The measurement of NOx in remote air is very challenging, in particular because of the difficulty of accurately determining the NO2 artefact of photolytic convertor-CLD (P-CL) measurements, the current gold standard technique for accurate NOx measurements.

This manuscript, while not entirely novel as the authors point out in terms of presenting an alternative quartz glass converter for P-CL measurement of NO2, is very useful especially because of the discussion of  laboratory experiments to investigate the instrumental background produced by the photolytic converter in the NOc channel and the characterisation of an improved convertor.

I recommend publication after the following points have been addressed:

Pg 5.  A NO₂-> NO conversion efficiency of 14% (or even 20% for the original convertor) is very low (i.e. Andersen et al. 2020 report CEs of >50% ).  I suggest the authors mention that a higher CE is desirable for improved accuracy and perhaps suggest ways this could be implemented.   

Pg 5.  “Therefore, a pre-chamber measurement is operated for 20 seconds every 5 minutes where ozone is added to the sample gas flow”.   What is the efficiency of the pre-chamber volume (i.e. how much of the added NO from the calibration gas reacts with O3) ?  It should be >98% or so.

Pg 5 Ln 141.  The “constant temperature of 25oC” in the convertors is not monitored, and so could presumably be a lot higher when the LED lights are on.  The authors rightly point out that accurate determination of this temperature is critical for the calculations of the NO₂ artefact.  It would also be highly beneficial to perform measurements of e.g. PAN degradation to confirm the artifact calculations (and, indirectly, indicate the temperature in the chamber).

Ln 178. “Please note that the instrumental background for the NO data was determined by nighttime measurements of NO instead of zero air measurements ...”    How often was night-time NO determined and what was the variability?

Page 7.  An calculation of uncertainty for both NO and NO2 measurements is missing from the Experimental section.  

Ln 193 “Please note that these data (OH and HO2) are still preliminary”    Are final data yet available?  This would be highly desirable since HO₂ and OH are required for the calculation of [PNA], and CH₃O₂ is calculated from HO₂ and required to calculate MPN. 

Figure 5.  Please include all data in the figure legend (including BG) and explain the orange dotted lines in the caption.  The word “exemplarily” is not needed in the caption.

Lns 352 onwards.  The authors demonstrate convincingly that memory effects of the porous convertor coupled to water vapour changes are a strong driver of changes in the instrumental NOc background.   However, the adsorbing/desorbing of NO molecules will likely also be affected by pressure as well.  Could the authors comment on this?

Ln 400 onwards.  I congratulate the authors on their much improved photolytic convertor and its apparent stability and insensitivity to varying humidity and lack of memory effects.  I would recommend also that experiments are conducted with varying pressure to evaluate pressure-dependence of the background.  

Ln 470 onwards.  In the Conclusions section, the authors could consider adding recommendations on airborne NO2 measurements by P-CL, i.e. avoiding constant altitude changes in flight, which will inevitably change the background, and ensuring sufficient background measurements at each altitude change.  This would be useful for the community.

The authors describe an implementation of a photolytic NO2 converter demonstrating it’s’ use in airborne atmospheric research – specifically high altitude aircraft measurements. The selective photolysis of NO2 using a narrow band UV source illuminating a quartz cell, followed by detection of the resulting NO by chemiluminescence has been the reference method of NOx determination since the early 2000s’.

With some comments addressed the manuscript can make a valuable contribution to AMT.

General comments:

The authors don’t give a clear rational for modifying of the Droplet Measurement Technologies (also Air Quality Design, and now Teledyne API) Blue Light Converter in the way they have, especially given the sub-optimal results.

Why reuse the low-powered 1 W, 395 nm, UV Hex, Norlux Corp. LEDs when much more powerful units are available – and in fact are used in more recent BLCs (see: https://doi.org/10.5194/amt-9-2483-2016)?

What is the rational for moulding the PTFE around the quartz envelope? Were no alternatives tried? Similar aircraft implementations from NOAA, NCAR and FAAM, plus the paper cited by Andersen et al., 2021 use quartz cells wrapped in baking foil! Vapour deposition of optical silver has also been used in the past with no benefit over simply wrapping.

I realise in advance the answer to the two previous questions may be due to certification hurdles of HALO/DLR.

The authors should note that their PLC/BLC implementation has remarkably similar characteristics to a once commercially available unit also marketed by Droplet Measurement Technologies – a glass envelope, shrouded in PTFE, with arrays of Norlux UV-LEDs at either end. In this case the volume is ~ 115 cm3 which is the main difference. It is well described in the paper by Pollack et al., 2010 which the authors cite. These NO2 converters were previously operated by NCAR and FAAM on their aircraft, though are long since retired.

There is a marked drop in photolysis frequency between the two BLC modifications (0.66 to 0.46) using presumably the same LEDs. The authors should discuss why this is the case; is it the design, or aging of the LEDs, change in sample gas temperature etc.

There is no description of the aircraft inlet from which the NOx instrument samples. Is that heated? What is the residence time to the instrument? What is the sample line/cabin temperature? When discussing the uncertainty of airborne measurements these must be taken into account.

