The study of Richter and Tsukamoto entitled “Investigation of quartz ESR residual signals in the last glacial and early Holocene fluvial deposits from the Lower Rhine” presents original results that are very important for the ESR dating community.  The manuscript is well written. The main message is that Al-h as well a Ti-Li and Ti-H signals in quartz are not fully reset in the case of fluvial sediments. By analyzing Holocence samples that were securely dated by OSL for ESR dating the authors report a mean value of 1350 ± 120 Gy for the residual of Al centre, 610 ±60 Gy for the lithium-compensated Ti centre (Ti-Li), 170 ± 20 Gy for the hydrogen-compensated Ti centre (Ti-H), and 470 ±50 Gy for the signal originated from both the Ti-Li and Ti-H centres (termed Ti-mix), concluding that fluvial sediments carry a significant residual dose, and therefore the subtraction of residual dose using a modern analogue is highly recommended to obtain reliable ESR ages. I fully concur with the main message of the paper and I think it is worth being published.

However, besides this, the study presents yet another problem, namely the poor dose recovery results. The way the dose recovery experiment was conducted is problematic in my view and leads to circular reasoning. If the authors cannot prove that my observation is wrong, the interpretation provided now needs to be revised. The approach of the dose recovery test was to use the “natural” sample that holds this significant residual, irradiate it with 1000 Gy and as the authors explain “The dose recovery ratio was calculated by subtracting the natural De from the recovered dose and the difference of the natural + 1000 Gy and the natural De was then divided by the added dose of 1000 Gy.” The problem is that by doing so, it is inherently assumed that the determined De is correct. While this would not be a problem in a luminescence experiment with a residual of say 10 Gy compared to the given dose of 1000 Gy, here it is, because the magnitude of the residual, in this case the De is of the same order of magnitude as the given dose. The authors then state that “The dose recovery ratio for the Al signal is high with 1.74 ± 0.16, which indicates a sensitivity change due to thermal treatment during SAR protocol, therefore the reported residual doses may be overestimated.” But to my understanding, first it was assumed accurate based on your experiment or am I missing something? And you subtracted this value in the numerator of your ratio. The way this experiment stands defies in my view the idea of a dose recovery test as the given dose should be known and the equivalent dose is by definition an unknown. The only way in which this dose recovery could work is if the “sensitivity” is the same in the De (or residual dose here) measurement and during the dose recovery experiment. Which brings me to the point when I have to mention that I find the term “sensitivity change” to be extremely difficult to be conceptually integrated in ESR dating. Anyway, conceptually the problem is that you draw a conclusion on the accuracy of the measurement protocol (namely that it is not very accurate) by feeding into your protocol a value measured by the same protocol that is assumed to be accurate.

My second problem concerns figure 5 where the residual of Al-h is plotted as function of that of Ti-Li. While I fully agree with the logic of this approach and I consider it a very elegant way to extract the unbleachable part of the Al-h signal, even by eye one can see that that line does not fit the data. I plotted the data myself and obtained an intercept of 134 ± 208 Gy by simple linear fitting so I cannot understand how that regression was obtained. The result I have quoted above is of course very concerning, as I understand that the authors know from experience, by the simple examination of the signals measured on modern analogues from aeolian environments that Al-h is not fully bleachable. I fully support this view. I think that the fact that the intercept is close to zero and the scatter in your data simply reflects the problems associated with the measurement protocol, as mentioned in my comments above and below.

While one should not expect this problem to be solved in the framework of this particular study the observation “The De value of the Al centre was initially decreased by the preheat at 160 °C, but shows a steady increase in De with increasing preheat temperature. At 220 °C no De calculation was possible, because all regenerated signal intensities were below the natural” is extremely concerning. I suggest the authors show some of the measured signals and the constructed dose response curves in the supplementary file. At this point one has to ask himself why one needs to preheat in ESR dating? How come a single type of defect can have a part that is not thermally stable and one that is not, as mentioned in Toyoda and Ikeya, 1994 QG, and explained here as “a preheat plateau test was carried out to assure only stable signals are used”? And how does that relate to the values of E and s quoted in studies such as Richter et al., 2020, QI? ESR is selective, as such only one type of defect is targeted. Does this mean that we acknowledge that both localized and delocalized transitions are at play? If the localized route is the one that leads to an apparent instability, then one should expect the degree of instability to be dose dependent (see Benzid and Timar-Gabor 2020, AIP Advances 075114). If indeed there is a degree of instability that is dose dependent that might explain the observation I am referring to above.

