- Email: [email protected]
Batch-to-batch variation in domestic paints: Insights into the newly commercialized recycled paints
Forensic Science International 303 (2019) 109946
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Batch-to-batch variation in domestic paints: Insights into the newly commercialized recycled paints Samuel Quevillona,1, Nicholas Toupina,1, André Tremblayc, René Fournierc , Cyril Muehlethalera,b,* a b c
Université du Québec à Trois-Rivières, Département de Chimie, Biochimie, et Physique, Trois-Rivières, Canada Laboratoire de Recherche en Criminalistique, Trois-Rivières, Canada Laboratoire des Sciences Judiciaires et de Médecine Légale, Direction de la Chimie, Montréal, Canada
A R T I C L E I N F O
A B S T R A C T
Article history: Received 26 July 2019 Received in revised form 29 August 2019 Accepted 11 September 2019 Available online 12 September 2019
A few recent studies attempted to evaluate the differentiation of paints at a production batch level and reported results depending largely on the paint type. The discrimination from production batches is much more random than brands and/or models levels and subject to many unknowns, which suggests that a particular production batch can suddenly present a substantially different composition than the one produced right before or right after it. To add to this existing complexity, most of the paint companies now propose a range of recycled paints among their products. These recycled paints are composed of wastes collected by recycling plants, sorted by their color and binder type (i.e. latex, alkyds), and mixed together in large tanks to form the basis material for future formulations. Quality controls on these recycled batches are voluntarily less precise, and a higher variation is expected in esthetic and chemical properties of the paint. In this project, we collaborated with a North American paint producer that gave us access to its samples, paint formulations recipes, and a summary of the quality controls and corrections they performed on each production batches. The whole study was conducted blindfold and a final verification was made with the manufacturer to evaluate the accuracy of the results. The data set comprised two models of regular paints and two models of recycled paint. Five different production batches were collected per model for a total of 20 samples analyzed by Microscopy, Infrared spectroscopy, Raman spectroscopy, and Pyrolysis GC/MS. © 2019 Published by Elsevier B.V.
Keywords: Architectural paint Chemical analysis Production batches Forensics Microscopy Fourier transform infra-red spectroscopy (FTIR) Pyrolysis-GC/MS Raman spectroscopy
1. Introduction Population and discrimination studies represent a significant sub-field of all the articles about forensic paint analyses that were published ([1–8]). These studies not only permit to justify the choice of correct sequences of analysis that maximize discriminating power, but they also evaluate the heterogeneity in the population (uniqueness of the composition vs class characteristics), and finally assess at which level of discrimination one could/should evaluate a source attribution (i.e. brand, model, batch). These studies were important resources for paint interpretation, but also to provide general knowledge to better understand the market and the relative complexities of paint
* Corresponding author at: Université du Québec à Trois-Rivières, Département de Chimie, Biochimie, et Physique, Trois-Rivières, Canada. E-mail address: [email protected] (C. Muehlethaler). 1 Both authors contributed equally the paper. http://dx.doi.org/10.1016/j.forsciint.2019.109946 0379-0738/© 2019 Published by Elsevier B.V.
formulations (i.e. automotive vs household vs spray) (Caddy [9]. For the most common paint types and in most forensic cases, the differentiation at brand and model levels has often been the desired outcome. A few studies have tried to evaluate the utility of differentiating at a production batch level with very mixed results, showing an important influence of the paint type ([7,8,10]). A survey of the literature published to date shows that the differentiation at a production batch level is much more arbitrary and subject to many unknowns, such as (1) the quality controls and adjustments the paints are undergoing, (2) the varying origins of raw materials, or (3) the manufacturer’s decision to make changes in the paint recipes for improvement or discontinuity of raw materials. All of these reasons suggest that a particular production batch can suddenly present a substantially different composition than the one produced right before or right after it. Batch to batch variation is not systematic, but can suddenly present a drastically different chemical signature, possibly even individual, that would allow differentiating it from other production batches. The previous studies have shown that colored paints have a much
2
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
higher chance of showing a different composition due to the increased pigment(s) complexity. Black and white paints, on the other hand, showed no or little differences in their composition ([7,8]). While forensic scientists are still undergoing studies to understand the potential of discrimination between paints, industrials have taken it a step further. Most of the paint companies now propose a range of recycled paints among their products. These recycled paints are composed of wastes collected by recycling plants. First sorted by their color and binder type (i.e. latex, polyester) all the leftovers are mixed together in large tanks to form the basis material for future formulations. This lot can be used as is for lower quality paints or undercoats, or be refined and completed to required specifications. Quality controls on these recycled batches are obviously less precise, and a higher variation is expected in esthetic properties of the paint. While not monitored directly by the producers (i.e. not part of the systematic end-user quality controls), the variation in chemical composition should follow the same trends, if not greater. Most of the discrimination studies to date have been blind performed, without access to the manufacturer formulation recipes and unknowingly of the paint adjustments that were made. For this project, we collaborated with a North American paint company that shared with us all the details about their paint formulations. This participation allows us to bring a new viewpoint to forensic paint comparison, and the possibility to validate our analytical hypotheses. In this article, we present the differentiation of 5 batches of 4 different domestic paints measured by a standard sequence of analytical measurements (i.e. microscopy, infrared spectroscopy, Raman spectroscopy, and Py-GC/MS). By having a direct access to the manufacturing company we have compared our analytical findings with their paint formulation recipes, and evaluated if the quality control adjustments they have made do show up as significant modifications in the chemical composition. Two of the paints analyzed are respectively grey and white recycled paints, which permit us to bring a few insights into their chemical heterogeneity for analysis in forensic cases. 2. Material and method 2.1. Sampling Four different models of paint were obtained from the manufacturer and named A to D. For each model, five samples from different production batches were compared. Two of these models are regular paint (A and B), while the other two are recycled (C and D). The color, origin and production date of the samples are presented in Table 1. Reference microscopic slides were prepared for each of the production batches, forming a data set of 20 slides in total (Fig. 1). Every sample was, according to the quality assurance protocols of the manufacturer, mechanically homogenized for 20 min seven days prior to preparation of the slides. A second homogenizationwas done manually for one minute just before applying the paint. All slides had been previously washed with acetone, dried, and manipulated with gloves at all times. Paint from each sample was applied simultaneously on two glass slides with a rubber scraper, in order to create duplicates of equal thickness. One slide was reserved for pyrolysisGC/MS and manipulated as little as possible, whereas the remaining one would be used for all the other techniques which are less prone to detect contaminations due to manipulation. All paints slides have been dried a whole week prior to analysis. 2.2. Methods Four techniques were employed following a usual forensic paint examination: microscopy, infrared spectroscopy, Raman spectroscopy,
