The origin of life has always been one of the most intriguing questions throughout human history. Biomolecules delivered to early Earth by asteroids, meteorites or comets during the period of heavy bombardment about four billion years ago have been proposed to play a role in the origin of life1,2. Similar processes may apply to rocky exoplanets (or their moons). Analysis of meteoritic material led to the identification of amino acids, sugars and nucleobases, among other complex organic molecules of extraterrestrial origin3,4. The simplest amino acid, glycine, has been discovered in comets5. The widespread hypothesis of the formation of organic molecules in space suggests that they are synthesized in the icy mantle that covers the refractory particles of cosmic dust6,7. At later stages, these dust grains form asteroids and comets. These bodies may have a liquid phase at low temperatures, where water mixes with gases such as CO2 and NH3 (refs. 8,9,10). The chemistry in the liquid phase may further enhance the molecular complexity. Experimentally, the formation of various amino acids, and even their dimers (as well as other organic molecules) has been detected following energetic processing of different molecular ices11,12. More recently, non-energetic pathways were also found for small organic molecules13,14 and glycine15,16. However, the formation of biopolymers under space conditions has not been demonstrated, although several chemical pathways in terrestrial conditions have been widely accepted17,18,19.

It is generally assumed that the prebiotic synthesis of peptides occurs in two stages. The first stage is the formation of amino acids, and the second is their polymerization20. The polymerization process requires the condensation of amino acids accompanied by the loss of water. This has a high energy barrier and therefore proceeds only at high temperatures or requires energetic processing of the material21,22. Consequently, each of these stages has a relatively low probability. Instead of first synthesizing amino acids to subsequently dehydrate them for the polymerization process, we suggest a much simpler and direct formation of the aminoketene (NH2CH=C=O) and its polymerization forming peptides under astrophysically relevant conditions. Quantum chemical calculations predict that the \({\mathrm{CO}} + {\mathrm{C}} + {\mathrm{NH}}_3{\mathop { \to }\limits^{{\mathrm{surface}}}}\) NH2CH=C=O reaction is barrierless, and therefore occurs on the cold dust grains without involving external energy23. Although isolated aminoketene is theoretically predicted to be stable24,25, it has never been observed in experimental studies to the best of our knowledge. This can be related to the extremely high reactivity of the aminoketene, which could also lead to its efficient polymerization. The polymerization of NH2CH=C=O via the nucleophilic attack of the nitrogen on the carbonyl carbon and an intramolecular proton transfer from NH2 to CH results in the formation of peptide chains. Therefore, in contrast to the polymerization of amino acids, the formation of polyglycine via polymerization of aminoketene is a much simpler process.

To test this reaction pathway experimentally, we performed the co-deposition of CO, C and NH3 on the surface of both Si and KBr substrates cooled to 10 K and placed inside an ultrahigh vacuum (UHV) chamber to mimic the conditions in dense molecular clouds.

Chemistry at 10 K

Infrared (IR) absorption spectra of the ice produced after the co-deposition of CO + NH3 and CO + C + NH3 at low temperature (T = 10 K) are shown in Fig. 1. The additional control experiments, involving only two reactants C + CO and C + NH3, are shown in Extended Data Figs. 1 and 2.

Fig. 1: IR absorption spectra of the ice mixtures produced at 10 K.
figure 1

The red line shows the IR spectrum of CO/NH3 (1:1) ice and the black line the spectrum of CO/NH3/C (1:1:0.1) ice. Stretching and bending vibrations are denoted ν and Δ, respectively.

