150mm Paper

This paper presents the results of analysis of the first two 150mm Si detectors. UMN run 60 was a test of S1501, a bare 150mm diameter, 30 mm thick Si crystal with two contact-free electrodes. UMN run 65 tested S1502, a 150mm diameter, 25.4 mm thick Si crystal with a 5Q mask on one side and contact-free electrode on the other.

Overleaf Edit Link

arXiv Link

150mm Si paper v2.4

150mm Si paper v2.3

This is a very interesting paper. I would like to congratulate the authors. I have little to say apart some specific comments.

-Page 2, line 18 : Luke phonon production : The actual name for this process is the Neganov-Trofimov-Luke phonon generation. An appropriate reference for the first two names would be JETP Lett. 28 (1968) 378. The name “Neganov-Trofimov-Luke” has now been adopted in all mentions in the paper. The JETP reference discusses low temperature calorimetry, but does not discuss phonon amplification. The “Otkryt.Izobret” reference (available: http://inspirehep.net/record/1416918) is the standard one used by CDMS in such journals as PRL, PRD as it specifically discusses the transfer of charge carrier electrical potential energy to measurable heat when drifted across a cryogenic semiconductor detector.

- P 8, L 147 + Fig. 5 caption : Carrier diffusion should not reduce the measured signal, the signal reduction being due to trapping. Perhaps, diffusion, by increasing the charge path, could induce some more trapping. I suggest you mention precisely what you mean in the text. Indeed, increased path length means there are more chances for trapping, so this phrase was somewhat redundant. Modified text to read, “As the bias voltage is increased, carriers are less likely to trap or recombine leading to increased CCE and thus larger signal amplitudes for a given interaction energy.”

- P 9, L 169 : “The full amount of charge is measured and the CCE is close to 100%” but I note that the signal is still increasing with the bias and should apparently increase some more. You do not seem to have reached 100% collection. But the signal on fig. 6b has already reached and even got above the calibration value of 100%. Why is that so?Fig. 6b data is normalized to an absolute CCE (derived from readout circuit properties and vacuum gap) which has a large systematic error (orange band). With this error, the values are still consistent with CCE=1 That being said, effects such as impact ionization of shallow trapped states may allow CCE to be greater than 1, however, we don't see strong evidence for that here. Modified text to read, “At high crystal bias fields, the full amount of liberated charge is measured and the CCE approaches 100\% within uncertainty.”

- P 8 L 189 : “the initial charge cloud” I would write “the initial charge cloud recombination”Done

- P 8 last paragraph and Table 1 : Please provide the expression of the trapping model if possible.It's rather long, but we have included the full expression as it has been slightly modified from that used by Penn et. al. to include a term due to the vacuum gap. We still retained the paragraph briefly describing the different loss mechanisms and parameterizations so the reader can understand what the model accounts for without having to read the entire Penn et. al. article.

- P 10, L 218 : “recombine” should be replaced by “be collected”Done

- P 10 L 224 : “remove the bias field” why not invert the field to make the electrons and holes from the LED light to drift in the direction of the carriers trapped at the crystal surfaces? And how is it that producing more eh pairs from the LED does not simply increase the trapped charge in the detector? How is it that the totality of the carriers created by the LED plus the previously trapped charge are all removed?We neglected to mention that the bias is not only removed, but the two electrodes of the detector are both grounded while flashing occurs. This means that the only field present in the crystal is due to charges trapped in the crystal bulk or which had accumulated at the vacuum gap crystal surface. So when the LED light produces e/h pairs, they naturally drift to where they are needed to neutralize the crystal or to the grounded electrode. In this manner, all trapped charges are neutralized and the detector is “reset”. I have included the following lines to clarify this point: “The loss of neutralization was mitigated by periodically removing the bias field, grounding the electrodes, and illuminating the detector with … The photons produce free charge carriers in the crystal which then diffuse to overcharged regions neutralizing any bound space charge. If the crystal is already neutralized, the carriers are simply collected by the grounded electrode. In this manner, trapped charges are neutralized and the detector is reset.”

- P. 11 L 256 : the ionization energy in Si is about 3.62 eV. Please provide some references more recent than Pehl et al. although the Pehl et al. paper is certainly an excellent reference. At cryogenic temperatures, the required ionization energy increases with decreasing temperature to ~3.8 eV. There have been very few measurements of this since Pehl. The best I can find recently is https://doi.org/10.1016/S0168-9002(96)00822-4 which is still at higher temperature than Pehl and rather higher than the temperatures at which our experiment was conducted.

