PyXRD v0.6.7: a free and open-source program to quantify disordered phyllosilicates using multi-specimen X-ray diffraction profile fitting

Dumon, M.; Van Ranst, E.

doi:https://doi.org/10.5194/gmd-9-41-2016

Articles | Volume 9, issue 1

https://doi.org/10.5194/gmd-9-41-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/gmd-9-41-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 9, issue 1

Model description paper

|

15 Jan 2016

Model description paper |

| 15 Jan 2016

PyXRD v0.6.7: a free and open-source program to quantify disordered phyllosilicates using multi-specimen X-ray diffraction profile fitting

M. Dumon and E. Van Ranst

Download

Final revised paper (published on 15 Jan 2016)
Supplement to the final revised paper
Preprint (discussion started on 05 Mar 2015)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC C616: 'Review of Ms', Anonymous Referee #1, 05 May 2015
- AC C896: 'Answer to Reviewer #1', Mathijs Dumon, 29 May 2015
RC C630: 'Interactive comment 2 on “PyXRD v0.6.2: a FOSS program to quantify disordered, layered minerals using multi-specimen X-ray diffraction profile fitting” by M. Dumon and E. Van Ranst', Anonymous Referee #2, 06 May 2015
- AC C898: 'Answer to Reviewer #2', Mathijs Dumon, 29 May 2015

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Mathijs Dumon on behalf of the Authors (29 May 2015) Author's response

ED: Referee Nomination & Report Request started (15 Jun 2015) by Lutz Gross

RR by Anonymous Referee #1 (26 Jun 2015)

