Biodegradable and Compostable Shopping Bags under Investigation by FTIR Spectroscopy

Round 1

Reviewer 1 Report

The paper is written in good English and the results are clear and accessible for a large audience of readers (even for the non-scientific audience). One of the rare papers, where the scientific discussion covers well with the general public discussion on the subject. I strongly recommend the publication of this paper in the Journal of Applied Sciences at the MDPI. 

Author Response

Comments and Suggestions for Authors

The paper is written in good English and the results are clear and accessible for a large audience of readers (even for the non-scientific audience). One of the rare papers, where the scientific discussion covers well with the general public discussion on the subject. I strongly recommend the publication of this paper in the Journal of Applied Sciences at the MDPI. 

The authors warmly thank the Reviewer for his strongly positive judgment. Thanks.

Author Response File: Author Response.pdf

Reviewer 2 Report

Dear authors,

an improvement of the manuscript must be done.

There is a lack of information in some sections of the manuscript: 

1. line 47 - HLDPE - this is a mistake, this is not an abbreviation for high density polyethylene. Furthermore, LDPE is more often used for production of shopping bags than HDPE.
2. line 48- the authors write about the harmful monomer ethene. It is not clear how it affects the organisms in the sea. It is not possible to have the large quantities of the unreacted ethene captured in polyethylene in the shopping bags. It is highly important to clarify this.
3. chapter 2.1. Instruments - please, explain the measurements. Is there ATR used?
4. chapter 2.2. Why starch is used? It is not clear. PLA is produced by polymerization of lactic acid. Please, explain this. 
5. chapter 2.3.1. it is needed to present FTIR spectrum of polyethylene
6. The authors state that the ratio A/A% indicates PLA content - it is not clear, it is needed to explain this and to add the reference.
7. Figure 4 - it is not possible to detect the differences between the samples, there are to many samples on the same figure.

Author Response

Comments and Suggestions for Authors

Dear authors, an improvement of the manuscript must be done.

There is a lack of information in some sections of the manuscript:

line 47 - HLDPE - this is a mistake, this is not an abbreviation for high density polyethylene. Furthermore, LDPE is more often used for production of shopping bags than HDPE.

We thank the Reviewer for having reported the error in the acronym for high-density polyethylene, which has been removed (line 55). We also thank the Reviewer for his appropriate comment. The sentence that contained the incorrect abbreviation was reformulated accordingly, and it was moved to a more appropriate part of the manuscript, to make the speech more continuous and fluid. Please see lines 55-57 (removed) and lines 33-35 (inserted).

line 48- the authors write about the harmful monomer ethene. It is not clear how it affects the organisms in the sea. It is not possible to have the large quantities of the unreacted ethene captured in polyethylene in the shopping bags. It is highly important to clarify this.

We thank the Reviewer for his comment, and we agree with the Reviewer. The sentences concerning ethene have been removed (lines 57-61). Moreover, instead of speculating on the dangerousness of ethene for the marine environment, which is not significantly documented, a brief discussion was included on the danger and pollution due to the accumulation of plastic debris. Please see lines 45-52 and the new reference [Ref. 4].

chapter 2.1. Instruments - please, explain the measurements. Is there ATR used?

We thank the Reviewer for his request, because from it, we realized that we had reported an incorrect information concerning the mode of the measurements of FTIR spectra. FTIR spectra can be measured in transmission mode, in reflection mode or, as the Reviewer noted, in Attenuated Total Reflection (ATR) mode. The sentences reporting incorrect information have been corrected. Moreover, the modality of the acquisitions has been inserted in other parts of the manuscript (lines 140, 169, 303 and 308). ATR was not used, because our instrument model does not allow such acquisition mode.

chapter 2.2. Why starch is used? It is not clear. PLA is produced by polymerization of lactic acid. Please, explain this.

We thank the Reviewer for his comment and request, because from it we realized that additional clarifications concerning our decision to include starch in the study are necessary. We agree with the Reviewer concerning the common production of PLA, but PLA for bioplastics should derive from biomass resources and renewable raw materials, including starch.  

