Cited by 132Plasmonic and Surface Plasmon ResearchGold and Silver Nanoparticles Synthesis and ApplicationsSpectroscopy Techniques in Biomedical and Chemical Research

The DOI matches its Crossref record; 64 of the 64 references it cites were found in a registry or an index; 9 of the 53 sources its claims cite were checked for support.

Wavelength-multiplexed hook nanoantennas for machine learning enabled mid-infrared spectroscopy

Open access

Abstract

Infrared (IR) plasmonic nanoantennas (PNAs) are powerful tools to identify molecules by the IR fingerprint absorption from plasmon-molecules interaction. However, the sensitivity and bandwidth of PNAs are limited by the small overlap between molecules and sensing hotspots and the sharp plasmonic resonance peaks. In addition to intuitive methods like enhancement of electric field of PNAs and enrichment of molecules on PNAs surfaces, we propose a loss engineering method to optimize damping rate by reducing radiative loss using hook nanoantennas (HNAs). Furthermore, with the spectral multiplexing of the HNAs from gradient dimension, the wavelength-multiplexed HNAs (WMHNAs) serve as ultrasensitive vibrational probes in a continuous ultra-broadband region (wavelengths from 6 μm to 9 μm). Leveraging the multi-dimensional features captured by WMHNA, we develop a machine learning method to extract complementary physical and chemical information from molecules. The proof-of-concept demonstration of molecular recognition from mixed alcohols (methanol, ethanol, and isopropanol) shows 100% identification accuracy from the microfluidic integrated WMHNAs. Our work brings another degree of freedom to optimize PNAs towards small-volume, real-time, label-free molecular recognition from various species in low concentrations for chemical and biological diagnostics.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

The paper introduces wavelength-multiplexed hook nanoantennas (WMHNAs) that use loss engineering and broadband 6–9 μm plasmonic resonance to enhance mid-infrared molecular fingerprint detection. Combined with PCA/SVM machine learning, the platform identifies 1% methanol, ethanol, isopropanol, and their mixtures in water with 100% accuracy.

Key findings

  • Hook nanoantennas tune radiative loss by changing the geometric difference ΔL between arms; optimal sensitivity occurred at ΔL = 0.65 μm for transmission (extracted loss ratio f = 0.644) and ΔL = 1.25 μm for reflection (f = 1.834), close to theoretical optima of f = 0.5 and f = 2, respectively.
  • The 16-element WMHNA provides continuous ultra-broadband operation from 6 μm to 9 μm, matching mid-infrared fingerprint absorption; it captured 15 absorption peaks of methanol, ethanol, isopropanol, and water.
  • WMHNA enhanced water scissoring absorption at 6.0 μm by 34.36 times and stretching absorption at 2.95 μm by 4.17 times; compared with direct thin-film measurement, WMHNA and HNA showed significant enhancement of absorption spectra by 3 orders of magnitude.
  • PCA extracted multi-dimensional features from refractometric, spectroscopic, and antenna loading effects; SVM classification of 1% alcohol solutions and mixtures achieved 100% accuracy on both training and test sets.

Methods

  • Design and simulation: temporal coupled-mode theory (TCMT) modeled plasmon–molecule coupling; finite-difference time-domain (FDTD, Lumerical) simulated far-field spectra and near-field distributions of hook nanoantennas.
  • Fabrication and measurement: gold HNAs were fabricated on CaF₂ by electron-beam lithography and lift-off; FTIR microscopy with an MCT detector measured reflection/transmission spectra, and a PDMS microfluidic chamber enabled liquid sensing.
  • Machine learning: PCA reduced WMHNA spectra to principal components; SVM with an RBF kernel (C = 1.0, gamma = 0.1) classified analytes. The dataset contained 50 spectra per molecule, split into 40 training and 10 testing spectra.