There is no schematic of the instrument given so I must assume it is identical to F2 given in Tadic et al., 2020. Several elements of the design shown there may skew measurements, especially during your discussion of the effects of humidity.

Firstly, the sample flows through mass flow controllers in the high pressure side of the inlet system which a) increases hold-up/lag, but also provides plenty of stainless steel surfaces to form layers of water on. Secondly, the NO and NOx channels appear to have different volumes due to there only being an NO2 converter on the NOx channel and no dead cell on the NO channel. Necessarily, there are different surface areas between the two, and different volumes, thus data must be offset between the two channels to compensate for the different residence times which themselves must either be very carefully measured or modelled. I doubt for instance that the true residence time if the BLC is 0.34 seconds – this is more likely 1 e-folding time. Lastly, and most crucially, the flow of humidified ozone is switched between reaction and pre-chamber of the CLD, with the sample (of varying humidity) constantly passing through the pre-chamber. This results in wildly fluctuating humidity within the pre-chamber. Many airborne CLDs follow the scheme in Pollack et al. 2010 whereby the humidified ozone constantly passes through the pre-chamber and the sample is switched – this also acts to decrease the response time of the instrument by removing dead volume.

Ultimately, I don’t think flaws in the instrument design can account for the humidity effects described, though they should be considered.

The whole discussion on possible mechanisms for NO/H2O selective/competitive sorption is highly speculative and hard to follow at times.

A typical test flight when commissioning a new NOx instrument is to fly whilst adding an amount of NO well above ambient to the inlet, performing profiles, orbits, in-cloud, boundary layer, and free troposphere runs – a system which is performing well will show no deviation throughout the entire flight envelope.

Specific comments:

Line 86: this was also the conclusion of Reed et al., 2016 which is cited.

Line 117: ‘commercially available’ – all four example of the CLD 790 SR were built for DLR on special order, no?

Line 123: State the reason why the photolysis cell is operated at 110 mb i.e. this is a pressure height of ~50kft which is the service ceiling of HALO/G550.

Line 124: The stated wavelength is 398 nm of the UV LEDs – the design wavelength is 395 nm – is this a typo or was it measured (and not shown/described)? If the latter then a lot of energy is being wasted outside of the quantum yield of NO2 which drops rapidly at ~400 nm.

Line 132: please state the j value along with the conversion efficiency i.e. 0.656 s-1

Line 138: please state the j value along with the conversion efficiency i.e. 0.457 s-1

Line 138: presumably the gas flow doesn’t contact the LEDs either in the design depicted in F1b? This would in-turn lead to much less sample heating and have a large impact on any thermal artefacts.

Line 178: limits of detection are only useful when an averaging time is stated, please add this, how many standard deviations are included in the determination of LOD and uncertainty? e.g. 5 pptv averaged over 10 seconds, 3 sigma uncertainty of 6% etc.

Line 146: What is the residence time of the pre-chamber; same as the reaction chamber? What is the efficiency? What is the material?

Line 282: ‘…monitoring system for pressure…’ in an airborne system the pressure must always be known and/or controlled otherwise the conversion efficiency of NO2 to NO is unknowable, regardless of potential artefacts or not!  

Line 370: I’m not sure it is true that there are no trends in the NO channel signal – I think the scale may be helping here – perhaps fit the trends to remove any doubt or adjust the scale.

Line 411 -414: the logic in this statement is flawed – you can only measure NO with a CLD, therefore you only saw NO in your experiment. You could do the same experiment with the BLC connected to a direct NO2 measurement, or a PAN-GC or a CIMS for that matter and would likely see may compounds desorb.

Sect 2.5: please define all the acronyms for the NOy species (MPN,PAN…) at their first use.

This manuscript describes a photolytic converter for airborne measurements of NO₂. The focus of the paper is measurements of NO₂ in remote areas where mixing ratios are sub 1 ppb. The authors describe interferences and artefacts associated with a commercial photolytic converter that complicate these sub-1 ppb measurements and suggest modifications to the commercial converter that reduce the effects of artefacts. There is not an overabundance of publications about the nuances of NO₂ photolytic converters, and thus I believe this manuscript can be a value contribution to the atmospheric community.

I agree with the assessments and comments posted by the other referees. I had many of the same concerns. Thus, in an effort to streamline the review process, I have only included my additional thoughts here. I hope they are helpful.

Comments:

Title: I do not think that the word “new” is appropriate for the title. This is because there is not anything particularly novel about the modified converter. The use of a fully enclosed quartz cell with and without a reflective Teflon shroud is not new among the airborne research community. However, these PCL systems are typically custom built (e.g., Pollack et al., 2010, Jordan et al., 2020). I wonder if a better title could be “Modification of a commercial photolytic converter for improved aircraft measurements of NO₂ via chemiluminescence”.

Abstract: Please add another sentence or two to the abstract about the aircraft measurement findings related to the NO₂ reservoir species. This is first and foremost in the results section but seems to be lacking mention in the abstract. Also, the abstract is a bit misleading in that it highlights the memory effect as the key phenomenon. Yet, the observations from CAFÉ are likely a combination of phenomena that also include an artefact from the subtraction of two signal channels and a changing background.