Table 1 Color, origin and production date of every sample. Sample
Color
Origin
Production date (month/year)
A1 A2 A3 A4 A5 B1 B2 B3 B4 B5 C1 C2 C3 C4 C5 D1 D2 D3 D4 D5
White White White White White Beige Beige Beige Beige Beige White White White White White Gray Gray Gray Gray Gray
New New New New New New New New New New Recycled Recycled Recycled Recycled Recycled Recycled Recycled Recycled Recycled Recycled
01/18 12/17 11/17 11/17 11/17 12/17 10/17 10/17 10/17 08/17 11/17 11/17 10/17 10/17 09/17 11/17 11/17 10/17 09/17 08/17
and pyrolysis GC–MS. Experimental details about each technique and instrument are presented in Table 2. Further details about the sample preparation are described here-below: 2.2.1. Microscopy One fragment from each of the five production batches were flattened onto a glass slide side-by-side, close enough for each of them to be observed in the same field of view through a Nikon Eclipse Ci with a 4x objective. No mounting liquid or cover slip were used and three similar slides were prepared for each model. Samples were observed under episcopic bright and dark field, polarized light, and fluorescence lights (Nikon cubes DapiHoechst, FITC, and TITC). 2.2.2. Fourier transform infrared spectroscopy (FTIR) The FTIR Spectrometer was used in transmittance through diamond compression cell. Seven replicates were taken per production batch. In each case, a paint fragment collected with a scalpel blade was deposited onto the diamond cell and flattened. Transmittance spectra in the range of 4000-650 cm 1 were measured with a resolution of 4,0 cm 1 and 16 co-added spectra ([7,8]). Discrimination between samples was made visually using both qualitative differences (presence or absence of certain peaks) and semi-quantitative differences (relative intensity of major peaks). A PCA statistical analysis was then employed to confirm the exclusions that had been obtained visually. 2.2.3. Raman spectroscopy Raman analyses consisted of three replicates for each production batch. Measures were taken directly on the glass slides with a x20 objective. Paints being heterogeneous at such high magnification, analyses were restricted to the most homogeneous parts of the paint binder (Fig.2). Wavenumbers ranging from 3390200 cmS 1 were collected with a Full-range grating with both 532 nm and 780 nm lasers. A polystyrene standard was analyzed before each session or following laser change to validate the method’s accuracy. An automatic baseline correction was conducted with OMNIC 9,1 (Thermo Scientific, USA). Discrimination between samples was made visually using both qualitative differences (presence or absence of certain peaks) and semiquantitative differences (relative intensity of major peaks). A PCA statistical analysis was then employed to confirm the exclusions that had been obtained visually.
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
3
Fig. 1. Illustration of the five different production batches (1–5) for the four different paint models (A–D).
Table 2 Techniques and instrument parameters. Technique
Instrument
Microscopy Fourier transform infrared spectroscopy
Nikon Eclipse Ci (4x, 20x, 40x): Light field/ dark field/ polarized light / fluorescence (Dapi-Hoechst, FTIC and TITC filters). Thermo Nicolet iS10, 4 cm 1 resolution, 4000-650 cm 1 interval, DTGS KBr detector, 16 scans, software: OMNIC 9,1 (Thermo Scientific, USA) Thermo DXR Raman, lasers 532 nmFR, 780 nmFR, 780 nmHR, 3390-100cm 1 FR, 1800-100 cm 1 HR, 35 scans, software: OMNIC 9,1 (Thermo Scientific, USA) Autosampler CDS 5250 pyrolyser, Pyrolysis temp: 700 C, 10 seconds. Gas chromatograph : Agilent 7890A, DB-5 column (30 m length, 250 mm diameter, 0,25 mm thickness), injecter temp : 280 C, oven prog: 40 C/2 min; 290 C by an increase of 9 C/min (27,78 min); 290 C/33,7 min, mass spectrometer detector Agilent 5975C, mass domain 35-500 m/z
Raman spectroscopy Pyrolysis gas chromatography coupled to a mass spectrometer
Fig. 2. Illustration of the pigments distribution and surface’s rugosity between samples A1, B3, C1 and D4 when observed with Raman under dark field and a x20 objective.
2.2.4. Pyrolysis gas chromatography coupled to a mass spectrometer (Py-GC/MS) For each sample, paint fragments of equal length and width were collected using a scalpel under a stereomicroscope. These fragments were inserted into pyrolysis tubes. Prior to the sample’s insertion, the tubes had been prepared and cleaned to avoid any kind of contamination by introducing a quartz rod and approximately
2 mm of quartz wool before being pyrolyzed at 1200 C twice for 30 s. A polystyrene standard was analyzed between each sample of paint to ascertain the method’s accuracy. Polystyrene and each sample were then pyrolyzed at 700 C for 10 s. Every production batch was analyzed in triplicate. Differentiation between batches was based first on their qualitative aspects, namely the absence or presence of peaks. A semi-quantitative discriminationwas also possible based on
4
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
important variations in the relative intensities of peaks. Only peaks that were at least three times as high as the background noise were considered during the analysis. 2.3. Principal component analysis (PCA) Unscrambler X10 software (CAMO Norway) was used to compute Principal Component Analyses (PCA) on the infrared spectra. The spectra were first pretreated by SNV. PCA is a widely used statistical analysis technique that reorganizes and summarizes multidimensional data under new latent variables whilst minimizing information loss. The result is a limited set of data which is laid out in a way allowing for an easier visual interpretation. The final discrimination is, however, still subject to human interpretation ([7,11,8]) 3. Results 3.1. Microscopy No difference was visible between the batches from A, B and D models regardless of the microscopy lightings used. Some differences, however, were observed in the case of the paint C. Two groups were isolated: (C1;C5) and (C2;C3;C4). These differences
were visible in the fluorescent behavior of the samples, C1 and C5 being highly fluorescent through FITC and TRITC filters (Fig.3). Those same variations were slightly visible under dark field and in double polarization. Some samples also contained fluorescing spots whose amount varied between replicates. However, these could not be used for further discrimination since their amount varied as much within the replicates of each sample as between the samples themselves. The observed differences were reproducible between the triplicate preparations. No other batch was discriminated using microscopy, whether by bright and dark field, polarized light microscopy or fluorescence using the Dapi-Hoechst, FITC and TRITC filters. 3.2. Fourier transform infrared spectroscopy (FTIR) No significant differences were observed in the spectra of the A, B and D paints production batches (Fig. 4). Paint C, on the other hand, could be separated in three different groups: (C1), (C5) and (C2;C3;C4). Variations are visible between 900-1500 cm 1. They are also corroborated by the PCA statistical analysis, which distinguishes clearly between those three groups (Fig. 5). The peaks’ identification revealed the presence of acrylic resin (1730 cm 1), polyvinyl acetate (1025 cm 1) and calcium carbonate (1450 cm 1) for each of the four models.