- P. 13 L 296 and ff.: The UV light will not penetrate the Si enough to cause any charge neutralisation. We agree that the UV light seemed to not penetrate enough. This was not obvious to us a-priori because the UV light still produces e/h pairs as it is absorbed. It was hoped that these would be able to drift to the overcharged areas of the crystal and neutralize them. However, our results indicate that was not the case.

Dear Editor, Dear Authors, the manuscript satisfies the criteria of impact, innovation and interest for the reader of NIM-A. The text is properly written, clear and comprehensive, the figures support the text by displaying interesting data. The research calls for a publication and further continuation.

I recommend publication of this paper in NIM-A after the authors address the following minor comments:

- The abstract includes repeating phrases and needs to be slightly revised (shortened) accordingly. I have emphasized some words in the following extract: “In order to SCALE UP the sensitive mass … These devices were operated with CONTACT-FREE BIAS ELECTRODES … One detector was instrumented … using a single CONTACT-FREE READOUT ELECTRODE… The results show promise for … CONTACT-FREE ELECTRODE ARRANGEMENTS for SCALING UP … detector mass” Limited to a single occurrence of both “scale up” and “contact-free” in the abstract. Also eliminated an unnecessary sentence and streamlined some of the wording. It now reads, “Cryogenic semiconductor detectors operated at temperatures below 100 mK are commonly used in particle physics experiments searching for dark matter. The largest such germanium and silicon detectors, with diameters of 100 mm and thickness of 33 mm, are planned for use by the Super Cryogenic Dark Matter Search (SuperCDMS) experiment at SNOLAB, Canada. Still larger individual detectors are being investigated to scale up the sensitive mass of future experiments. We present here the first results of testing two prototype 150 mm diameter silicon ionization detectors. The detectors are 25 mm and 33 mm thick with masses 1.7 and 2.2 times larger than those currently planned for SuperCDMS. These devices were operated with contact-free bias electrodes to minimize leakage currents which currently limit operation at high bias voltages. The results show promise for the use of such technologies in solid state cryogenic detectors.”

- In Sect. 4.2.1. a model is developed to describe the charge signal loss over time by accounting for neutralization losses. The model was used to predict the change in the amplitude of the calibration signal, shown in Fig. 9. I was wondering why a prediction from this model was not shown in comparison to the data in Fig. 8 nor measured data points added to Fig. 9. Does data and model fit together? Fig. 9 was remade with measured data points and error bars have been included. 50V data and model were used so as to include a large sample of measured data points. The figure caption was amended with ,“Shown for comparison are measured 60 keV peak signals.” Tangentially related to this is that Fig. 10 was also remade to include data measured at 2 additional bias voltages which had been mistakenly omitted from the first version (one of which was used as the example shown in Fig. 9). The new data points fall on the same trend and does not qualitatively change the figure or the conclusions drawn from it.

- Fig. 5: the red, green, and blue lines are almost not visible on the right side of the histograms when overlaid with the gray spectrum. The authors should try to add transparency to the gray histogram and improve the contrast by the appropriate choice of colors and hatches.Plotted only 3 of the histograms and improved fill and hatching styles and colors for clarity as suggested.

- Fig. 11 needs larger axis labels and titles.Done

- The proposed flashing of the detector to restore the electric fields seems to be a time-consuming process when taking the heat-up and cool-back into account. What would be the duty cycle in a real detector operation for dark matter searches? What is its overall deadtime (or downtime) compared to existing detectors? Is the reduced duty cycle compensating for the increased sensitive mass and bias voltage? The authors should comment on this issue in their conclusions.

