the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Comment to "Detecting supramolecular organic nanoparticles during heat wave by Zhang et al."
Abstract. Atmospheric new-particle formation is a major source of aerosol particles that influence air quality, cloud properties, and climate. Understanding the molecular mechanisms governing the initial steps of particle formation is therefore essential for accurately representing aerosol formation and its climatic effects in atmospheric models. Zhang et al. (2026c) recently proposed that hydrogen-bond–driven self-assembly of neutral carboxylic acids is a spontaneous and ubiquitous atmospheric new-particle formation mechanism. If correct, this would represent a fundamental shift in the current understanding of atmospheric nucleation.
In this comment, we show that the proposed nucleation mechanism is not adequately supported by the observations or theoretical analysis presented in Zhang et al. The reported particle-composition measurements lack sufficient validation to establish the proposed molecular composition of the smallest particles and do not directly constrain the molecular processes responsible for the earliest stages of particle formation. Furthermore, using the thermodynamic data reported by Zhang et al., we demonstrate that cluster evaporation overwhelmingly exceeds growth by molecular collisions, resulting in negligible particle-formation rates under the reported atmospheric conditions. These kinetic and thermodynamic analyses demonstrate that hydrogen-bond–driven clustering of neutral carboxylic acids cannot explain the reported observations and is unlikely to represent an atmospherically relevant new-particle formation mechanism.
- Preprint
(1003 KB) - Metadata XML
- Corresponding article
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-3785', Jonas Elm, 07 Aug 2026
-
AC1: 'Reply on RC1', Nanna Myllys, 10 Nov 2026
We thank the referee for the positive feedback.
We have incorporated the suggestions into the text as 1) added “studied” neutral carboxylic acids and 2) “And at higher temperatures such as 311 K, clusters are even less stable due to increased entropy penalty as dG=dH–TdS.”
Citation: https://doi.org/10.5194/egusphere-2026-3785-AC1
-
AC1: 'Reply on RC1', Nanna Myllys, 10 Nov 2026
-
RC2: 'Comment on egusphere-2026-3785', James Brean, 10 Aug 2026
In February 2026, Zhang et al. published a study in Science reporting frequent NPF during a central Texas heatwave. Using size-resolved measurements of aerosol particles down to 3 nm they measured multifunctional acids dominating the mass, with trace sulfuric acid and no detectable amines except in larger particles. On the basis of quantum chemical calculations, they proposed that hydrogen-bond driven self-assembly of these multifunctional acids drove their observed nucleation. This pathway was argued to have no free-energy barrier, and was not volatility-limited, allowing it to proceed in the hot Texan summer.
They further proposed this as a common mechanism for new particle formation across diverse tropospheric conditions. If correct, this would throw the acid-base and HOM pathways established by a large body of field and laboratory work into question.
This present comment addresses the core limitations of Zhang et al. Zhang et al. base their claim of spontaneous self-assembly on the negative ΔG for successive monomer additions, computed at a reference pressure of 1 atm. However, atmospheric monomer partial pressures are eight to ten orders of magnitude below that, and Myllys et al. show that once the free energies are corrected to ambient concentrations a barrier appears. Zhang et al. also compute forward collision rates without ever computing the reverse rates that their own thermochemistry implies. In this comment, Myllys et al. give this more rigorous treatment, and find that evaporation beats collision by roughly three orders of magnitude for every aggregate from A2 to A6.
The authors also raise some important comments about the experimental evidence, although I feel this section of the comment could be fleshed out slightly. I recommend this comment for publication. I provide a number of comments on the present text and also a couple of extra points for consideration below:
General comments
1) The comment quotes Zhang et al.'s dimerisation free energy as -7.7 (DFT) and -7.6 kcal/mol (DLPNO//DFT). The erratum (Zhang et al., 2026b) revises the value to -6.1 kcal/mol. The comment acknowledges the erratum but does not appear to propagate it, and the -7.6 value is still quoted at line 32 without qualification. I don’t think this would change the major conclusions of this comment if either energy was used, but it is important for consistency.
2) In Zhang et al., only A2 received a DLPNO-CCSD(T) single point correction; A3 to A6 are ωB97X-D only. The larger species were constructed by adding monomers to the previously optimised species. Is it possible to comment on this approach for larger species for us non-specialists in these calculations?
2) Lines 158 of the comment onwards make a very important point, which is that a measurement of composition at 3 to 10 nm carries almost no information about nucleation. This could do with a short quantitative treatment. Taking a critical cluster of mobility diameter 1.5 nm and assuming spherical particles of uniform density, the fraction of the mass of a larger particle contributed by the original cluster is extremely small. Approximately 10% for a 3 nm particle, and below 1% for a 10 nm particle. In other words, 90 % of the mass of Zhang et al.'s smallest measured particles, and >99 % of the mass at 10 nm, was acquired after nucleation. Even a perfectly accurate composition measurement at 3 nm is therefore overwhelmingly a measurement of the condensed growth material, not of the nucleating species. More damning is that that the supplement states that the nano-DMA was run at low resolution, with 5 SLPM sample against 15 SLPM sheath, passing a roughly normal distribution of standard deviation equal to 0.3 times the selected size. Because mass scales with the cube of diameter, that means that the mass-weighted effective diameter of the measured 3 nm particles is in fact higher than 3 nm, pushing these fractions down. These would go down even further accounting for the appropriate mobility-to-mass-diameter calculations!
3) Myllys et al. highlight correctly that the authors use CO3-/CO4- to detect acids. This is an ionisation scheme that is highly selective for acids, and Zhang et al. then find that their particles are made of acids, and conclude that nucleation is dominated by acids. This is perhaps slightly circular reasoning. This, as well as the claim that this was unambiguously verified using H3O+ ionisation (which is itself highly selective), could be discussed in more detail.
Additional considerations
The following go beyond the scope of a review of this comment, and stray into reviewing Zhang et al. themselves. I include them because they may be useful during the open discussion, or if the authors choose to expand their treatment of the experimental evidence. None are conditions of publication.
1) A major consideration is that Zhang et al. propose a new mechanism without sufficiently disproving the much more obvious mechanism of sulfuric acid-base nucleation. The authors state that: “For the three strongest NPF events … the average temperature … was] 32°” and then later “An unexpected outcome from our field observations is the high temperatures (approaching 40°C) at which frequent NPF events take place. Such a finding cannot be explained by presently known theories for NPF”. This is simply untrue.
- First, H2SO4-base nucleation can proceed at these temperatures and is routinely observed to do so. The obvious example is Beijing, where summertime NPF frequently occurs at >30 degrees, with formation rates very similar to those observed by Zhang et al. (e.g., Deng et al.: https://doi.org/10.1021/acs.est.0c00808).
- Second, there are already reports of H2SO4-amine-OOM nucleation in Houston. These measurements can be found at Tiszenkel et al. https://doi.org/10.1038/s43247-025-02310-4.
- The authors argue that their observed SO2 was below 0.5 ppb (the detection limit of their instrument), but 0.5 ppb is sufficient to produce 1e6 to 1e7 /cm3 H2SO4 at the low CS and high insolation that they observe, which is once again in-line with Deng et al.'s summertime Beijing observations.
- In their particle phase measurements they measure H2SO4 comprising roughly more than zero, but less than 20% of total particle mass. This is once again completely consistent with the H2SO4:organic ratios observed in the most relevant TDCIMS dataset of Li et al. in Beijing (https://doi.org/10.1021/acs.est.2c01566), where, once again, the nucleation is well known to proceed through an acid-base mechanism, with HOM condensation dominating the growth.
2) Zhang et al. flash-heat their sample to 350 degrees in the span of three seconds. An additional consideration is therefore thermal desorption artefacts. It is well documented in the FIGAERO literature that accretion products and other very low volatility species are detected as higher volatility compounds after heating, because they decompose (e.g. Lopez-Hilfiker et al. 2015; doi:10.5194/acp-15-7765-2015). At 350 °C I am not sure that Zhang’s measured species survive intact. As the collected mass was evaporated into the ID-CIMS within ~3 s, there is no temperature ramp data, so no thermograms exist and the usual diagnostic for separating direct desorption from decomposition is unavailable. Further to this, Myllys et al. highlight that the sensitivity calibration was performed by depositing the small acids onto the collector, which itself provides no information about whether these are instrumental artefacts. I would caveat that we would not necessarily expect larger species to produce multiacids upon heating, but this is still a major uncertainty.