Fig. 3. Paint C, production batches 1–5, observed under x4 magnification through FITC (left) and TRITC (right) filters.
Fig. 4. Example of infrared spectra for batches 1–5 (top to bottom) of paint B where no difference, either qualitative or semi-quantitative, were observed.
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
5
Fig. 5. Infrared spectra of batches 1–5 for paint C (top to bottom), and the results of the Principal Component Analysis (bottom) showing three discriminated groups (C1, C5 and C2-C3-C4).
3.3. Raman spectroscopy Raman analysis, although highlighting the difference in composition between the paints (A–D), did not permit any further distinction between the production batches (Fig.6). Titanium oxide in its rutile form was detected in each paint (A–D) and carbon black was identified in recycled paint D (Fig.7) [12]. Paint D presented a Raman spectrum of poor quality due to the fact that the laser had to be set on its lowest intensity since the dark tone of paint D amplified its absorbance. 3.4. Pyrolysis gas chromatography coupled to mass spectrometry (PYGC/MS)
groups is shown in Fig. 9. A close-up of the region around the retention time at 15 min shows a minor peak for C1, whereas C5 has a much more important peak with a slightly lower retention time. The three other batches do not possess peaks intense enough to be considered in the 3:1 signal-to-noise ratio. Fig. 10 illustrates the mass spectra differences for the minor peak before 15 min for both C1 and C5. This peak was later identified as being butyl benzoate, with its major peak at 105 m/z corresponding to carboxylic acid. The C5 spectrum was identified as a propenoic acid (2-methyl, 3-hydroxy-2, 4-trimethylpentyl ester), with its major peak at 71 m/z corresponding to the propenoic acid function itself. 4. Discussion
A, B and D paints all yielded chromatograms that were comparable between batches, except for very slight differences in relative intensities which were insufficient to pronounce an exclusion, and fell within the replicates (intra- or within-sample) variability. Mass spectra were checked for all considered peaks that presented a signal-to-noise ratio of at least three times. No meaningful differences were visually observed between the overlaid spectra of each batch (Fig. 8). The analysis of paint C, however, made it possible to distinguish between the groups (C1), (C5) and (C2;C3;C4). This discrimination was based on many differences in minor peaks that were consistent among replicates. An example illustrating the three
4.1. Discriminating power of the techniques The discriminating power has been calculated for each model of paint and each technique used in this study [13]. The choice of differentiating two samples (or not) is based on all available replicates for each pair (i.e. 3 microscopy, 7 FTIR, 3 Raman, 3 Py-GC/ MS). The decision was taken by careful comparison of their qualitative and semi-quantitative differences and the overlap of their between- (inter-) and within- (intra-) variation. In the case of the C paint, FTIR and pyrolysis-GC/MS techniques achieved a discriminating power of 0.7 between
6
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
Fig. 6. Production batches 1–5 for paint C (from top to bottom: C1, C2, C3, C4, and C5), measured by a 532 nm laser and full range gratings.
Fig. 7. Comparison of B2 Raman spectra with rutile references obtained by a 532 nm full range laser (A) and by a 780 nm full range laser (B). Comparison of D4 Raman spectrum and a carbon black reference spectrum analyzed respectively with a 532 nm full range laser (experimental, left) and a 5145 nm full range laser (reference, right) (C). Reference spectra were obtained from the RRUFF database [14].
samples, whilst microscopy allowed a slightly lower discrimination of 0.6. Raman spectroscopy left all C paint samples unseparated. All techniques have a discriminating power of zero for A, B and D paints. The overview of the discrimination for each model (seen in Table 3) shows that only the samples of paint C were partially discriminated. All other samples remained undifferentiable.
4.2. Verification with the manufacturer The paint characterization and analytical measurements were conducted without any prior knowledge of the paint composition or production batches adjustments that were made. The only information available was that paints C and D were recycled while A and B were not. After confirmation, the
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
7
Fig. 8. Major peaks identified in the paint A pyrograms of the 5 different batches.
Fig. 9. Comparison of the chromatograms from 5 batches of paint C with an enlargement of the zone surrounding 15 min.
manufacturer’s recipes (i.e. the precise amount of each raw material that was used in production) for paints A and B remained identical through all the batches that we received. In addition, every adjustment made to these batches during the quality control was made with the same raw materials, meaning that there exist no qualitative differences between those batches. Semi-quantitatively these production batches also were comparable, as would be expected for a production lot meeting the quality control requirements. This result is also consistent with previous results demonstrating that paints that
are uncolored (i.e. white, grey, black, . . . ) tend to be less differentiable ([7,8]). Recycled paints on the other hand, gave surprising results. Our initial hypothesis was that recycled paints, due to the many different origins of its raw materials, would be highly variable between production batches, both semi-quantitatively and qualitatively. While this was the case for the C paint that could be discriminated in three different groups (C1 vs C5 vs C2, C3, C4), the D paint remained indistinguishable over all the production batches. These results, although surprising, were however justified by the manufacturer’s
8
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
Fig. 10. Mass spectra of paints C1 (top) and C5 (bottom) corresponding to peaks illustrated in Fig. 9. The C1 mass spectrum corresponds to butyl benzoate, with its major peak at 105 corresponding to carboxylic acid. The C5 spectrum corresponds to a propenoic acid (2-methyl, 3-hydroxy-2, 4-trimethylpentyl ester), with its major peak at 71 corresponding to the propenoic acid function itself.
Table 3 Summary of the separation and discriminating power (DP) calculation for all samples considering all techniques. Paint
A (DP = 0)
B (DP = 0)
C (DP = 0,7)
D (DP = 0)
Discrimination
A1, A2, A3, A4, A5
B1, B2, B3, B4, B5
C1, C2, C3, C4, C5
D1, D2, D3, D4, D5
explanations. Paint C is a recycled product made by the manufacturer itself within its factory (Fig.11 (a)). Remains of subpar quality batches (i.e. failed quality control) as well as excess materials from various production lines are recovered and mixed together in order to create
the recycled material. Larger deviations from the desired attributes are accepted since this product is not intended for sale to the public but rather for their own use, or for other companies looking for cheaper raw materials. It is therefore reasonable to expect important
Fig. 11. (a) Recycling scheme for paint C: Remains of subpar quality or failed quality controls paints are recycled by the company and mixed together to produce a final product commercialized as a recycled paint. (b) Recycling scheme for paint D: wastes of paint from different origins (all brands and models) are collected through recycling centers and processed by a first company which mixes them together in large 2000 L vat, and sell them to a second company as~160 L barrels. The second company processes each barrel similarly to each of their other products. The final recycled product is homogeneous for each barrels within the same 2000 L vat, but expected to be much more variable for different 2000 L vat a few months apart.
