We present a nonlinear optical crystal (NLOC)-based terahertz (THz) microfluidic chip with a few arrays of split ring resonators (SRRs) for ultra-trace and quantitative measurements of liquid solutions. The proposed chip operates on the basis of near-field coupling between the SRRs and a local emission of point like THz source that is generated in the process of optical rectification in NLOCs on a sub-wavelength scale. The liquid solutions flowing inside the microchannel modify the resonance frequency and peak attenuation in the THz transmission spectra. In contrast to conventional bio-sensing with far/near-field THz waves, our technique can be expected to compactify the chip design as well as realize high sensitive near-field measurement of liquid solutions without any high-power optical/THz source, near-field probes, and prisms. Using this chip, we have succeeded in observing the 31.8 fmol of ion concentration in actual amount of 318 pl water solutions from the shift of the resonance frequency. The technique opens the door to microanalysis of biological samples with THz waves and accelerates development of THz lab-on-chip devices.

Microfluidic devices such as micro-total analysis systems have attracted significant attention as promising tools for medical diagnosis and biological analysis.1–5 These devices generally consist of integrated chemical equipment connected with microfluidic channels with dimensions of tens to hundreds of micrometers. One of the advantageous properties of these devices is that only picoliter volumes of liquid solution is required because the volume of the microfluidic channel is much smaller than that required by conventional chemical instruments. Therefore, significant progress has been made for cellular and nucleic acid studies by using various ultra-trace and quantitative analytical techniques such as microchip electrophoresis, enzyme-linked immunosorbent assay (ELISA), surface plasmon resonance (SPR), and electrochemical quartz crystal microbalance (EQCM).6–9 

Terahertz (THz, 0.1-10 THz) region gives very important information to clarify biological reaction dynamics such as hydrogen bonds and hydrophobic interactions, which is comparatively lower energy than that of infrared absorption.10–12 THz time-domain spectroscopy (THz-TDS) is one of the powerful techniques for label-free measurement of such interactions, and new physical insights attributed to the functional expression and structural change of water,13–15 biopolymer,16 and DNA17 can be revealed. Combination of such THz techniques and microfluidic devices will attract attention not only for the development of compact THz sensors but also for the possibility to build new analytical THz devices with a high-sensitivity. However, it is challenging for ultra-trace measurements with THz waves due to the spatial resolution of far-field THz waves and the strong absorption into polar solvent.

There are several research studies addressing the issue of development of high-sensitive THz sensors for ultra-trace measurements of liquid solution by using, e.g., resonators,18,19 dielectric waveguide systems,20 and meta-surfaces,21 and majority of which focused on the use of the far-field THz waves. The laser THz emission microscope (LTEM) has also attracted considerable research interest as a direct measurement tool of various samples by using the local emission of THz waves. Unlike other methods, the resolution in LTEM is not limited by the THz wavelength but by that of the optical source, allowing THz-emission spectroscopy and -emission imaging of various samples of sub-micrometer scale achievable.22–25 In the recently developed system, a nonlinear optical crystal (NLOC) is employed as a 2D THz emitter, and THz waves are locally generated in the process of optical rectification at the irradiation spots of femtosecond (fs) laser beams.26 Since the fs laser beams are focused at the output surface of the NLOC, the beam diameter of the generated THz waves is approximately the same size as that of the focused beam spot which overcomes the diffraction limit of THz waves.26 By setting samples in the vicinity of the THz generation spot on the 2D-THz emitter, THz-TDS and THz-imaging have been successfully achieved on the samples of sub-THz wavelength scale, such as dielectric thin-film,27 a human hair strand,28 metallic small apertures,29 THz meta-atom,30 and nanoparticles.31–33 Therefore, it is interesting to see if this technique can be also applied to the measurements of trace-amount of liquid solutions.

In this paper, we propose NLOC-based THz microfluidic chips and demonstrate their use for ultra-trace and quantitative THz-TDS measurements of liquid solution set inside a microchannel. Figure 1(a) shows a schematic drawing of the concept of the THz microfluidic chip. The chip consists of a THz radiation point source, a single microchannel, and a few arrays of meta-atoms, elementary units of metamaterials. The THz radiation is generated by optical rectification in the NLOC close underneath the microchannel and couples to the meta-atoms. The chip then determines the solution concentrations based on changes in the resonant frequency and peak attenuation of the THz transmission spectrum.

FIG. 1.

(a) A schematic drawing of experimental geometry of trace amounts of liquid solution and (b) an optical microscope image of fabricated microfluidic chip. (c) THz time-domain waveforms of devices with a single meta-atom and an array of 5 × 5 meta-atoms. An enlarged drawing around the first peak is also shown.

FIG. 1.