3) The measurements of Zhang et al. used a triple quadrupole CIMS instrument. They do not state the mass resolution of the instrument, but a unit-mass resolution instrument is implied by their survey scans followed by single ion monitoring. If so, then how can the Zhang et al. be so sure of their molecular assignments? CIMS measurements will often show many overlapping peaks, even in this <300 m/Q region. If there is indeed major thermal fragmentation of the measured molecules, then many of these fragmentation products will overlap with the masses of the multiacids.
Technical comments:
In this comment there are three Zhang references which are often incorrect, and out of order. E.g., the statement “Zhang et al. (2026a) concludes that”… points towards the Zenodo dataset.
Line 48: “For GlA–ToA complex,…” is a broken sentence.
Line 60: “In the DLPNO calculations”?
Line 61: “even Zhang et al. performed some…” should this say “even though Zhang.”…? Similarly it then goes on to say “we found 0.6 kcal/mol lower DFT free energy”. Should this say “We found a structure 0.6 kcal/mol lower in energy” or similar?
Line 66: Zhang et al conclude rather than concludes?
Line 100: I think the kD units are wrong.
Line 108: “The original article” rather than “an original article”?
Line 127: “which leads to overestimated collision and NPF rates”. This could be clearer as it’s not clear whether the overestimation is relative to measurement or a physically more justifiable calculation. Perhaps “which yields an upper bound…”?
Line 141: Maybe it would be better to state that it was done at lower detection limits, rather than higher sensitivities than previous achieved.
Line 147: A small comment that does not majorly change any conclusions: here the authors presume a density of 1 g/cm3, whereas Zhang presume 1.23 g/cm3. This is a generous choice, so I’d state that explicitly as you do for the evaporation calculations.
Line 148: where does the 10% charging and 10% penetration come from? Zhang et al. state that their polonium charger produces efficiencies of 10 to 70%. Presumably 10% is for the smallest diameters as charging efficiency decreases with decreasing size. What about the penetration?
Citation: https://doi.org/10.5194/egusphere-2026-3785-RC2 -
AC2: 'Reply on RC2', Nanna Myllys, 10 Nov 2026
We thank the referee for the positive and comprehensive reply.
General comments:
- Thank you for pointing this out, the revisited DLPNO value is updated. Furthermore, we note that the dimer evaporation rate corresponding to the higher-level theory DLPNO//DFT ΔG value of -6.1 kcal/mol is approximately an order of magnitude higher than the DFT-based evaporation rate, suggesting that the dimer is even less stable (on line 105 and Fig. 2).
- The applied computational workflow includes various issues. Most importantly, the configurational sampling does not take into account the monomer flexibility, and therefore, none of the structures seem to be a global minimum, some of them not even a minimum. Thus, using DFT values for A3 to A6 is not the major issue. While DFT indeed is a lower level of theory, applying a higher-level electron correlation method does not improve the thermochemistry predictions when minimum structures have not been found.
- We thank the reviewer for these important points about measuring nucleation vs. condensation. We agree that there are several caveats regarding the analysis of ambient nanoparticles that were not well-described by Zhang et al. These include poor charging efficiencies, high diffusional losses, exceptionally short lifetimes of a nm-sized particles in urban settings (on the order of tens of minutes) and the broad DMA transfer function that skews sampled mass towards larger particles. All of these should have been addressed in Zhang et al., and we implicitly included these concerns in our original manuscript when asking for more details on the "transmission through the inlet and classification device.” Regardless of these concerns, we also agree with the reviewer that an irrefutable fact is that information on 3-10 nm particles may say little about the nucleation process. Following the reviewer’s suggestion, we have made the following change to the text (line 186): "More than 90% of the compounds detected in 3–10 nm particles, and more than 99% in 10 nm particles, may have accumulated through condensation after the initial nucleation process. Particle-phase chemistry may also have altered the particle composition after condensation.”
- We thank the reviewer for this comment to use CO3-/CO4- to detect acids. While we mention this somewhat in the original text with the statement “The analytical selectivity of the employed TD-ID-CIMS approach toward acidic compounds complicates interpretation of the measurements in terms of volatility and nucleation chemistry,” we agree that the issue deserves more detail. We have added a paragraph to the revised manuscript. It explains that the CO3⁻/CO4⁻ scheme is selective for acids, so an acid-dominated particle composition is partly built into the detection method. It also explains that the H3O⁺ gas-phase measurements cannot confirm the result independently, because they have a similar bias against low-volatility, highly oxygenated organics.
Additional considerations:
- We agree that more comprehensive analysis of the possible NPF mechanisms could be made. Although a detailed evaluation of these mechanisms is beyond the scope of our comment, we have added relevant references to the manuscript to provide additional context for the interpretation of the observations. We specifically note a recent study by Halonen (PNAS, 2026), which showed that frequent NPF under heat-wave conditions can be explained by high concentrations of sulfuric acid–dimethylamine heterodimers.
- We agree that the high evaporation temperatures introduce additional uncertainty in the interpretation of the measured particle composition. As discussed in our response to comment 3, we have expanded the manuscript to address the potential thermal decomposition of labile functional groups during particle evaporation, which could alter the molecular composition prior to ionization. Additional characterization would be needed to establish unambiguously the abundance and composition of acidic compounds in the nanoparticles.
- We agree with point concerning triple quadrupole CIMS instrument and added the following statement to the comment: “The mass resolution of the triple quadrupole is not reported, but the mode of operation suggests unit-mass resolution. Furthermore, no MS/MS spectra are presented to support the molecular assignments. Therefore, the assignment of a signal to a specific molecular structure cannot be established from the measured mass alone. At unit-mass resolution, each signal is the sum of all ions at a given nominal m/z, including structural isomers and compounds of different elemental composition. Assigning a signal to a specific acid therefore cannot be justified from the measured signal alone.”
Technical comments:
- We thank for suggestions to improve the text. The manuscript has also been revised according to given technical comments.
Citation: https://doi.org/10.5194/egusphere-2026-3785-AC2
-
CC1: 'Comment on egusphere-2026-3785', Steven Girshick, 02 Sep 2026
Publisher’s note: this comment is a copy of RC3 and its content was therefore removed on 4 September 2026.
Citation: https://doi.org/10.5194/egusphere-2026-3785-CC1 -
AC3: 'Reply on RC3', Nanna Myllys, 10 Nov 2026
We thank the reviewer for their positive feedback on our manuscript and for recognizing the importance of the thermodynamic analysis and the discussion of the experimental limitations of Zhang et al. We appreciate the reviewer's support for publication.
- The ‘corrected quantum-chemical data’ refers to Gibbs free energies that are calculated for global minimum monomer structures. The DLPNO single-point energies can be computed on top of the DFT structures to provide more accurate Gibbs free energies. However, Zhang et al. have only calculated this correction for the dimer structure. The revisited DLPNO value is updated in line 32.
As the evaporation rate constants (Eq. 3) depend on the difference between the Gibbs free energies of the evaporating and the resulting cluster, they must apply the same quantum chemistry method for all clusters. In the kinetic simulation, we can thus only apply the DFT values as no DLPNO corrections are provided for clusters consisting of three to six monomer units; combining evaporation rates obtained by different levels of theories would be simply erroneous. However, we have added the DLPNO-corrected dimer evaporation rate in Fig. 2 to demonstrate that the used DFT-based dimer evaporation rate can be considered a lower-bound estimate that favors aggregate formation.
As the computational execution in Zhang et al. was problematic (none of the structures was a global minimum), optimization failed for some structures (vibrations include imaginary frequencies), and single point energies were originally calculated for wrong structures, we believe that ‘technical problems’ is a warranted criticism. The last point on wrong structures is the issue that was corrected by the authors in the erratum, after we pointed out that the presented values are not consistent (personal communication).
