Modeling Chemical Tests and Fiber Identification of Paper Materials Using Principal Component Analysis and Specular Reflection FTIR Data

Round 1

Reviewer 1 Report

1.       In Introduction: The literature review should be done in more detail. For the first ”chapter”, until ”Spectroscopy data collection methods”, there is no citation of literature, while a lot has been done in the past on the filed of non-destructive methods for studying the cultural heritage on paper.

2.       On page 4 and in lines 172–173: area size of 125 m ´ 126 m” and again the same mistake is repeated on page 7 and in line 275. Is the area ment to be expressed in [m] or [mm]?

3.       On page 4 and in line 180: ”The papers were subjected to chemical test ...” Why was not the paper filler included in ste study. After all, it is the second most represented component of any papaer, after the cellulose fibers.

4.       On page 4 and in line 190: ”The reference materials ...” How were the materials included in the presented research. From writen in article one can asume that they were select upon the kindness of provider and not upon the scientific goals.

5.       On page 8 and in line 301: ”The information about the fiber types ... was not included in this model.” Why?

6.       On page 12 and in line 381: ”... good confidence.” What is good? Express it in number and in reference to other materials.

7.       On page 12 and in line 388: ”... microscopy, ...” Which type of microscopy? The same mentioned in line 393, i.e. optical microscopy?

8.       Overall I am missing in discusion a bit more expression in numbers, values, and not just overlaping, etc.

Author Response

Reponse to Reviewer 1

Thank you for your clear and thoughtful feedback on the manuscript we submitted.  We have attempted to address all of your points in the revised manuscript.

 In Introduction: The literature review should be done in more detail. For the first ”chapter”, until ”Spectroscopy data collection methods”, there is no citation of literature, while a lot has been done in the past on the filed of non-destructive methods for studying the cultural heritage on paper.

Thank you for pointing out this oversight.  Additional references to nondestructive analysis of paper have been added.

2. On page 4 and in lines 172–173: area size of 125 m ´ 126 m” and again the same mistake is repeated on page 7 and in line 275. Is the area ment to be expressed in [m] or [mm]?

The Greek letter mu did not make it through the formatting process.  This has been replaced by the word microns written out.

3. On page 4 and in line 180: ”The papers were subjected to chemical test ...” Why was not the paper filler included in the study. After all, it is the second most represented component of any papaer, after the cellulose fibers.

The results of the chemical tests were taken from the datasheets provided with the Legacy press collection.  The authors did not perform the chemical tests themselves.

4. On page 4 and in line 190: ”The reference materials ...” How were the materials included in the presented research. From writen in article one can asume that they were select upon the kindness of provider and not upon the scientific goals.

The decision to use the Legacy Press Collection of papers was made because it is a curated collection of papers that many art conservation labs have, can be referenced and studied by other researchers. Using known collections allows for comparison of test results and adds to shared understanding of these papers which are studied in art conservation training programs. This project took advantage of the testing already conducted for sizing, lignin, and fiber composition which is recorded with the collection. No additional testing was performed on these papers such as filler content.

The Legacy press collection did not contain alum sizing which the authors considered another common sizing material that needed to be included in the study. 

The pure fiber type samples were included to prove the classification of fiber types using reflection FTIR data and PCA was possible without having complicating discussion about all the other components that go into paper.

5. On page 8 and in line 301: ”The information about the fiber types ... was not included in this model.” Why?

The authors were exploring if a reliable PCA model based on reflection FTIR could be built on just the sizing chemistry.  This turned out not to be the case.  The fiber types needed to be included too in the calibration classes for the model to be robust.

6. On page 12 and in line 381: ”... good confidence.” What is good? Express it in number and in reference to other materials.

Principal Component Analysis is a classification technique.  It is not really designed for quantification, hence the authors felt offering numbers would be misleading.  “good confidence” has been changed to “reasonable confidence”.

7. On page 12 and in line 388: ”... microscopy, ...” Which type of microscopy? The same mentioned in line 393, i.e. optical microscopy?

Cross polarized optical microscopy.  The text has been edited for clarity.

8. Overall I am missing in discusion a bit more expression in numbers, values, and not just overlaping, etc.

Principal Component Analysis (PCA) is a classification method.  To provide more in the way of numbers a regression method would be needed. 

Author Response File: Author Response.pdf

Reviewer 2 Report

This is an interesting paper, that is dealing with the application of PCA to FTIR measurements in order to answer content question regarding a very delicate material, that is paper. I strongly believe that the manuscript fits to the aim of “Heritage”, but its potentiality for publication should be considered after taking into account some points that listed in the following. Apart from that, I should mention that the manuscript is well written and easy to follow, although some commas here and there would make the reading of extensive sentences easier.