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
variations between batches of this product. First, the paint collected after a failed quality control is by definition subject to high variation because it does not match the color reference, even though the raw materials at its origin should in theory all be the same. Second, the need for less quality control on the new batch of recycled paints allowed for even greater variations. In addition, Microscopy, FTIR, and Pyrolysis-CG/MS independently achieved the same separation of samples. Finally, production documents confirmed that these batches were made consecutively and that two additional recipe changes occurred over the course of their production. Those changes happened between C1 and C2 as well as between C4 and C5, which matches the separation achieved during the study. The production batches from the second recycled paint model, paint D, were not discriminated. The fact that it was indeed a recycled paint made this result unexpected at first, but a satisfying explanation was provided after a meeting with the manufacturer. The raw material collection and initial mixing from all of the recycled wastes was processed in a different factory (Fig. 11 (b)). The recycled paint binders are obtained by collecting large amounts of leftovers of household paint, mixing them together in a 2000 L vat and processing them for commercialization. The content of this 2000 L vat was distributed among 12 barrels, each being sold exclusively to our manufacturer and used to make paints D. Each barrel served as the basis raw material for production in a second factory, and therefore followed the same process (i.e. adding more binder, pigments, . . . ) and the same quality control as other non-recycled paints, however based on the raw recycled materials they received. In principle, each load of materials coming from the recycling company should be unique, since it is a complex and highly variable mix. The fact that no difference was observed between each of the production batches was therefore surprising. Upon consultation of the production documents, we discovered that all the batches of D paint received for the study had been made consecutively and using barrels of materials that all came from the same 2000 L vat from the recycling company. These materials account for most of the volume of the final product. According to the production documents, the rest of the recipe also remained identical throughout the production of these batches. In short, there are no qualitative differences expected between the samples of D paint, which is consistent with our results. In summary, the separation of samples of the C paint and the absence of observable variations within the other models (A and B) using our methodology seems in accordance with the market. Although limited to 4 types of paints and 5 different production batches each, our results do validate the current state of practice. Our results can help appreciate the amount of variation that can be typically expected between production batches of architectural paint, which may be useful when comparing a trace and a reference sample. Batch to batch differences are, however, only rarely considered in practice due to the improbability of differences being present. In the context of a real investigation, assumptions can hardly be made as to whether the examined paint was properly mixed before its application. Therefore it is difficult to assess if observed differences in relative concentrations are due to differences between production batches or merely to the heterogeneity of the paint during its application. In order to better comprehend the difficulties associated with the recycling of leftovers, an ongoing study is pursuing the analysis on a much larger selection of recycled paints. 5. Conclusion This study aimed to measure the variability of different production batches of architectural paints, including some recycled models. An absence of discernible differences between samples of non-recycled models was demonstrated. The
9
discriminating power appears to be higher amongst recycled paints as observed by optical microscopy techniques, Fourier transform infrared spectroscopy, and Pyrolysis-GC/MS, each providing meaningful information about the binder composition. Raman spectroscopy, much more sensitive towards pigments identification, did not permit to highlight any discrimination between the production batches. All of the study’s conclusions were corroborated by a verification with the manufacturer. Recycled paints were the only ones to show discernible differences between batches and the potential for discrimination. Each of the observed differences could be explained by consulting the manufacturer production documents. Future studies on recycled products are required for forensic scientists, as these are likely to become more and more common on the market in the years to come. CRediT authorship contribution statement Samuel Quevillon: Investigation, Methodology, Formal analysis, Software, Visualization, Writing - original draft. Nicholas Toupin: Investigation, Methodology, Formal analysis, Software, Visualization, Writing - original draft. André Tremblay: Conceptualization, Methodology, Supervision, Writing - review & editing. René Fournier: Conceptualization, Methodology, Supervision, Writing - review & editing. Cyril Muehlethaler: Conceptualization, Funding acquisition, Project administration, Methodology, Supervision, Writing - review & editing. Acknowledgements We would like to thank Myriam Desrosiers from the Laboratoire des Sciences judiciaires et de médecine légale de Montréal for her precious help with instruments and manipulations. We would also like to thank our partner manufacturer for lending us their samples and allowing us to consult their production documents. CM acknowledges a NSERC Discovery grant (RGPIN-2019-04827). References [1] J. Zieba-Palus, M. Trzcinska B, Application of infrared and raman spectroscopy in paint trace examination, J. Forensic Sci. 58 (2013) 1359–1363. [2] C. Muehlethaler, G. Massonet, P. Esseiva, Discrimination and classification of FTIR spectra of red, blue and green spray paints using a multivariate statistical approach, Forensic Sci. Int. 244 (2014) 170–178. [3] F. Govaert, M. Bernard, Discriminating red spray paints by optical microscopy, Fourier transform infrared spectroscopy and X-ray fluorescence, Forensic Sci. Int. 140 (2004) 61–70. [4] B. Stuart, Elemental analysis, Forensic Analytical Techniques, John Wiley & Sons, 2013. [5] J. Zie˛ba-Palus, G. Zadora, J.M. Milczarek, P. Koscielniak, Pyrolysis-gas chromatography/mass spectrometry analysis as a useful tool in forensic examination of automotive paint traces, J. Chromatogr. A 1179 (2008) 41–46. [6] D.T. Burns, K.P. Doolan, The discrimination of automobile clear coat paints indistinguishable by Fourier transform infrared spectroscopy via pyrolysis-gas chromatography-mass spectroscopy, Anal. Chim. Acta 539 (2005) 157–164. [7] C. Muehlethaler, L. Gueissaz, G. Massonnet, Forensic paint analysis, Encycl. Forensic Sci. (2013) 265–272. [8] C. Muehlethaler, et al., Survey on batch-to-batch variation in spray paints: a collaborative study, Forensic Sci. Int. 229 (2013) 80–91. [9] B. Caddy (Ed.), Forensic Examination of Glass and Paint, CRC Press, 2002. [10] C. Roux, et al., Intra-sample vs. Inter-sample variability in architectural paint, Proceedings of the NIJ/FBI Trace Evidence Symposium (2007) 13–16. [11] C. Muehlethaler, G. Massonet, P. Esseiva, The application of chemometrics on Infrared and Raman spectra as a tool for forensic analysis of paints, Forensic Sci. Int. 209 (2011) 173–182. [12] P. Miroslawa, J.-N. Rousaud, S. Duber, Raman Microscopy Characterization of Carbon Blacks: Spectral Analysis and Structural Information, 84, Elsevier, 2015, pp. 479–490. [13] K.W. Smalldon, A.C. et Moffat, The calculation of discrimination power for a series of correlated attributes, Sci. Justive 13 (1973) 291–295. [14] B. Lafuente, R.T. Downs, H. Yang, N. Stone, The power of databases: the RRUFF project, in: T. Armbruster, R.M. Danisi (Eds.), Highlights in Mineralogical Crystallography, W. De Gruyter, Berlin, Germany, 2015, pp. 1–30.
