Error-compensating measurement and data-processing approaches for transient-based methods involving physico-chemical systems: Studies of membrane-based oxygen-selective and enzyme-based reactor/sensor devices

Christopher Emeka Uhegbu, Purdue University

Abstract

Alternate measurement and data-processing methods are evaluated for direct measurement or prediction of equilibrium-based signals from physico-chemical systems. Multipoint data from the transient regions of responses are used with suitable models and curve-fitting methods to predict the signal that would be measured for a system at equilibrium. The resulting equilibrium response usually is much less dependent on experimental variables than the transient responses used to predict it. The approach was evaluated for a membrane-based oxygen-selective electrode and an enzyme-based reactor/sensor device. To reduce the effects of experimental variables that affect steady-state measurement approaches with the oxygen-selective electrodes, a curve-fitting method was used to predict electrode response expected when analyte concentration at the electrode surface inside the membrane is the same as that in the bulk sample solution outside the membrane. Results show that the predictive method has a much lower dependence on membrane thickness than does the steady-state method. The relative error coefficients for membrane thickness between 25 $\mu$m and 50 $\mu$m and an oxygen concentration of 0.26 mmol L$\sp{-1}$ were 38% and 0.3% for the steady-state and predictive methods, respectively. This corresponds to an improvement of 125-fold for the predictive method. Analogous results were attained for the effect of stirring rate. In case of the enzyme-based reactor/sensor device, similar error compensation was achieved by using the device in a thin-layer flow mode such that the enzymatic reactions are the rate-limiting step. The total change when the substrate in the thin-layer is completely reacted was computed by integrating the current vs. time starting from the time, flow to the device was stopped down to the baseline. The total charge which is proportional to the amount of sample was predicted by using the curve fitting method in a suitable model to fit the charge vs. time data. This measurement/data-processing approach gave a linear range that was four-fold higher and pH dependence that was two-fold lower, relative to the conventional steady-state methods.

Degree

Ph.D.

Advisors

Pardue, Purdue University.

Subject Area

Analytical chemistry

COinS