The IR spectrum representing the deposition of all three reactants (CO, C and NH3) on the substrate at 10 K reveals several new absorption features, which are absent in reference spectra involving only two reactants (for example, C + CO, C + NH3 and CO + NH3). Therefore, these bands have to be assigned to a product formed by reactions of all three reactants at 10 K. In addition, the warming up of the CO/C/NH3 ice layer to room temperature leaves a residue on the substrate, which is also observed for the reaction of C atoms with only CO or NH3, but in much smaller amounts. We conclude that the product formed in the reaction of all three reactants at 10 K is required for the formation of the non-volatile products at 300 K. As C atoms are the limiting reactant (that is, the fluence of C atoms was at least ten times less than that of NH3 or CO), the reactions of C atoms with other C atoms or with its reaction products are negligible. Consequently, the observed product is expected to originate from the reaction of CO + C + NH3 and contains CH, CN, C=O and NHn groups, as shown in the IR spectrum (black line in Fig. 1). Given that NH3 does not react with CO at 10 K, the addition of C atoms must act as a chemical trigger to initiate such solid-state reactions. The formation of the C=C=O functional group is demonstrated by the appearance of the absorption bands in the ~1,965–2,275 cm−1 range. It implies the transformation of the CO triple bond (C\(\equiv\)O) to a CO double bond (C=O) species by attaching a C atom to the C atom of carbon monoxide. Similarly, the detection of CH and NHn (2,500–3,000 cm−1) and CN (1,320–1,500 cm−1) absorption bands also suggests covalent bond formation between the added C atom and ammonia: C-atom addition to N atoms of NH3 followed by H-atom rearrangement. On the basis of the above considerations, we can conclude that O=C=CHmNHn molecules are formed.

To better understand the chemistry involving all three reactants at 10 K, we performed quantum chemical calculations (results are shown in Fig. 2 and Extended Data Fig. 3). As shown in Extended Data Fig. 3, C atoms initially prefer to react with ammonia molecules, forming the prereactive complex CO + CNH3, which corresponds to the first energy well in Fig. 2. The C + NH3 reaction pathway has been studied experimentally16 using the recently developed calorimetric technique26. The proton transfer from nitrogen to carbon was observed to form NH2CH before the energy dissipation16. This mechanism also applies to the reaction investigated here, as shown by the appearance of NHn and CH absorption bands (~3,048 and ~2,820 cm−1, Fig. 1) after C-atom addition. The formed NH2CH product subsequently reacts barrierlessly with CO, leading to the formation of NH2CH=CO. There are thus two possible products of this reaction: NH2CH=C=O and CH2NH + CO. The latter is the same as the product of the reaction with only two reactants C + NH3 → CH2NH. Therefore, the reaction product observed upon the addition of all three reactants at 10 K can only be NH2CH=C=O, the functional groups of which are observed in the IR spectrum. The formation of this molecule was confirmed by temperature-programmed desorption in combination with quadrupole mass spectrometer analyses monitoring the ions with masses of 57 and 56 unified atomic mass units (u) as shown in Extended Data Fig. 4. At the same time, C atoms could react with either CO or NH3, and a smaller number of CCO or H2CNH, or even NH2CH2NH2, molecules can also be formed.

Fig. 2: Energy level diagram for the reaction involving CO, C and NH3 reactants.
figure 2

The reaction starts with the triplet state. The transition states (TS) and the singlet states (S) are labelled and the dashed lines show the possible reaction pathways, with the most probable pathway based on computational and experimental results in red.

Chemical transformation during temperature rise

After deposition, the substrate was heated at a rate of 2 K min−1. The IR spectra measured at selected temperatures are shown in Fig. 3. A comparison of the spectra at 10 K and 70 K shows that the sublimation of CO, occurring mainly from 25 to 40 K, does not initiate a new chemistry. However, with the sublimation of ammonia (above 110 K), the products of the C-atom reactions can finally encounter each other and react. We observed the formation and growth of new IR absorption bands, while the intensity of the νNHn and νC=O bands as shown in Fig. 3 decreased significantly. With the mass spectrometer, we did not detect any considerable desorption from the substrate besides CO and NH3. Therefore, the disappearance of these IR bands was due to chemical transformation rather than sublimation. The decrease of band intensities in the 2,800–3,050 cm−1 range is a common indicator of glycine polymerization27. This occurs due to the transformation of the NH3+ or NH2 groups of glycine into the NH groups present in peptides. Simultaneously with the transformation of NHn groups, we observed the formation of the peptide I band at 1,670 cm−1 and peptide II band with a peak maximum ranging from 1,548 to 1,569 cm-1 (Fig. 3 and Extended Data Fig. 5), which are the characteristic features of peptide bonds. The peptide bond is therefore formed at 100–120 K simultaneously with the decrease in the intensity of the IR bands associated with the molecules present in the ice at 10 K (Extended Data Fig. 6). The IR spectrum did not change substantially above 150 K; the residual material present on the substrate after warming up to 300 K (R300K) must therefore have formed at low temperature during ammonia sublimation.