Suggestions for revision or reasons for rejection

The revision version of the Ms “PyXRD v0.6.5: a FOSS program to quantify disordered phyllosilicates using multi-specimen X-ray diffraction profile fitting” submitted by M. Dumon and E. Van Ranst for publication in Geoscientific Model Development has been significantly reshaped compared to the initial submission.
As outlined in the review of the original submission, the program described in the Ms clearly represents a step forward in the right direction for the quantitative structural determination of complex clay minerals assemblages. As such I think it would be a good thing to get this article ultimately published in Geoscientific Model Development and the program available to the clay community.
The revised version of the Ms suffers however major weaknesses, part of which being related to comments made on the initial submission, and thus requires significant additional improvements before it could be formally accepted for publication.
- Additional data provided to support the ability of the provided calculation routine to reproduce calculations performed with other simulation programs is far from being convincing. With provided illustrations it is difficult to assess the reliability of calculations for mixed layers. Results reported in Table 2 do not suggest major discrepancy for the mixing function.
On the other hand, intensity distributions calculated for periodic structures (illite and talc) show significant differences with those calculated with Sybilla. These differences reach 8% for illite 005 peak and 13% for talc 001. From the authors’ words, the origin of these differences is unknown (l. 297-298)!!! The proposed explanation (different scattering factors for the ionic species) appears unlikely as to my knowledge both programs use scattering factors for partially ionized species (Si2+, Al1.5+, …) thus minimizing differences related to these parameters. In addition, differences in scattering factors should influence calculated XRD patterns in a systematic manner as a function of diffraction angle, which is not the case in the two comparisons shown in Figure 3.
In my opinion reported differences ARE significant and a convincing explanation should be sought. If all parameters are identical, identical results must be obtained before going any further.
- The authors declined providing an example with a known natural mixture not to make their Ms too long. It is a pity but it is their choice. I think however that such an example would be much more useful than Figures 5-7 that all show the same thing (one is probably enough, and would even fit in a Supplementary data).
- Overall, the conclusions as to the ability of the proposed refinement algorithm to determine relevant structural (and quantitative) parameters from a single XRD pattern has been modified. The conclusion still remains essentially unchanged in the Abstract and in the Summary however, and this should be modified. In addition, I think that modifications brought to the Introduction section make the resulting text messy and imprecise. I do NOT agree that added sections help “to avoid any confusion”.
The first reason for this is the emphasis set by the authors on QPA (quantitative phase analysis) of clay-bearing samples which appears to be their main objective (l. 50-51). I have no problem with this emphasis, but an obvious question one may ask is the validity of the sole analysis of oriented clay-size fractions for this purpose, compared to the analysis of bulk samples.
Next, the authors claim that computation of XRD data corresponding to a structure model is the best solution for the proposed QPA, which is correct. What is not is that “structural and compositional information” is not a by-product of these calculations as suggested (l. 59-60) but the primary information derived from these calculations. To make things clear, there is no way to obtain a relevant QPA if a correct identification (the structural and compositional information) is not performed first: there is no way you can quantify accurately quartz and calcite in a sample if you are modeling its XRD pattern with K-spars and dolomite… With this respect, the description of the multi-specimen method as dedicated “originally […] for the qualitative evaluation of XRD patterns”(l. 102-103) is again misleading. The objective of this method was (and still is) the determination of structural models for investigated samples. Consistently, the first step of these models is to determine the nature of the different phases (periodic and/or interstratified) present in the sample (and the multi-specimen approach is especially well-suited for this purpose to remove possible ambiguities arising from the analysis of a single pattern). The next step is to refine the composition (chemical composition, proportion of the different layer types and their stacking mode). Finally, relative proportions of the different phases are refined to fit the data. Collation of structural models derived from different specimens of the samples (including relative proportions of the different phases) allows assessing the validity of the proposed models.
Finally, wording used throughout the Ms brings additional confusion. For example “programs” and “models” are used to describe the same thing (l. 72, 73). On the other hand “model” is used to describe the calculation routines (l. abstract, 73, 81, 318, …), structural models (l. 101, 110, 114, 343, 345, …), specific parameters of the structural model such as the set of junction probabilities (l. 276, 287, …).
- More generally, I think that the text, and specifically the Abstract, Introduction, Summary, and Conclusions sections should be extensively edited aiming at a precise and unambiguous writing. I am giving several examples of ambiguous, erroneous, imprecise sections below, but it is the author’s responsibility to produce a high-quality Ms. With this respect, reading the Ms before submission could avoid a reference list with references not cited in the text (Molina Ballesteros for example), not sorted (the 1st 4 citations), incorrect figure captions (Captions of Figure 6 & 7 should refer to Figure 5 not 3), not to mention typos.

Reference list remains essentially restricted to recent (>2000) articles owing to the development of XRD profile modeling over the last two decades as a result of the increased availability of computing resources and calculation routines. Most of the latter had however been developed and used in the 1970’s, and it would be reasonable to cite at least some of these pioneering works.

Minor remarks:
l. 1: Remove comma after ‘disordered”
l. 15: “calculation routine” rather than “model”. More genrally the authors should try to make a consistent use of “model”, “program”, … throughout the Ms to avoid ambiguities.
l. 16: “formalism” rather than “background”
l. 19: allows modelling
l. 24: “one refinement…and another employing…”
l. 28: all minerals present
l. 65: Modelling preferred orientation is one of the key to an accurate QPA from randomly oriented powders. Assumption of a perfectly random orientation is almost never made.
l. 67: “occurs mainly” as again orientation is not perfect, the deviation being accounted for by the sigmastar parameter.
l. 71: Only part of the stacking (dis)order (that along the c* axis) is accounted for.
l. 74-77: I understand that this is just a list of examples, but modelling of XRD patterns from interstratified samples has been developed in the late 1960’s and early 1970’s by several groups. It would be a good idea to acknowledge these pioneering works. More recently codes such as DiffaX have been used also for clay minerals.
l. 95 “limits, or by choosing starting values close…”
l. 106-107: Incorrect assertion. Hydration heterogeneity has been shown to be intrinsic to “swelling” clay minerals and not to directly depend on layer charge.
l. 132-133: “multispecimen”
l. 220-222: incomplete sentence
l. 263: “13 … phases”
l. 266: “this paper’s supplement” or “supplementary data”
l. 397-399: If relative proportions of the different smectite layers have a limited influence on calculated XRD patterns, additional iterations will not improve the convergence, that will remain hampered by this low sensitivity, and thus by the related large esd’s on these parameters.
l. 457: meaning unclear. What is an experiment? A sample? An additional XRD pattern from the same sample?
l. 457-462: I don’t agree. Listed parameters (essentially occupation factors or event atomic coordinates for interlayer species) will always be adjustable parameters, as in all fits to XRD data. The multispecimen approach never aimed at improving this aspects but rather at providing additional constraints to the junction probability parameters.
l. 490-491: references should formatted, sorted, and the presence of relevant in-text citations checked.
Table 1: The set of parameters listed for phase #8 is not sufficient to describe it.
Figures 6-7: Captions should be modified to take into account additional figures 3 & 4.

