Extension study of a statistical age prediction model for acrylic paints

https://doi.org/10.1016/j.polymdegradstab.2020.109263Get rights and content

Highlights

  • Feasible age prediction model for different acrylic paint brands and pigment colours.

  • Predictive robustness in acrylic paints exposed to slightly different aging conditions.

  • Application to the dating of relatively old contemporary artworks.

Abstract

In this work, the robustness and potential applicability of statistical age prediction models applied to the dating of different acrylic paints were studied. The FTIR-ATR analysis of three acrylic colours (Hansa yellow, phthalocyanine green and ultramarine blue) from two manufacturers (Liquitex® and Vallejo®) subjected to accelerated ageing was carried out. The acrylic paints were characterised and the modifications of their ATR spectra throughout ageing were studied. Predictive models developed with the Liquitex® brand containing phthalocyanine green pigment were then applied to other colour and brands of acrylic paints and their robustness and feasibility were studied based on calculated accuracy error values. The influence of the pigment on the ageing of the paint components, the type and quantity of additives present in the acrylic paint as well as the ageing conditions to which it was subjected were decisive in the short-term predictive model, which explains the low accuracy values obtained for all the acrylic paints analysed. However, the slower degradation processes taking place in the longer term and the stabilisation of the acrylic paints at higher stages of ageing made them fit successfully into the long-term model, obtaining an error of between 14 and 23%. Thus, the predictive statistical model is robust and feasible to be used for different colours of the same brand of acrylic paint as well as for acrylic paints of different brands that have been long-term aged under slightly different conditions of accelerated ageing. In conclusion, this methodology could be a promising tool in the field of dating contemporary artworks of a certain age.

Introduction

The contemporary art market is experiencing an unprecedented level of buying and selling activity that generates millions of dollars worldwide. As art prices rise, art-related crime is on the increase, making counterfeiting a multi-million dollar enterprise. The uncovering of fake artworks has a negative impact on the entire art market. That is why new technologies are being used to detect counterfeits that have already entered the art market as well as to prevent their sale and acquisition [1,2].

Non-destructive techniques such as Raman spectroscopy [[3], [4], [5], [6]], Fourier transform infrared spectroscopy (FTIR) [3], IR hyper spectral imaging (IR-HIS) [7,8], FTIR-attenuated total reflection (FTIR-ATR) [6,9], IR reflection [10] and scanning electron microscopy-energy dispersive X-ray spectroscopy (SEM-EDX) [6,9] are usually used as first option as they do not damage the work of art [3,6,9]. Among the techniques, FTIR-ATR spectroscopy is widely used for the characterisation of artwork materials [6,9,11] and for the study of the influence of pigments on the photodegradation of acrylic paints [[12], [13], [14], [15]].

In some cases, non-invasive techniques are not enough, so analytical techniques combined with statistical methods, such as multivariate chemometrics, have been widely used to extract the maximum and most representative information from the acquired data in an interpretable manner. Partial Least Squares (PLS) is an X→y regression methodology typically used for either predictive purposes or control monitoring. Despite the applications of partial least squares (PLS) for quantification of paint components in the restoration of artworks [16] as well as for the determination of their properties in quality control [17], there are not many studies that apply this method to the dating of works of art, although the applicability and effectiveness of this tool has been demonstrated in other forensic fields [18,19]. An important challenge that the PLS has to face is the eventual lack of robustness of the models when they want to be used for universal applications [20]. Since the setup of an optimal FTIR-ATR multivariate model is costly, once it has been developed it is expected to be valid for a long period of time.