Limitations

  • The paper does not explicitly state limitations.
  • Evident from the text: the molecular recognition demonstration is a proof-of-concept limited to three chemically similar alcohols at 1% concentration in water and their mixtures, with a small dataset of 50 spectra per molecule.
  • Evident from the text: the WMHNA design shown covers 6–9 μm, while the paper notes that covering all IR fingerprint ranges would require further engineering of antenna length for specific molecular targets.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41467-022-31520-z

Cites

64 of 64 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Working mechanisms of liquid molecular identification by wavelength-multiplexed hook nanoantenna array (WMHNA). a Schematic drawing of WMHNA on CaF2 platform with PDMS microfluidic chamber. Inset SEM image of one unit cell of WMHNA. b The far-field spectra of WMHNA with and without water measured in the reflection mode of the FTIR microscope. c The reference IR absorption spectra of alcoholic liquid (methanol, ethanol, and isopropanol) in the IR fingerprint region match with the broadband response of WMHNA. d Illustration of multi-dimensional information in WMHNA system. (i) Refractometric effect (RE): wavelength shift caused by the refractive index of analytes; (ii) Spectroscopic effect (SE): intensity drop due to the absorption of analytes; (iii) Antenna loading effect (ALE): the peak difference brought by the wavelength mismatch between analytes vibration and antenna resonance. e Machine learning process to extract multi-dimensional sensing information for recognition of 1% alcohols and their mixtures in water.
Fig. 1 Working mechanisms of liquid molecular identification by wavelength-multiplexed hook nanoantenna array (WMHNA). a Schematic drawing of WMHNA on CaF2 platform with PDMS microfluidic chamber. Inset SEM image of one unit cell of WMHNA. b The far-field spectra of WMHNA with and without water measured in the reflection mode of the FTIR microscope. c The reference IR absorption spectra of alcoholic liquid (methanol, ethanol, and isopropanol) in the IR fingerprint region match with the broadband response of WMHNA. d Illustration of multi-dimensional information in WMHNA system. (i) Refractometric effect (RE): wavelength shift caused by the refractive index of analytes; (ii) Spectroscopic effect (SE): intensity drop due to the absorption of analytes; (iii) Antenna loading effect (ALE): the peak difference brought by the wavelength mismatch between analytes vibration and antenna resonance. e Machine learning process to extract multi-dimensional sensing information for recognition of 1% alcohols and their mixtures in water.CC BY · PubMed Central
Fig. 2 Design principle of hook nanoantenna. a The derived intensity changes in transmission and reflection mode with the perfect match of molecular absorption and antenna resonance by TCMT. b The summary of methods affecting the coupling strength between molecules and antenna by previous works. c The proposed hook antenna for sensing optimization by tuning the radiative loss and the effect of loss ratio to intensity change in transmission and reflection mode. d The simulated far-field spectra in transmission (i), reflection (ii), and absorption(iii) of hook antenna by changing the ΔL, indicating the tuning of radiative loss. e The electric field enhancement distribution of hook nanoantenna devices at resonance wavelength with different ΔL of 1.6 μm (i), 0.8 μm (ii) and 0 μm (iii). (f) The extracted maximum electric field enhancement of hook nanoantenna devices with different ΔL.
Fig. 2 Design principle of hook nanoantenna. a The derived intensity changes in transmission and reflection mode with the perfect match of molecular absorption and antenna resonance by TCMT. b The summary of methods affecting the coupling strength between molecules and antenna by previous works. c The proposed hook antenna for sensing optimization by tuning the radiative loss and the effect of loss ratio to intensity change in transmission and reflection mode. d The simulated far-field spectra in transmission (i), reflection (ii), and absorption(iii) of hook antenna by changing the ΔL, indicating the tuning of radiative loss. e The electric field enhancement distribution of hook nanoantenna devices at resonance wavelength with different ΔL of 1.6 μm (i), 0.8 μm (ii) and 0 μm (iii). (f) The extracted maximum electric field enhancement of hook nanoantenna devices with different ΔL.CC BY · PubMed Central
Fig. 3 Experiment characterization of hook nanoantennas sensing performance. a, b The testing transmission (a) and reflection (b) spectra of hook nanoantenna devices with different ΔL for PMMA thin film sensing. The carbonyl stretching vibration in 5.76 μm or 1736 cm−1 is used to characterize the sensitivity of hook nanoantennas. The colors of the line refer to the different ΔL in e, f. c The extracted absorptive and radiative loss of hook nanoantennas by fitting the experimental spectra with TCMT. d, e The baselined corrected spectra difference in transmission(d) and reflection(e). f The calculated sensitivity of hook nanoantenna devices with different ΔL. The sphere symbols refer to reflection mode and square symbols represent transmission mode.