Hide

RR by Anonymous Referee #2 (28 Jul 2015)

Suggestions for revision or reasons for rejection

General comment

The authors have answered to the first comments and additional parts have been added in the revised version of the manuscript. The major suggestions of this review focus on the new parts added in the manuscript in the aim to strengthen these parts. Additional minor comments concern the other parts of the manuscript.

Major comment

Results
3.1 Model validation using discrete phases

The title of this paragraph should be changed because this paragraph deals with the comparison of calculated XRD patterns from PyXRD and Sybilla©. In addition, the comparison does not only consider discrete phases, comparison of calculated mixed layer minerals is also performed. The title could be: “Comparison between Sybilla© and PyXRD results: calculated 00l reflections for single discrete clays and MLMs”.

Among the clay minerals which were chosen for this comparison, the MLMs and their structural characteristics should be modified in order to be more relevant. Indeed, authors have used the default setting of the different illite-smectite MLMs that are proposed in Sybilla© and in some cases studied, the interest of these examples is still limited. For instance, XRD calculated pattern of R1 illite-smectite (Wi = 0.6; Pss = 0.5) is very closed to the R0 illite-smectite (Wi = 0.6; Pss = 0.4) and also does not largely differ from the R0 illite-smectite chosen (Wi = 0.5; Pss = 0.5). In addition, the choice of the default setting from Sybilla© is also limited for MLMs with higher Reichweite (R2 and R3) because of the too low mean CSDS of 4 that limits the influence of the stacking order (R3 or R0 MLMs illite-smectite have quasi-similar XRD calculated profiles with mean CSDS of 3). Thus the parameters of the MLMs chosen need to be modified and I suggest presenting R1, R2 and R3 MLMs illite-smectite with mean CSDS of 10 or more with MPDO (maximum possible degree of ordering) for R1. In addition, I would suggest exposing the calculated XRD patterns for these MLMs in a figure. The XRD patterns calculated with PyXRD (solid line) could be plotted over the XRD patterns calculated using Sybilla© (crosses) with the difference plot below. The figure 4 presented in the revised manuscript could be also improved (I do not obtain with Sybilla© the same calculated XRD pattern for the R2 illite-smectite MLM presented in figure 4 that motivate the presentation of calculated XRD patterns).

In the Table 1, the authors give the residual error (Rp) obtained between XRD patterns calculated with PyXRD and Sybilla©. Although quite low, they remain significant, particularly for the talc with a Rp equal to 6.4%. This value is higher than Rp obtained for the comparison of experimental and calculated XRD profiles on natural samples and consequently need more explanation in the text or a new calculation. The explanation in the manuscript l294-298 does not support this difference otherwise such difference should be present on each XRD calculation exposed. In addition, I suggest to the authors to also calculate the Rwp (weighted profile R-factors; Howard and Preston, 1989) that could be more relevant in the present case.

