Deconvolution based separation of overlapping terahertz wave for multilayer thin film thickness measurement

This study examines the effectiveness of signal processing techniques in improving the accuracy of multilayer thin film thickness measurement using terahertz (THz) waves. Three signal separation methods such as deconvolution, Hilbert based deconvolution, and Wavelet based deconvolution were applied to extract individual reflection signals from overlapping THz waves and evaluate their performances by estimating the thickness of polyethylene (PE) separator films with thicknesses of 5, 10, and 15 µm, respectively. The findings revealed that both deconvolution and Hilbert based deconvolution achieved the same level of accuracy for PE separator films regardless of thickness, with an average film thickness error of -0.073%. However, Hilbert based deconvolution demonstrated improved performance in air layer thickness estimation, achieving a lower error rate of -1.06% compared to 4.92% for deconvolution despite requiring additional signal processing resources. Conversely, Wavelet based deconvolution failed to separate overlapping signals, preventing accurate time delay extraction and making it unsuitable for multilayer thin film thickness estimation. These results confirm that both deconvolution and Hilbert based deconvolution are effective for PE separator films; however, deconvolution is the most suitable method for THz based thin film measurement owing to its accuracy and efficiency. Hilbert based deconvolution provides additional phase based analysis capabilities and improves air layer thickness estimation but requires more computational resources. The limitations of Wavelet based deconvolution highlight the need for hybrid approaches combining amplitude and phase information for enhanced performance. Future work will focus on experimental validation and exploring the industrial applicability of THz measurement techniques for multilayer film characterization. • Deconvolution and Hilbert-based deconvolution offer similar accuracy for PE films. • Hilbert-based deconvolution improves air layer thickness estimation performance. • Wavelet-based deconvolution is unsuitable for multilayer thin-film thickness. • Hilbert-based deconvolution requires more resources for additional analysis.