Fig. 3: IR absorption spectra obtained during annealing of the material produced by co-deposition of CO + C + NH3.
figure 3

The labels for the vertical dashed lines indicate the positions of the peaks appearing during heating at the temperatures indicated. All CO is sublimated at 70 K and most NH3 is sublimated at 120 K.

Characterization of the room-temperature residue

The in situ characterization was performed by IR spectroscopy. After removing the substrate from the UHV chamber and exposing it to air, we did not observe any noticeable modification of the IR spectrum of R300K. Therefore, an ex situ analysis could also be performed. The high-resolution ex situ mass spectrum of R300K is shown in Fig. 4. To better characterize the mass spectrum, we selected a threshold shown by the horizontal dashed line in Fig. 4. The peaks with intensities higher than this threshold were analysed to determine the series. We could identify many series of mass peaks, which are separated by the same mass of about 57.0215 u. This value exactly matches the mass of the NH2CH=C=O molecule, which was found to be the main product formed at 10 K. The intensities of the peaks in these series drop exponentially with increasing mass. A series is formed if the same molecule is continuously added to an initial species. Therefore, the presence of these series with such an intensity distribution unambiguously indicates the polymer formation with the mass of the monomer unit equal to the distance between the peaks. The complete list of all series found is given in Extended Data Fig. 7. As shown in Fig. 4, most of the analysed mass peaks belong to the series.

Fig. 4: Ex situ mass spectrometric analysis of the R300K residue.
figure 4

Top: the mass spectra of R300K obtained from ex situ analysis using an ESI-Orbitrap mass spectrometer. The horizontal orange dashed line shows the threshold used to select the analysed peaks. The peaks that belong (or do not belong) to the series are displayed in red (blue). Black peaks lie below threshold and were not analysed for series. The molecular structures display the Gl4 peptides detected in the mass spectra as protonated species [analyte + H]+. Bottom: the intensities of the bands derived from two observed series corresponding to glycine peptides are plotted as a function of the numbers of monomeric units (n). Error bars are standard deviations of corresponding peak intensities in the measured mass spectra.

We can therefore conclude that the polymerization of the NH2CH=C=O molecule formed at 10 K, is the main chemical pathway leading to the formation of R300K. The formation of a variety of different series (Extended Data Fig. 7) can be understood assuming that other molecules formed at 10 K could be added during the polymer formation.

The IR spectrum of R300K is shown in Fig. 5 (see also details in Extended Data Fig. 5). It highlights the absorption features characteristic for the peptide bonds. As the NH2CH=C=O molecule does not contain the peptide bond, the appearance of this bond in the formed polymer indicates the type of the bond between the monomeric units. Upon sublimation of volatile species, the monomeric units of NH2CH=C=O are assembled through peptide bonds, resulting in glycine oligomers, which differ from canonical peptides only by the groups at their C termini. This proceeds by the polymerization of NH2CH=C=O molecules, involving proton transfer from nitrogen to the carbon atom of the CH group. However, our density functional theory calculations show that the NHCH2C=O (as well as NH2CHCO–NHCH2C=O) molecules are not stable; they are expected to fragment with the loss of the CO molecule. Therefore, peptide chains produced by this mechanism are terminated by NH2 on one side, which corresponds to the canonical structure of the peptide. The other side of the peptide chain is terminated by a CH2 group instead of a classical COOH group. However, the observed series is not the most intense one. The most intense peak (189.0983 u) shown in Fig. 4 suggests that the favoured oligopeptides are terminated by amino groups on both sides, which assumes an efficient involvement of ammonia in the polymerization process. In spite of the differences in length and terminal functional groups between the canonical Gl3 peptide and the peptides identified in our experiment, there is a good match between the two spectra shown in Fig. 5. The main discrepancy is the width of the IR absorption bands; this is expected owing to the formation of oligopeptides with a wide distribution of lengths. The band broadening is caused by the change in the positions of the IR absorption bands with the number of units in the glycine oligomers28.

Fig. 5: Comparison between the IR absorption spectra obtained from R300K and Gl3.
figure 5

The vertical dashed lines are drawn to visualize the coincidences between the peaks in the two spectra. The Gaussian profiles beneath R300K spectrum show the deconvolution of the experimental spectrum (Extended Data Fig. 5). The spectrum of triglycine (Gl3) is adopted from ref. 28.