Lines 164-174. As long as FTIR is the only applied method, it should be extensively justified, eg number of scans, resolution, background measurement (most probably golden mirror)

Line 172. I think that the micron symbol went missing regarding the measurement area. The same in line 275.

Lines 180-181 & 201-209 & Table 1. The papers under study were subjected to tests regarding their protein (gelatin, G), starch (S) and lignin (L) content. This information is scattered in two different sections and a table. In terms of clarity and easiness to follow, this content should be placed within the same paragraph, perhaps “Sample Material and Classification”.

Lines 209-211 and Figure 1. The collected FTIR spectra show indeed the extended similarity between the samples. Nevertheless, as long as FTIR is the main experimental technique, the authors should proceed to the full characterization of the collected spectra, and not only the attribution of amide and OH bands. Moreover, what about the bands of cellulose?

Figure 1 and lines 165-174. Reflectance measurements of samples that present texture are quite difficult, as apart from the specular factor, the diffusely scattered light is also collected, causing the collection of an extremely low in intensity spectrum, and a severe broadening of the bands which leads to a “strange” spectrum appearance. Although in the microscope’s reflectance mode mainly the specular factor is collected, have the authors considered using Kubelka-Munk conversion to deal with this?

Line 243 & 255: data are/were weighted

Line 275. “the small sampling area” I understand that the statement has to do with the comparison between the sampling area and the heterogeneous nature of the samples. Apart from that, the authors should consider rephrasing, as this kind of area is not considered small for FTIR microspectroscopy, and it can be misinterpreted. Finally, I believe that the authors refer to “125x125”

Author Response

Response to Reviewer 2

Thank you for your clear and thoughtful feedback on the manuscript we submitted.  We have attempted to address all of your points in the revised manuscript.

This is an interesting paper, that is dealing with the application of PCA to FTIR measurements in order to answer content question regarding a very delicate material, that is paper. I strongly believe that the manuscript fits to the aim of “Heritage”, but its potentiality for publication should be considered after taking into account some points that listed in the following. Apart from that, I should mention that the manuscript is well written and easy to follow, although some commas here and there would make the reading of extensive sentences easier.

Lines 164-174. As long as FTIR is the only applied method, it should be extensively justified, eg number of scans, resolution, background measurement (most probably golden mirror)

Thank you for pointing out this oversight.  Requested details have been added to the text. 

Line 172. I think that the micron symbol went missing regarding the measurement area. The same in line 275.

Yes, the Greek letter mu did not make it through the formatting process.  It has been replaced by microns written out.

Lines 180-181 & 201-209 & Table 1. The papers under study were subjected to tests regarding their protein (gelatin, G), starch (S) and lignin (L) content. This information is scattered in two different sections and a table. In terms of clarity and easiness to follow, this content should be placed within the same paragraph, perhaps “Sample Material and Classification”.

Thank you for this note.  The text has been rearranged as suggested.

Lines 209-211 and Figure 1. The collected FTIR spectra show indeed the extended similarity between the samples. Nevertheless, as long as FTIR is the main experimental technique, the authors should proceed to the full characterization of the collected spectra, and not only the attribution of amide and OH bands. Moreover, what about the bands of cellulose?

With data science classification techniques, such as PCA, it is not necessary to try to parse and assign each peak in a spectrum. The shape of the spectra can be viewed as a wholistic signature of a paper’s specific chemical composition.

Figure 1 and lines 165-174. Reflectance measurements of samples that present texture are quite difficult, as apart from the specular factor, the diffusely scattered light is also collected, causing the collection of an extremely low in intensity spectrum, and a severe broadening of the bands which leads to a “strange” spectrum appearance. Although in the microscope’s reflectance mode mainly the specular factor is collected, have the authors considered using Kubelka-Munk conversion to deal with this?

Spectral processing techniques such as the Kubelka-Munk or Kramers-Kronig transforms are not necessary if the PCA model is built with data that is collected in the same method as the data from the unknown objects. 

Line 243 & 255: data are/were weighted

Corrected the language.

Line 275. “the small sampling area” I understand that the statement has to do with the comparison between the sampling area and the heterogeneous nature of the samples. Apart from that, the authors should consider rephrasing, as this kind of area is not considered small for FTIR microspectroscopy, and it can be misinterpreted. Finally, I believe that the authors refer to “125x125”

Thank you for noting the confusing text.  The text has been rephrased for clarity.

Author Response File: Author Response.pdf