Howard, S.A. and Preston, K.D. (1989) Profile fitting of powder diffraction patterns. In D.L. Bish and J.E. Post, Eds., Modern Powder Diffraction, 20, p. 217–275. Reviews in Mineralogy, Mineralogical Society of America, Chantilly, Virginia.
3.2 Model validation using assemblages

As for the part 3.1, I would like to suggest a title. The title could be: “Comparison between Sybilla© and PyXRD results: calculated 00l reflections for mixtures of discrete clays and MLMs”.

The authors mention that Sybilla© “does not have an easy way to calculate a pattern for a certain mixture of phases (or in any case this feature is unknown to the authors), while PyXRD does” (l305-306). This is possible using Sybilla© however, I agree with the authors that an easy way does not exist. To create this type of file with Sybilla©, you need to first have a .txt file with the angular range of your XRD pattern (for example a XRD pattern calculated with PyXRD as performed in the manuscript). After opening the XRD pattern and writing the diffractometer parameters in the dialog box (fitting>add gene phases>select phases) and selecting the phases (for example kaolinite and illite as for the first mixture), you just have to save the calculated pattern (fitting>save all). You have saved an .xml file with the structural parameters of the mixture. Open this file with notepad and then you will able to modify in this file all that you need (for example, phase content in the present case). After performing the modifications, save this file as an .xml with another name and open this file in Sybilla© with fitting>add from input file. The mixture modified should appear. Save the fit pattern in .csv file (fitting>save fit pattern). Open the .csv in a spreadsheet program and you have the file needed.
N.B.: I have added a .xml file and a .csv file that give an example for the first mixture kaolinite, illite.
Using this procedure, the authors should directly compare the output from Sybilla© with the output of PyXRD for mixture and improve the validation test for the weight fractions. In addition, I suggest to the authors to give in Table 2 the Rp and Rwp values between the XRD patterns of mixtures calculated with Sybilla© and PyXRD. Moreover, I suggest adding a figure that show the comparison between the calculated XRD patterns from the two software as mentioned above for the part 3.1.

Finally, the authors should also add the Rp and Rwp values for the four assemblages calculated. These R-factor values will help the reader to estimate the order of magnitude of the misfits that could remain in the theoretical cases exposed.

Minor comment

L31-32: “… might be sufficient”, precise for the cases studied.
L34: remove “very” before useful
L101-102: add references, other authors have used this approach or precise “and reference therein”
L118: “… or at least more robust)…”. This part of the sentence is not necessary. I suggest to remove it.
L130: replace “model” by software
L251: add a space between strategy and algorithm
L291: replace “diffrerent” by different
L384-385: “these results…excellent results”. I suggest to remove this sentence that is not necessary.
L398: remove a space after “strategy”
L460-462: “If…more parameters”. This sentence is unclear. Remove this sentence or precise the idea.

Referee Report: PDF

Hide

ED: Reconsider after major revisions (04 Aug 2015) by Lutz Gross

AR by Mathijs Dumon on behalf of the Authors (16 Oct 2015) Author's response Manuscript

ED: Referee Nomination & Report Request started (20 Oct 2015) by Lutz Gross

RR by Anonymous Referee #3 (11 Nov 2015)

RR by Anonymous Referee #1 (11 Nov 2015)

ED: Publish subject to minor revisions (Editor review) (23 Nov 2015) by Lutz Gross

AR by Mathijs Dumon on behalf of the Authors (30 Nov 2015) Author's response Manuscript

ED: Publish as is (06 Dec 2015) by Lutz Gross

AR by Mathijs Dumon on behalf of the Authors (07 Dec 2015) Author's response Manuscript

Short summary

This paper presents a FOSS model called PyXRD used to improve the quantification of complex mixed-layer phyllosilicate assemblages using X-ray diffraction. The novelty of this model is the ab initio incorporation of the multi-specimen method, making it possible to share phases and their parameters across multiple specimens. We present results from a comparison of PyXRD with Sybilla v2.2.2 and a number of theoretical experiments illustrating the use of the multi-specimen set-up.