NH•, respectively; abstraction on the CH3 site has a smaller barrier and is more exothermic. The H-addition to either moiety NH2 or CH3 of CH3NH2 results in the rupture of the C−N bond to form CH3 + NH3 (the most exothermic) or CH4 + NH2 (involving the largest barrier). Both radicals can proceed with further H-abstraction without barrier to form closed-shell methylene imine CH2NH (light green background). The H-addition to •CH2NH2 to reproduce CH3NH2 is barrierless, and that to CH2NH to form •CH2NH2 has a small barrier (~13 kJ mol−1).

Fig. 2: Potential-energy scheme of H-addition and H-abstraction of CH3NH2 calculated with the CCSD(T)/aug-cc-pVTZ//B3LYP/aug-cc-pVTZ method.
figure 2

Energies corrected for zero-point vibrational energy (ZPVE) are in kJ mol−1; those calculated with the B3LYP/aug-cc-pVTZ methods are listed in parentheses for comparison.

Full size image

The observation of •CH2NH2 (group A) and CH2NH (group B) agrees with the predicted minimal paths for consecutive H-abstraction of CH3NH2. The H-abstraction of •CH2NH2 to form CH2NH is barrierless, whereas that of CH3NH2 to form •CH2NH2 has barrier ~33 kJ mol−1, small enough for the tunneling reaction to occur even at 3.2 K. Our previous observations of H-abstraction of methanol25, formamide26, methyl formate27, acetamide28, acetic acid29, and glycine30 by H atoms showed predicted barriers of 36, 26, 41, 41, 43, and 29 kJ mol−1, respectively. Although the calculations showed that H abstraction on the amino hydrogen to form CH3NH• has a barrier ~40 kJ mol−1, we did not observe CH3NH•, which can be readily identified at 3241.5, 2828.9, 2795.5, 1365.8, 1025.2 cm−1 in solid p-H2 as reported by Ruzi and Anderson24.

The destruction of •CH2NH2 to form CH2NH upon irradiation at 460 nm also agrees with the TD-B3LYP/aug-cc-pVTZ calculations showing an absorption band near 450 nm corresponding to HOMO (α) → LUMO (α) for •CH2NH2 (Supplementary Figs. 4 and 5 and Supplementary Table 3), but no absorption for CH3NH2 at wavelength greater than 250 nm. We observed also the formation of a small proportion of NH3 (~8% of •CH2NH2), indicating that H addition to CH3NH2 produced CH3 + NH3. We observed also CH3Cl, likely produced from the reaction of CH3 with Cl.

Isotopic Substitution reaction

We performed also experiments on partially deuterated methyl amine, CD3NH2. The representative spectra at various stages are depicted in Supplementary Fig. 6. The IR difference spectrum on secondary photolysis at 460 nm of the UV/IR-irradiated matrix after being maintained in darkness for 10 h is compared with IR stick spectra predicted for •CD2NH2 and CD2NH in Supplementary Fig. 7. Lines in group A′ match with IR lines predicted for •CD2NH2. Similarly, lines in group B′ match with those predicted for CD2NH. In addition to lines in group A′ and group B′, we observed lines at 2125.2, 996.2, 937.6, 838.7, and 769.7 cm−1, denoted group C′, that can be assigned to CD2HNH2, as presented in Supplementary Fig. 8. This observation confirms that H addition to •CD2NH2 produced CD2HNH2. A comparison of experimental results and calculations for •CD2NH2, CD2NH, and CD2HNH2 is listed in Supplementary Tables 4−6. This isotopic experiment provides not only spectral confirmation of •CH2NH2 and CH2NH, but also direct evidence of the conversion of •CH2NH2 back to CH3NH2 by H-addition, consistent with the report by Oba et al. that H−D substitution of solid CH3NH2 (and its isotopologues) is more rapid in the CH3 moiety than in the NH2 moiety when CH3NH2 reacts with H or D atoms under astrophysically relevant conditions31. Although we performed no experiment on CD3ND2 to confirm that H + CD2ND → •CD2NDH occurred, we expect this reaction to occur because of a small barrier.

