the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Aug 2026
Highly oxygenated organic molecule formation from 2,5-dimethylfuran oxidation by O3 and OH: an experimental and computational study
Abstract. The gas-phase oxidation of 2,5-dimethylfuran (2,5-DMF) by ozone (O3) and hydroxyl radicals (OH) was investigated in flow reactors at atmospheric pressure and room temperature using nitrate chemical ionization orbitrap mass spectrometry. At a residence time of 0.8 s, highly oxygenated organic molecules (HOM) with up to nine oxygen atoms were detected in the ozonolysis channel, and eight oxygen atoms in the OH channel. These observations demonstrate that sequential intramolecular hydrogen-shift autoxidation is sufficiently rapid to form products with up to nine oxygen atoms on sub-second timescales. At 7 s, C10–C12 accretion products form through different combinations of C5 and C6 alkyl peroxy (RO2) radicals. Quantum chemical calculations using density functional theory and coupled-cluster methods identified a distinct Criegee intermediate geometry (Anti-CI-2A) that provides a plausible route to HOM species containing up to nine oxygen atoms. Its different methyl/hydrogen orientation relative to the conventional Syn and Anti conformers enables a rapid 1,6-H shift, facilitating autoxidation toward O9 formation and suggesting that primary ozonide decomposition in structurally complex ozonolysis systems may access reactive Criegee intermediate geometries beyond the conventional Syn and Anti forms. For OH-initiated oxidation, the proposed mechanism accounts for HOM monomer formation up to O6, with rapid termination and radical recycling limiting further autoxidation. Overall, 2,5-DMF produces low-volatility oxidation products from both ozonolysis and OH-initiated pathways with implications for secondary organic aerosol formation.
- Preprint
(2174 KB) - Metadata XML
-
Supplement
(1477 KB) - BibTeX
- EndNote
Status: open (until 24 Sep 2026)
- RC1: 'Comment on egusphere-2026-4450', Anonymous Referee #1, 03 Sep 2026 reply
-
RC2: 'Comment on egusphere-2026-4450', Anonymous Referee #2, 03 Sep 2026
reply
Reviewer report on Asgher et al. “Highly oxygenated organic molecule formation from 2,5-dimethylfuran oxidation by O3 and OH: an experimental and computational study”
Asgher et al. present a combined experimental and computational study on the formation of highly oxygenated molecules from the reactions of 2,5-dimethylfuran with O3 and OH. Furans and alkylated furans are emitted in large quantities from biomass burning, but the oxidation mechanisms of these compound classes are still rather incomplete. Consequently, the present work is a valuable contribution and of broad interest to the readership of ACP.
Overall, I think the manuscript requires minor revisions before it can be published. I have the impression that, given the range of experimental conditions described in the Methods section, the authors could present and discuss the experimental results in greater detail. This would help the reader to better follow and assess the authors’ conclusions.
General comments:
1.) When I first read the manuscript, I found myself somewhat confused by the way the authors use terms such as “product”, “accretion product” and “HOM monomer”. While “accretion product” is clear, it was not clear to me whether “HOM monomer” is restricted to closed-shell species or whether it also includes the corresponding peroxy radicals. I would therefore encourage the authors to distinguish more clearly between closed-shell products and peroxy radicals, and to explicitly indicate whenever a molecular formula refers to a radical species. I think this would make it much easier for the reader to follow the authors’ reasoning throughout the manuscript.
2.) The discussion of the experimental results addresses the competition between unimolecular and bimolecular radical reactions and the shift in product distribution with increasing reaction time. This is associated with a larger contribution of OH chemistry and higher RO2 concentrations. Regarding this discussion, I have two questions that might be worth addressing:
a.) Do the authors have an estimate of the peroxy radical concentrations in their experimental set-up, perhaps based on the kinetic modelling they performed for POZ formation and decomposition? If not, would it be possible to estimate radical concentrations in the flow tube? As a reader I would be curious to know at least a rough estimate of the RO2 concentration and how they change under the different experimental conditions, since RO2 concentrations are used to explain the change in product distribution.
b.) The authors have varied initial reactant concentrations by more than a factor of ten in the “O3 only” and “O3 + TME” experiments. If the longer reaction time leads to a clear change in product distribution due to higher RO2 concentrations, should a similar effect not also be observed when varying the initial reactant concentrations over such a large range? It might therefor be worth providing more details on how the product distribution changes as function of the initial reactant concentration.
