Second Review report for manuscript se-2019-205
I read carefully the answers to the two review reports and the new version of the manuscript.
A lot of efforts have been done that have greatly improved the scientific quality of the manuscript. However, I believe that there are still a few unanswered questions. This is mostly due to the fact that some explanations provided by the authors are not satisfactory or are missing. These points must be clarified.
My answers are inserted in the following “comment/answers” text.
Comment 1: The results about partitioning are equivalent to those of a previous published study by another group (nothing new) and the discussion about one low pressure result (at 0.2 GPa) is not convincing (see specific comments below). The results about bromine speciation in high pressure fluids are new and they deserve to be published, unfortunately the partitioning of Br is not measured
for the same chemical system (haplogranite, HPG) than its speciation (Na2Si2O5, NS2), which makes any comparison difficult. Therefore I would recommend to delete the part about partitioning, or at least to provide convincing explanation (see specific comments), and to focus on the speciation results.
Answers:
• While the in-situ Br partitioning experiments are not the first of their kind, they provide a
unique opportunity for cross-checking experimental reproducibility and thus, we believe they
deserve to be included in the manuscript. References to previous work by Bureau can be found
throughout the manuscript and the favourable comparison between their and our studies
further supports the reliability of the in-situ measurements. It should be stressed that in-situ
measurements as those reported here and in rare previous work are extremely challenging, but
the only reliable way to assess element partitioning and speciation at extreme P-T conditions,
and thus any new data would be a valuable contribution to the field. Therefore, we prefer to
keep the partitioning experiments as part of the current manuscript.
Reviewer’s answer
OK
• Partitioning experiments involved haplogranite melts (Si, Al, Na, K), while Br speciation in
melts could only be determined for sodium disilicate (Na2Si2O5) due to insufficient Br
concentrations in the haplogranite melt (400-2000 ppm Br). Yet, both XANES and EXAFS
analyses (Figures 4 and 5; Table 3) show that Br local environment is very similar in the
haplogranite and NS2 glasses. Thus, it can be expected that Br incorporation mechanism in
both melts is similar and controlled by the presence of alkalis, either Na or K, and that all
peralkaline silicate melts will have affinity to incorporate high amounts of Br under high P-T
conditions. A note has been added in Lines 479-482 to clarify this point and emphasize the
similarities between the haplogranite and NS2 systems.
Reviewer’s answer
It has been experimentally demonstrated that the solubility of Br in silicate melts is highly dependent of the composition of the silicate (SiO2, Al/alkalis ratio, see Bureau Metrich, GCA, 2003), it is not convincing to claim that the speciation of Br should be the same for haplogranite and for NS2. Results obtained on glasses (Figure 4) cannot be used to predict the speciation of Br in melts.
• We believe that underlining the differences between our low P DBr (4.8 at 800 °C, 0.2 GPa)
and those of Bureau et al. and Cadoux et al (17.5-20.2 at 900 °C, 0.2 GPa) is of relevance to
this study to highlight that significant amounts of Br (and Cl) may be retained in
degassed lavas, as reported in natural context. The discussion has been modified to
highlight this point (Lines 356-364).
Reviewer’s answer
I recognize that the authors have followed my recommendations about the discussion of Df/m at low pressure, and this is a good point, however what is proposed in the new discussion from line 354 to 361, is wrong. As explained in Balcone-Boissard et al., 2010, in some cases, the eruptive style allows a very fast decompression and the consequence is that degassing is not at equilibrium. In these cases, Br and Cl may be retained in significant amounts in the lavas, whereas if degassing is at equilibrium, they are totally washed out from the silicate melt. However, it cannot apply to the present experimental study, because the “experimental” degassing (i.e. decompression) is not fast enough, pressure was decreased slowly, from a step to another one in order to allow in situ measurements. Furthermore, in previous study, performed at the same conditions and with the same chemical composition it is shown that bromine is totally degassed. This is probably what the authors would have found if they would have analyzed the quenched samples.
Comment 2: Br partition coefficients (D) are measured in situ for HPG system within the range 0.2 –
1.7 GPa and 592 – 840 °C: they are ranging from 4.1 to 15.3, they fall in the same range than those
from Bureau et al., 2010, for similar conditions (0.66 –1.7 GPa, 590-890◦C, D from 2.18 to 9.2).
However, if one plots all results in a diagram D versus pressure, data exhibits a lot of dispersion and
no real relationship, as it should be expected (i.e. an increase of D with decompression due to
degassing). This is not discussed at all.
Answer:
• Referee 1 was right to point out the dispersion of the data and the apparent lack of correlation
with P. A similar lack of pressure dependency (as well as density, or composition) has
previously been reported for vapour/brine partitioning of some metals (Cu, Au, Ag). It was
suggested that large differences in the speciation of these elements in both phases could be
responsible for such a behaviour (Pokrovski et al., 2008). We thus suggest that the large
differences in Br chemical and structural environment in between the coexisting fluid and melt
phase could as well explain the scattered DBr in our study and the apparent lack of simple
trends with P, density or dissolved silicate content, since the physical-chemical controls of Br
in the two phases are very different due to the different speciation.
Reviewer’s answer
If this would be right it should have been noticed in previous measurements of partitioning for the haplogranite-water-Br system. This is not the case. Br behavior cannot be compared to transition elements Cu, Au, Ag, having chemical properties far to those of halogens.
We would also like to point out that with a single exception at 0.2 GPa, all our data were recorded at P > 0.5 GPa, where the existing studies have also shown that DBr values do not change significantly with pressure and exhibit a similar degree of scattering, between 1 and 10 (Bureau et al., 2010).
