Review of Jensen et al.

Jensen et al provide a comprehensive and useful analysis of best practices for interpretation of high-resolution PTR data. The manuscript provides a detailed discussion of factors contributing to PTR sensitivity and its variability in the field. The manuscript will be a helpful asset to the community and should be published following the authors attention to the following comments:

General Comment:

The authors do an excellent job discussing how fragmentation of a parent ion impacts its sensitivity. That is, fragmentation reduces the expected signal at MH+ as some fraction of the molecules fragment to smaller ions. This impacts the retrieved sensitivity and the comparison of the measured and expected sensitivity. The value of f for a molecule can be determined from the GC. There is less discussion about the positive bias that fragmentation can invoke. For example, at 69 m/Q (or the exact mass of C₅H₉⁺), some fraction of the ions detected here are protonated isoprene (you know this fraction from your C₅H₉⁺ chromatogram) and some fraction is fragmented larger molecules. This can be significant. From what I can tell the toolbox here does not address this issue of fragmentation. I appreciate that this is tricky. If the authors do not want to tackle this, I think that is fine, but it would be helpful to provide a short statement about how this could impact this analysis.

Specific Comments:

Line 115: Please confirm whether the entire inlet or just the Vocus subsampling inlet was overflowed for calibration and zero.

Section 2.3: Please confirm if the inlet for the HC trap and the catalytic zero source were drawn from room air or from ambient air.

Line 175: This equation (E3) holds, so long as another (larger) molecule does not fragment into the detected ion [RH+]. I agree that E3 is correct in isolation, but in the atmosphere if a large fraction of the measured signal at RH+ is not from R but from a larger molecule that fragments, the sensitivity could not be applied to [RH+] to deduce [R] without knowing the fraction of the signal at [RH+] that is from R. Take for example isoprene, only 40% of isoprene is retrieved at RH+ (per your table S1), but the signal at RH+ is comprised on many other molecules beyond isoprene. This could be extracted from the chromatogram as well for the ambient data. Perhaps I missed it, but how is this side of fragmentation being accounted for?

Line 200: These are Tables S1 and S2, not Tables 1 and 2 (took me a while to find them).

Line 205: I’m a bit confused by this sentence. Why does it matter if the transmission function is different for the fragments. Is this because you need to know T(m/Q) to accurately determine f (i.e. if the transmission of the fragment is not accounted for and it is smaller, you would overpredict the actual value of f?) Otherwise, isn’t the value of T(m/Q) in equation 3 specific to RH+? Sorry, if I’m turned around on this a reader may be as well, so it wouldn’t hurt adding a sentence or two here to more fully describe this.

Line 215: It would be interesting to add how many k(PTR) values are known, calculated, vs estimated based on parameterization.

Line 215: I can understand how this procedure is applied to ions that are the protonated parent molecule (RH+), but how/when do we know that is true and how is this applied to a measured ion that could be a combination RH+ and fragments? (related to the question above). For example, at 69 m/Q, some fraction of this is protonated isoprene (you know this from your chromatogram) and some fraction is fragmented larger molecules. It might help the reader to walk through your procedure for an example like this on how you would extract [isoprene].

Line 248: There is some strange formatting here with the inserted symbols.

Line 265: What is the physical reason for transmission to decline at high mass? I would have expected this to be operating as a high (mass) pass filter?

Line 380: You have used the term “spectral interference” a few times. I did not see it defined. Since there could be a few different interpretations of this, it would be helpful to clarify this at first use. My apologies if I didn’t catch it.

Line 410: What is the y-intercept in the slope that is not constrained by the zero. It looks quite large. Were lower concentration calibration points done to fill in the gap between the 1-3 ppb region to assess this further?

Line 416: If diffusion is important, do the residuals scale with the diffusion constants as expected?

Section 4.5: It would be helpful to include in Table S1 (or elsewhere) the average zero values for these ions. I appreciate that it could be back calculated from the LOD, but I think it would be helpful for Vocus users to be familiar with what zero (cps) can be achieved with these sources. Or perhaps add a panel to Figure 6 that has a characteristic zero spectra for the catalyst that everything is referenced to.

