Prediction of the thermal decomposition of organic peroxides by validated QSPR models

Highlights

• QSPR models were developed for thermal stability of organic peroxides.

• Two accurate MLR models were exhibited based on quantum chemical descriptors.

• Performances were evaluated by a series of internal and external validations.

• The new QSPR models satisfied all OCDE principles of validation for regulatory use.

Abstract

Organic peroxides are unstable chemicals which can easily decompose and may lead to explosion. Such a process can be characterized by physico-chemical parameters such as heat and temperature of decomposition, whose determination is crucial to manage related hazards. These thermal stability properties are also required within many regulatory frameworks related to chemicals in order to assess their hazardous properties. In this work, new quantitative structure–property relationships (QSPR) models were developed to predict accurately the thermal stability of organic peroxides from their molecular structure respecting the OECD guidelines for regulatory acceptability of QSPRs. Based on the acquisition of 38 reference experimental data using DSC (differential scanning calorimetry) apparatus in homogenous experimental conditions, multi-linear models were derived for the prediction of the decomposition heat and the onset temperature using different types of molecular descriptors. Models were tested by internal and external validation tests and their applicability domains were defined and analyzed. Being rigorously validated, they presented the best performances in terms of fitting, robustness and predictive power and the descriptors used in these models were linked to the peroxide bond whose breaking represents the main decomposition mechanism of organic peroxides.

Introduction

Organic peroxides are reactive compounds, containing the −O−O− bond [1], [2], that can be formed naturally by auto-oxidation with oxygen in certain solvents, such as diethylether. They can lead to highly explosive peroxidic residues, requiring specific safety precautions, like addition of oxidation inhibitors to prevent the formation of undesirable organic peroxides [3], [4], [5], [6]. Organic peroxides are more or less stable due to a relatively low −O−O− bond energy (20–50 kcal/mol) [7]. Since they generate instable radicals during their decomposition, organic peroxides are commonly used as catalyst and as radical polymerization initiators. The decomposition of organic peroxides can nevertheless be dangerous and can lead to serious effects [8], [9], [10]. To reduce the risk of incidents and of accidents, their hazards are intensively studied. In order to avoid the accidents by the customers and users, the organic peroxide producers provide information concerning the properties of commercial organic peroxides and give recommendations for safe handling and use of organic peroxides [11], [12]. General documents also describe the characteristics of organic peroxides and the safety rules to apply to safely handle them at laboratory scale [3].

Concerning the regulations, organic peroxides belong to a dedicated division (Division 5.2), as described in the UN Dangerous Goods Transportation Recommendations [13], or in GHS (Globally Harmonized System of classification and labelling of chemicals) [14] with 7 different classes (types A to G) related to their hazardous potential and leading to different amounts authorized for transport.

The decomposition of organic peroxides can be triggered and accelerated by heat, mechanical shock or friction and by various contaminants [15], [16]. To produce, transport and provide safely the numerous organic peroxides, the industry generally commercializes them in low concentration diluted in variable solvents.

In order to improve and homogenize the knowledge of the marketed chemicals, the European Union regulation REACH (Registration, Evaluation, Authorization and Restriction of Chemicals) [17], [18] requires the evaluation of physico-chemical, toxicological and eco-toxicological properties for all chemicals produced or imported by more than one ton by year in Europe. To help industry to meet the requirements of REACH regulation as far as the chemical safety assessment is concerned, ECHA technical guidances have been published [19], [20] and a general testing strategy for physico-chemical properties was proposed to consider the order of testing.