Finally, to confirm the proposed molecular structure of the polymers, we performed the higher-energy C-trap dissociation29 of the most abundant ion from Fig. 4. The result of this measurement is shown in Fig. 6. This analysis explicitly confirms the suggested molecular structure of this ion with mass 189.0984 u. The efficient loss of two ammonia molecules demonstrated by peaks b3 and b3* clearly shows the terminal location of both amino groups. The a3 peak shows that the side amino group is linked with the CO unit. The b2 and y1 peaks demonstrate that the side H2NCO unit is connected to the CH2NH unit. The same analysis can also be performed from the other side, where the terminal amino group is joined to the CH2CO group, as demonstrated by the y2 fragmentation peak. The fragmented ion (189.0983 u) belongs to the series of peaks that considerably exceeds the intensities of other series. Therefore, the mass peaks from this series can be separated by both mass and intensity. This shows that the formation of the ions from this series happens via polymerization of the NH2CH=C=O molecule. All of this unambiguously confirms the suggested molecular structures of the ions from this series. Moreover, it also strongly suggests that ions from other series shown in Fig. 4 have a similar molecular structure and are modified peptides. We evaluated the entire mass spectrum, comparing the integrated intensities of mass peaks in series and the total integrated intensities of all analysed peaks. This analysis determined a polyglycine concentration in our sample of about 85%, given the same ratio between the polyglycine peaks and other mass peaks among non-analysed peaks below the threshold. If all non-analysed (weak) peaks did not belong to polyglycine, the value would be reduced to 19%. Considering that the R300K material is completely solvated in water and that the ratio of the intensities of the peptide bands towards other IR bands in the spectrum of our sample is comparable to that in the spectrum of pure oligoglycine, we concluded that the predominantly formed molecules in our experiments are polyglycine.

Fig. 6: The higher-energy C-trap dissociation of the 189.0983 u ion.
figure 6

The fragmentation peaks are annotated based on the accepted nomenclature for the peptide fragments29,44. The loss of an additional ammonia molecule is indicated with an asterisk. The inset shows the molecular structure of the investigated ion. The coloured marks (a, b, y) determine the fragments that correspond to the mass peaks identified in the spectrum.


The reactants used in this work are among the most abundant species present in the interstellar medium (ISM). The fractional abundances of NH3 and CO in the ice mantles covering refractory dust particles are 10% and 40%, respectively30. In the ISM, more than half of all carbon exists in the form of atomic gas31. The formation of dense clouds, where new stars and planets are formed, goes through a stage of translucent clouds, where a notable portion of carbonaceous dust is expected to be formed due to the accretion of C atoms. In translucent molecular clouds, the dust temperature is low enough to allow the formation of molecular ices while carbon is dominantly present in the atomic or CO forms32. Therefore, reactions between accreted CO, C and NH3 could be very common in this region. At the low temperatures of dust in translucent molecular clouds (T = 10–20 K)33, this leads to the formation of organics34. As demonstrated here, a portion of these organics could be in the form of peptides. At later stages, such dust becomes the building blocks of comets or meteorites. The formed organics could therefore have been delivered to the early Earth during the period of heavy bombardment. The survival of organics at all delivery stages is thus a relevant question. In this sense, comets have the advantage of being less mechanically stable than asteroids, because their fragmentation after entering the atmosphere may allow strong impacts to be avoided due to the loss of kinetic energy in the atmosphere as the meteor breaks up2. Comets might deliver intact organics produced in the ISM at a rate of at least 106–107 kg yr−1 (ref. 2). However, even the direct impact of larger objects may allow the partial survival of molecules, including amino acids35,36. The biggest issue with this extraterrestrial scenario is the stability of the newly formed organics in the harsh energetic environment of the ISM, filled with X-ray and ultraviolet photons. The non-energetic formation of peptides may be advantageous in this regard: if gas-to-solid phase transitions and reactions occur in regions with a low ultraviolet flux, the formed peptides could survive long enough to be incorporated into bulk solids, which would shield them from further destruction. This is in line with the potential detection of proteins in meteorites37,38. Peptides also play a key role in the origin of life20. The delivery of peptides to Earth should therefore expand the opportunities for evolutionary chemistry and biology. Moreover, the peptides with two amino terminals dominantly formed in our experiments are capable of efficient self-assembly39, which offers the prospect of new and interesting opportunities for abiogenesis.


