Multiparametric water quality sensor based on carbon nanotubes: Performance assessment in realistic environment

Good quality water is crucial to most developing nations' sustainability. However, there is a clear lack of affordable and reliable solutions to monitor water quality. According to the WHO 2022 Sustainable Development Goals report, about 3 billion people do not have information on their water quality. While off-line measurements are commonly practiced, the availability of in-situ monitoring solutions is considered critical to the generalization of water monitoring, but current technologies are bulky, expensive and usually do not target  a sufficient number of quality parameters. [1]

To meet this challenge, the LOTUS project (https://www.lotus-india.eu/) brings forward a low-cost, compact, versatile multiparametric chemical sensor aiming at real-time monitoring of chlorine, pH, temperature and conductivity in potable water. The proposed solution –a tube of 21.2 cm in length by 3.5 cm in diameter – is composed of a replaceable sensor head incorporating the sensing elements and a sensor body containing the acquisition and communication electronics. The sensor head integrates a 1cm² silicon chip with 2 temperature sensors (serpentine-shaped thermistors), 3 conductivity sensors (parallel electrodes in a 4-probe configuration) and a 10x2 sensor array of multi-walled carbon nanotube (CNT) chemistors. The CNT are arranged in random networks between interdigitated electrodes and are either non-functionalized or functionalized with a dedicated polymer. [1]

We evaluated the performance of 7 units of this solution in Sense-city facility (located at University Gustave Eiffel, France - https://sense-city.ifsttar.fr/ ),  exploiting its 44m potable water loop with 93.8-mm PVC pipes. The system was operated at 25 m³/h and 1 bar, at temperature ranging between 15°C and 20°C, conductivity between 870 µS/cm and 1270 µS/cm; and chlorine between 0 and 5 mg/L. Because of the high-level of electromagnetic interferences in Sense-City and limited shielding of the acquisition system, the sensor signal is severely noisy and various steps of denoising are required. From the initial dataset were extracted a small number of devices and time periods with both sufficient variations in the target parameters and manageable level of signal-over-noise ratio. 

For chip 141, over 150hours of testing, CNT-based chemistors showed sensitivity to pH and active chlorine (HClO) with differentiated response between functionalized and non-functionalized devices. However, pH and chlorine can only be estimated with MAE respectively 0.17 and 0.18mg/L due to the high noise level. Over 400h, with chip 141, the real-time temperature of the water can be estimated with an MAE of 0.4°C in flowing water and 0.1°C  in static water. The chip 141 dataset did not feature enough conductivity variation to assess performances. This was achieved on chip AS001 with an MAE of 176.2 µS/cm over 80 hours.

Overall, these results provide a preliminary proof of operation of the solution in realistic environment, with the high noise level being a major limitation. A new version of system is being designed to reduce the noise, to be tested in Sense-City in 2023.

[1] Cousin, P. et al. (2022). Improving Water Quality and Security with Advanced Sensors and Indirect Water Sensing Methods. Springer Water. https://doi.org/10.1007/978-3-031-08262-7_11