Reviewer’s answer
The major problem is the extreme dispersion of the partitioning coefficients with respect to pressure that is still not explained or even discussed by the authors: we expect a decrease of D with pressure, which is logical as close to total miscibility the D should be equal to unity.
• We also note that additional discrepancies may arise from uncertainties on the estimation of
the fluid composition, which will affect the calculated DBr. We, for instance, recognize that the
large DBr value obtained at 1.7 GPa is clearly off the trend, probably due to the fact that the
fluid composition was calculated using the albite solubility data of Wohlers et al. (2011)
instead of Anderson and Burnham (1983), to take into account the higher P conditions in this
experiment.
Reviewer’s answer
Br concentration in all phases and Partition Coefficients can be calculated by using the PyMCA soft (see additional comment). This would provide an answer to that question.
• We have added an additional discussion in the revised manuscript, both in the Results section
(Lines 316-342) and in the Methods section, where we provide additional details about
uncertainties on fluid and melt compositions and how they translate to the DBr (Lines 179-186
and 240-258). We also agree with this referee that the discussion of the temperature effect in a
single experimental run was of weak relevance in terms of partitioning behaviour, and hence
we have removed it from the revised manuscript.
Reviewer’s answer
OK
• The conclusions drawn from our partitioning experiments, however, remain unchanged: we
confirm that although Br preferentially partitions into the aqueous fluid over silicate melt, high
amounts of Br can yet be incorporated in hydrous granitic melts. To strengthen this argument,
we added an estimation of Br concentration range in the high P-T melts of this study,
calculated using the in-situ DBr and initial phase proportions (Lines 339-340).
Reviewer’s answer
It is not new that the amounts of Br in haplogranitic melts may be high at high pressure, see solubility data, but this is not what you have measured here. It would be useful to explain how you calculate the Br content of the melt, and why you have not used this to calculate the Br contents of the aqueous fluid and the partition coefficients.
Comment 3: About the same value of D is obtained at 0.9, 0.8, 0.65, 0.2 GPa, respectively 4.4, 4.2,
4.1, and 4.8. High values of D are obtained at high investigated pressure (15 at 0.9 GPa and 9.7 at 1.7
GPa) where unity would be expected due to imminent total miscibility. Why? Such a discrepancy may
be due to the pressure determination. The authors use the diffraction of gold and its equation of state. However, it is well known that gold diffraction is not a good tool to for low pressure determination, as an example see Heinz and Jeanloz, JAP, 1983, where the first measurements are performed at 4.42GPa. Furthermore, in a recent intercomparison of the use of EOS for pressure determination (including Au), Ye et al., JGR, 2017, it is conclude that at high temperature, accuracy cannot be better than 1 GPa, from 2.5 GPa up to 140. For that reason, the discussion from line 278 to 294 is not relevant and should be suppressed.
Answer:
• The choice of gold as the in-situ pressure calibrant was motivated by its chemical stability in
high P-T fluids and melts. Although we agree that the absolute accuracy of this method may
not be better than 1 GPa (Ye et al., 2017), the actual relative precision is much better and is
likely to be within 10% of the P value, as reported in Louvel et al. (2014). Moreover, the unit
cell volume of Au displayed systematic changes as a function of increasing P-T in the HDAC,
thus demonstrating that Au is sensitive to relatively small pressure changes during the run, and
special care was taken in the HDAC alignment/centering to preserve the sample-detector
distance to ensure the reliability of the unit cell volume variations. Therefore, we are confident
that the relative pressure variations during the run are captured by the Au pressure calibrant.
To further support the appropriate pressure determination, we emphasize that the phase
relations, including miscibility, in the haplogranite-H2O system are within the P-T range
reported for other alkali silicate systems (e.g. Paillat et al., 1992; Stalder et al., 2000).
Reviewer’s answer
This is not convincing. The authors cannot recognize in a first sentence that “the absolute accuracy of this method may not be better than 1 GPa” and then write, “the actual relative precision is much better and is likely to be within 10% of the P value”. Which would mean an order of magnitude less than what is well known from decades. If the authors are really sure about their accuracy they should provide a calibration. Otherwise it is not correct to calculate pressure with an accuracy of 0.1 GPa, and to use it to calculate the composition of the silicate content of the fluids (see comments/answers from/to reviewer 2)
One explanation to justify the dispersion of Df/m data for a same pressure and the absence of relation between pressure and Df/m, is that the pressure values may be so inaccurate that you cannot see it.
A text must be added about pressure determination, with a realistic accuracy.
• As mentioned above, additional discussion on the uncertainties of our calculations is now
added both in the methods and result sections. Note that all fluid and melt properties
(composition, density and effective transmission) were calculated assuming an uncertainty on
pressure determination of 10% (Lines 183-186 and 245-246).
Reviewer’s answer
This may cause strong mistakes in your calculations.
Additional comments:
• The title has been updated to specify the nature of fluids (ie., aqueous) and melts (ie., silicate).
• Supplementary materials have been incorporated into the main manuscript text to better
provide the reader with as many details as possible and to better address the reviewers’
concerns (Lines 103-108; 115-118; 121-130).
• A set of dashed lines was added to Figure 5 to underline the shift of EXAFS oscillations with
change in composition. The caption of the figure was changed accordingly
Reviewer’s additional comment: in their answer to reviewer 2, the authors state that it is not possible to calculate the Br contents in the fluid phase and in the melt phase. This is not true. A specific software was developed for that purpose: PyMCA, http://pymca.sourceforge.net/
The soft is in open access and user’s friendly. If the experiments were done carefully it should not be a problem for the authors to process their data with it.
Process the data with PyMCA would provide another determination of the partition coefficients, the concentrations of all the phases (aqueous and silicate) and would also allow mass balance ca |
- Printer-friendly version
- Supplement