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Batch-to-batch variation in domestic paints: Insights into the newly commercialized recycled paints Samuel Quevillona,1, Nicholas Toupina,1, André Tremblayc, René Fournierc , Cyril Muehlethalera,b,* a b c
Université du Québec à Trois-Rivières, Département de Chimie, Biochimie, et Physique, Trois-Rivières, Canada Laboratoire de Recherche en Criminalistique, Trois-Rivières, Canada Laboratoire des Sciences Judiciaires et de Médecine Légale, Direction de la Chimie, Montréal, Canada
A R T I C L E I N F O
A B S T R A C T
Article history: Received 26 July 2019 Received in revised form 29 August 2019 Accepted 11 September 2019 Available online 12 September 2019
A few recent studies attempted to evaluate the differentiation of paints at a production batch level and reported results depending largely on the paint type. The discrimination from production batches is much more random than brands and/or models levels and subject to many unknowns, which suggests that a particular production batch can suddenly present a substantially different composition than the one produced right before or right after it. To add to this existing complexity, most of the paint companies now propose a range of recycled paints among their products. These recycled paints are composed of wastes collected by recycling plants, sorted by their color and binder type (i.e. latex, alkyds), and mixed together in large tanks to form the basis material for future formulations. Quality controls on these recycled batches are voluntarily less precise, and a higher variation is expected in esthetic and chemical properties of the paint. In this project, we collaborated with a North American paint producer that gave us access to its samples, paint formulations recipes, and a summary of the quality controls and corrections they performed on each production batches. The whole study was conducted blindfold and a final verification was made with the manufacturer to evaluate the accuracy of the results. The data set comprised two models of regular paints and two models of recycled paint. Five different production batches were collected per model for a total of 20 samples analyzed by Microscopy, Infrared spectroscopy, Raman spectroscopy, and Pyrolysis GC/MS. © 2019 Published by Elsevier B.V.
Keywords: Architectural paint Chemical analysis Production batches Forensics Microscopy Fourier transform infra-red spectroscopy (FTIR) Pyrolysis-GC/MS Raman spectroscopy
1. Introduction Population and discrimination studies represent a significant sub-field of all the articles about forensic paint analyses that were published ([1–8]). These studies not only permit to justify the choice of correct sequences of analysis that maximize discriminating power, but they also evaluate the heterogeneity in the population (uniqueness of the composition vs class characteristics), and finally assess at which level of discrimination one could/should evaluate a source attribution (i.e. brand, model, batch). These studies were important resources for paint interpretation, but also to provide general knowledge to better understand the market and the relative complexities of paint
* Corresponding author at: Université du Québec à Trois-Rivières, Département de Chimie, Biochimie, et Physique, Trois-Rivières, Canada. E-mail address: [email protected] (C. Muehlethaler). 1 Both authors contributed equally the paper. http://dx.doi.org/10.1016/j.forsciint.2019.109946 0379-0738/© 2019 Published by Elsevier B.V.
formulations (i.e. automotive vs household vs spray) (Caddy [9]. For the most common paint types and in most forensic cases, the differentiation at brand and model levels has often been the desired outcome. A few studies have tried to evaluate the utility of differentiating at a production batch level with very mixed results, showing an important influence of the paint type ([7,8,10]). A survey of the literature published to date shows that the differentiation at a production batch level is much more arbitrary and subject to many unknowns, such as (1) the quality controls and adjustments the paints are undergoing, (2) the varying origins of raw materials, or (3) the manufacturer’s decision to make changes in the paint recipes for improvement or discontinuity of raw materials. All of these reasons suggest that a particular production batch can suddenly present a substantially different composition than the one produced right before or right after it. Batch to batch variation is not systematic, but can suddenly present a drastically different chemical signature, possibly even individual, that would allow differentiating it from other production batches. The previous studies have shown that colored paints have a much
2
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
higher chance of showing a different composition due to the increased pigment(s) complexity. Black and white paints, on the other hand, showed no or little differences in their composition ([7,8]). While forensic scientists are still undergoing studies to understand the potential of discrimination between paints, industrials have taken it a step further. Most of the paint companies now propose a range of recycled paints among their products. These recycled paints are composed of wastes collected by recycling plants. First sorted by their color and binder type (i.e. latex, polyester) all the leftovers are mixed together in large tanks to form the basis material for future formulations. This lot can be used as is for lower quality paints or undercoats, or be refined and completed to required specifications. Quality controls on these recycled batches are obviously less precise, and a higher variation is expected in esthetic properties of the paint. While not monitored directly by the producers (i.e. not part of the systematic end-user quality controls), the variation in chemical composition should follow the same trends, if not greater. Most of the discrimination studies to date have been blind performed, without access to the manufacturer formulation recipes and unknowingly of the paint adjustments that were made. For this project, we collaborated with a North American paint company that shared with us all the details about their paint formulations. This participation allows us to bring a new viewpoint to forensic paint comparison, and the possibility to validate our analytical hypotheses. In this article, we present the differentiation of 5 batches of 4 different domestic paints measured by a standard sequence of analytical measurements (i.e. microscopy, infrared spectroscopy, Raman spectroscopy, and Py-GC/MS). By having a direct access to the manufacturing company we have compared our analytical findings with their paint formulation recipes, and evaluated if the quality control adjustments they have made do show up as significant modifications in the chemical composition. Two of the paints analyzed are respectively grey and white recycled paints, which permit us to bring a few insights into their chemical heterogeneity for analysis in forensic cases. 2. Material and method 2.1. Sampling Four different models of paint were obtained from the manufacturer and named A to D. For each model, five samples from different production batches were compared. Two of these models are regular paint (A and B), while the other two are recycled (C and D). The color, origin and production date of the samples are presented in Table 1. Reference microscopic slides were prepared for each of the production batches, forming a data set of 20 slides in total (Fig. 1). Every sample was, according to the quality assurance protocols of the manufacturer, mechanically homogenized for 20 min seven days prior to preparation of the slides. A second homogenizationwas done manually for one minute just before applying the paint. All slides had been previously washed with acetone, dried, and manipulated with gloves at all times. Paint from each sample was applied simultaneously on two glass slides with a rubber scraper, in order to create duplicates of equal thickness. One slide was reserved for pyrolysisGC/MS and manipulated as little as possible, whereas the remaining one would be used for all the other techniques which are less prone to detect contaminations due to manipulation. All paints slides have been dried a whole week prior to analysis. 2.2. Methods Four techniques were employed following a usual forensic paint examination: microscopy, infrared spectroscopy, Raman spectroscopy,