In this regard, explanations were inserted both in chapter 2.2., as requested (lines 146-147) and in section 3.1. (lines 228-232). In addition, also a new reference [Ref. 22] has been included.

chapter 2.3.1. it is needed to present FTIR spectrum of polyethylene

Very good suggestion. The FTIR spectrum of pure polyethylene has been included as Figure 3a, close to Figure 3b (PLA), for comparison purposes. Moreover, following the Reviewer's comment, the location of Figure 2 (unrevised manuscript) and of some parts of the paragraph 2.3.1., were reconsidered and deemed more suitable for the Discussion Section than for the Experimental Section. So, they were deleted in the Experimental Section (lines 173-176 and 179) and moved to the Discussion Section. In particular, old Figure 2 (now Figure 3) with relative discussion has been inserted in section 3.2. (lines 310-313 and line 315 (Figure caption). Another portion has been moved to Section 3.4. (lines 455-470). More information about the choice of the bands to obtain A/A% has been also included, to satisfy the Reviewer request at the next point.

The authors state that the ratio A/A% indicates PLA content - it is not clear, it is needed to explain this and to add the reference.

We thank the Reviewer for his suggestion. Explanations have been included in Section 3.4., as reported in the previous point. Please, see specifically lines 455-462 and 465-466.

Figure 4 - it is not possible to detect the differences between the samples, there are to many samples on the same figure.

We agree with the Reviewer and thank him for his suggestions. Figure 4 has been completely remade by removing 12 samples, whose presence was not essential for understanding the difference between the samples. Figure 4 resulted improved and more clear. Also the main text has been slightly modified, accordingly. Please see lines 316-318, 320-322 and 323-336.

He hope to have satisfied the Reviewer, and thank him again.

Author Response File: Author Response.pdf

Reviewer 3 Report

See attached review.

Comments for author File: Comments.pdf

Author Response

Review of Applied Science – 1059621

1. General comments.

1. The authors have investigated FTIR methods for determining whether shopping bags are biodegradable and compostable. Using samples of known plastics that are not in compliance and samples which are in compliance, FTIR Spectroscopy was used to examine shopping bags from different manufacturers and define a methodology for determining whether they met compliance standards prescribed by TÜV-Italy.
2. This work has merit from an ecological standpoint – the development of a simple method by which plastics manufacturers can quickly do quality control and measure whether or not their products meet government biodegradation/compostability standards.

2. Editorial comments. The article is generally well-written with only a few instances of incorrect word usage or grammatical errors. Selected examples are listed below:

399. Page 13, Lines 397-399. Awkward sentence structure and improper use of plural form. As Reads, “However, also other bands appeared to affect strongly the samples separations, such as those in the range 1500-750 cm-1 and 3500-3000 cm-1, thus establishing that not only PLA content was discriminant for detecting compostable samples.”

Recommend rewording: “However, other bands appeared to affect sample separations, such as those in the range 1500-750 cm-1 and 3500-3000 cm-1, thus establishing that sample content other than PLA content was a discriminant for detecting compostable samples.”

We thank the Reviewer for his comment and suggestion. We agree with the Reviewer concerning an improper use of plural form, therefore the sentence was reformulated, but in a slightly different mode, in order not to change its meaning too much. Please see lines 441-446.

492. Page 16, Line 492. As reads, “At all level of the present…” Recommend rewording: “At all levels of the present…”

The suggested change has been done. Line 548.

504. Page 16, Line 504. As reads, “A special thank go to…” Recommend rewording: “Special thanks go to…”

The change requested has been applied. Line 560.

Recommendation: That the authors ensure a thorough review is accomplished by an English editor prior to publication.

As recommended, the manuscript has been reviewed by a mother tongue professor of English. Please see the new insertion in the acknowledgments (lines 561-562).

3. Content comments.
4. Figure 6a is of poor quality and unreadable – there are too many overlapping data labels in the graph. Recommend spreading the labels our in the graph and drawing lines from the labels to the actual data points. Alternatively, identify the samples in Table 2 by a numeral which could then be used in Figure 6a to make the figure more readable.