The hold time of these devices should go as t_hold~V_tot/(Ionization Rate). (Using Eq. 4 with sigma=(Ionization Rate)*time/Area) At a surface test facility such as at UMN, the background rate is several orders of magnitude higher than that at a low-background WIMP-search site (Soudan, SNOLAB). So, operating this device at ~75 V bias, which provides ~1 hr hold time at the surface, should provide >10 hours hold time in an underground, low-background facility such as SNOLAB. The cooling period depends on the fridge cooling power, but currently is no different from previous, smaller CDMS detectors (~20 min at the UMN test facility). The resulting duty cycle is comparable to the duty cycle of previous CDMS detectors. Even without contact-free biasing, some charges become trapped and accumulate in the crystal bulk and flashing has been regularly used to ensure detector neutralization in all previous SuperCDMS operations. e.g. the CDMSlite experiment at Soudan regularly flashed every 11 hours, with a ~20 min cooling period thereafter. This ~90% duty cycle could be increased in contact-free detectors by increasing the bias voltage, assuming impact ionization does not begin to occur. Added the following paragraph to the end of Sec. 4, “The flashing and cool down process for these devices generally took 30 minutes. Given hold times of 10 min - 1hr, this was a significant component of the detector's operational duty cycle at the test facility. The hold time is driven by charge accumulation on the contact-free crystal surface(s) due to events in the detector. If operated in a low-background environment, such as that for a dark matter search where the energy deposition rate is several orders of magnitude lower, the detector hold time would be proportionally longer. Such a detector could easily achieve duty cycles of >90% if operated with V_{tot} >~ 50 V in a low-background facility such as SNOLAB.”

150mm Si paper v2.2

Please include a reference to figure number or line number when adding comments.

- Title: should we include Mark Platt, engineer from TAMU? Nader says yes.

- line 43 or so - should we mention how the contacts were fabricated for the concentric electrodes? Did they use a standard recipe for it? Maybe give a brief description? Added “The electrodes were fabricated on one crystal face by depositing a layer of amorphous Si followed by a layer of aluminum while the opposite crystal face was left bare. The five channel geometries were then defined by wet-etching.”

- line 49 - I would reference Fig 1 again here Done

- line 109 - you should clarify what you mean with “absolute calibration”. Maybe add a descriptor like “from first principles” or “absolute calibration based on the readout circuit design”. My concern is that you can get absolute calibration using a source as well, so we should clarify this point. Changed to “Unfortunately, this prevented an absolute calibration of the S1501 readout circuit gains from first principles.” Then later “This allowed an absolute calibration of the charge signals based solely on the readout circuit design for all 5 channels of S1502.”

- line 114 - related to the above comment, maybe say “ …of each pulse using an external gamma source for calibration.”We actually just normalized the seagull plot in the end for this device. Changed it to say “This allowed each pulse to be fit with an average template constructed from characteristic events, thereby estimating the relative amplitude of each event.”

- line 125 - explain why you used different sizes for collimators (relating to event rate etc)Added “The center channel source had a different geometry and required a larger collimator hole to achieve a similar event rate to the others.”

- line 128: “mean free path of 60 keV gammas in Si is ~ 3 cm” (adding in Si) Done

- line 151 - are you worried about cross-talk between channels at all? Maybe add a comment why this is not a concern After the description of the single channel cut, added “This ensured that we only considered events in which all carriers were drifted to a single channel.” This addresses “cross-talk” due to an event being split across channels. Capacitive cross-talk does change the absolute calibration, but is probably addressed by earlier comments about the readout circuit being used to get the abs calibration

- line 153 - “few minutes” - could you be more quantitative,e.g. 5 minutes changed to “\sim 2–10 minutes”

- line 179: “…cloud and trapping while drifting the charge across the crystal.” Changed to “charge loss was due to a combination of loss in the initial charge cloud and trapping while drifting the charge across the crystal.” The initial cloud may also have substantial recombination losses.

- line 186 (roughly) - it would be good to introduce the model briefly, it is hard to understand what the model parameters really mean without seeing the model. It would make the paper more complete, and would not rely on having the reader search the reference paper… Added the following, “Briefly, the model includes diffusion in the initial charge cloud with a length scale, $L$. All carriers that diffuse less than a field-dependent radius, parameterized as $bE^{-n}$, are trapped before the bias field erodes the cloud. The population of carriers which survive the initial charge cloud are then drifted across the crystal, attenuated by a field-dependent capture length, $cE^m$.”

- line 192 - how did you use channel A? Qouter cut? no, we just made sure it was consistent with noise for all the CCE data as we described earlier.

