Publication Detail

In-Situ Quantification of Major Ions in Water via a Novel Electronic Tongue Architecture

A. V. Mueller, Harold F. Hemond
2012
1 pp.
MITSG 12-53
$5.50 DOM / $7.50 INT
ORDER HARDCOPY / DOWNLOAD

A novel in situ instrument is developed using a commercial sensor hardware (primarily ion-selective electrodes (ISEs)) paired with a novel artificial neural network (ANN) processor designed to take advantage of a priori chemical knowledge about the system. A method is demonstrated for the training and use of the novel ANN architecture to process raw ISE outputs for the simultaneous quantification of all major ionic concentrations in fresh waters, including Na+, K+, Ca2+, Mg2+, NH4+, Cl-, NO3-, H+, OH-, and SO42-.
Custom-built interface electronics allow simultaneous measurement of high impedance signals from a suite of ISEs using a single reference electrode, and signals are digitized in real time. Implementation of a standardized protocol for determination of equilibrium electrode potential [1] allows asynchronous identification of steady state potential on each channel (each with different response times) without the need for a trained technician to oversee measurements.
A suite of 14 sensors (11 solid state or glass ion selective electrodes, pH, temperature, and electrical conductivity) was used to characterize 75 environmentally-representative (i.e., of northeastern United States waters) lab-created water samples and 65 single-salt standards; resulting signals were used to train a range of ANN architectures for prediction of analyte concentrations. Inclusion of mathematical constraints based on conductivity and charge balance relationships into a novel extended ANN architecture further improved estimation capabilities. Final results demonstrate over an order of magnitude decrease in relative error as compared with use of ISEs as stand-alone sensors: concentration-dependent bias was strongly reduced, and useful estimates were made even for analytes for which no specific ISE exists (SO42-, Mg2+, HCO3-). Simultaneous un-biased quantification of all eight target ions is achieved with <20\% error on most channels including NO3- ([NO3-] <100 uM) and ~50\% un-biased error for NH4+ ([NH4+] <100 uM), however it is also demonstrated that unbiased errors of ~10\% are achievable for N-species ions even at low concentration (~1-10uM) if slightly higher uncertainties on other channels can be tolerated [2].
References:
[1] Mueller, A.V. and H.F. Hemond, 2011. Towards an automated, standardized protocol for determination of equilibrium potential of ion-selective electrodes. Analytica Chimica Acta 590: 71-78.
[2] Mueller, A.V. 2012. PhD thesis: “Development of a Combined Multi-Sensor/Signal Processing Architecture for Improved In-Situ Quantification of the Charge Balance of Natural Waters”, MIT, Cambridge, MA.

type: Presentations

Parent Project

Project No.: 2011-R/RCM-33
Title: Combating nitrogen-driven coastal eutrophication: a selective ion array approach to rapid in-situ measurement of nitrate and ammonium

Other publications by this author (as Lead)