3.) I would encourage the authors to address the impact of NO on HOM formation more thoroughly, based on the following comments and questions:
a.) Table 1 states that the NO mixing ratios were varied in the range of 20–500 ppbv in these experiments, but only results obtained at 100 ppbv are presented. Is there a specific reason for choosing this condition, or were the results largely independent of the NO mixing ratio?
b.) Did the authors perform any kinetic modelling to assess the relative contributions of 2,5-DMF ozonolysis and the reaction of 2,5-DMF with OH under these experimental conditions? I am wondering whether, based on the information provided in Table 1, the O3 loss is dominated (by about 95%, if I am not mistaken) by its reaction with NO. In this context, can the authors rule out that NO3 formation and subsequent reaction of 2,5-DMF with NO3, which proceeds close to the collision limit, play a role and potentially interfere with the organic nitrate formation monitored here? I am not questioning the experimental results, but I think this point would be worth addressing.
c.) The authors state that the RONO2 species observed in the mass spectra are C6H9O4–9NO and C6H7O7NO and assign these species to the different peroxy radicals formed in the system. However, I do not think that this assignment is entirely correct. If the dominant peroxy radicals formed are C6H9O4 following OH addition, and C6H9O6 and C6H9O8 following successive unimolecular isomerization, the corresponding RONO2 species should be C6H9O4NO, C6H9O6NO, and C6H9O8NO. In contrast, organic nitrates of the form C6H9OXNO, where X is an odd number, would require an intermediate alkoxy radical step.
d.) The authors initially state and later conclude that 100 ppbv NO suppresses autoxidation. However, this appears to be inconsistent with the results presented. If I understand the proposed reaction mechanism correctly, all RONO2 species except C6H9O4NO are formed following unimolecular isomerization. Could the authors clarify how these results support the conclusion that 100 ppbv NO suppresses autoxidation?
4.) The authors performed calculations indicating a strong preference for Syn-CI-2 formation (about 95 %) among the Syn conformers, as also shown in Fig. 7. This is an interesting result, but I find it difficult to assess its overall significance without any information on the Anti conformers and, consequently, on the relative importance of all possible conformers. This seems particularly important because, in Fig. 6, the authors highlight that key intermediates leading to HOM formation originate from Syn-CI-1 and Anti-CI-2, for which the branching fraction is either low (Syn-CI-1) or not provided (Anti-CI-2). If possible, can the authors comment on the relative importance of all CI conformers?
5.) The authors show that the dioxirane route appears to be the dominant loss pathway for the Syn-CI-1 conformer, although a 1,4-H shift is accessible. This is unusual and rather surprising. Nevertheless, the authors highlight the VHP formed via the 1,4-H shift as a key intermediate in Fig. 6. I assume this is because the VHP is a key intermediate in HOM formation, irrespective of the relative importance of this pathway? If this is the case, I would encourage the authors to clarify this point in the figure caption.
6.) The reaction of 2,5-DMF with ozone has been investigated experimentally in previous studies. The authors specifically refer to a chamber study (Illmann and Rösgen, 2024) in which product formation was monitored by FTIR and PTR-MS, resulting in a carbon balance below 50 %. While I agree with the authors that this low carbon balance might be due to the inability of these techniques to detect higher oxygenated species, I am wondering whether the authors could provide any quantitative information, such as an estimate of the HOM yield?
The authors state that “these low-volatility products […] may contribute to the >50% carbon balance deficit”. I think it would be useful, if possible, to provide the reader with an estimate of the magnitude of the formation of low-volatility products. I would have assumed that a HOM yield should be well below the 50 % deficit, particularly since one of the key routes to HOM formation identified by the authors involves a CI conformer whose branching fraction appears to be rather low. Could the authors comment on this and, if possible, provide a quantitative estimate of the contribution of HOM formation?
Specific comments:
Table 1: “concentration (ppb)” should be “mixing ratio (ppb)”
Lines 192-194: I would encourage the authors to rephrase this sentence, since the peroxy radicals differ not only in the number of hydrogen atoms, but also in the number of oxygen atoms. Perhaps it would be clearer to say that the oxidant channels produce structurally distinct RO2 radicals, resulting in two separable product families that differ by two hydrogen atoms?
Lines 229-231: The sentence starting with “The C10 composition” is somewhat difficult to follow. I would encourage the authors to reformulate it for clarity.