As far as organic peroxides are concerned, even if they belong to a dedicated regulated division or class, the knowledge of their explosive properties, as defined by the UN recommended tests [21], is of great importance. As entry data, their thermal stability (represented by the energy and temperature of decomposition) is a key property that is considered as a pre-selection criterion to identify substances that could undergo explosive reactions. So, it is used in the complex procedure of classification of explosives and organic peroxides [21]. Indeed, the UN regulation indicates that there is no need to perform this complex procedure when the decomposition heat (corresponding to the amount of energy released during the decomposition) is lower than 500 J/g. Measured by calorimetric analyses, notably by differential scanning calorimetry (DSC), the decomposition heat is estimated with measurement uncertainties of about 5–10% [22], [23] and less than 5 °C for the onset temperature [23].

For safety reasons but also for technical reasons, experimental tests can be difficult to implement for unstable substances like organic peroxides. As a consequence, the development of methods used for the prediction of data can be of great help at the research and development step and can help to accelerate and fulfil the next registration deadlines. As a simple prediction tool of reactivity hazards, the CHETAH software based on Benson's group contribution method was developed by ASTM [24]. Considering only six organic peroxides, Mohan et al. [25] demonstrated some correlations between CHETAH criteria (oxygen balance, the maximum decomposition heat, the difference between heat of combustion and decomposition heat) and explosive properties. In one recent study, Sato et al. [26] showed that there is a mutual correlation between CHETAH criteria and the explosibility of self-reactive substances except for organic peroxides and azo compounds. Nevertheless, it has to be noticed that the CHETAH software provides the maximum decomposition heat (considering that the available oxygen first oxidizes hydrogen to water and then carbon to carbon dioxide) and not the actual experimental decomposition heat.

Considering the REACH regulatory framework, Lewis et al. [27] advocated the use of powerful computer-aided ab initio techniques to generate predictions of key properties of broad classes of chemicals, without resorting to costly experimentation and potentially hazardous testing. Among these alternative methods to experimental testing, Quantitative Structure–Activity/Property Relationships (QSAR/QSPR) were clearly recommended in REACH and in technical guidances [19], [20] to obtain information data. Indeed, they represent powerful tools of prediction [28] used more and more for physico-chemical applications [29], [30]. Their applicability in an industrial context was also recently demonstrated by Patlewicz et al. [31]. To support the development and use of QSPRs, OECD drawn up 5 principles for the validation of QSAR/QSPR models for regulatory purpose [32].

Some recent reviews [33], [34] list the existing predictive models developed for the relevant properties of chemicals in the context of REACH. In particular, Dearden et al. [34] focused on the validation of models dedicated to physico-chemical properties according to OECD principles to favor the use of predicted property values in submissions to the European Chemicals Agency (ECHA).

Considering the prediction of thermal stability, some models exist for different families of compounds like nitroaromatic compounds [35], [36], [37], [38], [39], nitramines [40], [41], [42], ionic liquids [43] or polymers [44]. Nevertheless, to our knowledge, only one reference exists [45] to predict the reactivity hazards of organic peroxides using the QSPR approach. In this study, a limited database of 16 organic peroxides was used to derive models with no validation set, neither definition of applicability domain. Therefore, the robustness and the predictivity of these models must be at least validated and may be improved.

Consequently, the aim of this paper was to develop new robust QSPR models respecting OECD principles dedicated to the prediction of the heat and temperature of decomposition. To achieve this goal and considering that no large database exists in literature for organic peroxides (only 16 and 9 DSC data from the works of Lu et al. [45] and Ando et al. [22] respectively), an experimental database of 38 organic peroxides was built for these properties obtained in homogenous experimental conditions using DSC. The amount and quality of experimental data allowed to performing both internal and external validation methods to ensure the good performances of the multi-linear regression (MLR) models developed and then reach more robust and accurate models than the ones proposed by Lu et al.

Besides, the approach combined QSPR methodology with quantum chemical descriptors obtained with density functional theory (DFT) calculations. Indeed, a better chemical interpretation of the developed models can be expected using this type of descriptors as already demonstrated in previous works for nitro compounds [46], [47], [48]. To our knowledge, this work leads to the first completely validated QSPR models (including the definition of applicability domains) dedicated to the prediction of the thermal stability of organic peroxides.