Enhancing terahertz molecular fingerprint detection by a

Letter Vol. 45, No. 8 / 15 April 2020 /Optics Letters 2335

Enhancing terahertz molecular fingerprintdetection by a dielectric metagratingJinfeng Zhu,1,* Shan Jiang,1 Yinong Xie,1 Fajun Li,1 Lianghui Du,2 Kun Meng,2

Liguo Zhu,2 AND Jun Zhou3

1Institute of Electromagnetics and Acoustics, XiamenUniversity, Xiamen 361005, China2Microsystem and Terahertz Research Center, China Academy of Engineering Physics, Mianyang, Sichuan 621900, China3Terahertz Research Center, University of Electronic Science and Technology of China, Chengdu 610054, China*Corresponding author: nanoantenna@hotmail.com

Received 24 January 2020; revised 14 March 2020; accepted 16 March 2020; posted 19 March 2020 (Doc. ID 389045);published 9 April 2020

Terahertz (THz) sensing of molecular fingerprint enableswide applications in biomedicine and security detection.Conventional detection approaches face big barriers intrace analysis of analyte due to the difficulties of enhancingthe broadband molecular absorption. In order to achievestrong broadband wave–matter interaction for the analyte,we propose a method based on THz wave angular scanningon a dielectric metagrating. In virtue of the guided-moderesonance, one can strengthen the local electric field in vari-ous trace-amount analytes by tuning the polarization andincident angle, which leads to significant enhancement onthe broadband signal of molecular fingerprint. The studypaves the way for more applications of THz trace-amountdetection. ©2020Optical Society of America

https://doi.org/10.1364/OL.389045

Terahertz (THz) spectroscopy, defined in the electromagneticwave frequency interval from 0.1 to 10 THz, enables manynew applications and holds great promise for molecular finger-print detection in biological and security sensing [1–5]. This isbecause THz detection is nondestructive, and many complexmolecules have intra-/intermolecule rotational or vibrationalmodes in THz frequencies, which can be characterized by pro-nounced features in the absorption spectra [6]. The moleculesof analytes to be detected usually have extremely small sizescompared with THz wavelengths. This leads to very weak inter-action between the molecules and THz wave, so that the analytesare typically required to have large volumes with observable THzabsorption for molecular identification. However, in the detec-tion for trace-amount molecules, the wavelength-analyte sizemismatch and a lack of powerful sources hinder wide applica-tions of THz technology toward molecular fingerprint detection[7]. In order to solve this problem, many efforts have been madeon using THz antenna arrays, spoof surface plasmons, andmetamaterials in virtue of the subwavelength field confinement[8–10]. These methods not only suffer from large intrinsicdamping of metal, but also have the limitation of narrowbandsensing enhancement, which is not suitable for the broadbandfeature detection of molecular fingerprint absorption spectra.

Some investigations on dynamically tunable detectors have beenconducted to capture the broadband THz absorption features[11–13], but the enhancement of detection is very limited, andthe sensing configurations require the control of temperature,bias voltage, or static magnetic field, which is complicated withadditional noise in practice. Therefore, a flexible THz detectionmethod with the prominent enhancement of broadband signalis still quite in demand for molecular fingerprint sensing.

In this work, we design the dielectric metagrating withpolarization and angle manipulation of an incident THz wavefor the enhancement of molecular fingerprint detection. Ourscheme immensely eliminates the inherent optical absorptionthat would interfere the sensing performance. The combinedmultiplexed signals by dynamic angle scanning reflect the sig-nificant enhancement on broadband detection of molecularfingerprint for identifying the trace-amount analyte of only1 µm thick. The physical mechanism of enhancing detection isdemonstrated systematically, and the high sensing performancefor three analytes is illuminated by electromagnetic full wavesimulations.