(a) A schematic drawing of experimental geometry of trace amounts of liquid solution and (b) an optical microscope image of fabricated microfluidic chip. (c) THz time-domain waveforms of devices with a single meta-atom and an array of 5 × 5 meta-atoms. An enlarged drawing around the first peak is also shown.

Close modal

Metamaterials show a fascinating electromagnetic property and have a potential for unique optical and THz components34,35 to enhance the sensitivity. However, we need to reduce the number of meta-atoms to make the device compact, which requires the need to examine the array structure of the meta-atoms. In a previous report,30 we have studied such an issue and indicated that an array of 5 × 5 or larger is good enough to obtain high Q and a lattice constant can be set at around 120 μm, although here the meta-atom and array structures are not optimized yet. To make the fluid channel, we employ a combination of a 500-μm-thick (110) oriented GaAs as NLOC and a two-gap split ring resonator (SRR) which has a diameter of 84 μm with gaps of 20 μm and width of 10 μm. The array structure is periodically arrayed with a period of 120 μm and the number of 11 × 11 units (corresponds to the area of 1.2 mm × 1.2 mm) by a conventional photolithographic technique and radio-frequency sputtering of a 1-nm-thick adhesion layer of titanium on the GaAs wafer, followed by deposition of 0.2 μm of gold. The channel with the width of 26.5 μm, a depth of 10 μm, and a length of 2.2 mm is then fabricated along their gap regions by a wet etching process, as shown in Fig. 1(b).

We utilized a THz measurement system similar to a conventional THz-TDS using a fiber-coupled fs laser source at a wavelength of 1560 nm (see Fig. S1 of the supplementary material). The optical pump beam is focused at the central region in a center meta-atom of the array and THz waves are generated at the irradiation spot, which directly interact with the liquid solution set in the microchannel. A bowtie type photoconductive antenna (PCA) was used as a detector. Note that the system is equipped with a galvanometer which permits fast scanning of the pump beam. We used this feature to easily adjust the beam irradiation spot to the exact position by monitoring the reflected laser image before the measurement (see Fig. S2 of the supplementary material). This technique also enables us to obtain a high spatial resolution THz image of samples. In contrast to the standard THz-TDS, we can utilize the near-field interactions between the THz waves and samples; therefore, high-sensitive spectroscopy of sub-THz wavelength scale samples can be realized.26 The significant change of the sensitivity by using the proposed near-field approach is shown in detail in Fig. S3 of the supplementary material. To evaluate the chip performance, the THz transmittance T(ω) = Esamp(ω)/Eref(ω) was observed, where Esamp(ω) and Eref(ω) are the frequency-dependent THz amplitudes obtained at the microchannel with and without a liquid material of interest, respectively. Figure 1(c) shows typical THz time-domain waveforms of devices with a single meta-atom and an array of 5 × 5 meta-atoms. Multiple reflections are due to the THz wave traveling and reflection in GaAs. It can be seen that the waveforms near the main peak differ from each other.