Fig. 3 Experiment characterization of hook nanoantennas sensing performance. a, b The testing transmission (a) and reflection (b) spectra of hook nanoantenna devices with different ΔL for PMMA thin film sensing. The carbonyl stretching vibration in 5.76 μm or 1736 cm−1 is used to characterize the sensitivity of hook nanoantennas. The colors of the line refer to the different ΔL in e, f. c The extracted absorptive and radiative loss of hook nanoantennas by fitting the experimental spectra with TCMT. d, e The baselined corrected spectra difference in transmission(d) and reflection(e). f The calculated sensitivity of hook nanoantenna devices with different ΔL. The sphere symbols refer to reflection mode and square symbols represent transmission mode.CC BY · PubMed Central
Fig. 4 Experiment characterization of WMHNA to decrease loading effect with broadband absorption enhancement. a The testing reflection spectra of hook nanoantenna with PMMA thin film at different L. b The baseline-corrected reflection spectra from (a). c The normalized sensitivity of hook nanoantenna devices with different L indicates the loading effect. Sensitivity becomes maximum when the two wavelengths of molecular vibration and antenna resonance are matched. d the reflection spectra of WHHNA and hook nanoantenna devices (P1, P8, and P16) to form the supercell. e The sensing characterization of WMHNA by two types of thin films of PMMA and silk. f Baseline-corrected sensing signal of WMHNA compared with device P1, P8, and P16 by two analytes of PMMA and silk.
Fig. 4 Experiment characterization of WMHNA to decrease loading effect with broadband absorption enhancement. a The testing reflection spectra of hook nanoantenna with PMMA thin film at different L. b The baseline-corrected reflection spectra from (a). c The normalized sensitivity of hook nanoantenna devices with different L indicates the loading effect. Sensitivity becomes maximum when the two wavelengths of molecular vibration and antenna resonance are matched. d the reflection spectra of WHHNA and hook nanoantenna devices (P1, P8, and P16) to form the supercell. e The sensing characterization of WMHNA by two types of thin films of PMMA and silk. f Baseline-corrected sensing signal of WMHNA compared with device P1, P8, and P16 by two analytes of PMMA and silk.CC BY · PubMed Central
Fig. 5 Demonstration of broadband liquid dynamics monitoring of acetone and water using WMHNA integrated with microfluidics. a The schematic drawing of WMHNA chip integrated with PDMS microfluidic chamber. b The dynamic monitoring of broadband spectra of WMHNA with acetone injection to water. c The extracted absorbance spectra at different time. d The comparison of spectra of WMHNA with and without water. e The integrated absorbance at each absorption peak indicates the dynamic change of analytes.
Fig. 5 Demonstration of broadband liquid dynamics monitoring of acetone and water using WMHNA integrated with microfluidics. a The schematic drawing of WMHNA chip integrated with PDMS microfluidic chamber. b The dynamic monitoring of broadband spectra of WMHNA with acetone injection to water. c The extracted absorbance spectra at different time. d The comparison of spectra of WMHNA with and without water. e The integrated absorbance at each absorption peak indicates the dynamic change of analytes.CC BY · PubMed Central
Fig. 6 Demonstration of molecular identification of low concentration alcohols with multi-dimensional sensing signal from WMHNA by machine learning. a The broadband spectra of WMHNA under different analytes. All alcohol solvents are diluted to 1% in DI water. b The extracted absorption spectra of different alcohol solvents from WMHNA. c The corresponding second-order derivative of absorption spectra in (b). d The reflection spectra of sensing data with different analytes states. e The machine learning processed spectra of WMHNA after dimension reduction by principal component analysis. the first principal component (PC 1) represents the antenna loading effect of water absorption peaks at 6.0 μm. The second principal component (PC 2) represents the wavelength shift of WMHNA due to the refractive index of the analyte. The third principal component (PC 3) represents the fingerprint absorption of molecules. f The weight of scores of each spectrum in three-dimensional space after PCA for WMHNA. Each cluster indicates one type of molecule and its mixtures. g The confusion map for machine learning outcome indicates the 100% accuracy of molecular identification.
Fig. 6 Demonstration of molecular identification of low concentration alcohols with multi-dimensional sensing signal from WMHNA by machine learning. a The broadband spectra of WMHNA under different analytes. All alcohol solvents are diluted to 1% in DI water. b The extracted absorption spectra of different alcohol solvents from WMHNA. c The corresponding second-order derivative of absorption spectra in (b). d The reflection spectra of sensing data with different analytes states. e The machine learning processed spectra of WMHNA after dimension reduction by principal component analysis. the first principal component (PC 1) represents the antenna loading effect of water absorption peaks at 6.0 μm. The second principal component (PC 2) represents the wavelength shift of WMHNA due to the refractive index of the analyte. The third principal component (PC 3) represents the fingerprint absorption of molecules. f The weight of scores of each spectrum in three-dimensional space after PCA for WMHNA. Each cluster indicates one type of molecule and its mixtures. g The confusion map for machine learning outcome indicates the 100% accuracy of molecular identification.CC BY · PubMed Central

