North China Electric Power University Researchers Detail Uranyl Sensor

Researchers at North China Electric Power University published a study on 27 July 2026 in Sustainable Carbon Materials detailing a regenerable sensor that concentrates uranyl ions for Raman measurement and automates the identification of their spectral signature.

Building the FA@tPF Sensor Structure

To fabricate the sensing material, Zhenli Sun’s team coated magnetic iron oxide particles with silica and attached gold nanoparticles to their surfaces. They then added a covalent organic polymer layer. This combination forms the material designated as FA@tPF.

The magnetic core allows operators to collect the particles using a magnet. The polymer layer is responsible for capturing uranyl ions from water samples. Meanwhile, the gold nanoparticles amplify Raman signals from any material brought into close contact with their surfaces.

Microscopy and chemical analyses confirmed the layered structure of the particles. Measurements also verified that the finished particles retain their magnetic response after the coating processes.

Sensor Detects Uranyl Solutions Down to 1 × 10-7 Moles per Liter

In laboratory evaluations using standard uranyl solutions, researchers mixed the sensor with water, collected the particles magnetically, and measured the Raman spectrum using a portable instrument. A characteristic signal near 850 inverse centimeters remained detectable down to 1 × 10^-7 moles per liter after a 20-minute exposure.

The signal tracked concentration accurately across a tested range from 1 × 10^-7 to 1 × 10^-4 moles per liter. Uncoated particles lacking the polymer layer produced no identifiable uranyl peak even at the higher concentration of 1 × 10^-4 moles per liter.

A flow-cell experiment designed to test enrichment from moving water achieved the exact same detection limit after a 20-minute interval. Common coexisting ions introduced during interference tests had little effect on the target uranyl signal.

Machine Learning Model Achieves 100% Classification Accuracy

To test regenerability, the team released captured uranyl using a sodium carbonate solution. The characteristic peak disappeared entirely after cleaning and returned once the particles captured uranyl again.

The signal remained detectable through six full cycles, though its intensity declined slightly with each reuse. Following the spectral measurements, the researchers analyzed the data using principal component analysis paired with a convolutional neural network.

The machine learning model achieved 100% classification accuracy on the reported dataset. An interpretation method confirmed that the model’s classification decisions relied chiefly on the uranyl-associated peak located near 850 inverse centimeters.

Next Steps for Environmental Monitoring Platforms

The published findings demonstrate how chemical capture, magnetic enrichment, Raman measurement, and spectral analysis can operate within a single reusable platform. Performance across a wider variety of real environmental samples remains to be established.