- The Gibbs formation free energy of a cluster depends on the enthalpy and entropy as dG=dH – TdS. The hydrogen-bond formation is exothermic process leading to a negative dH value. However, the entropy is decreased when a cluster forms from two (or more) momoners , i.e. dS is also (strongly) negative. We note that the dS contribution comes mainly from the loss of translational and rotational degrees of freedom, and is thus roughly similar for all atmospherically relevant clustering reactions (with the same number of molecules involved). For example, Elm et al. (JAS, 2020) compared 45 different representative clustering reactions, and found that their dS varies between about –30 and –40 cal/Kmol. The higher the temperature, the larger the impact of the entropy term. Therefore, dG becomes substantially higher for any clustering process when temperature increases. (Using the dS range from Elm et al., a 10 K increase typically increases dG by about 0.3-0.4 kcal/mol.)
Since evaporation rate constants are exponentially dependent on the Gibbs free energies, they generally exhibit a significantly stronger temperature dependence than collision rates (which typically depend on the square root of the temperature). We agree that increasing temperature can enhance the gas-phase production of NPF precursors. In contrast to the very modest effect of temperature on collision rates, the temperature sensitivity of precursor production may indeed compete with the increasing evaporation rate. (This is also observed in the actual atmosphere – NPF rates are not typically the highest in the very coldest conditions.) However, assessing these effects is beyond the scope of the study. Instead, we apply the highest reported organic acid concentrations of ppb-levels to assess the upper bound of the theoretical particle formation rates (at a constant acid concentration).
The following clarification has been added (line 37): “And at higher temperatures such as 311 K, clusters are even less stable due to increased entropy penalty as dG=dH–TdS.” We have also modified the sentence in question to explicitly acknowledge that the results apply to constant OH (line 158):: “... the NPF rate (at constant [OA]) becomes even lower at elevated temperatures, including heat-wave conditions”
- Typo in line 108 is corrected.
- Higher concentrations and/or lower temperatures indeed make nucleation more efficient. Thus, there can possibly exist a scenario, such as an industrial accident, in which carboxylic acid nucleation might become locally non-negligible. Such event is not, however, addressed in the original study by Zhang et al. who instead suggest to have identified a widespread atmospheric NPF pathway.
Citation: https://doi.org/10.5194/egusphere-2026-3785-AC3
-
AC3: 'Reply on RC3', Nanna Myllys, 10 Nov 2026
-
RC3: 'Comment on egusphere-2026-3785', Steven Girshick, 03 Sep 2026
Review of Myllys et al., “Comment to ‘Detecting supramolecular organic nanoparticles….”
The article presents convincing arguments as to why the conclusions reported by Zhang et al. are of doubtful validity. The most important point is made in Fig. 3, which shows that the Zhang et al.-calculated standard Gibbs free energies of formation, when corrected to account for an upper-bound value of the actual concentration of organic acid molecules in the atmosphere, indicate that the free energy changes for cluster growth up to at least size 6 are all positive, i.e. a thermodynamic barrier exists for nucleation, contrary to the assertion of Zhang et al. This implies that evaporation rates from dimers (as well as from the larger clusters shown) will exceed aggregation rates under relevant conditions, so Section 2.2.1 isn’t really necessary, though I do not object to its inclusion.
Section 2.4, a discussion of experimental limitations of Zhang et al., is also convincingly argued and makes an essential companion to the theoretical argument.
Especially considering that the Zhang et al. article was published in Science, the critique presented here is important, and I recommend publication.
I have only a few minor concerns and suggestions.
—In the caption of Fig. 4, it is stated that “the use of corrected quantum-chemical data would further reduce the simulated NPF rates.” Does this refer to the corrected dimerization energy reported in the erratum published by Zhang et al? If so that should be referenced; if not it should be clarified. And I am puzzled as to why the corrected value published in the erratum was not used in the calculations in this article. It is stated on line 60 that “there seem to be technical problems in DLPNO calculations which have been corrected in an erratum.” Perhaps, but publication of an erratum does not necessarily imply that there were technical problems in the original calculation, so that statement is an unwarranted criticism.
—In lines 135-136 it is stated that “Because cluster stability decreases with increasing temperature…the [new particle formation rate] becomes even lower at elevated temperatures, including heat-wave conditions” (which are emphasized in the Zhang et al paper). However this statement is not necessarily correct, for several reasons: the temperature-dependence of the Gibbs free energy of carboxylic acid is not explored in this work; increasing temperature increases collision rates; and increasing temperature might increase the rate of carboxylic acid formation by chemical reactions, thereby increasing its concentration. Therefore I recommend deleting the statement, unless it can be more strongly supported.
—Very minor, but in line 108, “in contrast to the claim made in an original article,” the indefinite article “an” should obviously be changed to the definite article “the”.
—Not required, but the authors might consider addressing the question, based on Zhang et al.’s calculation of the standard free energy of forming the GIA-ToA complex, how high would the carboxylic acid concentration have to be to make nucleation barrierless? If I am not mistaken, from equation (5) one finds that a concentration of ~2 ppm would be required, as opposed to the ~2 ppb value (line 101) which evidently is the highest reported total organic-acid concentration in the atmosphere. Might there be conditions (for example, an industrial accident) that could produce ~ppm concentrations of OA, and if so would the mechanism proposed be Zhang et al be plausible under those conditions?
Citation: https://doi.org/10.5194/egusphere-2026-3785-RC3 -
AC3: 'Reply on RC3', Nanna Myllys, 10 Nov 2026
We thank the reviewer for their positive feedback on our manuscript and for recognizing the importance of the thermodynamic analysis and the discussion of the experimental limitations of Zhang et al. We appreciate the reviewer's support for publication.
- The ‘corrected quantum-chemical data’ refers to Gibbs free energies that are calculated for global minimum monomer structures. The DLPNO single-point energies can be computed on top of the DFT structures to provide more accurate Gibbs free energies. However, Zhang et al. have only calculated this correction for the dimer structure. The revisited DLPNO value is updated in line 32.
As the evaporation rate constants (Eq. 3) depend on the difference between the Gibbs free energies of the evaporating and the resulting cluster, they must apply the same quantum chemistry method for all clusters. In the kinetic simulation, we can thus only apply the DFT values as no DLPNO corrections are provided for clusters consisting of three to six monomer units; combining evaporation rates obtained by different levels of theories would be simply erroneous. However, we have added the DLPNO-corrected dimer evaporation rate in Fig. 2 to demonstrate that the used DFT-based dimer evaporation rate can be considered a lower-bound estimate that favors aggregate formation.
As the computational execution in Zhang et al. was problematic (none of the structures was a global minimum), optimization failed for some structures (vibrations include imaginary frequencies), and single point energies were originally calculated for wrong structures, we believe that ‘technical problems’ is a warranted criticism. The last point on wrong structures is the issue that was corrected by the authors in the erratum, after we pointed out that the presented values are not consistent (personal communication).
- The Gibbs formation free energy of a cluster depends on the enthalpy and entropy as dG=dH – TdS. The hydrogen-bond formation is exothermic process leading to a negative dH value. However, the entropy is decreased when a cluster forms from two (or more) momoners , i.e. dS is also (strongly) negative. We note that the dS contribution comes mainly from the loss of translational and rotational degrees of freedom, and is thus roughly similar for all atmospherically relevant clustering reactions (with the same number of molecules involved). For example, Elm et al. (JAS, 2020) compared 45 different representative clustering reactions, and found that their dS varies between about –30 and –40 cal/Kmol. The higher the temperature, the larger the impact of the entropy term. Therefore, dG becomes substantially higher for any clustering process when temperature increases. (Using the dS range from Elm et al., a 10 K increase typically increases dG by about 0.3-0.4 kcal/mol.)
Since evaporation rate constants are exponentially dependent on the Gibbs free energies, they generally exhibit a significantly stronger temperature dependence than collision rates (which typically depend on the square root of the temperature). We agree that increasing temperature can enhance the gas-phase production of NPF precursors. In contrast to the very modest effect of temperature on collision rates, the temperature sensitivity of precursor production may indeed compete with the increasing evaporation rate. (This is also observed in the actual atmosphere – NPF rates are not typically the highest in the very coldest conditions.) However, assessing these effects is beyond the scope of the study. Instead, we apply the highest reported organic acid concentrations of ppb-levels to assess the upper bound of the theoretical particle formation rates (at a constant acid concentration).