As a preliminary investigation on the development of such THz microfluidic chip, we first developed a simple THz chip as shown in Fig. 2(a). The chip consists of a pair of circular holes that are formed in a GaAs surface. Each hole has a diameter of 200 μm and a depth of 40 μm and is fabricated at an interval of 1 mm. Figure 2(b) shows the transmittance spectrum through the empty hole. As you can see, the transmittance dip around 0.75 THz appears near the cut-off frequency (fd) as described in fd = c/2d, where c and d are the velocity of light and hole diameter, respectively. These transmission dips are also observed by using different holes (see Fig. S4 of the supplementary material) and are quite similar to the resonance effect observed in a single rectangular hole.36 It is considered that the hole used here is nearly equivalent to single hole waveguides and worked as a kind of high-pass filter.37,38 In contrast to standard far-field THz measurements, this technique shows strong filtering effects even by using a single circular hole because the THz source is smaller than the hole diameter and therefore all the THz waves propagate through the hole without any loss. The reason for the increase in transmittance below the cut-off frequency can be due to the decrease of the signal to noise ratio (SNR) below the system noise dynamic range.39 To understand the interaction between THz waves and the hole structure, examples are shown in Fig. 2(c) as THz frequency domain images extracted at 0.4 THz (<fd), 1.1 THz (≈fd), and 1.5 THz (>fd). It is seen that THz transmission is not observed at the hole below the cut-off frequency fd, while it begins to be observable around fd and eventually it is clearer at the frequency higher than fd. Especially for the frequency band >fd, we could observe highly focused THz beam spot inside the hole. Similar results have been reported by using smaller metallic apertures.29 This result indicates that we can utilize efficient interaction between THz waves and samples set in the hole and the chip can be used as a high sensitive sensor in this frequency band. To demonstrate the sensitivity of the device in detecting trace amounts of liquid solutions, 3 kinds of mineral water solutions with different hardness of 60 mg/l, 88 mg/l, and 304 mg/l were measured. In the measurement, one of the holes was filled with the sample and the other one was filled with reference liquid (distillated water with 0 mg/l of minerals), each with a volume of ∼50 nl by using an ultra-micro-syringe pump system (0.5BR-7BV: World Precision Instruments, Inc.). The holes were covered with a 580-μm-thick quartz plate to prevent evaporation of solutions. All samples were preliminarily prepared in the experimental room at 23 °C and relative humidity of 36% before the measurement and carefully avoiding any drastic changes in the environment of the experimental room during the measurements. As for the evaluation, we employed the frequency dependent differential THz absorption Δα(ω), as described in Δα(ω) = αsample(ω) − αref(ω), where αsample(ω) and αref(ω) are frequency dependent THz absorbances of the sample and reference liquid, respectively. Figure 2(d) shows the differential THz absorption spectra of the 3 kinds of solutions prepared. Here, the THz time waveforms were measured five times for each sample, and Δα(ω) was calculated using the average of these waveforms. It is found that each spectrum shows a specific valley structure and the differential absorption intensity increases with an increase in their hardness around 1.1 THz. By increasing the minerals in the solution, the THz absorbance is decreased, which means that the THz waves are more easily transmitted through the solution [see also Fig. S10(a) of the supplementary material]. Therefore, this technique could be used for a high-sensitive measurement of the volume of the solution. As a result, we could detect the amount of solute which is 75 pg, 110 pg, and 380 pg of minerals in the actual solution volume of 1.25 nl, for the cases of 60 mg/l, 88 mg/l, and 304 mg/l, respectively. On the other hand, despite a significant change of the hardness from 88 mg/l to 304 mg/l, those spectra exhibit almost similar tendencies. Since the chip detects the minute changes of the amplitude of THz absorption, it is susceptible to the minute changes of the solution volume in the process of sample setting. Based on these preliminary investigations, this chip has a potential for measurements of trace amount of liquid solutions and the sensitivity can be further improved by using more quantitative measurement approach. As regards to this point, it can be said that the use of a microchannel is more appropriate for measuring trace amount of liquid solutions.

FIG. 2.

(a) A schematic drawing of experimental geometry of small aperture THz chip. (b) A THz transmittance spectrum and (c) frequency domain images of the empty chip. (d) THz differential spectra of mineral water with different minerals.

FIG. 2.

(a) A schematic drawing of experimental geometry of small aperture THz chip. (b) A THz transmittance spectrum and (c) frequency domain images of the empty chip. (d) THz differential spectra of mineral water with different minerals.

Close modal

Considering the fabrication of such a NLOC-based THz chip, it is important to understand the nonlinear and resonance responses of THz waves through the NLOC and the meta-atom, respectively. This will determine the optimal device structure, such as the direction of the microchannel with respect to the GaAs crystal structure. To optimize the electrical coupling among the nonlinear THz generation and resonance response of meta-atom, we examined the polarization dependence of the localized emission of THz waves on resonance response of meta-atom by rotating the NLOC and the meta-atom structures as shown in Figs. 3(a) and 3(c), respectively. First, we evaluated the nonlinear THz response from the bare GaAs and chose the best direction of the polarization of the normally incident optical pump by changing the azimuthal angle θ0 from [001] axis in GaAs. Figure 3(b) shows the measured THz waveforms of GaAs versus θ0. The polarization of the optical beam was parallel to the PCA detector axis. It is found that the amplitude of the nonlinear THz radiation reached the positive/negative maximum at θ0 = 54°, 120°, 235°, and 300°. These values agree well with those of calculated zinc blende semiconductors described by ITHzθ0=3/4ITHzmaxsin2θ043sin2θ0.40 Then, we fixed the polarization of the optical beam E0 with its angle of θ0 = 54° and evaluated the resonance response by orienting the structure of the SRR at an angle of φ with respect to E0. Here, a reflected laser image was used to find φ (see Fig. S2 of the supplementary material). Figures 3(d) and 3(e) show measured and numerically simulated THz transmittance spectra [finite-difference time domain (FDTD) solution: Lumerical Solutions, Inc.] through SRR, respectively. As you can see, we could obtain good agreement in the resonance response; no resonance response was observed when E0 is perpendicular to the direction of the gap [indicating g in Fig. 3(c)], while the resonance response was observed around 0.58 THz in the measurement and was increased with an increase in the angle between E0 and the direction of the gap. Eventually, it shows the maximum when E0 is parallel to the direction of the gap. Therefore, the microchannel was positioned over this area.

FIG. 3.

(a) A schematic drawing of angular dependence of THz radiation from GaAs(110). (b) THz peak amplitudes from the bare GaAs(110) vs azimuthal angle, θ0, with the optical pump beam polarization. (c) A schematic drawing of angular dependence of THz radiation from a two-gap SRR. (d) Measured and (e) simulated THz transmittance spectra of the two gap SRRs fabricated on GaAs with various angles φ. The polarization of the optical pump beam is fixed at θ0 = 54°, as shown by a dotted round mark in (b).