My last comment concerns the interpretation “The opposite tendency of the heat treatment for the Ti-Li and Ti-H centres in the preheat plateau test (Fig. 2) suggests that some charge transfer between the Ti-Li and Ti-H centres is likely.” Personally, I am unable to grasp what model is proposed. Please be more specific. What is the transferred charge?

Technical comment: line 70-reference to table 3 should be made instead of table2.

This is an interesting study that provides new data around the behaviour of the ESR signals in quartz from young (Late Pleistocene to Holocene) sediment. In particular, the authors evaluate whether the ESR signals associated to the Al and Ti centres are actually fully reset during sediment transport and prior to deposition. This is done by comparing the ESR dose estimates with those independently obtained from OSL measurements. As a matter of fact, the ESR dose values are significantly (one or two orders of magnitude) higher (between 150 and 1350 Gy depending of the centre considered) than the OSL ones (15 to 35 Gy). This indicates that the samples have not been fully bleached during transport, resulting in a massive dose (and age) overestimation if the unbleached component of the signal is not taken into account. Given these results, the authors recommend the use of modern-analogue samples to evaluate whether the ESR signals of Pleistocene samples have been fully reset.

The paper is very well written, easy to read and some of the conclusions are supported by the data displayed in this work, although the general take-home message should be in my opinion somewhat moderated (and especially the last sentence of the abstract: ‘The results of this study suggest that fluvial sediments carry a significant residual dose, and therefore the subtraction of residual dose using a modern analogue is highly recommended to obtain reliable ESR ages.’). The conclusions may be valid for this specific set of samples, but should not be universally extrapolated to all samples in any sedimentary context. The present study is based on a limited number of samples, which all come from the same sedimentary environment, and supposedly have the same bleaching and deposition history. It is therefore not surprising that they all provide consistent results. The sedimentological characteristics of the sediment are known to possibly impact the level of optical bleaching achieved during transport (e.g., Voinchet et al., 2015). Moreover, there are many works in which ESR dating (using Al, Ti-Li and/or Ti-H signals) provided consistent results with independent age control based on Luminescence, Cosmogenic, Palaeomagnetism or Ar-Ar methods (see an overview in Duval et al., 2020b), for chronologies ranging from the Early Pleistocene to the Late Pleistocene. These should be considered as empirical evidence indicating that the ESR signals have indeed been fully reset (i.e., zeroed for the Ti signals, and reset to the unbleachable level for the Al signal) in frequent occasions, and that the so-called 'residual dose' is not significantly impacting the results. My point here is not to minimize the importance and relevance of the present results, but just to put them in a broader perspective. Whether the ESR signal is fully reset may depend on many factors, including the transport conditions, and nature and composition of the raw sediment. The present study is based on samples from a single environment that is perhaps simply not favourable to ensure complete bleaching of the ESR signals. Finally, there are several studies for which the De values derived from either Al or Ti centres are <100-150 Gy (see an overview in Bartz et al., 2020), i.e., below the dose estimates obtained in the present work. This indicate that the magnitude of this so-called unbleached (residual) component is sample- and environment-dependent. This aspect would deserve to be mentioned in the present work.

Additionally, although the  ‘residual dose’ may be due to an incomplete bleaching of the ESR signals prior to deposition, other sources of uncertainty should also be explored, as they may impact the dose evaluation. For example, there is no mention at all of the real uncertainty associated to the ESR measurements (see point #2 in General comments), as well as of the potential bias simply induced by the way the ESR intensities are measured (see point #1). If the authors could explore and discuss these questions in the manuscript, that would surely bring added value to the study.