In previous research by L. Ortiz-Herrero et al. [21], a non-invasive paint dating methodology was developed for the authentication of acrylic paints used in contemporary works of art. A multivariate chemometric method (orthogonal partial least squares (OPLS)) was applied to FTIR-ATR spectroscopic data acquired from Liquitex® brand aged paints containing the phthalocyanine green pigment (PG7) to develop age prediction models split into two temporal ranges (short- and long-term). The applicability of the predictive models was tested in artworks created with the Liquitex® brand by internationally recognised contemporary Basque artists. The correlation obtained between natural ageing and its equivalent to the accelerated one may have possible applications in the dating of artworks for up to at least 22 years preserved under study conditions and created with the same paint formulation. In addition, the loadings graph allows studying which variables influence the predictive models. From this graph, the ageing processes of the paint constituents by which the predictive models were characterised were studied. The short-term model was influenced by the degradation of the acrylic binder due to chain-scission reactions and oxidation phenomena as well as by the non-ionic polyethylene oxide surfactant due to the excision of the hydrocarbon chains. The long-term model was only characterised by the modifications undergone by the acrylic binder due to the disappearance of oxidized species and cross-linking reactions. Thus, it was concluded that the predictive models could potentially be applied to other brands of acrylic paints with the same n-butyl acrylate-methyl methacrylate (nBA-MMA) copolymer as well as to different Liquitex® brand paint colours, as pigment modifications during ageing seemed not to be a determining factor in the predictive models.

The aim of this article was to corroborate the above facts and to study the robustness and potential applicability of the predictive models in different acrylic paint colours and brands. For this purpose, three Liquitex® paint colours and a Vallejo® paint that shared the same organic pigment and acrylic resin as the paint used in the previous research were selected. The paints were exposed to slightly modified conditions of accelerated ageing, analysed by FTIR-ATR and subsequently applied to the predictive models to determine their robustness and feasibility based on the accuracy error values obtained for each paint sample. The acrylic paints were also characterised and the evolution of their ATR spectra throughout ageing was studied.

Section snippets

Paint samples and preparation

The choice of manufacturers and paint colours was made on the basis of publications and interviews with contemporary artists to find out the most studied and widely used materials. Different pigments were selected to check whether the type and colour of the pigment influenced the chemical behaviour and stability of the acrylic paints, and therefore whether it could affect the applicability of the predictive model. Commercial acrylic paints from Liquitex Acrylic Paint® (U.S.A.) containing Hansa

Characterisation by FTIR-ATR

The characterisation of FTIR-ATR spectra of Liquitex® paints containing PY3 and PB29 pigments was performed (Fig. 1). The characterisation of Liquitex® paint with PG7 pigment was previously carried out by L. Ortiz-Herrero et al. [21]. The two Liquitex® paints gave a characteristic nBA-MMA copolymer profile. A strong band at around 1727 cm−1 ascribed to the carbonyl stretching vibration was observed. The absorption bands at about 2955, 2934 and 2874 cm−1 were attributed to C–H bond stretching

Conclusions

In this study, four commercial acrylic paints from two different manufacturers were characterised by FTIR-ATR. Each acrylic paint showed characteristic modifications of its ATR spectrum throughout the accelerated ageing in spite of sharing the same paint constituents such as the acrylic binder, the surfactant and in two of the cases the organic pigment.

Pigments were found to influence the short-term stability of acrylic paints when exposed to accelerated ageing, as they may act as promoters or

Funding

This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.

CRediT authorship contribution statement

L. Ortiz-Herrero: Investigation, Methodology, Validation, Formal analysis, Data curation, Writing - original draft, Visualization. I. Cardaba: Resources, Investigation. L. Bartolomé: Conceptualization, Supervision, Project administration, Writing - review & editing. M.L. Alonso: Writing - review & editing. M.I. Maguregui: Conceptualization, Supervision, Project administration, Writing - review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

The authors would like to thank the General Research Services (SGIker) of the UPV/EHU for the technological support. Authors also thank to Omaira de la Hera of the Faculty of Science and Technology of the UPV/EHU for her useful discussion. L. Ortiz- Herrero thanks UPV/EHU for the pre-doctoral fellowship.

References (32)

Cited by (4)

  • Multivariate (O)PLS regression methods in forensic dating

    2021, TrAC - Trends in Analytical Chemistry
    Citation Excerpt :

    Since multivariate regression models are built from samples prepared and aged under controlled conditions, we would expect each forensic scenario to have its own tailored model. However, this is not always the case, as there are multivariate regression models that have been successfully applied to samples different from those used in the construction of the model and exposed to moderately different conditions, as Ortiz-Herrero et al. state in their study [31]. Ortiz-Herrero et al. [31] studied the robustness and feasibility of the short- and long-term ageing OPLS models of the Liquitex® paint brand when applied to date different brands and colours of acrylic emulsion paints.

View full text