We agree with the Reviewer, but the Figure 6a cannot be changed, since it was provided as such, directly by the software. In addition, the software does not allow to insert numbers for identify the samples, but only letters. However, in order to satisfy the Reviewer and to improve its quality and readability, Figure 6a was resized.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The previous comments are answered and the manuscript is improved.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The authors presented a novel way to check the quality of plastics bag. It would definitely further the research in this field.

Author Response

Reviewer 1

The authors presented a novel way to check the quality of plastics bag. It would definitely further the research in this field.

The authors warmly thank the Reviewer for his strongly positive judgment. Thanks.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this manuscript, the authors describe a facile method for detecting contaminant-ridden plastic bags through FTIR-PCA and predictive linear regression model development. The motivation behind the research is thoroughly shown, and the experimental tool of choice is appropriate in my opinion. The data reduction and analysis, however, needs significant improvement by selecting a more mathematically useful quantity for spectrum representation and probing further into what the principal components signify, respectively. I would recommend that the authors revise the manuscript according to the following comments:

- The authors represent the spectra in transmittance vs. wavenumber. In terms of multivariate analysis, since the linearly combinable quantity under Beer-Lambert’s law is absorbance (negative log of transmittance), you should perform all analysis in absorbance vs. wavenumber, as in most of the FTIR-PCA are done. Additionally, in some of the spectra the absorbance is over 1.5 (transmittance of < 3%) in regions of interest, such as the carbonyl peak. Depending on the instrumentation, the signal to noise ratio deteriorates, and the Beer-Lambert’s law is not applicable anymore. To avoid this, you may need to consider thinning the film mechanically. The “peak area” ratios should be really compared in the absorbance plots.

- The authors may be thinking that since the PCA categorizes the samples in three groups, it is still a valid approach. This is true within a certain context that the impurities (other than PE and PLA) do not produce very similar FTIR spectrum. If these are the only two components, most likely a simple linear decoupling is possible. In reality, there are other polyesters with varying biodegradability (PCL, PET, PHA, …) which show a very spectrum around 3000cm-1 and around 1700cm-1, but varies in spectral features in other wavenumbers. These subtle variations may be detectable by performing PCA and trying to understand what the principal components are. In order to do this, you need to have the PC loading curves for the score plot axes you are showing. It would be interesting to see what the result of the PCA is when the absorbance is used for plotting.

- If you look at the x axis of the two plots in Figure 6, the component 1 limits are different but the values seem to be identical. This is reassuring in the sense that the two codes are equivalent, but there is no reason to show both plots.

- In order to perform if the predictive linear regression is working or not, you need to have an independent method of determining what the PLA contents are in each of your calibration sample set, compared to where they lie in PCA scores. I do not understand why the ratio of transmittance area under the curve is used to calibrate at 2 points (0% for PE, 40% for Vincotte_S732) for a linear curve, and was considered to be a reliable calibration. After the PCA is redone with the absorbance, the PCA score plot needs to be highlighted with a set of calibration samples with known amount of PLA with variable amounts of unknown contaminants, from which one of the PC scores can be used to predict the PLA content. The corresponding PC loading should show a very similar spectral feature to pure PLA.

Author Response

Reviewer 2

In this manuscript, the authors describe a facile method for detecting contaminant-ridden plastic bags through FTIR-PCA and predictive linear regression model development. The motivation behind the research is thoroughly shown, and the experimental tool of choice is appropriate in my opinion. The data reduction and analysis, however, needs significant improvement by selecting a more mathematically useful quantity for spectrum representation and probing further into what the principal components signify, respectively. I would recommend that the authors revise the manuscript according to the following comments:

- The authors represent the spectra in transmittance vs. wavenumber. In terms of multivariate analysis, since the linearly combinable quantity under Beer-Lambert’s law is absorbance (negative log of transmittance), you should perform all analysis in absorbance vs. wavenumber, as in most of the FTIR-PCA are done. Additionally, in some of the spectra the absorbance is over 1.5 (transmittance of < 3%) in regions of interest, such as the carbonyl peak. Depending on the instrumentation, the signal to noise ratio deteriorates, and the Beer-Lambert’s law is not applicable anymore. To avoid this, you may need to consider thinning the film mechanically. The “peak area” ratios should be really compared in the absorbance plots.