The proposed dielectric metagrating is illustrated in Fig. 1,which consists of an array of periodic subwavelength poly-thene (PE) gratings on a polymethylpentene (TPX) layer witha substrate. In the THz region, PE and TPX are assumed to benonmagnetic and lossless. The refractive index of PE and TPXare 1.54 and 1.46, respectively [14,15]. The simulation methodbased on the finite element method is used to investigate thesensing mechanism and optimize the sensing performance. Inthe simulation, all the materials are assumed to be isotropic, theunit cell of a subwavelength grating is adopted, and the Floquetboundary conditions are applied. In the proposed structure, thethickness of the analyte is extremely small compared with theTHz wavelengths. In view of this, the s -polarized incident waveis more advantageous than the p-polarized incident wave forenhancing wave–matter interaction when changing the incidentangle, according to our previous work [16–18]. This is becausethe electric field is always parallel to the ultrathin analyte for allthe incident angles at s polarization. Therefore, we mainly focuson the use of the s -polarized THz wave for the next discussion.

0146-9592/20/082335-04 Journal © 2020Optical Society of America

2336 Vol. 45, No. 8 / 15 April 2020 /Optics Letters Letter

Fig. 1. Schematic drawing of metagrating. The symbols p , w, t1,t2, and θ represent the period, the width of PE grating, the thickness ofPE, the thickness of TPX, and the incident angle, respectively.

In order to reveal the physics from a simplified view, we firststudy the angle-dependent THz response of a freestandingdielectric metagrating without a substrate. The reflectancespectrum as a function of the incident angle for a freestandingmetagrating is plotted in Fig. 2(a), which indicates that there isa narrowband unity reflectance corresponding to each incidentangle and the central frequency is monotonically decreasing asthe angle increases. Such a THz response can be explained bythe guided-mode resonance theory [19], in which an incidentTHz wave passing through the metagrating at a specific anglewill form a waveguide mode at the corresponding frequency. Inthis mode, the THz wave can be trapped and guided within themetagrating. In order to couple the incident wave to the guidedmode, the phase-matching condition must be met, which can bedescribed by the following equation:

β = k0

(n0 sin θ −

cfr p

), (1)

where β, k0, n0, c , and fr denote the propagation constant offirst-order guided mode, the free-space wavenumber, the refrac-tive index in the air, the vacuum light speed, and the resonancefrequency corresponding to the incident angle θ , respectively.The theoretical calculation based on Eq. (1) demonstratesa good consistency with the simulation result, as shown inFig. 2(b). When the angular scanning of incidence is adopted,each guided-mode resonance at a particular incident anglecould provide a better opportunity to enhance the wave–matterinteraction around the resonant frequency for the analyte to becoated on the metagrating. In addition, it also enables the cov-erage for a wide range of resonance frequencies, which is criticalfor broadband signal enhancement. The angular scanning ofthe incident THz wave establishes the important basis for theenhanced broadband fingerprint sensing. The use of the guidedmode under the angular scanning gives a better solution toenhance the interaction between the THz wave and the analyte,but a dedicated structure design of the metagrating is still quiteessential to enhance the THz absorption of the analyte for moreeffective sensing.

We next focus on the effects of coating the trace-amount ana-lyte, and further reveal the mechanism of sensing enhancement.The uniformly coated material α-lactose monohydrate of 1 µmthick is used for this next investigation. We first study the fre-quency range around 0.53 THz, where α-lactose monohydratehas the largest optical extinction coefficient. We compare theguided-mode effects for a metagrating without and with a metalsubstrate. As shown in Fig. 3(a), the freestanding metagrating

Fig. 2. (a) Reflectance spectra of a freestanding metagrating with-out the analyte for various θ , where p = 400 µm, w= 180 µm,t1 = 85 µm, and t2 = 100 µm. (b) The resonance frequency as afunction of θ .

has a maximum absorbance of 28.9% in the analyte, while themetagrating with a metal substrate has a much higher maxi-mum absorbance of 56.3%. This can be explained by the fielddistributions shown in the gray dashed box of Fig. 3(b). Themagnetic field distributions for both the metagrating withoutand with a metal substrate clearly reflect the features of theguided-mode resonance. The use of a metal substrate inducesmuch stronger magnetic field enhancement inside the PE grat-ing, and leads to significant enhancement for the electric fieldparallel to the α-lactose monohydrate layer. The frequency-dependent THz absorption of the analyte can be described bythe following equation:

A( f )= 4π f nk∫

V|E |2dV , (2)

where n and k represent the real and imaginary parts of refractiveindex for the analyte, respectively. V is the volume of the ana-lyte. The THz absorption in the analyte is mainly determinedby the electric loss but not magnetic loss. In the guided-moderesonance, the magnetic field is mainly concentrated in themetagrating, but the electric field is mainly enhanced sur-rounding the surface of the metagrating. Due to the higherenhancement of the electric field, the frequency-dependentabsorption of the α-lactose monohydrate layer in the meta-grating with a metal substrate is elevated more considerablythan that in the freestanding metagrating, according to Eq. (2).Therefore, we use the structure with a metal substrate for thefollowing discussion.

