The authors present an observational study on PM2.5 aqueous extract photochemistry. They measured singlet oxygen and excited triplet states on PM2.5 collected in urban and rural sites in Hong Kong. Not much is known about singlet oxygen and excited triplet states observations outside of Europe and North America and I think that this manuscript add interesting observations on PM2.5 properties.

Abstract and Introduction

The abstract and introduction are clearly written and present well the singlet oxygen and organic triplet literature.

Methods section

The methods section is well documented, and the experiments well described. I have one small technical comment about the determination of the rate of light absorption (equation 2):

In equation 2, it is not clear what the authors used for the optical pathlength, is it the diameter of the quartz tubes or some average path length through the tubes ?  A side question on that point is if an incorrect pathlength in the solution would influence the singlet oxygen quantum yield determination. I am wondering about the presented singlet oxygen quantum yield numbers that are higher than in previous studies. The Rayonet reactor used by the authors has reflecting side walls and the effective path length could be longer than the measured one due to photon passing multiple time through the experiment’s solution.

Results and discussion

Reading the results and discussion, I was left a little wondering about the reasons for the observed seasonality. Maybe the authors could elaborate a little more. Here are some of my thoughts on the subject:

The authors did not see significant differences in the extracts between the three sampling sites (and attributed that to the brown carbon source being mostly not local) but observed a seasonal difference in extracts characteristics. Reading the article, I understand that the authors attributed the winter brown carbon to mainland China sources. I was left wondering about what the summer brown carbon sources are. Are the authors attributing the summer provenance to local sources, marine emissions or on lands further apart from Hong Kong ? It would be worth being clearer about this point.

This could have some implications if the summer aerosols are older than the winter ones and could explain some of the seasonal differences. Literature on photobleaching indicate that light exposure induces a loss of sensitizing properties (Water Research Volume 66, 1 December 2014, Pages 140-148 Photobleaching-induced changes in photosensitizing properties of dissolved organic matter) and a loss in absorbance (Environ. Sci. Technol. 2021, 55, 13152−13163).

The authors observed an increased singlet oxygen quantum yield for the fall and winter extracts. If the summer extracts were older and more exposed to sunlight that could explain part of the observed seasonal difference.

A last thought on the high singlet oxygen quantum yield observed is that ozone exposure may induce an increase in singlet oxygen quantum yield (Environ. Sci. Technol. 2019, 53, 5622−5632). If the authors think that their extracts were more exposed to ozone than extracts from other (north American and European studies), that could be a possible explanation.

Here are now some more specific comments than need to be addressed:

1) Line 254 and Lines 268-269: “due to the presence of aromatic compounds (e.g., polycyclic aromatic hydrocarbons) from local vehicle emissions” and “These results implied that the water-soluble BrC in PM2.5 was weakly influenced by local emission sources near the sites.” These two sentences look to be saying first that local emission sources are important and second that it is not important.

2) Line 332: “On average, Rabs was about 20 times higher than the sum of Rf,1O2 and Rf,3C∗ . This indicated that majority of the (photo) energy absorbed by the illuminated extracts in the photochemical experiments were dissipated by non-reactive pathways” this paragraph is misleading. Rf,3C∗ is a very small subset of the total triplets. The total triplets rate of light absorption can be estimated to be around 3 times Rf,1O2. The factor 3 coming from the estimate of the yield of triplet state conversion to singlet oxygen found in Environ. Sci. Technol. 2017, 51, 13151−13160. Also, the authors should not sum Rf,1O2 and Rf,3C∗ as singlet oxygen if formed from the triplet states.

The authors measured the formation of two oxidants – triplet excited states of brown carbon (³C*) and singlet molecular oxygen (¹O₂) – in aqueous extracts of particles collected from three sites in Hong Kong over the course of a year.  They report oxidant concentrations and quantum yields and examine how these parameters depend on solution characteristics such as light absorption and water-soluble organic carbon (WSOC).  There are only a few studies of ³C* in atmospheric samples, and somewhat more of ¹O₂, but few of these were performed on samples from Asia.  Thus the results here, which represent a significant amount of work - are a welcome addition to our limited knowledge base.  I am overall supportive of the manuscript, although there are some points that need to be addressed, as described below.

