It has been observed that during the transformation of minerals at higher temperatures (> 1000 degrees C), mineral species are formed containing a high number of oxygen molecules, i.e. gehlenite (Ca(2)Al(2)SiO(7)) mullite (Al(6)Si(3)O(15)), margarite (CaAl(4)Si(2)O(15)(OH)(2)) and almandine (Fe(3)Al(2)Si(3)O(12)). Results of the coal sources evaluated in this investigation indicated significant differences in mineral elemental composition, i.e. the CaO content varied between 5 mass % and 10 mass %, the Fe(2)O(3) content varied between 1.6 mass % to more than 5 mass %, as well as differences in the TiO(2), P(2)O(5) and MgO content. The coal sources producing the highest concentration of Ca-Al-Si species (CaAl(2)Si(2)O(8) anorthite and CaAl(4)Si(2)O(10)(OH)(2) margarite), which crystallized from the slag-liquid phase during the combustion stage, also contained the highest amount of acidic components or highest percentage of kaolinite. The highest concentration of mullite and free SiO(2) after the gasification reaction (before the combustion zone), also resulted in the highest concentration of Ca-Al-Si compounds forming during the oxidation phase. The free-SiO(2) in the mineral structure of the coal sources resulted then in the formation of mineral structures with Mg, Na or Ca when present in the mineral structure, to form new mineral compounds such as KAl(3)Si(3)O(10)(OH)(2) (Muscovite), Mg(5)Al(2)Si(3)O(10)(OH)(8) (clinochlore), or other high oxygen molecule-containing mineral compounds. Thus, if free-SiO(2), was not present after the gasification phase, and mostly taken up in the form of anorthite (due to high or higher CaO contents or Fe-contents in high Fe-containing coal sources), the concentration of Si-oxygen capture compounds are relatively low. An acceptable linear correlation between oxygen capture tendencies (increase in mineral matter content during the combustion phase) versus CaO-content was obtained with the South African coal sources evaluated. This confirmed the observations obtained based on HT-XRD and FactSage modelling. It can be concluded that the linear model to predict oxygen capture behavior from CaO-content is acceptable and can be used as a predictive tool. The SiO(2) content, for example, has an inverse affect on oxygen trends up to a specific concentration of CaO in the coal. However, this model is only valid for the coal types tested (South African Highveld coal sources), and additional test work will have to be conducted for other coal types, i.e. northern hemisphere coal. (c) 2008 Elsevier Ltd. All rights reserved.