Table 1 Color, origin and production date of every sample. Sample
Color
Origin
Production date (month/year)
A1 A2 A3 A4 A5 B1 B2 B3 B4 B5 C1 C2 C3 C4 C5 D1 D2 D3 D4 D5
White White White White White Beige Beige Beige Beige Beige White White White White White Gray Gray Gray Gray Gray
New New New New New New New New New New Recycled Recycled Recycled Recycled Recycled Recycled Recycled Recycled Recycled Recycled
01/18 12/17 11/17 11/17 11/17 12/17 10/17 10/17 10/17 08/17 11/17 11/17 10/17 10/17 09/17 11/17 11/17 10/17 09/17 08/17
and pyrolysis GC–MS. Experimental details about each technique and instrument are presented in Table 2. Further details about the sample preparation are described here-below: 2.2.1. Microscopy One fragment from each of the five production batches were flattened onto a glass slide side-by-side, close enough for each of them to be observed in the same field of view through a Nikon Eclipse Ci with a 4x objective. No mounting liquid or cover slip were used and three similar slides were prepared for each model. Samples were observed under episcopic bright and dark field, polarized light, and fluorescence lights (Nikon cubes DapiHoechst, FITC, and TITC). 2.2.2. Fourier transform infrared spectroscopy (FTIR) The FTIR Spectrometer was used in transmittance through diamond compression cell. Seven replicates were taken per production batch. In each case, a paint fragment collected with a scalpel blade was deposited onto the diamond cell and flattened. Transmittance spectra in the range of 4000-650 cm 1 were measured with a resolution of 4,0 cm 1 and 16 co-added spectra ([7,8]). Discrimination between samples was made visually using both qualitative differences (presence or absence of certain peaks) and semi-quantitative differences (relative intensity of major peaks). A PCA statistical analysis was then employed to confirm the exclusions that had been obtained visually. 2.2.3. Raman spectroscopy Raman analyses consisted of three replicates for each production batch. Measures were taken directly on the glass slides with a x20 objective. Paints being heterogeneous at such high magnification, analyses were restricted to the most homogeneous parts of the paint binder (Fig.2). Wavenumbers ranging from 3390200 cmS 1 were collected with a Full-range grating with both 532 nm and 780 nm lasers. A polystyrene standard was analyzed before each session or following laser change to validate the method’s accuracy. An automatic baseline correction was conducted with OMNIC 9,1 (Thermo Scientific, USA). Discrimination between samples was made visually using both qualitative differences (presence or absence of certain peaks) and semiquantitative differences (relative intensity of major peaks). A PCA statistical analysis was then employed to confirm the exclusions that had been obtained visually.
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
3
Fig. 1. Illustration of the five different production batches (1–5) for the four different paint models (A–D).
Table 2 Techniques and instrument parameters. Technique
Instrument
Microscopy Fourier transform infrared spectroscopy
Nikon Eclipse Ci (4x, 20x, 40x): Light field/ dark field/ polarized light / fluorescence (Dapi-Hoechst, FTIC and TITC filters). Thermo Nicolet iS10, 4 cm 1 resolution, 4000-650 cm 1 interval, DTGS KBr detector, 16 scans, software: OMNIC 9,1 (Thermo Scientific, USA) Thermo DXR Raman, lasers 532 nmFR, 780 nmFR, 780 nmHR, 3390-100cm 1 FR, 1800-100 cm 1 HR, 35 scans, software: OMNIC 9,1 (Thermo Scientific, USA) Autosampler CDS 5250 pyrolyser, Pyrolysis temp: 700 C, 10 seconds. Gas chromatograph : Agilent 7890A, DB-5 column (30 m length, 250 mm diameter, 0,25 mm thickness), injecter temp : 280 C, oven prog: 40 C/2 min; 290 C by an increase of 9 C/min (27,78 min); 290 C/33,7 min, mass spectrometer detector Agilent 5975C, mass domain 35-500 m/z
Raman spectroscopy Pyrolysis gas chromatography coupled to a mass spectrometer
Fig. 2. Illustration of the pigments distribution and surface’s rugosity between samples A1, B3, C1 and D4 when observed with Raman under dark field and a x20 objective.
2.2.4. Pyrolysis gas chromatography coupled to a mass spectrometer (Py-GC/MS) For each sample, paint fragments of equal length and width were collected using a scalpel under a stereomicroscope. These fragments were inserted into pyrolysis tubes. Prior to the sample’s insertion, the tubes had been prepared and cleaned to avoid any kind of contamination by introducing a quartz rod and approximately
2 mm of quartz wool before being pyrolyzed at 1200 C twice for 30 s. A polystyrene standard was analyzed between each sample of paint to ascertain the method’s accuracy. Polystyrene and each sample were then pyrolyzed at 700 C for 10 s. Every production batch was analyzed in triplicate. Differentiation between batches was based first on their qualitative aspects, namely the absence or presence of peaks. A semi-quantitative discriminationwas also possible based on
4
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
important variations in the relative intensities of peaks. Only peaks that were at least three times as high as the background noise were considered during the analysis. 2.3. Principal component analysis (PCA) Unscrambler X10 software (CAMO Norway) was used to compute Principal Component Analyses (PCA) on the infrared spectra. The spectra were first pretreated by SNV. PCA is a widely used statistical analysis technique that reorganizes and summarizes multidimensional data under new latent variables whilst minimizing information loss. The result is a limited set of data which is laid out in a way allowing for an easier visual interpretation. The final discrimination is, however, still subject to human interpretation ([7,11,8]) 3. Results 3.1. Microscopy No difference was visible between the batches from A, B and D models regardless of the microscopy lightings used. Some differences, however, were observed in the case of the paint C. Two groups were isolated: (C1;C5) and (C2;C3;C4). These differences
were visible in the fluorescent behavior of the samples, C1 and C5 being highly fluorescent through FITC and TRITC filters (Fig.3). Those same variations were slightly visible under dark field and in double polarization. Some samples also contained fluorescing spots whose amount varied between replicates. However, these could not be used for further discrimination since their amount varied as much within the replicates of each sample as between the samples themselves. The observed differences were reproducible between the triplicate preparations. No other batch was discriminated using microscopy, whether by bright and dark field, polarized light microscopy or fluorescence using the Dapi-Hoechst, FITC and TRITC filters. 3.2. Fourier transform infrared spectroscopy (FTIR) No significant differences were observed in the spectra of the A, B and D paints production batches (Fig. 4). Paint C, on the other hand, could be separated in three different groups: (C1), (C5) and (C2;C3;C4). Variations are visible between 900-1500 cm 1. They are also corroborated by the PCA statistical analysis, which distinguishes clearly between those three groups (Fig. 5). The peaks’ identification revealed the presence of acrylic resin (1730 cm 1), polyvinyl acetate (1025 cm 1) and calcium carbonate (1450 cm 1) for each of the four models.
Fig. 3. Paint C, production batches 1–5, observed under x4 magnification through FITC (left) and TRITC (right) filters.