The author thank the Reviewer for its suggestions. We make kindly note to the Reviewer that although the FTIR spectra in the original manuscript were represented in transmittance vs. wavenumbers, the spectra were acquired also in absorbance and were exploited to compute the areas of bands, which were used to obtain the ratios A/A. Please, see the sentences already included in the manuscript at lines 160-163 (highlighted in yellow) and following reported.

”… the ratio between the areas of bands (A/A%) in the range 1600-1800 cm^-1 and 3037-2800 cm^-1, (considered indicative of PLA % content), were measured in the absorbance scale automatically by using the Turbo chrome workstation (version 6.1.1.) equipment software.”  

Consequently, the “peak area” ratios were already compared in the absorbance plots.

Anyway, all the FTIR analysis were re-performed in absorbance vs. wavenumber after having thinned the films mechanically, to low the values of absorbance, as suggested. Consequently, PCAs were remade on absorbance data and by pretreating data (column centering), obtaining better separation and similar but more interesting results. Please see the modified Sections 3.2 and 3.3 (lines 273-441). The authors decided to report only the information obtained with R and the original Figure 6a was removed. Additional Figures 6b, 7b and 8b have been included reporting the loading plots associated to the PCAs and the discussion was significantly improved.

- The authors may be thinking that since the PCA categorizes the samples in three groups, it is still a valid approach. This is true within a certain context that the impurities (other than PE and PLA) do not produce very similar FTIR spectrum. If these are the only two components, most likely a simple linear decoupling is possible. In reality, there are other polyesters with varying biodegradability (PCL, PET, PHA, …) which show a very spectrum around 3000cm-1 and around 1700cm-1, but varies in spectral features in other wavenumbers. These subtle variations may be detectable by performing PCA and trying to understand what the principal components are. In order to do this, you need to have the PC loading curves for the score plot axes you are showing. It would be interesting to see what the result of the PCA is when the absorbance is used for plotting.

As reported in the previous point, the PCs loading curves have been included in the revised manuscript. Please, see the added Figures 6b, 7b and 8b reporting the loading plots associated to the PCAs. A relative discussion have been inserted. Please see lines 338-345 and 356-441.

- If you look at the x axis of the two plots in Figure 6, the component 1 limits are different but the values seem to be identical. This is reassuring in the sense that the two codes are equivalent, but there is no reason to show both plots.

We agree with the Reviewer. Original Figure 6a which is the only one reporting PCA results obtained with MatLab software is useless and consequently it has been removed.

- In order to perform if the predictive linear regression is working or not, you need to have an independent method of determining what the PLA contents are in each of your calibration sample set, compared to where they lie in PCA scores. I do not understand why the ratio of transmittance area under the curve is used to calibrate at 2 points (0% for PE, 40% for Vincotte_S732) for a linear curve, and was considered to be a reliable calibration. After the PCA is redone with the absorbance, the PCA score plot needs to be highlighted with a set of calibration samples with known amount of PLA with variable amounts of unknown contaminants, from which one of the PC scores can be used to predict the PLA content. The corresponding PC loading should show a very similar spectral feature to pure PLA.

We apologize to the Reviewer, but this time his comments are not so clear. However, also with the help of the comments of other Reviewers on the matter, they tried to answer. Concerning this specific comment: “I do not understand why the ratio of transmittance area under the curve is used to calibrate at 2 points (0% for PE, 40% for Vincotte_S732) for a linear curve, and was considered to be a reliable calibration”

We do not consider the PLRM a calibration curve, the term was never used in the manuscript. This model is a linear regression intended to be just a tool to verify if, following a production process, the PLA content in the final product respects the initial expectations, i.e. the quantities of PLA put among the ingredients. Furthermore, it was not built by setting 2 points, but as reported in the text at lines 450-459:

“….Since the area of the band in the range 1800-1600 cm-1 (C=O group of PLA) is indicative of the PLA content, it was selected as significant spectral data, but in order to minimize the error related to the difference in optical path of the sample aliquots pieces deposited between the KBr dishes, the values of the areas ratios (A/A%) between the A% of the band in the range 1800-1600 cm-1 (C=O group of PLA) and that of the bands in the range 3037-2800 cm-1 were considered as more reliable spectral data. The relevance of these bands was confirmed by the loadings profiles obtained by the PCAs (Figure 6b, 7b and 8b). Taking as reference the Vinçotte_S732 sample, having a minimum declared content of PLA of 40 % and an A/A % value of 1.52, the minimum PLA percentage content of all the film samples was calculated through the use of mathematic proportions..”