General comments

1. Identification of the ESR signals and measurement of ESR intensities.

There is still little consensus within the community around the way Ti ESR signals are identified, measured and reported. First of all, the Ti-H signal is made of a series of doublet, one of them being centred at g = 1.915. This one overlaps with the Ti-Li peak @ g = 1.913. Consequently, the peak reported as ‘Ti-Li’ by the authors on Figure 1B and throughout the manuscript  (called option D in Duval and Guilarte, 2015) is actually resulting from the mixed contribution from Ti-H and Ti-Li signals (e.g., Toyoda et al., 2000). Therefore, this should be reported as ‘Ti signal’ or ‘mix Ti’, but not as Ti-Li (unless the authors can demonstrate that the Ti-H component has no impact on the peak). The only way to measure the single contribution of the Ti-Li centre is to measure the peak intensity @ g= 1.979 (Option E in Duval and Guilarte, 2015; Absorption line 1 in Beerten et al., 2020). If option E gives the same result as D, then it would be fair to consider that the Ti-H component has no significant influence on the peak intensity measured @ g = 1.913.

The authors decide to use an unusual way of measuring the Ti signal (so-called option B in Duval and Guilarte, 2015; Absorption line 2 in Beerten et al., 2020). This one is not frequently employed for dating purpose (I only recall Beerten et al., 2020). Therefore, their results are not directly comparable with all previous studies that are cited in the MS, since they are mostly based on Option A (as per reported in Duval and Guilarte, 2015) (e.g., Timar-Gabor et al., 2020; Rink et al., 2007) or option D (Tissoux et al., 2007). This is why I would encourage the author to report the data from options A and E (at least in SM), so that : (i) the present work can be directly comparable with other studies, and (ii) we can evaluate whether the dose results significantly differ depending on how the signal is measured.

2. Measurement procedure and uncertainty associated to ESR intensities.

Unlike for tooth enamel, ESR measurements of quartz samples at low temperature are usually characterized by a non-negligible uncertainty (at least a few %) that may result from a combination of different factors, such as the stability of the ESR spectrometer, the heterogeneity of the quartz samples, the angular dependence of the signal, the position of the tubes in the cavity, etc. Although ESR measurements are usually performed under controlled conditions, measurement precision for the Al and Ti signals is typically of at least a few %. This has a non-negligible impact on the D_E values, which is why repeated measurements should be performed in order to evaluate the repetability of the D_E estimates (e.g. Duval et al., 2017, 2020a). A variability of at least 10% may be frequently achieved (at least with the MAAD method) and it may sometimes be >30%, casting thus in this case reasonable doubts on the reliability of the ESR data obtained.

In their measurement procedure, the authors take into account the angular dependence of the signal by rotating the tube 3 times in the cavity. So I understand that they derived a mean ESR intensity value and associated error, which was then used for DRC fitting and dose evaluation (correct? please clarify in the MS). So, I wonder:

• About the magnitude of the experimental error derived from the rotations in the cavity ? This error may significantly vary depending on the signal considered. It would be interesting if the authors could comment on this.
• In which extent this experimental error may impact the De value? This should be discussed.

Moreover, there is no Dose Response Curves (DRC) displayed. Can the authors provide some examples in the MS for the different signals measured ?

Apart from the angular dependence of the signal, there is no mention of any repeated ESR measurements, while it is known that the variability of the ESR intensities over successive days of measurements may be significant. And this variability may be even higher when using a HS cavity (compared with a ST cavity). This may be especially crucial for the Ti-H signal, whose ESR intensity is the weakest of all signals, resulting in significant experimental errors that may sometimes exceed 10%. In other words, a poor De repeatability could possibly partly explain the dose results obtained in this work, which is why this should be evaluated for the present set of samples.

Finally, only 1 single aliquot has been measured per sample, but again, we have no idea of how representative is this aliquot of the whole sample. It seems essential to measure a few aliquots per sample in order to evaluate the dispersion of the results. I imagine this would be a standard procedure in OSL dating, and I would recommend to do it here in order to evaluate in which extent the homogeneity of the sample may impact dos estimates.