Moreover, we study the absorbance spectrum as a functionof the incident angle of the THz wave. The change of incidentangle from 5◦ to 29◦ could excite a series of guided-mode reso-nances with the resonance frequency reducing gradually. Theseresonances induce the absorbance spectra with various peakvalues associated with the dispersive property of the α-lactosemonohydrate in the THz range [see the inset of Fig. 3(a)]. Forinstance, in the absorption band of α-lactose monohydratefrom 0.46 THz to 0.60 THz [shown in Fig. 3(a)], the incident

Letter Vol. 45, No. 8 / 15 April 2020 /Optics Letters 2337

Fig. 3. (a) Absorbance spectra of α-lactose monohydrate on themetagrating for various θ . (b) Field distributions. The ones in the graydashed line box are for the on-resonance metagratings. The ones in thepink dashed line box are for the metagrating with a metal substrate.

angle 26◦ leads to the highest absorption at f = 0.53 THz, theincident angles 23◦ and 29◦ generate relatively lower absorptionsurrounding the frequency 0.53 THz, and the use of θ = 5◦

indicates nearly no absorption in this THz band. These effectsare further illuminated by the analysis of field distributions inthe pink dashed box of Fig. 3(b). At the frequency of 0.53 THz,the on-resonance condition for θ = 26◦ demonstrates highelectric field enhancement on the analyte, and the electric fieldis significantly reduced as θ is changed to 23◦ and 29◦, respec-tively. The use of θ = 5◦ shows extremely small electric field inthe analyte, because the frequency point of 0.53 THz deviatesmuch farther away from the guided-mode resonance frequency.The above study implies that the angular scanning can enablea series of absorbance spectra, which reflects enhanced broad-band sensing to detect the THz fingerprint of the trace amountof analyte.

After revealing the enhancing mechanism, we continue toinvestigate the general sensing performance of molecular finger-print for the α-lactose monohydrate layer of 1 µm thick. As weknow, α-lactose monohydrate has larger values of extinctioncoefficient around the frequencies of 0.53 THz, 1.20 THz, and1.38 THz [20], as shown in Figs. 4(a) and 4(d). These frequen-cies correspond to strong THz absorption bands of this material.The absorption band goes over a very large frequency range from0.53 THz to 1.38 THz, and it is infeasible to cover such a largerange by a scanning of θ on one metagrating. Therefore, weadopt two metagratings with the angular scans in the ranges of0.45 THz–0.63 THz and 1.15 THz–1.65 THz, respectively. Asshown in Fig. 4(b), the gray dashed curves correspond to a seriesof absorbance spectra within the range of 0.45 THz–0.63 THzfor angles θ changing from 5◦ to 40◦ gradually, and the reddashed curve indicates an envelope formed by the maximum

Fig. 4. Optical properties for the coated α-lactose mono-hydrate. From 0.45 THz to 0.63 THz: (a) complex refractiveindex; (b) absorbance spectra, where p = 400 µm, w= 180 µm,t1 = 85 µm, and t2 = 100 µm for the metagrating; (c) broadbandabsorbance enhancement. From 1.15 THz to 1.65 THz: (d) com-plex refractive index; (e) absorbance spectra, where p = 173.7 µm,w= 86 µm, t1 = 47 µm, and t2 = 33.5 µm for the metagrating;(f ) broadband absorbance enhancement.

