Introduction
Air-drying coatings consist of at least three main components, a binder which forms the final coating, a solvent that aids in application and uniformity coating, and a drier which catalyzes the conversion of the binder into the coating (Webster 2020). Ongoing interest in sustainable and non-toxic coatings has led to an increasing use of seed oils as binders (Hofland 2012). Seed oils contain unsaturated fatty acids (Fig. 1) which polymerize in the presence of oxygen gas to form solid films. This process, known as autoxidation, is initiated by H-atom abstraction from sites adjacent to olefins generating C-centered radicals (Hubert et al. 1997). While seed oil autoxidation does occur in the presence of oxygen, it typically takes months for binders of this type to fully dry in the absence of a drier. Driers, typically oil-soluble transition metal compounds, catalyze H-atom abstraction allowing solid films to form within hours of application (Bieleman 2002).
Currently, most driers used commercially are based on cobalt. These cobalt-based driers are inexpensive and active at dosages as low as 0.01 wt. % (Charamzová et al. 2018). Recent concerns about potential toxicological and environmental effects of cobalt in driers (Leyssens et al. 2017) have led to efforts to transition from cobalt-based driers to less toxic substitutes, including driers based on manganese(Bouwman and van Gorkum 2007), iron (Křižan et al. 2017), and vanadium (Preininger et al. 2015). To accomplish this transition, a detailed understanding of the drying process is highly valuable.
Many mechanisms have been proposed for the role of the drier in autoxidation (van Gorkum and Bouwman 2005; Soucek et al. 2012; Honzíček 2019). One widely invoked hypothesis involves the formation and decomposition of organic peroxides through a Haber-Weiss-type mechanism (Haber and Weiss 1932). In this mechanism low-valent driers reduce organic peroxides by one electron, forming alkoxy radicals (RO) and hydroxide ions (OH-). Subsequently, oxidized driers interact with a second organic peroxide, reducing the metal to its original state and forming peroxy radicals (ROO) and protons (H+) (Fig. 2a). A second mechanism which has been proposed to contribute to radical initiation is peroxide homolysis catalyzed by Co(III) serving as a Lewis acid (Fig. 2b) (Spier et al. 2013). In addition to this mechanism, it has also been proposed that low-valent driers can interact directly with molecular oxygen to generate superoxide (O2•-) and high valent metal-oxo species (Fig. 2c) (Kumarathasan et al. 1992). Oxidized driers are then reduced to their original oxidation state by reacting with the binder.
Radical species formed through either mechanism, as well as high valent metal species that may be formed, then proceed to abstract H-atoms from bis-allylic methylene sites (Fig. 3, A ) in the binder. After H-atom abstraction, the olefins rearrange to a conjugatedcis -trans arrangement (Fig. 3, B ) and the C-centered radicals react with dissolved oxygen gas or other unsaturated sites in the binder, propagating the radical chain. Further interaction between these newly formed cis -trans conjugated species and propagating radicals leads to the formation of isolatedtrans -alkene groups in the binder (Fig. 3, C ) (Hubert et al. 1997). Details of these mechanisms, including the relative contribution of each of the possible initiation reactions, the importance of dissolved oxygen during propagation, and the role or roles of the drier in propagation and termination have been the subject of extensive computational modeling (Oakley et al. 2018). In order to test the veracity of these proposed mechanisms, more experimental data on the rates of autoxidation using different drier and oil combinations is needed (Orlova et al. 2021).
Drying oil autoxidation is typically studied with Fourier transform infrared spectroscopy (FT-IR) using the method developed by van de Voort (van de Voort et al. 1994). In a typical experiment, the area of the peak centered at 3010 cm-1 is measured as a function of time during the autoxidation process. This peak corresponds to an isolated cis -alkene stretching mode of the non-conjugated olefins initially present in the drying oil (Fig. 3, A ). During autoxidation cis -olefins isomerize, and the area of the peak at 3010 cm-1 decreases in intensity (de Boer et al. 2013). The rate of autoxidation is then determined by performing a linear regression on the peak area as a function of time (Pirš et al. 2014). While this method is generally reliable for determining the rate at which cis -alkene groups are consumed, it does not provide information on the other reactions taking place during autoxidation. In particular, this method does not provide information on the formation oftrans -alkene species (Fig. 3, C ) related to propagation and termination reactions.
As an alternative to monitoring a single peak in the IR region, we propose the use of global analysis to study drying oil autoxidation rates. Global analysis is a family of mathematical techniques used to find optimal fitting parameters between a model containing multiple unknown variables and the entirety of known experimental data. Global analysis typically begins by applying factor analysis to determine the number of variables contributing to the experimental data (Malinowski 2002). The results of this factor analysis are then combined with knowledge of the system to propose a model, and an iterative optimization algorithm identifies the model parameters which provide the best agreement between the experimental data and the proposed model. In chemistry and biochemistry, global analysis has been widely applied to spectroscopic data to improve the understanding of reaction kinetics in complex systems (Malinowski 2002; van Stokkum et al. 2004). Global analysis is particularly powerful in cases where: a) the spectrum of one or more components cannot be independently measured due to experimental limitations, b) the measured absorbances of one or more components are small relative to other components due to a low concentration and/or low absorptivities, and c) there is significant spectral overlap between the absorbances of different components in the system. Each of these factors is present in the time-resolved FT-IR spectra of drying oil autoxidation, making this process ripe for study using global analysis. In this manuscript, we report the results of our initial application of global analysis to linseed oil drying by cobalt driers.