- line 204 - maybe was instead of were? Done

- line 206 - “to develop over time during operation,….” Done

- line 214 - this line makes it sounds as if photons directly provide energy to trapped charges to make them escape the traps, which I don't think it correct. Rather, the photons create free charges when they interact in the crystal. These charges then propagate through the crystal toward the charged traps and neutralize them (which they can do because our bias field is off, so the dominant field is due to the charged traps). I would modify the language in this line. Changed to “The photons produce free charge carriers in the crystal which then diffuse to overcharged regions, neutralizing any bound space charge.”

- Eq 3: maybe I am missing something, but why is the surface charge potential (the second term in the brackets) scaled in the same way as V_tot? I understand that V_tot has to be split between 2 or 3 capacitors in series (the crystal itself and one or two vacuum gaps), but the surface charge potential is only across one capacitor (the crystal). Surface charge on the crystal faces induces equal image charges on the electrodes. The additional potential drop this causes across the gap turns out to be proportional to that same length ratio.

- line 243: maybe specify the rate of carrier production here? Assuming an average ionization energy of 3.8 eV / charge pair in Si \cite{PEHL196845}, we calculate the rate of carrier production in the crystal of $2.4 \times 10^7$ pairs/s.

- Fig 9: clarify why the blue curve does not start at 60 keV at t=0. Also, I don't understand the last sentence in the caption, please clarify? Changed to “The measured signal curve is due to 60 keV events with the effects of CCE applied.” Will consider remaking the plot with y axis labelled as “measured signal”, not just “signal”.

- Fig 10: the data points for negative biases are visibly below the model band, so it is not obvious that we can claim the model explains the data here. You may want to consider adding error bars on the data points, or at least mention in the caption the sources of errors and their magnitudes, so that we can strengthen the claim of consistency between data and model. I haven't been able to find a convincing explanation for this discrepancy. It almost certainly is due to the readout circuit, possibly a gain offset. Channel D was the only one read by a RevC board, the others were RevC.1. Unfortunately, later attempts to directly measure the gains have not had clear or consistent interpretations. However, the fact that channel D is so different from the 2 channels surrounding it is suspicious. To address this, I've added 2 min error bars to Fig. 10 (which is about right considering the line resolution and average peak slope) and a comment in the figure caption, “The discrepancy in channel D at negative biases is believed to be due to a readout electronics issue.” I have also added the following lines to the text “This simple model agrees well with most of the data, especially at high bias voltages. The only major discrepancy is for channel D at low, negative biases. The source of this discrepancy is not yet fully understood, but is likely related to the amplifier electronics on channel D, which were an earlier prototype of those used on the other channels.”

- line 314 (roughly): could the counter-bias surface charge have something to do with the failure of UV neutralization? I mean, if UV photons interact in or very close to the layer of surface charge, could it be that all charges produced in their interactions get neutralized right there (i.e. with the surface charge)? Do you have a sense of what happens to the surface charge during IR neutralization - is it neutralized or does it survive neutralization? If you flip the bias after neutralization, I think the surface charge gradually flips the sign during operation… May want to comment on this a bit more, we can also discuss… Generally softened claims relating to how/where we think charges are being neutralized. Changed the section title to “Detector Reset with LEDs” (i.e. removed “neutralization” for “reset”). Start the section with “After the detector signal has been degraded by the counter-bias field, it needs to be reset by eliminating overcharged areas and ``neutralizing'' the crystal.” To make it clear we are assuming neutral=reset. Changed “is due to the different penetration depths” to “may be related to…” Finally, end by adding “Perhaps” to the statement “This difference in photon absorption profile means carriers freed by UV must diffuse further to neutralize the crystal bulk, making the UV process less efficient.”

- line 319: where did 10nm come from? inverse of attenuation. 1/(1e6/cm) = 10 nm

- line 324: where did 100 um come from? oops, should be 10 um. Similarly it is inverse of attenuation.

- line 328-329 - I am not sure about this interpretation, see my comment on line 314 above Removed this line. Indeed, we haven't demonstrated where the trapped charges lie when the bias is removed.

- line 336 - do we know that this is true for “standard” detectors that don't have surface charge built-up? See my comment on line 314 again… Changed this to be more speculative “It was found that IR photons neutralized the detectors more efficiently than UV photons, perhaps due to the greater penetration of IR in cryogenic Si.”

- line 340 - remove one “LED” Done

150mm Si paper v2.1

Please include a reference to figure number or line number when adding comments.