Major Points

1. One of the difficulties with reporting ³C* and ¹O₂ concentrations in particle extracts is that the results depend on the extract concentration, i.e., the PM mass/liquid water mass ratio. In relatively dilute extracts, oxidant concentrations are proportional to the extract concentration, so that changes in dilution lead to significant changes in [³C*] and [¹O₂]. This complicates comparing oxidant concentrations, both across and within studies, as they will vary with the amount of water used for extract preparation as well as the ambient PM mass concentration. (Fortunately, the authors used a constant sampling flow rate and sampling time.)  Thus there’s not much meaning to statements such as “The range of [¹O₂]ss values is remarkably large…”  (Line 288).  From an environmental perspective, two aerosols with the same PM composition but very different PM mass concentrations (i.e., µg/m³) would have roughly the same concentration of ³C* or ¹O₂ in their particle water.  However, concentrations of the two oxidants in the PM extracts (assuming constant sampling time and solvent volume) would be very different. 

It would be helpful to discuss this issue at the beginning of the results section.  As part of this, the authors should report their PM mass/water mass ratios (more about this below) and explain where their extracts fall on the rain – fog/cloud – aerosol liquid water (ALW) continuum.  It would also be helpful to explain how much of the concentration variation that they report is due to differences in airborne PM mass concentrations or collected PM masses.

It would also be helpful to report DOC-normalized production rates, i.e., Rf(1O2)/WSOC and Rf(3C*)/WSOC, and how they compare to past work.  These are important parameters for estimating oxidant concentrations in ALW.

2. The meaning of BrC "quality" is unclear. On Line 425 it appears to be defined as "specific absorbance" (Line 425), but this term is a bit vague. It also seems that any definition of BrC quality should include the efficiency of oxidant formation, i.e., quantum yield.

3. Section 2.1.2 (Sampling and extraction protocols) needs more details, in part so there’s a record of sample dilution to aid with estimating oxidants under ALW conditions in the future. For example, how much Milli-Q water was used to extract a filter? When the consecutive filters were combined to make a sample, how much additional water was added to reach “an adequate volume”? If the dilution parameters were variable, the information should be put in Table S1.  It would also be helpful to include the PM mass/water mass ratio of each extract. Can this be estimated based on what was measured and/or from nearby ambient PM2.5 monitors?  Two other experimental methods questions that should be addressed: How long were filters vortexed?  How long were extracts stored in the refrigerator before illumination? 

4. Section 2.6. The disadvantage of SYR (and TMP) as a probe is that its decay can be inhibited by DOM and Cu, which leads to an underestimate of the oxidizing triplet concentration, as initially described in surface waters. Inhibition can be very important in PM extracts, especially if highly concentrated.  In our 2018 and 2019 work we didn't know this was an issue; the current manuscript seems to be in the same boat.  We discuss inhibition, how to correct it, and the original surface water references, in our more recent papers (Ma et al., 2023a and 2023b; also Ma et al., https://doi.org/10.5194/egusphere-2023-861).  In this third reference we report SYR inhibition factors for a year of samples: at DOC ~ 25 mg/L, the upper range of the extracts in the Lyu et al. manuscript, we measured inhibition factors (IF) as low as ~ 0.5.  While the IF depends both on DOC composition and concentration, as well as Cu concentrations, our result suggests that correcting for inhibition in the current manuscript would increase [³C*] by up to a factor of two.  The authors should add a discussion about inhibition and its potential impact on the current work.

5. Section 3.4. Earlier in the manuscript, the authors found that oxidant concentrations were strongly (for ¹O₂) or weakly (for ³C*) correlated with both WSOC and α(300). In section 3.4 they examine oxidant concentrations versus MAC(300) or SUV(254): these correlations are similar or slightly weaker to the cases with WSOC and α(300). The former correlations make more sense, in that both the oxidant concentrations, WSOC, and α(300) all depend on the concentration of the PM extract.  In contrast, the latter correlations are examining concentrations, which depend on extract concentrations, with absorbance measures (MAC(300) or SUV(254)) that should be independent of extract concentrations.  Given all of this, it seems better to show the WSOC and α(300) correlations in the main text and move the MAC(300) and SUV(254) results to the SI.  

6. There are a few opportunities to shorten the manuscript. Most significantly, the parameter [Ox]/WSOC is roughly an intermediate step between the previously examined [Ox] and QY(Ox), both in terms of the parameter as well as the results. At this point in the manuscript, [Ox]/WSOC feels repetitious and doesn't offer much that is new: I recommend moving the [Ox]/WSOC results and discussion to the supplement. 

There are other examples of repetition that should be removed, e.g., (1) the paragraph on lines 424 – 431 repeats ideas that were raised in the original discussion of Figure 3 and (2) comments 11 and 12 under Other Points below.

