Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks
Open accessThe DOI matches its Crossref record; 4 of the 5 references it cites were found in a registry or an index; none of the 5 sources its claims cite is checked for support yet.
Abstract
In this work, we present a novel approach to determine the refractive index n and the extinction coefficient κ from terahertz time-domain spectroscopy (THz-TDS) measurements using artificial neural networks (ANNs). We train the network in our experimental datasets to create a model that can accurately and efficiently calculate these material properties, the proposed neural network is superior to traditional analytical methods, which rely on approximations, and a faster and easier alternative to iterative root-finding algorithms. Our results demonstrate that this machine learning methodology solves common issues in THz-TDS data analysis, such as phase unwrapping, time-domain windowing, low computation rates, and high accuracy at low-frequency regions, effectively.
Overview
CoChat Summary
TL;DR
The paper proposes using artificial neural networks (ANNs) to extract refractive index \(n\) and extinction coefficient \(\kappa\) from terahertz time-domain spectroscopy (THz-TDS) measurements. It reports that the ANN results are close to initial approximate values and outperform the Drude-Lorentz method for a glucose sample, while avoiding phase unwrapping, time-domain windowing, and low-frequency issues.
Key findings
- The ANN is claimed to improve precision and computational cost compared with traditional analytical methods.
- For a glucose sample, the extracted refractive index and extinction coefficient from the ANN are “very close” to the initial estimated values \(n_a\) and \(\kappa_a\), whereas the Drude-Lorentz model results “are not in agreement despite their same behaviour.”
- The method is said to solve common THz-TDS analysis issues including phase unwrapping, time-domain windowing, low computation rates, and accuracy at low-frequency regions.
- No quantitative error metrics, numerical values, or statistical comparisons are reported in the text.
Methods
- The paper trains an ANN to calculate \(n\) and \(\kappa\) from THz-TDS data; the abstract and conclusion say training used experimental datasets, while the introduction says training used “large databases of simulated and theoretical interactions” and testing used experimental data.
- Initial approximations \(n_a\) and \(\kappa_a\) are calculated from the experimental data through the amplitude and phase of the transfer function, using the stated transfer-function formulation.
- Results compare ANN-extracted values with the standard Drude-Lorentz method and the initial approximate values for a glucose sample.
Limitations
- The paper states no explicit limitations.
- Evident from the text: no network architecture, training-set size, or training procedure details are given; no quantitative accuracy or computational-cost metrics are reported; only one material (glucose) is shown; and the description of training data is inconsistent between sections.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures
Verification
Claims the paper makes
0 Valid · 1 Not valid · 4 Not checked · 8 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We note that the results given by the ANN are very close to the initial estimated values”
§ Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks > 3 Results
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“where the those of the Drude-Lorentz model are not in agreement despite their same behaviour.”
§ Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks > 3 Results
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“The model surpassed common issues in materials characterization like phase unwrapping, time-domain windowing, or low SNR.”
§ Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks > 3 Results
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: unclear — It is unclear why the paper says it holds.
- Rests on an assumption not shown here.
None of the sources cited for this claim have been checked yet.
In the paper
“Neural networks (ANNs), process THz-TDS data and extract the material parameters n and κ bypassing these issues [3,4].”
§ Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks > 1 Introduction
What the sources say
- Not checked yetcites [3] Li Y., Shen X. A., Ewing R. L., and Li J., “Terahertz spectroscopic material identification using approximate entropy and deep neural network,” in Proc. IEEE Nat. Aerosp. Electron. Conf. (NAECON), Dayton, OH, USA, 2017, pp. 52–56, doi: 10.1109/NAECON.2017.8268744. source
- Not checked yetcites [4] Koch, 2023, Nat. Rev. Methods Primers source
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Thus, ANN is a powerful tool for maximizing the speed and accuracy of terahertz domain material characterization”
§ Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks > 3 Results
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
3 more claims
Create a free account to see all 13 claims and to verify the 5 references still open.
- It may state two separate claims as one. 4
- Probably not a substantial claim. 1
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks > 1 Introduction
“In this work we trained an artificial neural network using large databases of simulated and theoretical interactions and tested with experimental data.”
§ Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks > 4 Conclusion
“We used experimental data to train the neural network and developed a model significantly better than analytical and iterative methods in accuracy and computational efficiency.”
Citations
Cites
A reliable method for extraction of material parameters in terahertz time-domain spectroscopy
Duvillaret, L., Garet, F., & Coutaz, J.-L. (1996) · IEEE Journal of Selected Topics in Quantum Electronics · doi:10.1109/2944.571775 · Cited by 994
Terahertz time-domain spectroscopy
Koch, M., Mittleman, D. M., Ornik, J., & Castro-Camus, E. (2023) · Nature Reviews Methods Primers · doi:10.1038/s43586-023-00232-z · Cited by 410
Artificial neural networks for material parameter extraction in terahertz time-domain spectroscopy
Klokkou, N., Gorecki, J., Wilkinson, J. S., & Apostolopoulos, V. (2022) · Optics Express · doi:10.1364/oe.454756 · Cited by 34
Terahertz spectroscopic material identification using approximate entropy and deep neural network
Li, Y., Shen, X. A., Ewing, R. L., & Li, J. (2017) · 2017 IEEE National Aerospace and Electronics Conference (NAECON) · doi:10.1109/naecon.2017.8268744 · Cited by 14
Metadata
- Authors
- Kaouther, C. A., & Lazoul, M.
- Affiliations
- Polytechnic School of Algiers (DZ)
- Venue
- EPJ Web of Conferences
- Publisher
- EDP Sciences
- Published
- 2025
- Volume
- 335
- Pages
- 03046
- ISSN
- 2100-014X (ISSN-L)
- OpenAlex
- W4414393896
- Semantic Scholar
- 38e68fdb68e81244557429a46bb6cf47dd5e981a
- Cited by
- 0
- References
- 5
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- diamond
- doi.org
- Locations
- EPJ Web of Conferences (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Terahertz technology and applications
- Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Microwave and Dielectric Measurement Techniques
- Keywords
- terahertz time-domain spectroscopy, refractive index, extinction coefficient, material parameter identification, artificial neural networks, phase unwrapping, time-domain windowing, low-frequency band, machine learning
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-09-28, under version 3 of the claim definition (prompt extract-v3).
A second model labelled each claim on 2026-09-28 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks. https://app.cochat.ai/papers/10.1051/epjconf/202533503046
- The paper
- Kaouther, C. A., & Lazoul, M. (2025). Terahertz Time-Domain Spectroscopy Materials Parameters Extraction Using Neural Networks. EPJ Web of Conferences. https://doi.org/10.1051/epjconf/202533503046
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.
