Pyrolysis tests were conducted in an atmospheric thermobalance (a MK2 microbalance with a readability of 0.1 μg from CI Precision, UK) under nitrogen atmosphere (with a flow rate through the reaction chamber of 200 cm3 NTP min− 1). An unreplicated 2-level factorial design was adopted to study the effect of two factors: peak temperature (400–600 °C) and linear heating rate (5–40 °C min− 1). Three replicates at the center point (22.5 °C min− 1 and 500 °C) were performed to simultaneously estimate the experimental error and the overall curvature effect [32] and [33]. Table 2 shows the matrix of the experimental design. The initial sample weight was 250 mg, a relatively large amount compared to the most commonly used initial sample masses (1–4 mg), which are preferred to avoid serious systematic errors in the measurement of the sample temperature in a TGA device [16]. However, using small sample sizes can also lead to other potential problems, such as the inability to asses a representative sample and the possibility to magnify surface phenomena at the expense of intra-bed controlling processes [34].