values for all the absorbance spectra. According to the extinctioncoefficient shown in Fig. 4(a), this envelope illuminates theretrieval of the broadband molecular fingerprint for α-lactosemonohydrate, and demonstrates significant signal enhance-ment, compared with the reference absorbance spectrum (bluecurve) for the sample on an unpatterned substrate. For instance,at the frequency of 0.53 THz, the peak value of the referencespectrum is only 4.35%, but the peak value of the envelopeis up to 56.3%. In order to elucidate the broadband signalenhancement, we define an enhancement factor as below:

α = 10 lg

( ∫ f2f1

A( f )d f∫ f2f1

ARef( f )d f

), (3)

where A( f ) and ARef( f ) are the frequency-dependentabsorbance of the analyte on the metagrating and on the unpat-terned substrate, respectively. f1 and f2 represent the startingand ending points of a frequency interval, respectively. As shownin Fig. 4(c), the absorbance enhancement factors of α-lactosemonohydrate from 0.45 THz to 0.63 THz are all larger than9.67 dB. We adopt a similar scheme by an angular scan from10◦ to 30◦ for the detection from 1.15 THz to 1.65 THz. Asshown in Fig. 4(e), the envelope signal intensities at 1.195 THzand 1.38 THz are remarkably elevated from 1.21% and 9.03%(reference) to 16.1% and 84.5%, respectively. It is worth men-tioning that there is no obvious fingerprint peak for the referencespectrum at 1.195 THz, but this peak can be explicitly retrievedon the metagrating, which further indicates that the proposedmethod provides a powerful tool to capture the comprehensivefingerprint information of the analyte. In this frequency range,the absorbance enhancement factors are around 10 dB as shownin Fig. 4(f ), which demonstrates highly improved detectionperformance.

2338 Vol. 45, No. 8 / 15 April 2020 /Optics Letters Letter

Fig. 5. Optical properties for coated 2, 4-DNT: (a) complex refrac-tive index; (b) absorbance spectra, where p = 195 µm, w= 61 µm,t1 = 25 µm, and t2 = 35.5 µm for the metagrating; (c) broad-band absorbance enhancement. Optical properties for coatedRDX: (d) complex refractive index; (e) absorbance spectra, wherep = 230 µm, w= 60 µm, t1 = 116 µm, and t2 = 20 µm for themetagrating; (f ) broadband absorbance enhancement.

Finally, in order to evaluate the detection performance morecomprehensively, we investigate the sensing capabilities forthe trace amount of 2, 4-DNT and RDX, respectively. Thedetection of these two materials are of great importance, becausethey are widely used as the raw materials of explosives. As shownin Figs. 5(a) and 5(d), 2, 4-DNT and RDX have the maximumvalues of extinction coefficient at about 1.07 THz and 0.88 THz[21,22], respectively, which correspond to the central frequen-cies of THz absorption bands for these two materials. As shownin Figs. 5(b) and 5(e), the reference absorbance ratios (bluesolid curves) for the trace amount of these two materials on anunpatterned substrate are extremely small, on which conditionsthe measured data would be very unreliable due to the lowsignal-to-noise ratios in practice. According to the fingerprintfeatures of 2, 4-DNT and RDX, we design two kinds of meta-gratings with specific THz resonance properties. Comparedwith the reference spectra, the sensing envelope signals bythe angular scans on metagratings [10◦–56◦ for Fig. 5(b) and10◦–60◦ for Fig. 5(e)] are significantly enhanced. As observedfrom Figs. 5(b) and 5(e), the envelope signal intensities at1.07 THz for 2, 4-DNT and 0.88 THz for RDX are extraor-dinarily increased from 1.60% and 0.33% (reference) to39.9% and 8.15%, respectively. The enhancement fac-tors of absorbance in Figs. 5(c) and 5(f ) provide a detailedunderstanding for the broadband sensing performance by theproposed method. All the enhancement factors of absorbancein Figs. 5(c) and 5(f ) are above 12.2 dB for 2, 4-DNT andabove 13.9 dB for RDX, both of which demonstrate the highlyimproved sensing performance for broadband molecularfingerprint detection.

In summary, we propose the sensing scheme of angularmanipulation on dielectric metagratings for THz broadbandsignal enhancement on molecular fingerprint detection. The useof metagrating greatly enhances the wave–matter interactionfor THz sensing. By designing the specific guided-mode reso-nance structures and angular scanning ranges, one can detectthe molecular fingerprints for various trace-amount materials.The proposed method provides a powerful THz tool for ultra-sensitive trace analysis and will facilitate many new THz sensingapplications.

