Optimisation of saturation magnetisation of iron nanoparticles synthesized by hydrogen reduction: Taguchi technique, response surface method, and multiple linear and quadratic regression analyses

In this study, Taguchi method, response surface methodology and regression analyses have been applied to assess the effects of synthesis parameters on saturation magnetisation, M-s of iron nanoparticles produced by hydrogen reduction of iron oxide nanoparticles. The M-s values were obtained from the magnetisation loops measured by a vibrating sample magnetometer. Structural characterisations of the selected samples were done by X-ray diffraction technique and transmission electron microscopy. Orthogonal array L9 with three parameters (temperature, reaction time and H-2 flow rate) at three levels (low, medium and high) was used to obtain the experimental trials. Based on the signal to noise (S/N) ratio considering the condition larger is the better approach and the mean response, the highest Ms condition has been found when the temperature is 800 degrees C m/min, reaction time is 60 min and H-2 flow rate is 1000 ml/min. Analysis of Variance (ANOVA) is applied to find out the F-ratio and percentage contribution of each parameter by using experimental trials and S/N ratios. It was found that the temperature was the most significant parameter on the Ms of iron nanoparticles. A confirmation experiment has been carried out by using the experimental conditions obtained from Taguchi method. The Ms of the optimum sample was found to be 217.42 emu/g which was within 95% confidence level with the predicted optimal Ms of 214.03 emu/g. The quality losses for Ms obtained at the highest combinations were 70.7%. In addition, analyses of multiple linear and quadratic regressions were employed to derive the predictive equations of the Ms achieved via experimental design. The predicted values from the developed models and experimental values are found to be very close to each other justifying the significance of the models. Besides, the quadratic interactive regression model provided the best statistical performance with high R-2 and R-2(adj) values of 100 and 100%, respectively between the experimental and predicted values of Ms. More intensive predicted values were obtained by the quadratic regression models as compared to the multiple linear regression models. Taguchi prediction method was also very successful in the optimization of synthesis parameters for the highest Ms of nanoparticles within the prescribed limit.