Verification

Claims the paper makes

42 Own result · 21 Own interpretation · 19 No basis given · 12 Not checked yet · 1 Partly supported · 1 Mixed support · accuracy not yet measured

Verify all 44 references — free account
  • Probably not a substantial claim. 10
  • The quote only partly states this claim. 2
  • It may state two separate claims as one. 29
40 of 40
  • The authors state this as an assumption.

    There is a need for new nanoantenna structures for continuous broadband detection and new methods to distinguish chemically similar molecules from partially overlapped absorption spectra.

    “Therefore, new nanoantenna structures need to be developed for continuous broadband detection to capture multiple absorption peaks in fingerprint regions and new methods need to be conducted to distinguish the chemically similar molecules from the partially overlapped absorption spectra.”
  • The paper relies on this without support.

    Molecules of the same class have strong overlap in fingerprint absorption due to similar chemical structure.

    “However, since the same classes of molecules always have a strong overlap in fingerprint absorption due to the similar chemical structure, it is challenging to distinguish the same classes of molecules in small concentrations by monitoring single absorption peak of each molecules.”
  • The paper relies on this without support.

    In the IR fingerprint region, plasmonic nanoantenna signals of absorption changes and wavelength shifts always occur together.

    “Unfortunately, in the IR fingerprint region, the PNA signal of absorption changes and wavelength shifts always comes together, hindering the usage of complementary information about the analyte’s physical and chemical properties for molecule identification.”
  • The authors state this as an assumption.

    The temporal coupled-mode theory model accurately describes the plasmon-molecule coupling system.

    “To understand the design principles, we use the TCMT model52 to analyze the plasmon-molecules coupling system.”
  • The authors state this as an assumption.

    The Lorentz model accurately represents the optical properties of analytes.

    “To demonstrate the individual response caused by three effects, we use finite-difference time-domain (FDTD) methods to simulate the antenna response and manipulate the optical properties (refractive index, n and absorption coefficient, k) of analytes by Lorentz model, which has been used in our previous works19,51”
  • The authors state this as an assumption.

    In the TCMT model, the plasmonic resonance is treated as a bright mode coupled to incident light, and molecular vibration is treated as a dark mode whose coupling efficiency is much lower than the plasmonic nanoantenna and can be ignored.

    “We treat the plasmonic resonant (denoted as P) as a bright mode that is coupled to the incident light, while we treat the molecular vibration (denoted as M) as a dark mode, in which coupling efficiency is much lower than PNA and can be ignored in their coupling system.”
6 of 6

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

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.

4 of 4

Citations

Cites

Show all 64 references

Cited by

Sorted by citations

See all 132

Metadata

Authors
Ren, Z., Zhang, Z., Wei, J., Dong, B., & Lee, C.
Affiliations
National University of Singapore (SG)
Corresponding author
Lee, Chengkuo
Venue
Nature Communications
Published
2022-07-05
Volume
13
Issue
1
Pages
3859-3859
ISSN
2041-1723 (ISSN-L)
OpenAlex
W4283816337
Cited by
132
References
64
Citations by year
2026: 19 · 2025: 43 · 2024: 36 · 2023: 26 · 2022: 8
FWCI
8.20
Type
journal article
Language
en
Open access
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • National University of Singapore (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Research Foundation Singapore (NRF-CRP15-2015-02)
  • Agency for Science, Technology and Research (A18A5b0056)
  • Ministry of Education - Singapore (A-0005138-00-00)
  • National University of Singapore (A-0005947-06-00)
Topics
  • Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Plasmonic and Surface Plasmon Research
  • Physical Sciences › Materials Science › Electronic, Optical and Magnetic Materials › Gold and Silver Nanoparticles Synthesis and Applications
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Biophysics › Spectroscopy Techniques in Biomedical and Chemical Research
Keywords
mid-infrared plasmonics, plasmonic nanoantennas, loss engineering, spectral multiplexing, vibrational spectroscopy, alcohol identification, microfluidic integration
MeSH terms
Machine Learning, Spectrophotometry, Infrared, Peptide Nucleic Acids, Surface Plasmon Resonance, Nanostructures

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 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

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