The following clarification has been added (line 37): “And at higher temperatures such as 311 K, clusters are even less stable due to increased entropy penalty as dG=dH–TdS.” We have also modified the sentence in question to explicitly acknowledge that the results apply to constant OH (line 158):: “... the NPF rate (at constant [OA]) becomes even lower at elevated temperatures, including heat-wave conditions”
- Typo in line 108 is corrected.
- Higher concentrations and/or lower temperatures indeed make nucleation more efficient. Thus, there can possibly exist a scenario, such as an industrial accident, in which carboxylic acid nucleation might become locally non-negligible. Such event is not, however, addressed in the original study by Zhang et al. who instead suggest to have identified a widespread atmospheric NPF pathway.
Citation: https://doi.org/10.5194/egusphere-2026-3785-AC3
-
AC3: 'Reply on RC3', Nanna Myllys, 10 Nov 2026
-
CC2: 'Comment on egusphere-2026-3785', Renyi Zhang, 02 Oct 2026
Response to the comments by Myllys et al.
In the paper “Detecting supramolecular organic nanoparticles during heat wave”, we observed frequent new particle formation (NPF) under heatwave conditions (Zhang et al., 2026). The main results and findings from our work include: (1) sized-resolved chemical analysis of nanoparticles from 3 to 50 nm identified carboxylic acids (diacids and triacids) as the dominant constituents; (2) the mass fraction of sulfuric acid increased with size, while amines were detected only in larger particles (> 20 nm); (3) the mass fraction of non-acidic organics increased with size, including oligomeric products formed from heterogeneous reactions; (4) nucleation mode particle (< 25 nm) exhibited distinct physicochemical properties, notably with low hygroscopicity and density characteristic of organics-rich nanostructures; (5) the derived formation and growth rates for nucleation-mode particles based on the measured concentrations and size increase exhibited a correlation with the measured concentration of gaseous organic acids and the peak concentration of aromatics in the morning, indicating their critical role in NPF and their origins from photooxidation of anthropogenic and biogenic emissions; (6) we evaluated one plausible pathway using quantum chemical calculation, demonstrating the potential of diacids and triacids with multiple branches for cluster growth and stabilization; and (7) using a cluster model (i.e., ABCluster), we estimated the structures for nanoparticles close to 3 nm and derived a low density similar to our field measurements. On the basis of our field observations and theoretical evidence, we proposed that our observed NPF during heatwave can be explained by a mechanism in which stable supramolecular nanoparticles are produced from self-assembly of dominantly carboxylic acids via hydrogen-bond interactions.
In the comments by Myllys et al. (2026), the authors asserted that our proposed nucleation mechanism via self-assembly of organic acids was not supported by our theoretical analysis. Specifically, they argued that their calculated cluster evaporation overwhelmingly exceeds growth by molecular aggregations, and hydrogen-bond–driven clustering of carboxylic acids is unlikely to represent an atmospherically relevant NPF mechanism. They performed NPF simulation using the Atmospheric Cluster Dynamics Code (ACDC) for a pseudo-1-component system and derived a NPF rate of several orders of magnitude smaller than our measured rates. In addition, they raised several technical issues regarding our particle-composition measurements for the smallest particles and the constrain for the molecular processes responsible for the earliest stage of NPF.
Notably, the cluster-kinetics calculations by Myllys et al. contradicted not only our field evidence that organic acids were the dominant constituents of 3–30 nm particles but also other field measurements detecting gas-phase organic diacid dimers (with an average dimer-to-monomer ratio of ~0.25, Fang et al., 2020). While particle-phase composition for 3-30 nm particles and dimer observations do not uniquely identify the critical nucleus, those atmospheric measurements indicate that organic-acid clustering and growth are important under ambient NPF conditions. Below we address specifically the key aspects of the concerns by Myllys et al. (2026).
- Non-ideal behavior of organic acids. Historically, it is well established that organic acids exhibit strong intermolecular attraction and form hydrogen-bonded dimers in the vapor phase (Derawi et al., 2004). The detailed-balance framework and concentration-dependent Gibbs free-energy corrections used by Myllys et al. (2026) were derived for ideal gases. An evaluation of organic-acid-mediated NPF should consider cluster-specific interactions and effective activities. More importantly, field observation of dimer/cluster abundances is critical to constrain the modeling of NPF involving organic acids.
- Stability and evaporation of organic acid clusters. Myllys et al. evaluated our mechanism using a detailed-balance framework and ACDC, which assume thermalized clusters and derive the evaporation rate from the standard-state ΔG. This simplified evaporation treatment does not explicitly resolve energy accommodation, collisional cooling, or intramolecular vibrational energy redistribution during cluster formation, all of which influence whether a chemically excited aggregate survives long enough for further growth. Such omissions can especially bias evaporation for organic acid clusters with many vibrational modes. For example, the GlA-ToA and C6 (5 organic acids and one H2SO4) clusters possess 99 and 315 vibrational modes, respectively, while a H2SO4-NH3 cluster only has 27. Cluster size and vibrational states significantly influence unimolecular dissociation (Baer and Hase, 1996), and a master‑equation/Rice–Ramsperger–Kassel–Marcus (RRKM) treatment would be more appropriate than equilibrium‑derived evaporation. Molecular collision represents the driving force for unimolecular dissociation according to the Lindemann mechanism, and clusters of organic acids likely have a lower decomposition rate because the collision kinetic energy is distributed among many vibrational states. Also, the detailed balance framework calculated the evaporation rate at the high-pressure limit, where thermal equilibrium is assumed to be instantaneous. Without explicitly tracking the collision excitation of the clusters, this assumption may lead to overestimated evaporation rates for organic acid clusters. Finally, the presence of multiple monomer conformers under ambient conditions may further alter the cluster free energies and hence affect the calculated evaporation rates.
- Concentration-corrected free energy change for cluster formation. Myllys et al. converted the standard-state free energies to concentration-corrected ΔG using the ambient monomer vapor concentration (their Eqs. 4–5) to assess the clustering capability. However, this treatment omits independent constraints from cluster concentrations and does not reflect an open, non-equilibrium nature of atmospheric NPF, where clusters are sustained by continuous forward aggregations. For comparison, even for strongly stabilizing acid–base systems such as H2SO4–DMA with a standard ΔG of -13.2 kcal mol-1 (Chee et al., 2021), applying the same correction at typical vapor concentrations for H2SO4 (0.1 ppt) and DMA (5 ppt) yields ΔGactual of +3.4 kcal mol-1, respectively, emphasizing that such corrected ΔG values should not be interpreted alone to exclude cluster formation under non-equilibrium atmospheric conditions. Also, an evaporation rate exceeding the collision rate does not necessarily preclude a net NPF growth. Typically, only a small fraction of a monomer pool is needed, e.g., a survival probability of the order of 10−7 converts ~1010 cm−3 monomers into ~103 cm−3 freshly nucleated particles.
- Multiple plausible pathways for the initial stage of NPF. In our paper, we presented a qualitative assessment using quantum chemical calculation, i.e., initial dimerization via association between glutaric acid and toluic acid, followed by successive addition of three organic acids and one sulfuric acid. Our results showed that diacids and triacids, which possess multiple branches for cluster growth and stabilization, readily engage in double hydrogen-bond formation, a process that is jointly augmented by dipole-dipole interaction and electrostatic attraction. In reality, there likely exist other stable formation paths for the initial dimerization involving multifunctional organic acids. We agree that there is a need to systematically assess the formation and growth for molecular clusters of organic acids with distinct functionality (i.e., mono-, di-, and tri-acids with and without the carbonyl and hydroxyl groups), along with H2SO4, H2O and base species (NH3 and amines). Also, H2SO4 was detected at small mass fractions in 3–10 nm particles and increased with size, suggesting possible participation in stabilization under our measured conditions. Nevertheless, our observation evidence of dominate organic acids in 3-30 nm particles is consistent with the mechanism of molecular assembly of organic acids, along with other species such as H2SO4 and base species, via hydrogen-bond interactions.