Temporal profiles and reaction mechanism

The temporal evolution of the mixing ratios of each species is presented in Fig. 3 for two conditions with [H]0/[CH3NH2]0 ≈ 2.1 ([CH3NH2]0 ≈ 159 ppm) and [H]0/[CH3NH2]0 ≈ 7.2 ([CH3NH2]0 ≈ 186 ppm); details are discussed in Supplementary Note 1. •CH2NH2 was the dominant product, followed by CH2NH. In the H-deficient experiment, we observed an initial increase of •CH2NH2, followed by a decrease to reach a constant mixing ratio, consistent with the two-step mechanism for the formation from the first H abstraction of CH3NH2 and the destruction due to the second H abstraction; the anti-correlation between the mixing ratios of CH3NH2 and •CH2NH2 was clearly visible. In the H-rich experiment, •CH2NH2 and CH2NH increased significantly upon IR irradiation, and •CH2NH2 continuously decreased in darkness, indicating more significant H abstraction of •CH2NH2 due to the presence of more H atoms.

Fig. 3: Temporal evolution of mixing ratios of CH3NH2, •CH2NH2, CH2NH, NH3, and CH3Cl upon UV and IR irradiation of CH3NH2/Cl2/p-H2 matrices, followed by maintenance in darkness.
figure 3

a [H]0/[CH3NH2]0 ≈ 2.1 and [CH3NH2]0 = 159 ppm, b [H]0/[CH3NH2]0 ≈ 7.2 and [CH3NH2]0 = 186 ppm. The regions shaded with blue and red correspond to periods of UV and IR irradiation, respectively.

Full size image

The observation of consecutive H abstraction of CH3NH2 to form •CH2NH2 and CH2NH, and their H-addition to reform CH3NH2 and •CH2NH2, respectively, connects CH3NH2, •CH2NH2, and CH2NH via a dual-cycle mechanism shown in Fig. 4, similar to that among formamide HC(O)NH2, H2NCO and HNCO26. The first H-abstraction channel (reaction 1) depicted in Fig. 4 was reported by Garrod5 and Suzuki6 in their theoretical models. Hydrogenation of solid HCN and CH2NH at low temperature conducted by Theule et al.32 resulted in the formation of CH3NH2 directly, indicating the presence of two consecutive H addition (reactions 3 and 4), even though •CH2NH2 was not observed directly. In the present study, all guest molecules are well isolated in solid p-H2 at low temperature so that free radicals such as •CH2NH2 have much better chance to be trapped and maintained. Furthermore, because the hydrogenation experiments by Theule et al.32 were carried out by hydrogen bombardment, •CH2NH2 is expected to react readily with a second hydrogen atom to form the end product CH3NH2. Our observations of the formation of •CH2NH2 and CH2NH in darkness and the formation of CD2HNH2 from H reactions with CD3NH2 further support the dual-cycle mechanism.

Fig. 4: Dual-cycle mechanism of H-abstraction and H-addition reactions connecting CH3NH2, •CH2NH2, and CH2NH and the formation of CH3 and NH3.
figure 4

The area on the right with blue dotted boundary represents the major dual-cycle channels connecting CH3NH2, •CH2NH2, and CH2NH with two sets of H-abstraction and H-addition reactions. The area on the left with red dotted boundary represents minor reactions channels of the decomposition of CH3NH2 to form CH3 and NH3 upon H addition).

Full size image

We understand that our experimental conditions do not mimic the ISM conditions closely, so our results cannot be applied directly to the reactions in the ISM. For example, in the case of water ice environments, the interaction between water and the guest species might be stronger so that the stability and reactivity of radicals are different from the gaseous phase, as demonstrated by the simulations of radical-radical reactions on icy surfaces by Enrique-Romero et al.33 Nevertheless, our results clearly indicate that reaction of H with methylamine CH3NH2 produces •CH2NH2, an important radical precursor for the formation of glycine, directly supporting the mechanism, reaction (1), proposed by Ioppolo et al.19 for the formation of glycine under conditions similar to dark interstellar clouds with no need for UV irradiation or cosmic rays. The mobility of chemical reactants in the bulk ice is assumed through a swapping mechanism that was supported by laboratory work34,35 and theoretical investigations5; this mechanism likely brings H atom and CH3NH2 in close proximity to react.