Then, if we have understood the Reviewer, he asks that the model be validated, because otherwise it can only be “a self-validated” and not so reliable, model. And if we have understood, as not validated, such a model, may not be able to detect the presence of non-biodegradable pollutants, which have similar absorptions to PLA, in a new shopping bag film under study. The validation requested by the Reviewer is an excellent approach to create highly reliable models capable of identifying even subtle sophistications, and this is not within the scope of the authors and of this paper.

The developed model (PLRM) was created only to provide MecPlast with a quick method to estimate the PLA content of a new product, prepared with an initial nominal PLA content compliant with the laws, and not to highlight frauds. In our opinion, since even if the ingredients are always the same, it does not mean that the cake is always the same, it may therefore be useful to have a model, even if only self-validated, to be able to assess whether the finished product meets initial expectations. With our method, we do not want to identify shopping bags with contents intentionally different from those permitted by law, and used for fraud, but to evaluate shopping bags prepared in any case to comply with the law, and where substances whose absorption can overlap the fundamental absorption of PLA, have not been added. To detect fraud and award the conformity marks is the work of the authorized bodies. However, in order to satisfy the Academic Editor, the term “self-validated” has been added in lines 24, 472, 477 and 522, for greater clarity. In addition, discussion sentences on these issues and explaining the destination of use of PLRM have been included in lines 24-26, 116-120, 445-450, 476-477 and 524-526.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors describe in their manuscript the application of the FTIR spectroscopy to describe physicochemical properties of plastic-based shopping bags and confirm their characteristics with the conformity mark.
The collected FTIR spectra were further explored using principal component analysis and used to construct a calibration model for the prediction of poly-lactic acid content in plastic shopping bags. Results of this study indicate the presence of plastic shopping bags certified for regular use, but in fact, do not comply with the standard.

In my opinion, this article cannot be considered for publication in Applied Sciences in present form due to low scientific standards and pitfalls in the methodology. In particular, the authors use the PCA method in a very peculiar manner. I do not understand why to use three different models and two different software platforms? The text is not well-structured. There are repetitions. Even though the study has potential, the number of samples does not incorporate expected variability. The true intention is to construct discriminant/classification models, but regretfully it is not done.

General comments:
1) Why use three different software to perform the analysis (Matlab, R, and Excell)?
2) Regarding the PLA content – the authors do not use the reference method but foretell the PLA content directly from FTIR spectra with no further explanation.
3) Section 2.3.2: three different PCA models were constructed. They were surprisingly taking into account only 4 or 5 samples. Why?
4) The authors should avoid repetitions in the ‘Experimental section’ and ‘Discussion’.
5) The quality of the figures is rather low. I do not recommend to use lack background.
6) Why Figure 6 illustrates PCs obtained from Matlab and R? Is it really a loading plot? On the loading plot, you should see different wavelengths.

Author Response

Reviewer 3

The authors describe in their manuscript the application of the FTIR spectroscopy to describe physicochemical properties of plastic-based shopping bags and confirm their characteristics with the conformity mark.
The collected FTIR spectra were further explored using principal component analysis and used to construct a calibration model for the prediction of poly-lactic acid content in plastic shopping bags. Results of this study indicate the presence of plastic shopping bags certified for regular use, but in fact, do not comply with the standard.

In my opinion, this article cannot be considered for publication in Applied Sciences in present form due to low scientific standards and pitfalls in the methodology. In particular, the authors use the PCA method in a very peculiar manner. I do not understand why to use three different models and two different software platforms?

