Comparative study of electrochemical transducer fabrication methods with glucose oxidase as recognition layer

Abstract

Combination of reduced graphene oxide (rGO) with a conductive polymer, poly(3,4-ethylenedioxythiophene):poly(styrene sulfonate) (PEDOT:PSS) is promising as transducer material for electrochemical biosensors. However, fundamental research into this composite behaviour is essential for generating the necessary scientific understanding to realize a non-invasive glucose monitoring approach. In this study, the rGO-PEDOT:PSS modified electrodes were fabricated with four different methods (method A, B, C, and D) where the fabrication parameters such as reduction cycles, sequence and glucose oxidase immobilisation techniques were varied. The aim is to elucidate how these fabrication parameters can affect the electrochemical reversibility, mass transport properties, heterogeneous electron transfer rate constant (k⁰) and effective surface area (Aeff) of rGO-PEDOT:PSS materials in which can be determined from cyclic voltammetry (CV). This study also utilized machine learning algorithm to model the data from CV results where the most accurate model was used to analyse the interaction strength between the input and output data. From the electrochemical analysis, the ferri/ferrocyanide redox couple [Fe(CN)6]3-/4- shows quasi-reversible and diffusion-controlled behaviour on rGO-PEDOT:PSS-modified SPCEs of all fabrication method. In terms of k⁰, each fabrication methods generated different trend of k⁰ value with increasing reduction cycles. Overall, the range of k⁰ for rGO-PEDOT:PSS-modified SPCE are from 0.52 x10-5 to 4 x10-5 cm/s. We also found that the highest Aeff value with respect to fabrication method were obtained from different number of reduction cycles. Fabrication method A gave the highest Aeff when the composite was reduced for 5 reduction cycles (16.41 cm2). For methods B and D, the highest Aeff obtained was for 30 reduction cycles (17.48 cm2 for method B and 6.43 cm2 for method D) while for method C, the highest Aeff value was obtained for 15 reduction cycles (21.41 cm2). For SVM analysis, data from CV (ΔEp and Ipc) and the fabrication parameters were used to construct a prediction model. Linear and non-linear kernels were compared, and the best performance was showed by radial basis function (RBF) kernel with a perfect accuracy of 100%. The RBF kernel was then used to measure the interaction between input and output variables in which the kernel model demonstrated the strongest interaction between reduction cycles and ΔEp. In conclusion, this study reveals the effect of fabrication parameters on electrochemical characteristics of rGO-PEDOT:PSS-modified SPCEs and the capability of machine learning algorithm to model data and provide deeper insights on the fabrication process, opening a new route for development of non-invasive glucose biosensors.