• Line 26 has “(450 - 580..” has spaces around the dash, but line 151 (5-20) has no spaces, nor does Line 167. Make these consistent. Fixed
• Line 151, people might wonder why first 5 minutes are ignored. Meant to indicate that we used somewhere between the first 5 and first 20 minutes of data for each series. Just simplified it to “first few minutes of data”
• Line 181 says 'sim2' not ~2. Fixed
• Fig 8 the 100% mark doesn't appear to line up with 60 keV?Added “The times taken for a calibration peak to fall to 80\% and 60\% of its initial value are measured.
• Fig 9 I feel like I am missing something obvious, but why does the 60 keV peak signal start at ~48 keV? I don't think I can find a justification in the text. Added “The signal is from 60 keV events, but a lower signal is expected when charge collection efficiency is accounted for.”
• Line 212 ” It has been observed that not only the calibration peak, but the entire measured energy spectrum falls at the same time and rate..“ Is it worth presenting concrete evidence of this by showing the decay of Am and Co? Did S1502 ever get a nice data series with both sources decaying? I don't remember there being many good examples in S1501 data.There wasn't any Co data collected for S1502. This effect can be seen clearly when an energy vs. time plot is constructed with a logarithmic energy scale. Then both the 60 keV and muon events can be seen to decay together.
• […]surface charges versus the idea that IR does better and neutralizing because it reaches the bulk, as those ideas seem a little inconsistent to me. Good point, we should try to address this or at least speculate. It may just be that, when the detector is grounded for flashing, the charges don't stay at the surface, but start to diffuse back into the bulk. Then the argument about IR penetrating further holds. Added “As demonstrated in \ref{sec:counter_bias}, the accumulated surface charge determines hold time. This implies that, when the bias is removed, the surface charges must lie at a depth accessible to the IR but not the UV (i.e. $>$ 10 nm from the surface).”

• Abstract: DM and CNS are mentioned in the Abstract, but only DM is elaborated on in the Motivation. I suggest removing the CNS mention. I think the argument for scaling-up is clearer for DM where the cross-section is completely unknown. CNS removed from first abstract line
• 17: “High voltage” doesn't mean a lot in context - it's high compared to ZIPs and iZIPs, high compared to epsilon/e = 4V. Maybe don't use that expression and instead say “detectors biased at 50-100 V” Change made
• 33: Why ”(Si)“? Why not just “silicon”? For search purposes? silicon is mentioned ~15 times in the paper. 'Si' saves us a little space.
• 42: I'm not sure the description of the electrodes is complete enough. Are the outer ones “circular”? Maybe - or maybe the right word is “annular”. Maybe just say “five concentric electrodes”, and could also mention they are equal in area. One thing that might be confusing from the diagram is the segmented channels since we don't mention they are connected. So “concentric” would help clarify. Or better: in the caption itself mention that the segments of channels B and C are connected with wirebonds to form continuous annular electrodes. ”…five concentric electrodes..“ and added “The segments of channels B and C are connected with wirebonds so that each forms a continuous annular electrode.” to the caption.
• 105: I'd change “the ringing observed in the pulse tail” to “ringing in the pulse tail”. Done
• 112: I'd change “S1502 produced exponential pulses as expected” to “S1502 produced exponential pulses without ringing”Done
• 119: “down the crystal axis” - maybe “along the crystal axis” is better? Done
• 147: maybe replace “cuts” with “criteria”. Done
• 173: “There, it was found that,…” - remove comma after “There”.Done
• Fig 7: In the trapping model fit figure, wouldn't it be better for the legend to match the previous figure in terms of channel? i.e “B, C, D, E” rather than “QE, QD, QC, QB”. Calling them “Q” is a throwback to ZIPS where we distinguish charge and phonon channels. Done and added error bars to “Penn” data
• 188: “used to set cuts on other channels” - probably should be changed. Something like “used for data selection criteria for other channels”. Changed to “It was, however, used for data selection criteria for channels as described above.”
• 224: “constant rate of energy deposition in the detector” - I'd add the phrase “from ionizing radiation” Done
• 227: “mainly from background events” - I'd prefer “mainly from background radiation” Done

”…there is one thing that you do that you shouldn't: you forward reference. In other words, you ask the reader to refer to a section you haven't covered yet. You should avoid doing this as it means the reader needs to keep jumping around to keep up with you. If you are referring to a concept when you describe something, you should have already covered that concept earlier in your article.“ Left an early reference to flashing section, but noted that it would be discussed later in depth. Also reorganized the neutralization section to discuss flashing at the beginning. Also tried to handle in the same way the early reference to counter-bias effects in the CCE section. Just say there are “time-dependent” changes which must be removed when measuring CCE and which are discussed later.