Conclusions

The IR spectrum of •CH2NH2 is previously unreported; it provides a unique tool to probe this important intermediate, a precursor of glycine. We showed also the order of the consecutive H abstraction of CH3NH2 to be on the CH3 moiety first (to produce •CH2NH2), followed by the NH2 moiety (to produce CH2NH); the presence of the back reaction (H addition to radicals) was confirmed in experiments of H + CD3NH2 to produce CD2HNH2. This dual-cycle mechanism provides an explanation that CH2NH and CH3NH2 might be chemically connected. CH2NH was also considered to be a precursor of glycine via reactions with CO + H2O or CO2 + H2 in hot core or cold molecular clouds36,37,38.

Methods

Experimental details

A description on the IR absorption matrix-isolation system using p-H2 as a matrix host is available elsewhere22,39,40. A nickel-coated copper plate at 3.2 K served as a cold substrate for matrix samples and also for reflection absorption spectra. A closed-cycle helium refrigerator system was used to cool the substrate. A gaseous mixture of CH3NH2/Cl2/p-H2 (1/10/10000) was deposited, typically over a period of 9 h at a flow rate ~7 STP cm3 min−1 (STP indicates standard temperature 273 K and pressure 760 torr). The photolysis of Cl2 in the matrix at 365 nm (light-emitting diode, 5 W) produced Cl atoms; subsequent IR irradiation (unfiltered external SiC source) promoted the reaction Cl + H2 (v = 1) → H + HCl, resulting in the generation of H atoms. The matrix was maintained in darkness for 10 h to study H-tunneling reactions and was further photolyzed with light at 460 nm to distinguish lines of each species. To generate p-H2, normal H2 (99.9999%) was passed through a trap at 77 K before entering a converter containing an iron(III)-oxide catalyst cooled to 12.9 K with a closed-cycle helium refrigerator. Methylamine (CH3NH2, Sigma-Aldrich, purity ≥99%), having vapor pressure ~1400 Torr at 293 K, was used to prepare a gaseous mixture CH3NH2/p-H2 (1/10000). Methylamine-d3 (CD3NH2, Cambridge Isotope Laboratories, ≥98% isotopic purity) was used for isotopic experiments. A Fourier-transform infrared (FTIR) spectrometer equipped with KBr beam splitter and HgCdTe detector at 77 K was used to record IR absorption spectra covering a spectral range 600−4000 cm−1. A total of 200 interferometric scans at a resolution of 0.25 cm−1 were typically recorded at each stage of the experiment.

Quantum-chemical calculations

Geometry optimizations and vibrational analyses (wavenumbers and IR intensities) were calculated with the Gaussian16 program package41. Density-functional theory calculations were performed using B3LYP functionals42 and standard Dunning’s correlation-consistent basis set augmented with diffuse functions, aug-cc-pVTZ43. Furthermore, calculations on single-point electronic energies with the method coupled cluster with single and double and perturbative triple excitations, CCSD(T)44, were performed on geometries obtained with the B3LYP/aug-cc-pVTZ method; zero-point vibrational energies (ZPVE) were corrected according to the harmonic vibrational wavenumbers calculated with the B3LYP method. Both harmonic and anharmonic vibrational wavenumbers were calculated with the B3LYP/aug-cc-pVTZ method. To obtain scaled harmonic vibrational wavenumbers, plots of observed wavenumbers against calculated harmonic vibrational wavenumbers were employed for two separate regions. Two linear equations, y = (0.9810 ± 0.0126) x − (2.9 ± 16.9) and y = (0.8907 ± 0.0084) x + (233.0 ± 27.2), were derived for regions 800−1700 and 2900−3600 cm−1, respectively; y is the observed wavenumber and x is the calculated harmonic vibrational wavenumber. The average absolute deviation is 5.1 ± 3.8 cm−1 between experiments and scaled harmonic vibrational wavenumbers of CH3NH2 and 16.4 ± 22.7 cm−1 between experiments and anharmonic vibrational wavenumbers of CH3NH2. The same equations were employed to scale harmonic vibrational wavenumbers of all species considered in this work.