The experimental design presented seventeen combinations (Table 1), including five replicates of the central point in order to estimate pure error and to assess the lack of fit of the proposed models. All the experiments were performed randomly and in triplicate.