7. I also have two suggestions for future work that the authors are free to take or ignore. The first is simple: use simulated sunlight rather than a narrow wavelength band to obtain results that are more directly relevant to atmospheric conditions. The second suggestion is difficult: Strive to measure oxidant concentrations under particle water conditions. In Kaur et al. (2019) and Ma et al. (2023b), we estimated ALW concentrations of photooxidants by extrapolating from three series of dilutions of PM extracts.  But even the most concentrated of these extracts are far from ambient conditions, resulting in a large (enormous?) amount of uncertainty in the ALW estimates.  How can we as oxidant afficionados use different experimental methods to better determine particle photooxidants, whether it involves probes or other approaches?

Other Points

1. The title starts with “Efficient production”, but the quantum yields of triplets are low, indicating inefficient production of this oxidant compared to past samples.

2. A sentence summarizing the quantum yield results in the abstract would be helpful.

3. Line 123. It’s not clear what is meant by “bandwidth”. Is it the wavelength range for the lamp output?

4. Line 254. PAHs are likely a minor contributor to BrC in these water extracts. Is biomass burning, which emits more water-soluble aromatic BrC species, significant in the region? This seems a likely source of aromatic BrC in winter.

5. Line 263. WSOC and alpha(300) values depend on sample concentration, which will be influenced by the extent of dilution as well as PM mass collected (as described above). So it’s not clear that these parameters can be meaningfully compared across samples.

6. Section 2.6. (a) The correction procedure for "direct photolysis" of SYR is mathematically fine, but the description should be corrected: SYR shouldn’t undergo direct photodegradation since it does not absorb light in the range of their lamp. The loss of SYR in the filter blanks is likely due to background ³C* contamination by BrC species. (b) The current procedure uses the average rate constant for SYR with the four model triplets from Kaur and Anastasio (2018). How does this average compare with the rate constant for ³DMB*, which we used in our more recent work (e.g., Ma et al., 2023a). What are the implications for [³C*] based on this difference? (c) The top paragraph of page 8 suggests that Kaur and Anastasio (2018) used the average of the four model triplet rate constants, but this is not correct. We used two probes to assess the average reactivity of each sample’s triplets and then used a weighted rate constant specific for that reactivity. But in practice, the rate constant ended up being very close to the ³DMB* value for most samples.

7. Tables 1 and 2. (a) Uncertainties are 1 standard deviation? (b) The authors could simplify MAC units to m² g-C^–1 (since this is equivalent to 1E4 cm² g-C^–1). (c) Having 3 or 4 significant figures seems beyond the precision of the measurements. Is 2 sig figs a better choice? (d) Typo in Table 2 title: “-sate”

8. Line 284. “Since the ¹O₂ measurements were used to determine ³C* production…” It’s not clear what this means.

9. Line 304. I agree that the relatively short illumination wavelengths used here (compared to simulated sunlight) are probably a major reason for the higher ¹O₂ quantum yields, as past work has shown that photooxidant QYs tend to decrease with increasing wavelength. But on Line 345 the authors try to use the same lamp idea to also explain lower ³C* quantum yields. It seems unlikely that these two oxidants have the opposite dependence of QY on illumination wavelength.  Also, in response to Line 346, Kaur et al. (2019) saw that SYR and MeJA gave similar quantum yields for oxidizing triplets, so the use of only SYR in the current work doesn't seem to be the reason for lower ³C* QYs.

10. Line 305. What do you mean about "different methodologies"? Use of D2O versus simply using the FFA decay rate constant in water?

11. Lines 349 and 350. Don’t these two sentences say the same thing?

12. Lines 351 and 352. Don’t these two sentences say the same thing?

13. Figure 3 (and all violin plots). For panel a, which of the seasonal [¹O₂] means are statistically different from each other? To what extent are any of the seasonal differences driven by differences in PM2.5 mass concentration?

14. Line 386. It’s not clear what is meant by “Even after accounting for their spread…”

15. Line 389. “…due to their spread and standard deviations.” Isn’t this saying the same thing twice? I suggest you shorten this paragraph’s discussion of seasonal differences in quantum yields since there were no statistically significant differences.

16. Lines 506-509. This sentence is repetitious, including having the phrase "¹O₂ and ³C*" appear three times.

17. Figures S4 and S5. Column headings of the season over each column of panels (e.g., "Winter" over the first column) would help.

18. Figure S10. I don’t see any triangles, although they’re mentioned in the caption.

19. Table S1. The word "Set" here is confusing – doesn’t it represents a single filter (sampled for 72 hr)?  If so, it would be clearer to say "filter" rather than "set". As described earlier, it would be helpful to include in this Table the PM mass collected on each filter and/or the average PM2.5 mass concentration in air over each filter period.