- Forward association for organic acids. Estimating the forward association rate should account for anisotropic, long-range interactions (e.g., dipole–dipole and electrostatic), and orientation-dependent steric effects. Organic acids have substantial permanent dipoles and polarizabilities. The resulting attractions create anisotropic interaction potentials that increase the effective collision cross sections, adjusting molecular orientation, and enhance the association rate relative to the hard-sphere assumptions. Neglecting these effects likely underestimates the forward rate, while incorporating the kD value partially corrects for those effects.
- Steady-state assumptions. Nucleation rates (J) presented by Myllys et al. were calculated based on the steady-state assumptions for each cluster. Our measurements show that the early stage of NPF involves rapid evolution of particle size and number concentration, representing a multicomponent, open, and non-equilibrium state, with continuous production and removal of vapors, clusters, particles, and changes of other ambient conditions (Fig. 1 in Zhang et al.). Such conditions cast doubt on the applicability of steady-state assumptions as well as detailed balance framework to accurately assess the early stage of cluster formation and growth in the observed events.
- Uncertainty in nucleation-rate predictions. ACDC-derived nucleation rates are highly sensitive to the input evaporation rates (γ) and vapor concentrations. For example, a decrease of evaporation rate (γ) by 10 times results in an increase of J by 105 Also, doubling the acid concentration increases the J by 102 times. For organic acid clusters, estimates of γ may be biased by non-ideal behavior, energy accommodation, vibrational energy redistribution, detailed-balance assumption, high-pressure-limit approximations, hydration effects, conformer distributions, and alternative stable formation pathways. On the observational side, our gas-phase acid measurements likely represented a lower limit due to inlet wall losses, since larger multifunctional acids such as triacids were not detected (Zhang et al., 2026). Accordingly, we used the total measured gas-phase organic acids to correlate with formation rates rather than to predict absolute nucleation rates. Caution must be exercised when interpreting the ACDC outputs without assessing the propagation of these various uncertainties through the simulations.
- Detection of sub-10 nm particles. We would also like to clarify the TD-ID-CIMS measurements of sub-10 nm particles. First, the TD-ID-CIMS technique has been previously applied to the chemical analysis of 4–12 nm particles (Wang et al., 2010; Xu et al., 2014). The instrument in our study incorporated several features to improve the sensitivity toward nanoparticles, including rapid thermal desorption to increase the signal-to-noise ratio, multiple reagent-ion ionization methods to resolve different classes of compounds, and an ion drift tube to enhance ion-molecule reactions and ion transmission. Specifically, the combined use of negative (CO3-/CO4-) and positive (H3O+) reagent ions enable selective validation of acidic compounds as well as broader screening of organic compounds in the particle phase, while earlier studies relied primarily on O2- reagent-ion chemistry for sub-10 nm particles (Li et al., 2022). The detection sensitivity of sub-10 nm particles by TD-ID-CIMS in our measurements was determined from calibration using authentic organic acid samples, and the background signal was evaluated on non-NPF days with no observable organic-acids. These calibrations supported the capability of the instrument to resolve 3-nm particle composition. In addition, the TD-ID-CIMS operated at a relatively high sampling flow rate of 5 LPM and without a Nafion drying tube, whereas the SMPS system operated at 1.5 LPM. The measured particle penetration efficiency through the TD-ID-CIMS was higher than that of SMPS system. Moreover, to account for the rapidly evolving particle number concentration due to coagulation losses, the particle number concentration of freshly formed 3-nm particles collected by TD-ID-CIMS during an active NPF event was adjusted to be 10 times higher than that of 15-nm particle at peak number concentration detected by SMPS, as documented in the SI of Zhang et al. Therefore, the SMPS-derived number concentration likely represented a lower limit for the 3 nm particle concentration sampled by TD-ID-CIMS. Collectively, our combined evidence from the detection limits, mass closure, inlet transmission, event/non-event contrast, and consistent 3–30 nm measurements supported the conclusion that nucleation mode particles were dominated by organic acids.
In conclusion, our field observations provide strong evidence for self-assembly of carboxylic acids in the formation and growth of supramolecular nanoparticles. Myllys et al. (2026) presented a modeling analysis against our proposed nucleation mechanism without a thorough assessment of the validity and uncertainties of their thermodynamic framework for organic acids at atmospherically relevant conditions. Clearly, there is a need for further quantitative evaluation of the detailed molecular steps governing the cluster formation and growth for NPF. The traditional models should be used with caution when simulating organic-acid-mediated NPF. In our additional lab measurements, we produced sized-resolved nanoparticles from 3 to 300 nm from a mixture of multifunctional organic acid vapors using a flow reactor, supporting our field observations (a manuscript on this topic is under preparation). Molecular self-assembly is an important frontier in nanoparticle research because it is governed not only by volatility or bulk thermodynamic equilibrium, but also by molecular recognition and interactions, reactive-site availability, and detailed non-equilibrium kinetics (Wasio et al., 2014; Whitesides et al., 1991). Organic acids are well suited for self-assembly because the carboxyl groups form directional double hydrogen bonds and participate in dipole–dipole and electrostatic interactions, while diacids and triacids provide multiple sites for continued aggregate growth and stabilization. Atmospheric NPF occurs in an open, multicomponent, non-equilibrium environment with continuous photochemical vapor production, cluster formation/losses, and evolving particle-phase stabilization. Future work should therefore integrate laboratory experiments, field cluster/vapor measurements, and multicomponent non-equilibrium kinetic modeling to determine how the functionality, molecular interactions, temperature, and humidity regulate NPF.
References
Baer, T. and Hase, W. L.: The Dissociation of Small and Large Clusters, in: Unimolecular Reaction Dynamics, Oxford University Press, https://doi.org/10.1093/oso/9780195074949.003.0012, 1996.
Chee, S., Barsanti, K., Smith, J. N., and Myllys, N.: A predictive model for salt nanoparticle formation using heterodimer stability calculations, Atmos. Chem. Phys., 21, 11637–11654, https://doi.org/10.5194/acp-21-11637-2021, 2021.
Derawi, S. O., Zeuthen, J., Michelsen, M. L., Stenby, E. H., and Kontogeorgis, G. M.: Application of the CPA equation of state to organic acids, Fluid Phase Equilibria, 225, 107–113, https://doi.org/10.1016/j.fluid.2004.08.021, 2004.
Fang, X., Hu, M., Shang, D., Tang, R., Shi, L., Olenius, T., Wang, Y., Wang, H., Zhang, Z., Chen, S., Yu, X., Zhu, W., Lou, S., Ma, Y., Li, X., Zeng, L., Wu, Z., Zheng, J., and Guo, S.: Observational Evidence for the Involvement of Dicarboxylic Acids in Particle Nucleation, Environ. Sci. Technol. Lett., 7, 388–394, https://doi.org/10.1021/acs.estlett.0c00270, 2020.
Li, X., Li, Y., Cai, R., Yan, C., Qiao, X., Guo, Y., Deng, C., Yin, R., Chen, Y., Li, Y., Yao, L., Sarnela, N., Zhang, Y., Petäjä, T., Bianchi, F., Liu, Y., Kulmala, M., Hao, J., Smith, J. N., and Jiang, J.: Insufficient Condensable Organic Vapors Lead to Slow Growth of New Particles in an Urban Environment, Environ. Sci. Technol., 56, 9936–9946, https://doi.org/10.1021/acs.est.2c01566, 2022.
Myllys, N., Olenius, T., Kangasluoma, J., Stolzenburg, D., Kurtén, T., Rissanen, M., Smith, J., and Ehn, M.: Comment to “Detecting supramolecular organic nanoparticles during heat wave by Zhang et al.,” https://doi.org/10.5194/egusphere-2026-3785, 7 August 2026., 2026.
Wang, L., Khalizov, A. F., Zheng, J., Xu, W., Ma, Y., Lal, V., and Zhang, R.: Atmospheric nanoparticles formed from heterogeneous reactions of organics, Nature Geosci, 3, 238–242, https://doi.org/10.1038/ngeo778, 2010.