Fig. 4. Example of infrared spectra for batches 1–5 (top to bottom) of paint B where no difference, either qualitative or semi-quantitative, were observed.
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
5
Fig. 5. Infrared spectra of batches 1–5 for paint C (top to bottom), and the results of the Principal Component Analysis (bottom) showing three discriminated groups (C1, C5 and C2-C3-C4).
3.3. Raman spectroscopy Raman analysis, although highlighting the difference in composition between the paints (A–D), did not permit any further distinction between the production batches (Fig.6). Titanium oxide in its rutile form was detected in each paint (A–D) and carbon black was identified in recycled paint D (Fig.7) [12]. Paint D presented a Raman spectrum of poor quality due to the fact that the laser had to be set on its lowest intensity since the dark tone of paint D amplified its absorbance. 3.4. Pyrolysis gas chromatography coupled to mass spectrometry (PYGC/MS)
groups is shown in Fig. 9. A close-up of the region around the retention time at 15 min shows a minor peak for C1, whereas C5 has a much more important peak with a slightly lower retention time. The three other batches do not possess peaks intense enough to be considered in the 3:1 signal-to-noise ratio. Fig. 10 illustrates the mass spectra differences for the minor peak before 15 min for both C1 and C5. This peak was later identified as being butyl benzoate, with its major peak at 105 m/z corresponding to carboxylic acid. The C5 spectrum was identified as a propenoic acid (2-methyl, 3-hydroxy-2, 4-trimethylpentyl ester), with its major peak at 71 m/z corresponding to the propenoic acid function itself. 4. Discussion
A, B and D paints all yielded chromatograms that were comparable between batches, except for very slight differences in relative intensities which were insufficient to pronounce an exclusion, and fell within the replicates (intra- or within-sample) variability. Mass spectra were checked for all considered peaks that presented a signal-to-noise ratio of at least three times. No meaningful differences were visually observed between the overlaid spectra of each batch (Fig. 8). The analysis of paint C, however, made it possible to distinguish between the groups (C1), (C5) and (C2;C3;C4). This discrimination was based on many differences in minor peaks that were consistent among replicates. An example illustrating the three
4.1. Discriminating power of the techniques The discriminating power has been calculated for each model of paint and each technique used in this study [13]. The choice of differentiating two samples (or not) is based on all available replicates for each pair (i.e. 3 microscopy, 7 FTIR, 3 Raman, 3 Py-GC/ MS). The decision was taken by careful comparison of their qualitative and semi-quantitative differences and the overlap of their between- (inter-) and within- (intra-) variation. In the case of the C paint, FTIR and pyrolysis-GC/MS techniques achieved a discriminating power of 0.7 between
6
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
Fig. 6. Production batches 1–5 for paint C (from top to bottom: C1, C2, C3, C4, and C5), measured by a 532 nm laser and full range gratings.
Fig. 7. Comparison of B2 Raman spectra with rutile references obtained by a 532 nm full range laser (A) and by a 780 nm full range laser (B). Comparison of D4 Raman spectrum and a carbon black reference spectrum analyzed respectively with a 532 nm full range laser (experimental, left) and a 5145 nm full range laser (reference, right) (C). Reference spectra were obtained from the RRUFF database [14].
samples, whilst microscopy allowed a slightly lower discrimination of 0.6. Raman spectroscopy left all C paint samples unseparated. All techniques have a discriminating power of zero for A, B and D paints. The overview of the discrimination for each model (seen in Table 3) shows that only the samples of paint C were partially discriminated. All other samples remained undifferentiable.
4.2. Verification with the manufacturer The paint characterization and analytical measurements were conducted without any prior knowledge of the paint composition or production batches adjustments that were made. The only information available was that paints C and D were recycled while A and B were not. After confirmation, the
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
7
Fig. 8. Major peaks identified in the paint A pyrograms of the 5 different batches.
Fig. 9. Comparison of the chromatograms from 5 batches of paint C with an enlargement of the zone surrounding 15 min.
manufacturer’s recipes (i.e. the precise amount of each raw material that was used in production) for paints A and B remained identical through all the batches that we received. In addition, every adjustment made to these batches during the quality control was made with the same raw materials, meaning that there exist no qualitative differences between those batches. Semi-quantitatively these production batches also were comparable, as would be expected for a production lot meeting the quality control requirements. This result is also consistent with previous results demonstrating that paints that
are uncolored (i.e. white, grey, black, . . . ) tend to be less differentiable ([7,8]). Recycled paints on the other hand, gave surprising results. Our initial hypothesis was that recycled paints, due to the many different origins of its raw materials, would be highly variable between production batches, both semi-quantitatively and qualitatively. While this was the case for the C paint that could be discriminated in three different groups (C1 vs C5 vs C2, C3, C4), the D paint remained indistinguishable over all the production batches. These results, although surprising, were however justified by the manufacturer’s
8
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
Fig. 10. Mass spectra of paints C1 (top) and C5 (bottom) corresponding to peaks illustrated in Fig. 9. The C1 mass spectrum corresponds to butyl benzoate, with its major peak at 105 corresponding to carboxylic acid. The C5 spectrum corresponds to a propenoic acid (2-methyl, 3-hydroxy-2, 4-trimethylpentyl ester), with its major peak at 71 corresponding to the propenoic acid function itself.
Table 3 Summary of the separation and discriminating power (DP) calculation for all samples considering all techniques. Paint
A (DP = 0)
B (DP = 0)
C (DP = 0,7)
D (DP = 0)
Discrimination
A1, A2, A3, A4, A5
B1, B2, B3, B4, B5
C1, C2, C3, C4, C5
D1, D2, D3, D4, D5
explanations. Paint C is a recycled product made by the manufacturer itself within its factory (Fig.11 (a)). Remains of subpar quality batches (i.e. failed quality control) as well as excess materials from various production lines are recovered and mixed together in order to create
the recycled material. Larger deviations from the desired attributes are accepted since this product is not intended for sale to the public but rather for their own use, or for other companies looking for cheaper raw materials. It is therefore reasonable to expect important
Fig. 11. (a) Recycling scheme for paint C: Remains of subpar quality or failed quality controls paints are recycled by the company and mixed together to produce a final product commercialized as a recycled paint. (b) Recycling scheme for paint D: wastes of paint from different origins (all brands and models) are collected through recycling centers and processed by a first company which mixes them together in large 2000 L vat, and sell them to a second company as~160 L barrels. The second company processes each barrel similarly to each of their other products. The final recycled product is homogeneous for each barrels within the same 2000 L vat, but expected to be much more variable for different 2000 L vat a few months apart.