Line 610: MeSH/DMS should show a strong diel profile due to the large difference in DMS+OH vs MeSH+OH. I’d expect if you look at the nighttime correlation it will be even stronger.

Overview:

The authors have clearly spent a considerable amount of time working/learning/thinking about their GC-Vocus instrument and the interpretation of their data, and here they share important insights into how to analyse and calibrate GC-Vocus data. The authors ran a “test” (my words) campaign in Boulder, Colorado in the spring of 2021 and troubleshooted aspects of their instrument, particularly related to sensitivity and blanks, that had important impacts on their data collection. The authors wrote a tool kit for PTR sensitivity calculations that they are sharing with the community. Overall, this manuscript currently reads as a compilation of what the authors learned and how they resolved the complexities of calibrating GC-Vocus data. The authors did an excellent job referencing the literature. At the very end, the authors show some of the data they collected in Boulder with a focus on aldehydes, siloxanes and sulfur-containing compounds.

The work present is an important resource for the community despite this manuscript having a narrow (but growing!) audience of Vocus users, and likely more specifically GC-Vocus users. To improve this manuscript for publication, I would strongly encourage the authors to consider adopting a Standard Operating Procedure (SOP) style in order to teach the “whys” of their choices. The most likely future readers of this work will be graduate students and I expect they would greatly benefit from additional details and justifications throughout the text. I hope to have identified and described these points here as best as I can to help improve this manuscript as a resource. I anticipate this manuscript to be on the to-read list for any future/incoming (GC-)-Vocus users.

General Comments:

In the spirit of making this manuscript an SOP as well, I would encourage the authors to add more details so that the presented optimized data analysis could be repeated by a future GC-Vocus user:

• Show all their calibrations plots where the sensitivity was calculated
• Show sample time series of blanks, fast cal, long cal, GC over 2-3 cycles of their 2h procedure/ TPS script.
• Show background signals of (at least) their calibrants from which they calculated their LODs.
• Show zero time series. I suspect 2 mins was too short for many of the “stickier” compounds.
• Show the fragmentation patterns (like Fig S2) for all their calibrants.
• Show a fitted peak of m/z 19 (How did the authors integrate m/z 19? I am skeptical of what m/z 19 can tells us in the Vocus).
• Add at table with all the dates and times and data of the full calibrations pre- during and post-campaign. (This information would be useful for others planning their field campaign calibrations timing.)
• Give examples of graphical linear interpolations (to help substantial/illustrate the point on lines 127-129)

The authors present and discuss the value “f” (the fraction of signal attributed to the quantitative ion) (line 173) as a new parameter to be considered when calculating sensitivities. But I’d like to challenge the authors on the pros and cons of this parameter as I read their manuscript:

• On lines 314-316, the authors conclude that the fragmentation of the ions was constant throughout their campaign. So why go through the hassle of quantifying f then? One could just calibrate the sensitivity of the ion.
• What would be the error introduced if the authors calculated sensitivities based solely on their Vocus data (and did not have GC data available, so wouldn’t be able to calculate f). This discussion is likely very relevant for Vocus users without a GC add-on.
• Do the authors suspect that compounds may also be fragmenting on the GC column?
• Then can “f” only be calculated for known molecules that have been previously measured by PTR? What are the implications of a value such as f for untargeted analysis?
• Ionization sources like EI have large databases of mass specs of pure samples where the fragmentation pattern can be used to identify unknowns. In PTR, my impression is that the parameters of the instrument vary too much from one instrument to another for such a database to be useful. Do the authors anticipate having to re evaluate their “f” factor for every campaign they run?

Specific and Technical comments:

I encourage the authors to use additional subsections to help guide the reader to the section of interest. Each section can be constructed as a paragraph (with a topic sentence, details and summary sentence/transition sentence). For example, section 4.1 goes on for more than 3 pages without subsections, making it difficult to be used efficiently as a resource.