I realise that these extra measurements might be time consuming, but they can actually be carried out much faster with the SAR method than the MAAD with given the limited number of dose steps and irradiation times. My opinion is that all these issues should be thoroughly explored to see in which extent they may significantly (or not) impact the dose estimates.

3. Residual dose

This might go beyond the scope of the present paper, and would perhaps deserve a proper consensus within the ESR dating community, but I am personally struggling with the use of the term ‘residual dose’, for two reasons. First, I am not sure whether the term is actually correct: technically, it should rather be a ‘residual signal’, which can then be converted into a dose value. Second, I find the term ‘residual dose’ misleading, as it may correspond to either (i) the unbleachable component of the Al signal, or (ii) the remaining component of the Al and Ti signals due to incomplete reset prior to deposition, (iii) or sometimes even both.

Moreover, the comparison of residual dose values from the various signals may be biased by the unbleachable component of the Al signal (see specific comments #8 & #9). This one should be subtracted to the so-called ‘residual dose’ in order to make it directly comparable with those from the Ti centres. Actually, I am not sure why the magnitude of this residual unbleachable component of the Al signal has not been evaluated in the present study using a sumlight simulator, so that it can be subtracted from the signal of the Natural and Gamma-irradiated aliquots. And I wonder in which extent this unbleachable component of the Al signal may impact the dose recovery experiment (and the poor ratio obtained), since this component cannot be optically reset (but thermally yes).

Specific comments

1. Abstract:

• l.4: does the value of 1350 Gy obtained for the Al centre include the unbleacheable component of the signal? if so, this should be mentioned.
• l.8 to 10: I would suggest to add a sentence indicating that the dose recovery ratios obtained for Al and Ti-H signals suggest dose estimates may be significantly overestimated and underestimated, respectively.
• l.9 to 11: Again, I would suggest some moderation in the final sentence: the results obtained by the authors indicate that the fluvial sediment samples that have been investigated in this work carry a significant residual dose, but this conclusion should not be extrapolated to all samples in any context. And especially for the Al and Ti-H signals given their dose recovery ratios.

2. l.20-22: there is also significant variability among the samples for the bleaching kinetics of the Ti-H and Ti-Li signals (see Tissoux et al., 2007; Duval et al., 2017). This may be mentioned.
3. l.22-23: The two parts of the sentence are not incompatible. The signal may be fully bleachable and incompletely bleached due to insufficient exposure to sunlight. Please rephrase without the ‘although’.
4. l.24: studied or reported? All the publications from the ‘French’ group systematically provide the so-called ‘bleaching coefficient’ for the Al centre, which is the relative intensity of the unbleachable component (e.g., Voinchet et al., 2020; Duval et al., 2017, 2020a).
5. l.24: use ‘residual signal’ instead of ‘residual dose’
6. l.24, ‘So far very few studies have reported residual doses of the quartz ESR signals from sediments’: just to be clear, you mean residual component due to the incomplete reset of the ESR signal prior to sediment deposition?

7. l.26: again, technically, these are not residual doses’ but rather ‘residual signal that correspond to dose values of XXX’.
8. l.27-29: Question: does the so-called ‘residual dose’ reported by Tsukamoto et al. (2017) for the Al centre includes both unbleachable and unbleached (due to incomplete reset) components ? This should be clarified, otherwise these large numbers provided for the Al centre may be misleading.
9. l.28-32: I believe the same comment applies to the study by Timar-Gabor et al (2020): in order to be discuss values that are directly comparable for the Al and Ti centres, the author should report the so-called residual dose for the Al signal without the unbleachable component. Otherwise, this may be again misleading: the huge apparent residual dose for the Al centre is actually made of 2 components, with one of them that cannot be optically reset. The standard procedure in the dose evaluation associated with the Al centre does include the subtraction of the unbleachable component of the signal (e.g. Voinchet et al., 2020; Duval et al., 2017, 2020a)
10. l.34: again, what does residual dose mean here? The magnitude of the unbleachable component is systematically reported in all the studies mentioned (cf., ‘bleaching coefficient’ in the tables).
11. l.35: Toyoda and al do not report ESR any age estimates in that paper.
12. l.45: why for the first time? The authors report many previous studies with a similar approach (Beerten et al. 2006; Tsukamoto et al., 2017; Timar-Gabor et al., 2020
13. l.48, section 2. Samples:

• Just checking, the samples analysed in the present study are the same that have been analysed earlier by Lauer et al (2011): i.e., these are the exact same prepared quartz ? or do they come from a another sample preparation? This should be clarified in the MS.
• l.48: The authors should provide a basic description of the sedimentary environment (facies, grain size) for each sample: these parameters are known to impact the reset of the ESR signals (e.g., Voinchet et al., 2015)
• l.48: gain size of the samples dated ? (this can be added to Table 2)
• l.50-55: Unfortunately I could not access the paper by Lauer et al. (2011). Would be good if the authors could provide some additional info about how the OSL De values were calculated (single grain? Single aliquot? How many aliquots/grains measured to get the mean De values? etc.)

14. l.59, Section 3. ESR measurements:

• l.65: ‘modulation amplification’ or ‘modulation amplitude’?
• l.66: remove ‘of the spectra’
• l.68: what does ‘homogenous value’ mean? What about rephrasing as ‘to take into account the angular dependence of the signal’ instead ?
• l.69: a microwave power of 10mW for the Ti signals seems too much. Have the authors performed a microwave saturation curve ? Our data show that the Ti signals saturate at such power, which is why we usually measure using 5 mW.
• l.74: that is correct, and therefore the peak @ g=1.915 does not result from the single contribution of the Ti-Li signal, but from a mixture of both centres. This should be reported in this way throughout the manuscript.
• l.79: What is this reference ‘Tsukamoto, 2019, unpublished’? Is this a technical report? Not sure this should be mentioned here as it cannot be accessed.

18. l. 99, Subsection Equivalent doses, residual doses and ESR ages:

• l.109: I am not sure in which extent the OSL dose rate can be directly used for ESR. Maybe they should be adjusted. There are some specificities, like the alpha efficiency of 0.07 (Bartz et al., 2019b), that should probably be taken into account for a small (but non-negligible) internal dose rate and external alpha dose rate components.
• l.111: ‘These residual ages show how significant the effect of the residual dose may be in ESR dating of fluvial sediments.’ This is valid for the present set of samples. Not all samples behave the same.

19. Subsection Dose recovery test, l.121: in which extent the unbleachable component of the Al signal may impact the dose recovery ratio ?
20. l.127:, ‘The results clearly show that the ESR De for all samples are significantly larger than the OSL De of Lauer et al. (2011) and therefore residual subtraction is highly recommended’. 2 comments here: (1) if this residual component resulting from incomplete bleaching of the signal is not subtracted, then the ESR age results should be regarded as maximum possible burial ages; (2) residual subtraction would require the use of modern-analogue samples, but this should not be regarded as the perfect solution: it is based on the assumption that modern-analogues and Pleistocene samples that are being dated have experienced similar transport and bleaching conditions, which can not always be verified. Therefore, subtracting the ‘residual’ signal derived from modern-analogue samples may introduce another source of uncertainty in the equation, and does not ensure that the ESR ssignal of the dated sample has actually been fully reset.
21. Figure 1: I think there is an issue with the peaks reported for the Ti-H signal. There should be doublets for each g value, and there is actually an overlap on the Ti-Li peak @1.913 (e.g. Toyoda et al., 2000).
22. Figures 3 and 5: does the Al residual De values include the unbleachable component ? I believe it should be subtracted, in order to make the all data really comparable.
23. Table 1, footnote: *including unbleachable signal component

References

Beerten, K., Koen Verbeeck, Eric Laloy, Veerle Vanacker, Dimitri Vandenberghe, Marcus Christl, Johan De Grave, Laurent Wouters (2020) Electron spin resonance (ESR), optically stimulated luminescence (OSL) and cosmogenic radionuclide (CRN) dating of quartz from a Plio-Pleistocene sandy formation in the Campine area, NE Belgium. Quaternary International.