The PCA method has been applied to three different matrices (22x3251, 4x3251 and 5x2351) to obtain score plots able to highlight three different types of information. Firstly, it was obtained a score plot in which samples actually compliant with law and more similar to starch resulted well separated from plastic-based ones on PC1 (Fig. 6a). Secondly, it was obtained a score plot highlighting similarities/differences between the new films under study (M1, 2 and M3) and the MecPlast film, which received the conformity mark in the 2017 (S303) (Fig. 7a). Finally, it was obtained a score plot highlighting similarities/differences between the new films, sample S303 and a film with declared PLA minimum content of 40% and thickness minor than 15 µm (Fig. 8a). Concerning the use of two software, the authors agree with the Reviewer. The authors decided to report only the information obtained with R. Original Figure 6a has been removed.

The text is not well-structured. There are repetitions.

On the requests of other Reviewers and of the Academic Editor, the structure of the paper has been strongly modified and some redundant parts have been deleted.

Even though the study has potential, the number of samples does not incorporate expected variability. The true intention is to construct discriminant/classification models, but regretfully it is not done.

Apologizing to the Reviewer, we kindly point out that the real purpose of this article was not to construct discriminant/classification models, but as reported:

“Upon request and in collaboration with the company MecPlast, a packaging Italian industry (Savona, Liguria), certified by TÜV Italy, i.e. the Italian representative of TÜV Austria, a PCA-assisted FTIR study was conducted on populations of shopping bags opportunely assembled, to obtain information on the possible compliance with the standard, of three shopper films, recently extruded, before they were subjected to a periodic inspection.”

General comments:
1) Why use three different software to perform the analysis (Matlab, R, and Excell)?

MatLab was no longer used in the revised manuscript. Only R has been maintained, while Microsoft Excel has been used to create the matrices and to construct the linear regression model, by least square approach, not for PCA.

2) Regarding the PLA content – the authors do not use the reference method but foretell the PLA content directly from FTIR spectra with no further explanation.

What does the Reviewer mean with the “reference method”?

Anyway, the authors have not predicted “the PLA content directly from FTIR spectra with no further explanation”, but have used a reference sample with declared minimum content of PLA of 40% (S732) and have calculated consequently the minimum PLA content of all samples by using their A/A ratios (obtained from the FTIR spectra using a function provided by the FTIR software) and making mathematic proportions. Please see lines 456-459.

 
3) Section 2.3.2:  three different PCA models were constructed. They were surprisingly taking into account only 4 or 5 samples. Why?

In order to highlight the differences among specific samples of interest including those provided by MecPlast, which in the PCA on all the samples resulted clustered in the same population. Please see lines 347-354, 400-409 and 421-441.

4) The authors should avoid repetitions in the ‘Experimental section’ and ‘Discussion’.

All the manuscript was checked to detect repetitions which were removed.

5) The quality of the figures is rather low. I do not recommend to use lack background.

Figures 4, 5, 9 (revised paper) and 10 (revised paper) have been replaced with images having a white background.

6) Why Figure 6 illustrates PCs obtained from Matlab and R? Is it really a loading plot? On the loading plot, you should see different wavelengths.

The authors agree with the Reviewer concerning point 6. Consequently, only PCAs obtained from R were maintained in the revised manuscript. Figure 6a has been removed. Old Figure 6, as well as Figure 7 were not loading plots but score plots. Anyway, on suggestion of another Reviewer, all the FTIR analysis were re-performed in absorbance vs. wavenumber after having thinned the films mechanically, to low the values of absorbance. Consequently, PCAs were remade on absorbance data and by pretreating data (column centering), obtaining better separation and similar but more interesting results. Please see the modified Sections 3.2 and 3.3 (lines 273-441). Additional Figures 6b, 7b and 8b have been included reporting the loading plots associated to the PCAs and the discussion was significantly improved.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have appropriately addressed my comments through re-analysis of the absorbance data. 

Reviewer 3 Report

Dear Authors,

I have read responses to my comments and I have verified your improvements. Unfortunately, I cannot support this article for publication. My major objection arises with the construction of "naive" calibration model - based on single response information (a given amount of PLA) and the use of manually selected wavelengths. This is not the way how it is done in practice and in chemometrics.

Kind regards