Automated Processing of Nano-Impact Electrochemistry Signals Using Data-Driven Template Matching
Ziwen Zhao,Arunava Naha,Sagar Ganguli,Alina Sekretareva
DOI: https://doi.org/10.1149/ma2023-02572755mtgabs
2023-12-22
ECS Meeting Abstracts
Abstract:Nano-impact electrochemistry (NIE) is an advanced electrochemical technique that has gained significant attention in recent years due to its ability to investigate properties of individual entities, such as nanoparticles, cells, or even single molecules 1 . A typical NIE experiment involves investigating entities freely moving in solution and colliding with an ultramicroelectrode held at a specific potential, allowing individual entities to be electrochemically detected during the collision. The typical NIE responses are the staircase and blip types (spikes), 2 with the latter being more commonly observed in NIE experiments. Analysis of spike signals, however, is more complicated due to their irregular and asymmetric shapes. To analyze NIE signals, most studies rely on peak find functions using height threshold for spike identification 3 coupled to peak area calculation based on the assumption that the spike shape follows certain distribution functions 4 . This approach only partially captures the information present in the NIE signal and may result in an inaccurate analysis as demonstrated in Scheme 1, left. To address this issue, we propose a new algorithm, which leverages unsupervised machine learning and image template-matching technology. The algorithm consists of several steps. First, the raw data are denoised. This is done by applying the time-frequency analysis to the data to identify the spike frequency and remove higher-frequency noise using a bandpass filter. Background subtraction is then conducted to align the denoised curve to an approximate zero baseline. Further, some of the current spikes are sampled using a conventional height threshold method to generate templates. A set of templates is generated by clustering the corresponding feature parameters. Template matching is carried out by computing the normalized cross-correlation coefficient of each template on the offset signal. Finally, once the two end points of the spikes have been well-defined parameters characterizing the shape of the current spikes are extracted. Optionally, a new set of templates can be generated for the rematch with the data. The developed algorithm permit precisely detecting the two sides of the spike and extracting critical shape-related parameters, leading to a more accurate and detailed analysis of the NIE signals (Scheme 1, right). Furthermore, by generating templates that reflect the distinguishing features of spikes, observed current signals can be classified into different types, revealing underlying physical phenomena. This method can aid in the standardization of data processing and NIE signal interpretation across various experiments and applications. It can also be used as a pre-processing step when training classification neural networks, which can automate the spike identification process and develop NIE models. As a result, the proposed algorithm has the potential to provide valuable insights into the properties of single entities and advance our understanding of electrochemical reactions at the nanoscale. Reference 1. Baker LA. Perspective and Prospectus on Single-Entity Electrochemistry. J Am Chem Soc . 2018;140(46):15549-15559. doi:10.1021/jacs.8b09747 2. Kwon SJ, Zhou H, Fan FRF, Vorobyev V, Zhang B, Bard AJ. Stochastic electrochemistry with electrocatalytic nanoparticles at inert ultramicroelectrodes—theory and experiments. Phys Chem Chem Phys . 2011;13(12):5394-5402. doi:10.1039/C0CP02543G 3. Little CA, Xie R, Batchelor-McAuley C, et al. A quantitative methodology for the study of particle–electrode impacts. Phys Chem Chem Phys . 2018;20(19):13537-13546. doi:10.1039/C8CP01561A 4. Peak Finding and Measurement. https://terpconnect.umd.edu/~toh/spectrum/PeakFindingandMeasurement.htm Figure 1