Bartz, M., Arnold, L.J., Demuro, M., Duval, M., King, G.E., Álvarez Posada, C., Parés, J.M., Rixhon, G. and H. Brückner (2019a). Single-grain TT-OSL dating results confirm an Early Pleistocene age for the lower Moulouya River deposits (NE Morocco). Quaternary Geochronology 49, pp. 138-145.

Bartz M., Arnold L.J., Spooner N., Demuro, M., Campaña Lozano I., Rixhon G., Brückner H., Duval M. (2019b). First experimental evaluation of the alpha efficiency in coarse-grained quartz for ESR dating purpose: implications for dose rate evaluation. Scientific reports 9:19769, pp. 1-10. https://doi.org/10.1038/s41598-019-54688-9.

Duval, M. and Guilarte, V. (2015). ESR dosimetry of optically bleached quartz grains extracted from Plio-Quaternary sediment: Evaluating some key aspects of the ESR signals associated to the Ti-centers. Radiation Measurements 78(0): 28-41.

Duval, M., Arnold, L.J., Guilarte, V., Demuro, M, Santonja, M., Pérez-González, A. (2017). Electron Spin Resonance dating of optically bleached quartz grains from the Middle Palaeolithic site of Cuesta de la Bajada (Spain) using the multiple centres approach. Quaternary Geochronology 37, pp. 82-96.

Duval M., Voinchet P., Arnold L.J., Parés J.M., Minnella W., Guilarte V., Demuro M., Falguères C., Bahain J.-J. & Despriée J. (2020a). A multi-technique dating study of two Lower Palaeolithic sites from the Cher Valley (Middle Loire Catchment, France): Lunery-la Terre-des-Sablons and Brinay-la Noira. Quaternary International 556, pp. 71-87. DOI: 10.1016/j.quaint.2020.05.033.

Duval, M., Arnold, L.J., Rixhon G. (2020b). Electron Spin Resonance Dating in Quaternary studies: evolution, recent advances and applications. Quaternary International 556, pp. 1-10. Guest Editorial.  DOI: 10.1016/j.quaint.2020.07.044.

Rink, W.J., Bartoll, J., Schwarcz, H.P., Shane, P. and Bar-Yosef, O. (2007). "Testing the reliability of ESR dating of optically exposed buried quartz sediments." Radiation Measurements 42(10): 1618-1626.

Timar-Gabor, A., Chruścińska, A., Benzid, K., Fitzsimmons, K.E., Begy, R. and Bailey, M. (2020). Bleaching studies on Al-hole ([AlO4/h]0) electron spin resonance (ESR) signal in sedimentary quartz.Radiation Measurements 130: 106221.

Tissoux, H., Falguères, C., Voinchet, P., Toyoda, S., Bahain, J.J. and Despriée, J. (2007). Potential use of Ti-center in ESR dating of fluvial sediment. Quaternary Geochronology 2(1–4): 367-372.

Toyoda, S., Voinchet, P., Falguères, C., Dolo, J.M. and Laurent, M. (2000). Bleaching of ESR signals by the sunlight: a laboratory experiment for establishing the ESR dating of sediments. Applied Radiation and Isotopes 52(5): 1357-1362.

Voinchet, P., Toyoda, S., Falguères, C., Hernandez, M., Tissoux, H., Moreno, D. and Bahain, J.J. (2015). "Evaluation of ESR residual dose in quartz modern samples, an investigation on environmental dependence." Quaternary Geochronology 30, Part B: 506-512.

Voinchet, P., Pereira, A., Nomade, S., Falguères, C., Biddittu, I., Piperno, M., Moncel, M.H. and Bahain, J.J. (2020). "ESR dating applied to optically bleached quartz - A comparison with 40Ar/39Ar chronologies on Italian Middle Pleistocene sequences." Quaternary International.

Mathieu Duval, Burgos, 14/05/2021.

PS: not sure why, but the numbering of the sections and bullet points changes from the Edit to the Preview versions. Hope this can be fixed.