In situ experiments

The experiments were performed using the UHV INterStellar Ice Dust Experiment (INSIDE) set-up described elsewhere40. We performed the co-deposition of CO, C and NH3 on the surface of KBr substrates cooled down to 10 K and centred in the UHV chamber. The deposition of reactants was performed for about 1 h. Low-energy carbon atoms were generated by an atomic carbon source41. The source generated a pure flux of low-energy carbon atoms in the triplet ground state C(3PJ) and about 1% of CO/CO2 molecules. The background pressure inside the vacuum chamber (1 × 10−10 mbar) and the temperature of the substrate (10 K) allowed us to mimic the chemistry under dense molecular cloud conditions. We used approximately equal amounts of CO and NH3 molecules, while the number of C atoms was at least ten times smaller. The flux of C atoms from the source was estimated in previous work42. This small amount of C atoms allows us to exclude their reactions with each other, as well as with their reaction products. The gases were introduced through two separated leak valves. The ice thicknesses on substrates were monitored by infrared (IR) spectroscopy using a Fourier transform IR spectrometer (Vertex 80 v, Bruker) in the transmission mode. In situ mass spectrometry was performed using a quadrupole mass spectrometer (HXT300M, Hositrad) attached to the same UHV chamber. After the deposition, the substrate was heated at a rate of 2 K min1. As the material was warmed, IR absorption spectra and mass spectra were measured to monitor the residue and sublimated gas-phase species, respectively. The evolution of IR spectra during temperature rise (shown in Extended Data Fig. 6) was obtained by integrating the corresponding bands in the IR spectra.

Ex situ mass spectrometry analysis

For the mass spectrometry analysis, we used silicon substrates and performed a 4 h deposition of the reactants CO, C and NH3 under the same conditions as the in situ experiments. We obtained very similar IR spectra at all stages of the experiment, using both Si and KBr substrates. After warming up, the substrates were removed and analysed. The residue was extracted with a water–methanol mixture (70:30) and used for the mass spectral analysis. Complete solubility of the residue in water and water–methanol mixtures was observed. There was also no chemical reactivity of the residue with water, as shown by the unchanged IR spectrum of the material after prolonged exposure to wet air. The mass spectrometry was performed using the hybrid linear trap/Orbitrap mass spectrometer (Thermo Fisher QExactive plus mass spectrometer with a heated ESI source). The accuracy of the mass determination was higher than 5 ppm, which resulted in a possible error only in the fourth decimal place for the investigated mass range. Considering the limited number of elements used in the experiments (C, O, N and H), this high resolution allowed us to unambiguously determine the elemental composition of the detected ions in the mass range shown in Fig. 4. To ensure that the observed mass peaks were generated by the material formed in our experiment, we checked for the possible presence of contaminants (both on the substrate and in the mass spectrometer) using the same procedure with clean silicon substrates.

The same mass spectrometer was used for tandem mass spectrometry measurements. For this purpose, the quadrupole filter was set to preselect ions with the mass range of 188.9–189.3 u (the most abundant Gl3 ion observed in our mass spectra). This cation was then a subject for a higher-energy C-trap dissociation procedure, as described in ref. 29. The collisional energy was set to 30 eV. In test measurements with 10 and 50 eV collisional energies, we did not observe any principal difference in the fragmentation pattern. Two impurity peaks (189.0734 and 97.0373 u) were subtracted from measured spectra. These peaks were assigned to the impurities on the basis of the presence of this ion signal during the tuning of the collisional energies when all other mass peaks were missing, and the fact that ions with these masses cannot simply be fragments of the parent ion.

Quantum chemical computations

Quantum chemical calculations were performed using the GAUSSIAN16 software43. To find out the outcome of the reaction of C atoms landing on CO/NH3 ice, we performed a two-dimensional potential energy scan of the CO + C + NH3 reaction at the MP2/6-311+G (d, p) level. In this scan, we varied the C–C and C–N distances and allowed complete optimization of molecular geometries without any further restrictions. This computation allowed us to identify the formation of the first prereactive complex used for the calculation of the energy level diagram, which is shown in Fig. 2. The geometries of the molecules in the energy level diagram were determined at the B3LYP/6-311+G (d, p) level. The reaction energies were determined from the difference between the sum of the energies of the reactants and the energy of the product molecules with vibrational zero-point energy corrections.