1. Consider adding “with Contact-Free Electrodes” to the title. This may hint that the “first” is a large crystal w/ contact-free electrodes, not just that this is the first 150mm detector. However, a good part of the paper does involve the c.f. electrode. Changed to “Performance of the First 150 mm Diameter Cryogenic Silicon Ionization Detectors with Contact-Free Electrodes”

2. Author affiliations for Dan and I should be University of Minnesota. We may have left, but the work we did in relation to this paper was performed at the University of Minnesota. Hence, I do not think current affiliations matter. You may wish to add current affiliations as footnotes. Best to ask Vuk about this. Done

3. In the abstract, the sentence “The detectors are 25 mm and 33 mm thick with masses 1.7 and 2.2 times larger than those currently used or planned.” is a bit confusing. Perhaps replacing “currently used or planned” with “described above” or “for SuperCDMS described above” would be better. Changed to ”.. currently planned for SuperCDMS.“

4. Line 18, Ref. 1 is out of order. Ref. 1 was in the abstract. Removed reference in abstract to make the abstract stand-alone.

5. Lines 52 to 53, it might be clearer to use “Infrared” and “Ultraviolet” rather than the acronyms, as this is the first time you use these terms. I would suggest “Infrared (IR) at 940 nm and Ultraviolet (UV) at 310 nm Light Emitting Diodes (LEDs)…”. Done

6. Line 60, no need to put “to first order” in brackets (or parentheses in US-speak). Done

7. Line 69, “crystal” needs an article. Done

8. Line 101, might be good to add something like “in Eq. (1)” to the end of the sentence. Done

9. Line 116, “analysis” needs an article. Changed to ”…from the analysis.“

10. Line 120, replace “8 mil” with a decimal number. Most readers outside the United States are not familiar with “mil”. For example, I have often used it to mean “millimetre” in a similar fashion to how we use “micron” to refer to “micrometre”. Done

11. Line 124, ditto comment 10 above regarding the use of “mil”. Done

12. Line 124, in LaTeX you can use a double hyphen, which will look better. In other words type “B–D” rather than “B-D”. The single hyphen is too short, so rather than B to D, it implies B-D as being two words joined together. Hope that makes sense. Yep, it already is an “en-dash”

13. Line 127, try not to put a ~ sign before “half”. It doesn't look too good and reads poorly. Better to just spell it out in words. changed to ”… by about half“

14. Lines 128-134, it is a little unclear what you mean when you say “reversed”. One reading would be that the holes traverse the bulk of the crystal and thus generate most of the signal (try to avoid repeating “bulk” in the same sentence). However, I am not sure I have understood that correctly. It would be better to be a bit more explicit here to avoid any confusion. Indeed, this is hard to explain clearly. Changed to the following single sentence “In S1501 (S1502), the calibration sources were on the readout (bias) electrode, so positive bias field configurations preferentially involved electrons (holes) traversing most of the crystal thickness and thus generating the bulk of the induced signal.”

15. Figure 5 caption, you can't start a sentence with digits. Also, it doesn't read very well. I would suggest changing this to: “Shift of the 60 keV spectral line as measured by S1502 at various applied bias voltages. Fewer charges are collected at lower bias voltages because of trapping, recombination and diffusion effects, thus leading to lower charge collection efficiencies.” Something like that. Your figure caption should not require the reader to go back into the text to find out what you are trying to show. Used word-for-word, thanks!

16. Lines 136-144, I would suggest mentioning recombination in this paragraph too. Changed to “As the bias voltage is increased, carriers are more easily able to escape traps, spend less time diffusing, and are less likely to recombine. This leads to increased CCE and thus larger signal amplitudes for a given interaction energy.”