Title: the term GC could be included in the title since part of the novelty of this work is using the GC data.

Abstract: I’m left wondering at the end of the abstract about what was observed during the field campaign in Boulder 2021, and what the authors were aiming for as research questions during this campaign?

Introduction: I was confused whether the authors were aiming to discuss PTR technology overall or Vocus specifically. It would be worthwhile to have a paragraph discussing how the Vocus’ quadrupole on the drift tube uniquely changes the sensitivity, LODs and RH dependence for VOC detection. Such a discussion would set the stage better for the “why” this manuscript is timely. (for example line 46 should continue to discuss the different ion sources for H3O+)

Lines 41-42: I would argue that the logic is reversed here, and that it’s rather the technological advances that lead to new VOCs detected and new LODs/sensitivities achieved.

Line 47: change “functional group” to heteroatom, since a methyl group is considered a functional group on a molecule but wouldn’t have a proton affinity higher than that of water.

Line 48: I’m not sure I followed why fragmenting alkanes are “most notably” in the context of PTR.

Paragraph from lines 69-76 is subjectively redundant, and I would delete it to help make the writing more concise.

In their operating conditions, why did the authors run their IMR at 90 degs and a slightly higher pressure of 1.5 mbar (compared to 1.2 mbar). What would these parameters help optimize?

The authors share that their instrument had to be troubleshooted during their field campaign. I would suggest that the authors add a subsection called troubleshooting in their methods and provide more details on their ion source malfunction (for example on line 102) and their capillary clogging (and how did they unclog it?)

On the section of the clogged capillary: why would a change in flow impact the sensitivity? The mixing ratio does not change and the flow of water remains small compared to the inlet flow. In other words, what do the authors see as the implications of their discussion on lines 105-110?

Why would methanol (lines 119-120) be the only VOC here to have a water-cluster relevant for its quantification? I’m not sure I followed this argument, or the uniqueness of methanol in this case.

Thuner is an important part of Vocus data analysis and it isn’t mentioned anywhere in the text. I would encourage the authors to discuss this procedure in detail in their sensitivity discussion.

Methods, section 3: how do the authors calculate/determine the initial concentration of H3O+? Isn’t the Vocus blind to the reagent ion?

Methods, section 3: equations 1, 2 and 3 do not have consistent definitions. Would “S” be missing in Eq 1? S and Sinst cannot both be equal to [RH+]/[R]. It might be worth cleaning up this section, and providing a solution to these questions for one compound of choice (a compound discussed in section 4.6 for example?)

Lines 179-184 – could there be a picture of the interface/code that could be included here to support visually the contribution of the authors to developing the PTR-DT?

Line 319: why did the sensitivity increase?

Line 329 and 337: why did benzene and toluene have a different response to changing ion chemistry?

Line 366-367: specify which standards were included and which were omitted.

My understand of the Vocus technology (Krechmer et al. 2018) is that there is no RH dependence due to much larger concentration of H3O+ ions in the drift tube. I was therefore confused by the discussion on lines 371-373, since the authors go on lines 418 to specify there was no dependence on saturation vapour concentration (or do they mean ambient H2O?).

Section 4.3. based on this discussion, can the authors speak to the value of introducing an internal standard into the Vocus (which wouldn’t be present in ambient air).

I worry that the acetone artifacts discussed on lines 454-455 are related to acetone present internally inside in the instrument, and not due to the calibration…

Lines 479-484: I didn’t quite follow this discussion on ion mobility. Do the authors mean that as the flow reduced, the H3O+ flow made of a larger portion of the total flow?

Figure 1b: Could the authors clarify how/why many compounds were not included in the fit and how the authors went about deciding which compounds were excluded from the fit. This information would be important to avoid data manipulation.

Figure 1a: what did the post-field campaign comparison look like?

Figure 2: which compounds is this data representing?

Figure 3: the use of measured/derived values is confusing and should be rewritten with clarity.

Figure 6: what are the red traces?

According to Figure 7, the LODs of the standards would appear to be biases high compared to the 616 species quantified. Why might that be?