Wasio, N. A., Quardokus, R. C., Forrest, R. P., Lent, C. S., Corcelli, S. A., Christie, J. A., Henderson, K. W., and Kandel, S. A.: Self-assembly of hydrogen-bonded two-dimensional quasicrystals, Nature, 507, 86–89, https://doi.org/10.1038/nature12993, 2014.
Whitesides, G. M., Mathias, J. P., and Seto, C. T.: Molecular Self-Assembly and Nanochemistry: a Chemical Strategy for the Synthesis of Nanostructures, Science, 254, 1312–1319, https://doi.org/10.1126/science.1962191, 1991.
Xu, W., Gomez-Hernandez, M., Guo, S., Secrest, J., Marrero-Ortiz, W., Zhang, A. L., and Zhang, R.: Acid-Catalyzed Reactions of Epoxides for Atmospheric Nanoparticle Growth, J. Am. Chem. Soc., 136, 15477–15480, https://doi.org/10.1021/ja508989a, 2014.
Zhang, R., Li, Y., Zhao, J., Aridjis-Olivos, B., Zhao, L., Kowalewski, V., Kabir, M., Johnson, N. M., Nielsen, E. R., Brooks, S. D., Zhang, Y., Vedlitz, A., Porter, W., North, S. W., Li, W., Young, M. W., Seinfeld, J. H., Wang, Y., and Wang, Y.: Detecting supramolecular organic nanoparticles during heat wave, Science, 391, eady5192, https://doi.org/10.1126/science.ady5192, 2026.
Citation: https://doi.org/10.5194/egusphere-2026-3785-CC2 -
AC5: 'Reply on CC2', Nanna Myllys, 10 Nov 2026
We highly appreciate Prof. Zhang replying to our comment. We do not dispute that diacids may be involved in atmospheric NPF: we dispute only the specific claim that diacids alone can drive observed NPF. In cases where diacids are involved, the specific mechanism is likely related, for example, to the stabilization of sulfuric acid – base clusters.
Non-ideal behavior of organic acids:
The ideal gas approximation is only used in computing the translational entropy contributions (this same approach is also used in all studies computing clustering free energies, including the Zhang et al paper we are commenting on). No other component of the calculation requires or assumes ideal gas behavior. Indeed, the whole point of performing quantum chemical calculations on clusters is that they provide precisely the “cluster-specific interactions” mentioned here. “Effective activities”, on the other hand, is a macroscopic thermodynamic term not really appropriate or relevant to molecular-level investigation of cluster stabilities. Overall, we note that the detailed-balance framework (combined with quantum chemical free energies of sufficient quality, and reasonable estimates of collision rates) has been shown to yield NPF rates in good agreement with experiments, for example from the CLOUD chamber (see e.g. the Elm et al 2020 review, or the very recent review in Lian et al 2026). There are naturally several uncertainties and error sources related to this approach, but as discussed in more detail below, the vast majority of them tend to further decrease the NPF rates (not increase them).
Concerning dimer and cluster abundances, we completely agree that measured abundances of specific clusters are excellent constraints to NPF models. However, we note that the existence of organic acid dimers does not establish a sufficiently high rate of particle formation.
Stability and evaporation of organic acid clusters:
First, we note that the large number of vibrational modes is (for example based on the RRKM theory quoted here) an argument in favor of thermalization, not against it. Or in other words, it is precisely the very large number of vibrational modes mentioned here that make the “thermalized cluster” assumption valid. Having said that, prof. Zhang is naturally correct in their reply that energy non-accommodation and/or incomplete vibrational energy redistribution (in competition with “collisional cooling”) could in principle affect dissociation/evaporation rates, at least for the smallest (e.g. diacid dimer) clusters. However, these effects act unambiguously to increase evaporation/dissociation rates (see e.g. Kurtén et al. 2010, in which rather modest effects of energy non-accommodation were found for sulfuric acid dimers even with very conservative assumptions of energy redistribution). This is evident also from the original pressure-dependence work by Lindemann, as well as the more advanced approaches such as RRKM. See for example the discussion in section 2.2. of Wallington et al 2026 (“combination reactions”, please note that clustering reactions belong to this category). Quoting from there: “Starting from the high-pressure limit, the rate coefficients fall off with decreasing third body concentration [M]". Or in other words, the lower the pressure, the less effective the “combination reaction”, i.e. the clustering in our case. The high-pressure limit represents the maximum possible cluster formation rate.
Furthermore, the vastly different timescales of intramolecular vibrational energy redistribution and collisions of clusters with potentially condensing vapors imply that the clusters actually relevant for further growth must inevitably be thermalized. The mixing ratios of vapors relevant to NPF are, at best, on the order of ppb (with ppt or sub-ppt being more realistic estimate). Thus, clusters will collide with thermalizing bath gas molecules (N2 and O2) on the order of a billion times more frequently than with condensing vapors. Any effects of the type quoted in the reply will long have dissipated before e.g. a third diacid collides with a diacid dimer.
Concerning the inclusion of multiple monomer conformers, here we completely agree that it can affect free energies and evaporation rates. However, once again, the direction of the effect is the opposite of that implicitly argued in the reply: including multiple monomer conformers makes the cluster less stable. We note that including multiple cluster conformers does stabilize clusters. Thus, approximating the cluster free energy as a difference between the global minimum free energies contain some amount of error cancellation, and the sign of multiconfigurational correction depends on the amount and energies of local minima monomer and cluster structures. Attempts at quantifying this correction have typically yielded rather modest values, see e.g. Partanen et al. 2016, which is reasonable given that the free energy correction depends only logarithmically on the number of low-energy conformers.
Concentration-corrected free energy change for cluster formation:
We agree that the often-used “concentration-corrected” ΔG implicitly ignores the actual cluster concentration (as it involves correcting for the actual concentrations of the monomers, but not of the cluster product). The point of the correction is simply to illustrate that actual equilibria (or indeed non-equilibrium aggregations) depend also on reactant concentrations, not only on free energies (or the equivalent equilibrium constants) computed at a reference pressure (in this case, and in almost all similar studies, at 1 atm). Or in other words, a negative standard ΔG is a necessary, but very far from sufficient, criterion for effective NPF. The H2SO4-DMA case mentioned in the reply is actually a perfect illustration for this: in the given conditions (0.1 ppt H2SO4, 5 ppt DMA), the experimental NPF rate (e.g. from the CLOUD chamber) is on the order of 0.1...1 per cm3 and second: more than three orders of magnitude less than the kinetic collision rate of two H2SO4 molecules. Thus, even the strongly negative standard ΔG of -13.2 kcal mol-1 obviously does not guarantee kinetically limited NPF, demonstrating why a “concentration correction” is warranted.
We also completely agree that evaporation exceeding collision for the smallest clusters does not preclude NPF – this is, as demonstrated, the case for the above-mentioned H2SO4-DMA case, and indeed for all examples of actual “nucleation” in the literature (as the concept of nucleation by definition inherently involves a free energy barrier, meaning that the evaporation rate must exceed the collision rate for the initial clusters located behind the barrier). This is also not what we are claiming in our comment. Rather, we claim that for efficient NPF, the evaporation rate cannot exceed the collision rate by an arbitrarily large margin – some non-negligible fraction of clusters must survive for long enough to grow. We agree that the “1010 cm−3 monomers converting into 103 cm−3 freshly nucleated particles” values are roughly plausible estimates for the most abundant vapors involved in NPF processes occurring in the actual atmosphere. However, we note that the actual survival ratio of the clusters proposed in the Zhang et al article is considerably smaller than 10-7. As shown in Figure 2 in our comment, the survival ratio of each step from n=2 to n=6 molecules is smaller than a factor of 10-3, implying an overall survival ratio of much less than 10-12 just from n=2 to n=6, with no evidence of the odds improving much for the next few n values. Assessing the survival is the reason for performing the cluster kinetics simulations (Section 2.3), as the purpose of such model is to explicitly solve the net formation and growth for given aggregation and evaporation rates. As Fig. 4 demonstrates, the resulting formation rate of particles larger than ca. 1.5 nm is very low or negligible, demonstrating very low survival in the reported [OA] range.