S. Quevillon et al. / Forensic Science International 303 (2019) 109946
variations between batches of this product. First, the paint collected after a failed quality control is by definition subject to high variation because it does not match the color reference, even though the raw materials at its origin should in theory all be the same. Second, the need for less quality control on the new batch of recycled paints allowed for even greater variations. In addition, Microscopy, FTIR, and Pyrolysis-CG/MS independently achieved the same separation of samples. Finally, production documents confirmed that these batches were made consecutively and that two additional recipe changes occurred over the course of their production. Those changes happened between C1 and C2 as well as between C4 and C5, which matches the separation achieved during the study. The production batches from the second recycled paint model, paint D, were not discriminated. The fact that it was indeed a recycled paint made this result unexpected at first, but a satisfying explanation was provided after a meeting with the manufacturer. The raw material collection and initial mixing from all of the recycled wastes was processed in a different factory (Fig. 11 (b)). The recycled paint binders are obtained by collecting large amounts of leftovers of household paint, mixing them together in a 2000 L vat and processing them for commercialization. The content of this 2000 L vat was distributed among 12 barrels, each being sold exclusively to our manufacturer and used to make paints D. Each barrel served as the basis raw material for production in a second factory, and therefore followed the same process (i.e. adding more binder, pigments, . . . ) and the same quality control as other non-recycled paints, however based on the raw recycled materials they received. In principle, each load of materials coming from the recycling company should be unique, since it is a complex and highly variable mix. The fact that no difference was observed between each of the production batches was therefore surprising. Upon consultation of the production documents, we discovered that all the batches of D paint received for the study had been made consecutively and using barrels of materials that all came from the same 2000 L vat from the recycling company. These materials account for most of the volume of the final product. According to the production documents, the rest of the recipe also remained identical throughout the production of these batches. In short, there are no qualitative differences expected between the samples of D paint, which is consistent with our results. In summary, the separation of samples of the C paint and the absence of observable variations within the other models (A and B) using our methodology seems in accordance with the market. Although limited to 4 types of paints and 5 different production batches each, our results do validate the current state of practice. Our results can help appreciate the amount of variation that can be typically expected between production batches of architectural paint, which may be useful when comparing a trace and a reference sample. Batch to batch differences are, however, only rarely considered in practice due to the improbability of differences being present. In the context of a real investigation, assumptions can hardly be made as to whether the examined paint was properly mixed before its application. Therefore it is difficult to assess if observed differences in relative concentrations are due to differences between production batches or merely to the heterogeneity of the paint during its application. In order to better comprehend the difficulties associated with the recycling of leftovers, an ongoing study is pursuing the analysis on a much larger selection of recycled paints. 5. Conclusion This study aimed to measure the variability of different production batches of architectural paints, including some recycled models. An absence of discernible differences between samples of non-recycled models was demonstrated. The
9
discriminating power appears to be higher amongst recycled paints as observed by optical microscopy techniques, Fourier transform infrared spectroscopy, and Pyrolysis-GC/MS, each providing meaningful information about the binder composition. Raman spectroscopy, much more sensitive towards pigments identification, did not permit to highlight any discrimination between the production batches. All of the study’s conclusions were corroborated by a verification with the manufacturer. Recycled paints were the only ones to show discernible differences between batches and the potential for discrimination. Each of the observed differences could be explained by consulting the manufacturer production documents. Future studies on recycled products are required for forensic scientists, as these are likely to become more and more common on the market in the years to come. CRediT authorship contribution statement Samuel Quevillon: Investigation, Methodology, Formal analysis, Software, Visualization, Writing - original draft. Nicholas Toupin: Investigation, Methodology, Formal analysis, Software, Visualization, Writing - original draft. André Tremblay: Conceptualization, Methodology, Supervision, Writing - review & editing. René Fournier: Conceptualization, Methodology, Supervision, Writing - review & editing. Cyril Muehlethaler: Conceptualization, Funding acquisition, Project administration, Methodology, Supervision, Writing - review & editing. Acknowledgements We would like to thank Myriam Desrosiers from the Laboratoire des Sciences judiciaires et de médecine légale de Montréal for her precious help with instruments and manipulations. We would also like to thank our partner manufacturer for lending us their samples and allowing us to consult their production documents. CM acknowledges a NSERC Discovery grant (RGPIN-2019-04827). References [1] J. Zieba-Palus, M. Trzcinska B, Application of infrared and raman spectroscopy in paint trace examination, J. Forensic Sci. 58 (2013) 1359–1363. [2] C. Muehlethaler, G. Massonet, P. Esseiva, Discrimination and classification of FTIR spectra of red, blue and green spray paints using a multivariate statistical approach, Forensic Sci. Int. 244 (2014) 170–178. [3] F. Govaert, M. Bernard, Discriminating red spray paints by optical microscopy, Fourier transform infrared spectroscopy and X-ray fluorescence, Forensic Sci. Int. 140 (2004) 61–70. [4] B. Stuart, Elemental analysis, Forensic Analytical Techniques, John Wiley & Sons, 2013. [5] J. Zie˛ba-Palus, G. Zadora, J.M. Milczarek, P. Koscielniak, Pyrolysis-gas chromatography/mass spectrometry analysis as a useful tool in forensic examination of automotive paint traces, J. Chromatogr. A 1179 (2008) 41–46. [6] D.T. Burns, K.P. Doolan, The discrimination of automobile clear coat paints indistinguishable by Fourier transform infrared spectroscopy via pyrolysis-gas chromatography-mass spectroscopy, Anal. Chim. Acta 539 (2005) 157–164. [7] C. Muehlethaler, L. Gueissaz, G. Massonnet, Forensic paint analysis, Encycl. Forensic Sci. (2013) 265–272. [8] C. Muehlethaler, et al., Survey on batch-to-batch variation in spray paints: a collaborative study, Forensic Sci. Int. 229 (2013) 80–91. [9] B. Caddy (Ed.), Forensic Examination of Glass and Paint, CRC Press, 2002. [10] C. Roux, et al., Intra-sample vs. Inter-sample variability in architectural paint, Proceedings of the NIJ/FBI Trace Evidence Symposium (2007) 13–16. [11] C. Muehlethaler, G. Massonet, P. Esseiva, The application of chemometrics on Infrared and Raman spectra as a tool for forensic analysis of paints, Forensic Sci. Int. 209 (2011) 173–182. [12] P. Miroslawa, J.-N. Rousaud, S. Duber, Raman Microscopy Characterization of Carbon Blacks: Spectral Analysis and Structural Information, 84, Elsevier, 2015, pp. 479–490. [13] K.W. Smalldon, A.C. et Moffat, The calculation of discrimination power for a series of correlated attributes, Sci. Justive 13 (1973) 291–295. [14] B. Lafuente, R.T. Downs, H. Yang, N. Stone, The power of databases: the RRUFF project, in: T. Armbruster, R.M. Danisi (Eds.), Highlights in Mineralogical Crystallography, W. De Gruyter, Berlin, Germany, 2015, pp. 1–30.