Line 6 and throughout: Maybe it is just me, but I find the use of the word “conventional” to be a bit bothersome. This is because the photolytic converters typically used aboard aircraft do not use the porous Teflon material with ring channel for gas introduction. The word “conventional” seems more appropriate for ground-based applications that utilize commercial monitors and commercial converters. It might help to clarify the difference in the text.

Line 124: Can you add the year the BLC was purchased from DMT? This could help readers distinguish between the version of "conventional" BLC that you are using compared to other versions of "conventional" commercial BLCs. 

Line 134-142: I don’t think the use of the words “new” or “newly-developed” are appropriate here since several existing converters already separate the sample flow from direct contact with the porous Teflon surfaces. Maybe a better word for the converter shown in Figure 1b is “modified” or “updated”.

Section 2.2: I understand the elimination of a night flight (MF11), but why were only MF10 and MF12 through MF15 used in this study? Were MF01 through MF09 not good candidates, was NO₂ data not collected during those flights, or was this phenomenon not observed during those flights?

Line 170: Can you add a figure (either here or in the SI) that shows the J-curve for your converters? The conversion efficiencies of the “conventional” BLC (20%) and your “updated” converter (14%) are very low. This is likely a function of your very low cell pressure, which when combined with the high flow rate, results in a short residence time in the photolysis cell. It would be helpful to see how each converter (the conventional versus the updated BLC) behaves over a range of residence times. Regardless, a note should be included in the text to associate the low conversion efficiency with the low cell pressure, which is needed for high altitude measurements.

Line 180: Is it reasonable to utilize a nighttime NO concentration instead of zero measurements for determining c(NO) when the c(NO₂) is determined from the subtraction of the NO measurement from c(NOc) and c(background_NOc) determined from a zero? What is the magnitude of the difference between NO zeros at night versus NO zeros with an overflow of zero air? Has this difference been factored into an uncertainty calculation for c(NO) and c(NO₂)? What is the overall measurement uncertainty for NO and NO₂? Also, what was the concentration and uncertainty of the NO standard used for calibrations. What was the effective calibration mixing ratio after dilution into the sample flow? It would be helpful to add these details to the manuscript.

Line 305: I wonder if the changes in background can be more carefully characterized in a future flight by overblowing the instrument inlet with zero air for the duration of a test flight (aka. a “null” flight). The in-flight instrument performance can be evaluated from changes in the background signal levels during vertical profiles and maneuvers, which can inform about precision, detection limit, motion sensitivity, and fluctuations with pressure and temperature. It can also inform about lags in the recovery of background signals with these perturbations. For high altitude chemiluminescence applications, it might also be interesting to characterize the PMT dark counts versus background in a future test flight by periodically turning off the reagent O3 injection.

Line 329: How do the NO₂ measurements change if you assume a constant background signal per altitude level? My first instinct would be that subtracting an interpolated background signal would contribute a good bit to the negative excursions in NO₂. Since the CLD 790 SR has two separate channels, is the BG trace in Figure 5 meant to be the background signal of the NO₂ channel? How does the background of the NO channel differ from that of the NO₂ channel with the LEDs on and off? Does the NO channel background also change with altitude or only the background measured through the converter? Can you add the NO channel BG as a trace in the Figure?

Section 3.1.3: I admit, I found the logic of this section a little hard to follow. If I have this correct, the bulk of the discussion in this section is about the instances when NO2_CLD is enhanced yet there are no enhancements in NO2_PSS nor NO2_DOAS. The authors are claiming that the NO2_CLD enhancements are correlated with increases in water vapor as the aircraft descends. The authors associated the discrepancy to a hysteresis in the photolysis cell upon the introduction of water vapor. If this were the case, then I agree that a decrease in NO₂ back to baseline levels following the increase in water vapor with the lag time representing the memory effect time would be expected. However, the rising edge of the enhancement in NO₂ that starts to increase as a large step change in NO starts to decrease and that occurs earlier in time than the step change in water vapor concentration is not something that I would have expected from a memory effect phenomenon. This leads me to believe that the NO₂ peaks are more of an artefact of the NO channel subtraction, which is enhanced by a factor of 4 due to the correction for Ce, and less so from a memory effect of water vapor on the photolysis cell sampling surfaces.

From the manuscript (mainly the abstract, introduction, and conclusions), I am led to believe that the authors think the memory effect is the key phenomenon at play with the “conventional” converter. However, the results and discussion of the CAFÉ observations suggest that the signal subtraction, low Ce, fluctuations in background, and large changes in NO concentrations could also have been significant contributors to the observations. Thus, it seems a little misleading to only mention memory effects in the abstract and conclusions. It is my recommendation that the text in the abstract and conclusions be updated to reflect the observations and all possible factors that could have impacted the CAFÉ observations.

Section 3.2: The UV artefact (Figure S5) of 0.1 ppb seems substantial for a sub-1ppb ambient measurement. How does the UV artefact factor into your subtraction calculations (e.g., eq. 4) and into the overall measurement uncertainty? What does Figure S5 look like for the updated converter?