Finally, we note that the original Zhang et al manuscript repeatedly claims that the proposed clustering process is “spontaneous”, and motivate this for example by the decrease in (standard) ΔG as a function of cluster size. Even if the survival probability were as high as the 10−7 value given in the reply, the claim of “spontaneity” would clearly still be incorrect.
To clarify the meaning of the concentration-corrected ΔG values, we have added the following sentence to the manuscript: “While positive actual Gibbs free formation energies do not categorically rule out a clustering mechanism from contributing to NPF, they unambiguously disprove the claims of "spontaneity" proposed in Zhang et al (2026).“
Multiple plausible pathways for the initial stage of NPF:
We agree both that organic acids are likely to participate in the initial steps of NPF, and that of all different (non-basic) organic functional groups present in the atmosphere, carboxylic acids are probably the most relevant to NPF. We also fully agree that there may exist a large variety of potentially competitive multicomponent NPF routes in the atmosphere, and that further systematic assessments are needed. However, the pathway presented in the original paper is not one of the potentially competitive NPF routes.
Forward association for organic acids:
We agree, and thus use the kD values provided by Zhang et al. in our kinetic simulation, thereby including this effect in our evaluation. Yet, the evaporation rates greatly exceed the association rates kD.
Steady-state assumptions:
We agree that cluster concentrations may not be at a time-independent steady state at ambient atmospheric conditions. While atmospheric conditions often vary on time scales longer than those of the cluster kinetics, thus supporting the generally applied approximation of steady-state nucleation, explicit description of the time evolution of cluster concentrations at given conditions can indeed provide more insight to the formation and growth dynamics. Previous studies addressing the validity of the steady-state assumption indicate that overall, this approach tends to overestimate the formation rate, resulting in too strong NPF events with too early onset (Yu, 2003; Olenius and Riipinen, 2016; Olenius and Roldin, 2022). Therefore, also this model aspect contributes to our evaluation giving an upper-limit assessment for the theoretical formation rate.
We have added a sentence to the manuscript to clarify the effect of non-equilibrium and/or non-steady-state phenomena on NPF rates (including both incomplete thermalization discussed in previous paragraphs, and deviations from steady state): “Non-equilibrium effects such as incomplete thermalization of clusters, or deviations of the clustering process from a steady state, would also tend to further decrease the formation rate. “
Uncertainty in nucleation-rate predictions:
As discussed above, our simulations provide an upper bound for nucleation rates based on the data provided in the original manuscript. The error sources listed in the reply generally tend to decrease, not increase, NPF rates.
It is correct that the results of cluster kinetics simulations (performed by the ACDC program or another similar model) are sensitive to the input rate constants. It can also be noted that the sensitivities vary strongly with the simulated chemical system and ambient conditions, and there are no general dependencies for how much a simulated formation rate changes upon e.g. given change in evaporation rates or vapor concentrations. It is not clear to which simulated system and conditions the statement “a decrease of evaporation rate (γ) by 10 times results in an increase of J by 105” refers to. However, we welcome quantitative assessments of the theoretical or computational aspects that can be expected to decrease the evaporation rates e.g. by a factor of 10, and how this affects predicted formation rates at given ambient conditions.
Detection of sub-10 nm particles:
We thank the authors for the clarification. However, the response does not resolve our concerns, and in some respects reinforces them.
First, the response confirms the likely detection bias. It describes the CO3⁻/CO4⁻ and H3O⁺ reagent ions as enabling “selective validation of acidic compounds as well as broader screening of organic compounds.” Both schemes are selective, and neither is sensitive to highly oxygenated, low-volatility organic compounds. Agreement between the two ion chemistries therefore cannot independently verify that acids dominate. It only shows that both detect acids.
Second, the response does not explain how a measurable signal can be obtained at 3 or 5 nm. The authors state that penetration through the TD-ID-CIMS was higher than through the SMPS, but no penetration efficiencies are reported, in the original paper or in the response, so this claim cannot be evaluated. The statement that the 3 nm number concentration “was adjusted to be 10 times higher than that of 15-nm particle at peak number concentration” is also unclear. The SI gives only a 3 h collection time, and the adjustment procedure and resulting concentrations are not documented. A ratio of 10 also implies that a small fraction of larger particles reaching the collector would dominate the collected mass, because a single 15 nm particle has the mass of 125 particles of 3 nm.
Most importantly, the mass balance does not close. The reported 0.5 ng at 3 nm corresponds to about 3 × 10¹⁰ particles. With 10% charging and 40% transmission and 3h of 5 lpm sampling, this requires an ambient 3 nm concentration of about 8 × 10⁵ cm⁻³ sustained over the 3 h collection. That is about 1.5 orders of magnitude above the reported peak N3–25 of ~2 × 10⁴ cm⁻³, which covers the entire 3–25 nm range. The same discrepancy, though smaller, applies at 5 and 10 nm.
Finally:
We appreciate the detailed response and agree that the role of multifunctional organic acids in NPF deserves further investigation. We also agree that hydrogen bonding, dipole–dipole interactions, and other intermolecular forces can contribute to the formation and stabilization of organic acid clusters. However, none of the points in this discussion resolves our concern that the proposed mechanism cannot produce the observed particle formation rates under the reported atmospheric conditions.
Citation: https://doi.org/10.5194/egusphere-2026-3785-AC5
-
RC4: 'Comment on egusphere-2026-3785', Anonymous Referee #4, 06 Oct 2026
This comment presents a substantial and convincing challenge to the mechanism of new-particle formation proposed by Zhang et al. (R. Zhang et al., “Detecting supramolecular organic nanoparticles during heat wave”, Science, 391, eady5192, 2026). Its central point is that favourable cluster-association free energies at a standard pressure of 1 atm do not establish spontaneous particle formation at atmospheric vapour concentrations. When the authors account for molecular concentrations and competition between collision and evaporation, the predicted particle-formation rate is several orders of magnitude below that observed. This remains true despite assumptions that favour particle formation, including the use of the original study’s comparatively stable cluster free energies and the total organic-acid concentration in a pseudo-one-component calculation.
The paper also criticises the experimental measurements made by Zhang et al., arguing that measurements of organic acids in 3–10 nm particles may establish their contribution to particle composition or growth but do not directly identify the molecules and interactions responsible for initial nucleation. Further, the comment argues that the reported collected mass for 3 nm particles appears difficult to reconcile with the measured particle concentrations.
In my view, the central argument of the comment is scientifically sound, and I recommend publication after minor revision. I agree that Zhang et al. make claims that extend substantially beyond the evidence presented. In particular, measurements of organic-acid-rich particles at 3 nm and above do not identify the molecular species responsible for nucleation, while negative standard-state free energies do not establish spontaneous or barrierless cluster formation at atmospheric vapour concentrations. The inference by Zhang et al. that this represents a common mechanism under diverse tropospheric conditions is therefore especially surprising and weakly supported.
Minor revisions1. The manuscript should distinguish between the dimerisation free-energy value originally reported by Zhang et al., the value corrected in the subsequent erratum, and the authors’ own recalculation. It should also state which values are used in each kinetic calculation.
2. The authors should ideally explain how the chemically distinct cluster sequence was mapped onto the pseudo-one-component ACDC calculation, which evaporation pathways were included, and how the 1.7 nm formation rate was defined.
3. The simulated formation rate is defined at 1.7 nm, whereas the observational comparison appears to concern particles in the 3–25 nm range. The authors should identify the observational quantity used and explain how it relates to 1.7 nm particle formation rate . If losses during growth mean that the formation rate at 1.7 nm must exceed that observed at larger sizes, the comparison is conservative and this should be stated explicitly.
4. The calculations in the paper strongly rejects the proposed mechanism under the reported conditions and probably under comparable warm tropospheric conditions. The claim that it cannot operate under any atmospherically relevant conditions is broader than what has been demonstrated in the comment as it stands and should be moderated or justified further, although probably only briefly.
5. The section on experimental issues could be clearer and more fully developed. In particular, the assumed 10% charging and transmission efficiencies should be identified clearly as illustrative rather than measured or well-established values. The authors should explain more precisely what can be concluded from the resulting mass-balance estimate.