Funding. NSAF Joint Fund (U1830116); FundamentalResearch Funds for the Central Universities (20720190010);Key Laboratory of THz Technology, Ministry of Education,China.

Disclosures. The authors declare no conflicts of interest.

REFERENCES1. S. L. Dexheimer, Terahertz Spectroscopy: Principles and Applications

(CRC Press, 2007).2. X. Zang, C. Shi, L. Chen, B. Cai, Y. Zhu, and S. Zhuang, Sci. Rep. 5,

8901 (2015).3. C. Shi, X. Zang, L. Chen, Y. Peng, B. Cai, G. R. Nash, and Y. M. Zhu,

IEEE. Trans. Terahertz Sci. Technol. 6, 40 (2016).4. X. Zang, H. Ding, Y. Intaravanne, L. Chen, Y. Peng, J. Xie, Q. Ke,

A. V. Balakin, A. P. Shkurinov, X. Chen, Y. Zhu, and S. Zhuang, LaserPhoton. Rev. 13, 1900182 (2019).

5. X. Zang, H. Gong, Z. Li, J. Xie, Q. Cheng, L. Chen, A. P. Shkurinov, Y.Zhu, and S. Zhuang, Appl. Phys. Lett. 112, 171111 (2018).

6. M. R. Leahy-Hoppa, M. J. Fitch, X. Zheng, L. M. Hayden, and R.Osiander, Chem. Phys. Lett. 434, 227 (2007).

7. C. A. Schmuttenmaer, Chem. Rev. 104, 1759 (2004).8. D. K. Lee, J. H. Kang, J. S. Lee, H. S. Kim, C. Kim, J. H. Kim, T. Lee,

J. H. Son, Q. H. Park, andM. Seo, Sci. Rep. 5, 15459 (2015).9. S. Lin, K. Bhattarai, J. Zhou, and D. Talbayev, Opt. Express 24, 19448

(2016).10. T. S. Bui, T. D. Dao, L. H. Dang, L. D. Vu, A. Ohi, T. Nabatame, Y. Lee, T.

Nagao, and C. V. Hoang, Sci. Rep. 6, 32123 (2016).11. Z. Han, A. M. Soehartono, B. Gu, X. Wei, K. T. Yong, and Y. Shi, Opt.

Express 27, 9032 (2019).12. Y. Zhang, Y. Feng, B. Zhu, J. Zhao, and T. Jiang, Opt. Express 22,

22743 (2014).13. Y. Zhang, T. Li, B. Zeng, H. Zhang, H. Lv, X. Huang, W. Zhang, and

A. K. Azad, Nanoscale 7, 12682 (2015).14. M. Naftaly and R. E. Miles, Proc. IEEE 95, 1658 (2007).15. E. V. Fedulova, M. M. Nazarov, A. A. Angeluts, V. I. Sokolov, and A. P.

Shkurinov, Proc. SPIE 8337, 83370I (2012).16. J. Zhou, S. Yan, C. Li, J. Zhu, and Q. H. Liu, Opt. Express 26, 18155

(2018).17. J. Zhu, S. Yan, N. Feng, L. Ye, J.-Y. Ou, andQ. H. Liu, Appl. Phys. Lett.

112, 153106 (2018).18. J. Zhu, C. Li, J.-Y. Ou, and Q. H. Liu, Carbon 142, 430 (2019).19. S. S. Wang and R. Agnusson, Appl. Opt. 32, 2606 (1993).20. A. Roggenbuck, H. Schmitz, A. Deninger, I. C. Mayorga, J.

Hemberger, R. Güsten, and M. Grüninger, New J. Phys. 12, 043017(2010).

21. G. F. Liu, X. J. Ma, S. H. Ma, H. W. Zhao, M. W. Ma, M. Ge, and W. F.Wang, Chin. J. Chem. 26, 1257 (2008).

22. H. R. Park, K. J. Ahn, S. Han, Y. M. Bahk, N. Park, and D. S. Kim, NanoLett. 13, 1782 (2013).