17. Line 159-160, rephrase this to avoid repeating “uncertainty”. I suggest using “the error” instead of the second occurance of “uncertainty”. Done

18. Lines 165-166, you are correct that there should be no signal. However, Penn et al. have a point at 0 V. Do you know why? It's actually at 0.1 V/cm, they still see some small signal there.

19. Line 181, “sym2”. I assume you forgot the \ before “sym” in the LaTeX file. Done

20. Figure 7 caption, I think you mean “Ref. [24]”. Nice catch!

21. Line 210, probably best to say ”(both in the case of S1501“) instead of ”(both)“ to avoid any confusion.Done

22. Figure 8 caption, I know that S1502 is mentioned in the plot's title, but I would also mentioned it explicitly in the caption. Additionally, I would change “partly canceling” to “counteracting”. Finally, you should mention what you mean by “100%”, “80%” and “60%”.Done

23. Figure 9, the text box that states “V_tot = 5 V” and “90.3 MeV/s” seems a little out of place. You need to either explain them better on the plot itself or mention them in your caption. Changed to “An example of the modeled counter-bias development and resulting calibration signal for the case of a total detector bias of 5 V and energy deposition rate in the detector of 90.3 MeV/s.”

24. Lines 245-246, consider using “hold times” rather than “fall times” as Fig. 10 makes no mention of “fall times”.Changed this to discuss only “hold times”

25. Figure 10, it is difficult to make out the markers for Channel B. You may wish to change the markers to hollow rather than filled variants. Remade figure with open markers.

26. Figure 10 caption, this is the first time you mention a model of accumulating surface charge in a parallel plate capacitor. I haven't seen it mentioned in the text. At a cursory glance, it would appear as though you are using a different model to the one you describe in the text. Changed to “The results of a simple model of accumulating surface charge are seen to fit the data well.”

27. Figure 10, the referee might ask why the 1 sigma band is smaller at negative potentials. I think you should mention the reasons, if not in the text at least in the figure caption. Added “The model uncertainty is higher at positive biases due to larger variation in measured CCE curves at positive biases.”

28. Line 255, “factor of a few larger” is very vague. Too vague. Your definition of “few” may differ from mine, and it is all relative. State a number. It varies considerably by detector, but a factor of 10 is probably fair. Changed to reflect this, but kept the speculation about how this could be sensitive in low-background environments.

29. Line 265, “100 us pulses of at a rate of” should be “100 $\mu$s pulses at a rate of”. Done

30. Lines 268-270, you should mention that the detector is not under bias during cooling.Done

31. Line 275, I think you mentioned it above, but it might be worth mentioning the Si band gap energy again.Done

32. Lines 276-280, I know what you are trying to say, but the way you have phrased it, it sounds like you are contradicting yourself. You mention at the beginning of the paragraph that you tested neutralisation using various LED flashing schemes while the detector was grounded. Now you mention that you performed these tests at a bias of 10 V. I would add something like, “After flashing, a 10 V total bias was applied to the detector.” I think you can better phrase this sentence to avoid any confusion. Changed to “After each flashing, a 10 V total detector bias was applied for data collection. With a standard flash, this 10 V bias provides an 80\% hold time of 6 minutes.”

33. Lines 292-294 (and Figure 11 caption), you state that the UV LEDs did no better than grounding the detector without flashing, but in the following paragraph you state that grounding the detector for 25 minutes led to full neutralisation. So, would flashing the UV LEDs for 25 minutes (or 20 minutes followed by 5 minutes of cool down) achieve full neutralisation? I ask as this is what you have implied. I expect it would as the background would neutralise the detector. However, if you haven't made any measurements of how long it would take the UV LEDs to fully neutralise the detector, it might be best to leave this sentence out as you are just assuming that it is no better. For example, can you say for sure that flashing with UV LEDs for 15 minutes would not fully neutralise the detector? In other words that they have no effect whatsoever? Fair point. Added the disclaimer “For the range of flashing tested, the UV LEDs …“

34. Lines 305-319, I like your explanation but what do you mean by “penetration depth”? The reason I ask is that a penetration depth of 100 um would imply that the bulk of the crystal is not neutralised by flashing. In other words, anything deeper than this depth would not be affected by flashing and therefore only background events neutralise this area. You do answer this to the end of the paragraph, but it might be better to talk about this earlier in the paragraph to avoid confusion. Moved the sentence about freed carriers diffusing into the bulk to the beginning of the paragraph.