6. The manuscript contains a number of grammatical errors. For example, “Zhang et al. concludes” should be “Zhang et al. conclude”; “It should be still noted” should be “It should still be noted”; and “in an original article” should presumably be “in the original article”. I also struggled at points to determine which Zhang paper, erratum or repository was being referred to. The citations and associated labels should be checked and applied consistently. I suggest using Zhang et al. (2026a) for the original Science paper.
Finally, I note that this is an unusual paper to review because its principal purpose is to challenge the central conclusion of a prominent paper rather than to present a new study. This warrants particularly careful scrutiny of the comment’s calculations and claims. However, critical examination of published claims is also an important part of the scientific process, particularly where the original article is prominent and advances a broad new mechanism. I therefore support the publication of the comment and the approach taken by its authors. I have read the discussion associated with the comment on EGUsphere, but I do not consider it my role in this review to adjudicate that wider exchange. My assessment is confined to the scientific content and presentation of the submitted comment.Citation: https://doi.org/10.5194/egusphere-2026-3785-RC4 -
AC4: 'Reply on RC4', Nanna Myllys, 10 Nov 2026
We thank the reviewer for the positive response.
1. The problems in original calculations are discussed in Section 2.1 and the revisited DLPNO value is updated on line 32. However, as only DFT values are provided for clusters up to six monomer units, we use the provided data set in all kinetic calculations in Sections 2.2 and 2.3, as discussed above in the reply to a similar comment from Reviewer 3.
2. The Gibbs free energy values and collision (“capture”) rate constants from the original article for clusters up to six monomer units (Table S9 by Zhang et al.) were used to solve the cluster kinetics equations (Eq. (6)). All monomers were lumped as OA by setting the total vapor concentration to [OA] values corresponding to the observations. The formation rate shown in Fig. 4 is defined as the total formation rate of clusters that consist of at least seven OA monomer units. This corresponds to the largest size that can be assessed based on the ΔG and kD data by Zhang et al., assuming that these >7-acid clusters do not evaporate. The size which this formation rate corresponds to has been updated to be approximately >1.5 nm instead of >1.7 nm, following the corrected aggregate sizes in the erratum to the original article (as stated in e.g. Fig. 5A by Zhang et al.). For completeness, Fig. 4 presents theoretical formation rates obtained by (1) only including monomer (single-molecule) additions and evaporations, and (2) including also all coagulation and fission processes involving two clusters. For the latter, collision rate coefficients between two clusters are calculated according to the hard-sphere model (Section 2.3) as there are no dipole-dipole capture rate coefficients (kD) available for them in the original article. Fission rates are calculated from the Gibbs free energies according to Eq. (3).
3. Zhang et al. report the experimental formation rate for particles of sizes between 3 and 25 nm, based on the measured particle number concentration 𝑁3-25 for this size class. As discussed above, the theoretical formation rate can be obtained for >1.5 nm particles. The reviewer is correct that the formation rate is expected to decrease with cluster/nanoparticle size. During their growth, the newly formed particles are scavenged on the surfaces of larger aerosols. The smallest particles in the nanometer size range have highest scavenging rates due to their high mobility. Therefore, the formation rate of particles surviving to a given size decreases with size, and J1.5 nm is generally expected to be higher than J3-25 nm. This is clarified in the revised manuscript.
4. This has been modified to read “or generally in the lower troposphere at comparable or higher temperatures”.
5. Our manuscript focuses primarily on the computational limitations of the study by Zhang et al., since these can be evaluated directly from the information provided and the underlying thermodynamic framework. Our assessment of the experimental results is necessarily more limited. The field measurements cannot be reproduced retrospectively, and the TD-ID-CIMS instrument used by Zhang et al. is a specialized instrument for which independent validation is not readily available. Consequently, our evaluation of the experimental findings must rely on the information provided by Zhang et al. and the extent to which the analytical methods and measurements are adequately characterized. This is particularly important because the study reports particle compositions and concentrations that extend beyond those demonstrated in previous studies using comparable instrumentation. Although advances in analytical methodology have repeatedly enabled important breakthroughs in atmospheric chemistry, such advances are generally accompanied by detailed descriptions of instrument performance, calibration, and validation. In the case of Zhang et al., the experimental description relies substantially on an earlier publication from the same research group, without sufficiently documenting how the analytical approach supports the unusually ambitious compositional and mechanistic conclusions presented here. We therefore focus on identifying uncertainties and limitations in the interpretation of the experimental results, rather than claiming to demonstrate specific measurement artifacts or errors.
We nevertheless agree that the assumed charging and transmission efficiencies should be identified more clearly as illustrative estimates rather than measured or well-established values. We have revised the discussion accordingly. The purpose of the mass-balance calculation is to evaluate whether the atmospheric particle concentrations reported by Zhang et al. are consistent with the instrument detection limit, given reasonable assumptions about particle charging and transmission. (Under the illustrative assumptions of 10% charging efficiency and 10% transmission efficiency, the calculation indicates that the expected signals from 3 and 5 nm particles would be below the reported instrument detection limit. For 10 nm particles, the conclusion is less definitive. Assuming more favorable efficiencies of 70% charging and 80% transmission, an atmospheric concentration of approximately 1.6 × 10⁴ cm⁻³ would be required to reach the instrument detection limit. This is comparable to the peak concentration reported for N₃–₂₅. However, because the DMA selects only a fraction of the particles within this broader size distribution, the concentration of particles actually transmitted to the instrument would likely be insufficient to reach the detection limit.) We have clarified these assumptions and the interpretation of the resulting mass-balance estimate in the revised manuscript (see our comments above to RC2, Comment 2).
6. Thank you for the suggestions, the manuscript text has been modified accordingly.
We appreciate your time and effort to review this comment.
Citation: https://doi.org/10.5194/egusphere-2026-3785-AC4
-
AC4: 'Reply on RC4', Nanna Myllys, 10 Nov 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 682 | 249 | 60 | 991 | 48 | 41 |
- HTML: 682
- PDF: 249
- XML: 60
- Total: 991
- BibTeX: 48
- EndNote: 41
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of “Comment to Detecting supramolecular organic nanoparticles during heat wave by Zhang et al."
Myllys et al comments on the recent 2026 paper entitled “Detecting supramolecular organic nanoparticles during heat wave” by Zhang et al, published in Science. Zhang et al proposed that self-assembly (hydrogen bonding) of carboxylic acids is a spontaneous process leading to new-particle formation, even during extreme heat events. The submitted comment present evidence of misinterpretation of the experimental and theoretical results, essentially falsifying the conclusions of Zhang et al.
When I first read the study by Zhang et al, my immediate thought was that this paper completely ignores the immense knowledge obtained on organic enhanced cluster formation during the past two decades.It is by now well-established that it is most likely covalently bound dimers (accretion products) that drive organic nucleation at ombient conditions. In particular, Figure 5 in the paper by Zhang et al is highly misleading, as it only presents the standard free energies giving the impression that the clusters are stable. Considering the evaporation rates, or calculating the equilibrium cluster concentrations, would immediately demonstrate that the reported clusters are not sufficiently stable to explain the proposed mechanism. Hence, I am pleased to see the authors submit this critical comment, as I believe it exposes an important issue with the published paper. The comment is written by a strong team of highly qualified professionals. The critical analysis is technically sound, and the comment can essentially be published as is. I only have a few minor optional comments for the authors below.
Comments
Line 12: “These kinetic and thermodynamic analyses demonstrate that hydrogen-bond–driven clustering of neutral carboxylic acids cannot explain the reported observations and is unlikely to represent an atmospherically relevant new-particle formation mechanism.”
To be on the safe side, perhaps state that this refers to the “studied neutral carboxylic acids” in Zhang et al. We have recently shown that tricarboxylic acids might contribute to NPF, so there might be very special cases where neutral tricarboxylic acids can actually nucleate (if they exist).
Line 33: “These ∆G values appear unrealistically favorable, …”
I had the exact same thought when I first read the study. This is substantially more favourable than what have previously been found for carboxylic acid-carboxylic acid interactions. In addition, while addressed on page 7, it might also be worth pointing out already here that the free energy analysis is carried out at 298.15 K, while the paper states the mechanism to be relevant for up to 311.15 K (up to 38°C). This would further destabilize the studied carboxylic acid-carboxylic acid clusters.