Lines 246-248: I do not entirely understand this sentence. What do the authors mean by a competition between recombination and fragmentation pathways? I would have assumed that both arise from bimolecular reactions, in which case their relative contributions should not depend on the RO2 concentration. Could the authors please rephrase this statement or provide some additional information to clarify what is meant here?
Line 314: “Fig ??” should be “Fig 8”?
Citation: https://doi.org/10.5194/egusphere-2026-4450-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 71 | 43 | 10 | 124 | 21 | 15 | 12 |
- HTML: 71
- PDF: 43
- XML: 10
- Total: 124
- Supplement: 21
- BibTeX: 15
- EndNote: 12
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors provide a complementary set of experimental observations and theoretical analyses regarding the oxidation of 2,5-DMF by O3 and OH. The results are interesting and informative. They are definitely worth being published in ACP. However, there are a number of places where where further information and additional clarification would be helpful. Thus, I consider the manuscript to need minor revisions before it can be published.
The discussion of the symmetry factor in Eq. (1) is either too short or some errors were made. The symmetry factor should account for ratios of rotational symmetry numbers and numbers of enantiomers. For the reaction of 2,5-DMF with O3, the appropriate product and ratio of all the components (i.e., what the authors describe as the symmetry factor) should be eight, which correlates with the 8 ways that O3 can add (from above or below the DMF plane, to the left of right double bond of DMF, and with either of two O3 orientations). This 8 comes from 2 for the symmetry of DMF, 2 for the symmetry of O3 and 2 for the number of enantiomers in the TS. For the reaction with OH, the ratio should instead by 4.
It is hard to understand why LC-TST was used for the bimolecular reactions, while MC-TST was used for the unimolecular reactions. Some rationale should be provided for these different choices in methodology.
The discussion of the T1 diagnostic, which presumably is meant to inform the reader regarding the appropriateness and accuracy of the calculations, could benefit from significant additional details. For example, it is well known that O3 itself is a highly multireference species. Similarly, CIs are well known to have substantive multireference character. In the latter case, it is clear from many literature studies that CCSD(T) based evaluations of CI energies are generally in error by about 1 kcal/mol. Diradical pathways can also be significant in the decomposition of a POZ. If these pathways had been treated, they would have even larger multireference effects.
One of the challenges for readers who are not experts in theory is in understanding the expected accuracy of theoretical calculations. Without this understanding it is very hard to understand the utility and/or limitations of the theoretical analysis. For this reason, it would be incredibly valuable to the general audience to provide some indication of the expected accuracy of the present calculations. While this is not yet the publication norm, I believe it should be. From my experience, I would expect the two sigma accuracy of the energies to be about 1.5 to 2.0 kcal/mol. It is harder to know what kind of accuracy to expect in the partition function evaluations. I would guess that the uncertainties for that are somewhere between a factor of two and ten – at least for the MC-TST case. Meanwhile, the master equation treatment of the pressure dependence likely introduces an additional an additional factor of two to three uncertainty, at least when the pressures of interest are far from the high pressure limit.
It would also be useful to suggest an expected uncertainty for the M06-2X calculations used for the SI data. If this is 5 kcal/mol can all the SI pathways still be ruled out? What if it is 10 kcal/mol? In principle, you have an indication of the accuracy of the M06-2X energies for the fairly large set of stationary points where you also calculated the CCSD(T) energies.
It appears that the authors calculate the rate constant for O3 to add 25DMF. Those calculated rates should be compared to the literature rates. That comparison would provide some qualitative indication of the accuracy of their methods. Perhaps the calculated number is reported on line 41 of the SI, but that should be made more clear.
It is unclear to me how the green contribution in Fig. 2 b is obtained separately from the gray contribution. It would be helpful to have some specification of how that is done.
It seems odd to me that no mention is made in lines 217 to 224 of the increased C6H7O8 from the O3 pathway. On line 222, the authors mention an elevated C5H7O7 signal for the long timescale experiments. But the signals seem to be ~250 in Fig. 2a and ~200 in Fig. 2b. Thus, I am confused by that statement.
In Fig. 6, key intermediates are said to be highlighted in red. I am puzzled as to what makes those intermediates key? For example, the branching to VHP-A relative to the dioxirane channel next to it is less than 1%. How then is that a key channel? To make matters worse, they predict only a 5% branching to Syn-CI-1, which just adds to the confusion regarding how VHP-A can be a key intermediate.
Should there also be a diradical channel for the decomposition? Perhaps that channel might be 10 % of the flux to CI-1 and CI-2. But that would still make it more than the flux to VHP-A. Are the authors aware of some data that says a diradical channel would be irrelevant for 25DMF ozonolysis?
The IRC calculations connect to the syn CI-1 and CI-2 conformers. It would be helpful for the authors to say what they take this to mean for the formation of the anti CI-1 and anti CI-2 conformers? They show the product of Anti-CI-2 as being a key intermediate in Fig. 6. So apparently they believe that there is still some possibility for forming Anti-CI-2. Some statement should be made as to how the anti-conformers might be formed. Presumably some inspection of the TSs for forming the CIs would indicate that the torsional partition function likely contains some probability of being in a torsional state that would correlate with the Anti-CIs.
Some comment should be made as to whether or not the rates in Fig. 8 are fast enough to allow much production of the HOMs on the 0.8 s timescale or even on the 7 s timescale. I think the answer to that question depends to some extent on what the uncertainty in the rate predictions is. Do the simulations indicate that Anti-CI-2A is required in order to form any HOMs on the experimental timescales?
On p. 21, the contrast in lowest conformer rates and IRC connected rates suggests a possible issue. What if the lowest energy reactant conformer is connected to a much higher barrier conformer and the conformational dynamics is very slow. Then, if all the population is in the low enough reactant conformer, the rate could be much slower. This sort of issue could readily be resolved by simply determining the barriers and rates for the conformational dynamics in the reactant, which are probably rapid enough to remove any of these ambiguities. The contrast they present really does not resolve the issue.
Are C6H7O3-V and C6H7O3-VI different structures, or just different depictions of the same structure? If the latter, it would be better to just indicate the resonance with a standard … curved resonance bond across the two bonds (i.e., just like in their depiction of VHP-C) and then connect the subsequent V and VI species to that one resonance structure. To me, giving the same species two different labels is confusing. Furthermore, the terminal O is another part of that resonance and the resonance bond should extend to it. Similar comments could be made about a variety of other structures in their schematics. E.g., the C6H7O3 structures in Figs. 8a and 8b should include a resonance with the neighboring C=O.
In each of the VHP dissociations, the OH fragment arising from the OO fission could have reattached to the new C radical site (in one resonance form) in a roaming process to form hydroxyketones. This process was seen to be important in calculations for ozonolysis of alpha-pinene (J. Phys. Chem. A, 127, 10657, 2023) and directly observed in methyl ethyl CI (J. Am. Chem. Soc. 145, 19405, 2023). Some comments on the expected relevance of such roaming processes to the HOM formations would be helpful.
The discussion on the bottom of p. 21, presumes strong reliance on the appropriateness of IRC calculations in identifying final products. Transition state theory does not say that IRCs are followed out to products. The passage through the TS’s will occur over a distribution of torsional states. There are then two key questions. (i) Do different torsional values at the TS correlate with different conformers of the CIs (e.g., Syn-CI-2 vs. Anti-CI-2 vs. Anti-CI-2A)? (ii) If there are torsional values that correlate with Anti-CI-2A, does one expect a reasonable population of those torsional values? The first question could be reasonably addressed with simple optimizations that start from a specific torsional value and a slight kick in the direction of the products. Do you see different product CI conformers when you do that?
Technical Corrections:
For my own interest, I am curious why the authors use a stochastic solver rather than a traditional ODE solver to handle the kinetic simulations. I would have presumed that the latter was both simpler and more efficient.
The species “DICARBO2” is mentioned a few times. It would be helpful to give some form of proper chemical identification of this species when it is first mentioned.
On line 244, the authors refer to “Table 2, Fig. 3”. I don’t think Table 2 has anything to do with the TME experiments. Perhaps they mean Table 1, or some other Table.
On line 267, the phrase “Addition of 100 ppbv NO at delta t = 7 s” makes it sound like NO was added after the reactor had been reacting for 7 s. I doubt that is true and that the NO was part of the initial mixture. I presume the specification of the time was just to indicate which experiment is being considered. Since this description can be confusing it would be good to rephrase it.
On p. 17, and p. 22 Fig. ?? needs to be fixed.
The citation to Castaneda et al. needs to be fixed.
In the SI, the use of the same label for multiple structures in Figs. S4 and S5, gets a little confusing. Perhaps you could add a, b, c, etc. labels after the numerical labels.
The footnote to Table S3 refers to reaction Scheme 2. Should it instead refer to reaction Scheme 3?