FIG. 3.

(a) A schematic drawing of angular dependence of THz radiation from GaAs(110). (b) THz peak amplitudes from the bare GaAs(110) vs azimuthal angle, θ0, with the optical pump beam polarization. (c) A schematic drawing of angular dependence of THz radiation from a two-gap SRR. (d) Measured and (e) simulated THz transmittance spectra of the two gap SRRs fabricated on GaAs with various angles φ. The polarization of the optical pump beam is fixed at θ0 = 54°, as shown by a dotted round mark in (b).

Close modal

Another important issue in dealing with this near-field approach is how the number and interval of the meta-atoms affects their resonance responses. In conventional metamaterial research studies with THz waves, far-field THz waves are focused on a few thousands of resonators. Therefore, the resonance response is mainly affected by the shape of each resonator.41 On the other hand, meta-atoms that are excited with near-field emission of THz waves have not yet been investigated. We have already reported the direct measurement of single gap SRRs with near-field emission of THz waves;30 however, it is still incompletely understood for the case of double gap SRRs. In order to examine the impact of the nearest neighbor interaction on two-gap SRRs, we measured the THz transmittance spectra through the SRR with various periodic and super-cell size conditions. In all the measurements, only the central region of the center element of the SRR is excited with the localized emission of THz pulses. Figure 4(a) shows the measured THz transmittance spectra through the SRR when the period is fixed at 120 μm and the number of the meta-atoms is varied at 1 × 1, 3 × 3, and 5 × 5 units. It is noticed that the resonance peaks are observed at 0.54 THz for the cases of 1 × 1 and 3 × 3 units, while it is shifted at 0.49 THz for the case of 5 × 5 units. These frequencies can be considered as an evolution from an unperturbed resonance of a LC circuit to the inductive or capacitive coupling among the SRRs, causing a shift of the resonance frequency and eventually modifying the resonance linewidth. The intensities of the resonance are nearly identical after increasing the super-cell size from 3 × 3 units, which agrees with simulated results shown in Fig. 4(b). This is because the meta-atom excited with localized THz waves becomes a power feeding point and affects the neighboring meta-atoms.42 The dependence of the peak intensity of the generated THz waves, the number of the SRRs, and the incident pump laser power is shown in Fig. S9 of the supplementary material. Similar tendency was also observed when the periods of SRRs were varied (see Fig. S5 of the supplementary material). So it is considered that we can obtain enough resonance response with only 3 × 3 units of meta-atoms with its period of 120 μm in this case. To understand these neighboring effects of the SRRs, we simulated the electric field distribution of 7 × 7 units of SRRs for the cases of their periods with 120 μm at their resonance frequencies of 0.5 THz, as shown in Fig. 5. It is observed that the electric field from the excited SRR (the center SRR) mainly affects several units of its neighboring SRRs, while the outer SRRs are less affected. Additionally, the effects on the neighboring SRRs are lesser when the periods are 180 μm and 240 μm, which can be almost identical for the case of 1 × 1 unit (see Fig. S6 of the supplementary material).

FIG. 4.

(a) Measured and (b) simulated THz transmittance spectra of the SRR when the periods are fixed at 120 μm and the number of meta-atoms is varied at 1 × 1, 3 × 3, and 5 × 5 units.

FIG. 4.

(a) Measured and (b) simulated THz transmittance spectra of the SRR when the periods are fixed at 120 μm and the number of meta-atoms is varied at 1 × 1, 3 × 3, and 5 × 5 units.

Close modal
FIG. 5.

Simulated THz electric field distribution of 7 × 7 units of SRRs at the resonance frequency of 0.5 THz for the cases of their periods with 120 μm. The image is enlarged views of 5 × 5 arrays area for the clarity.

FIG. 5.

Simulated THz electric field distribution of 7 × 7 units of SRRs at the resonance frequency of 0.5 THz for the cases of their periods with 120 μm. The image is enlarged views of 5 × 5 arrays area for the clarity.

Close modal

To evaluate the sensitivity of the chip, we used distilled water (0 mg/l of minerals) and commercial mineral water (Contrex, 1468 mg/l of minerals) with different hardness as liquid solutions and measured their resonance characteristics. To avoid the effect of the ion composition change, we chose 1 particular brand of mineral water and prepared 5 kinds of sample solutions with different hardness by diluting it to 10 mg/l, 40 mg/l, 200 mg/l, 600 mg/l, and 1000 mg/l. The measurement condition is the same as in the preliminary investigation, except the sample setting. Here, the channel was covered by the quartz plate and the liquid solution of ∼50 nl was dripped with the microsyringe manually at one of the water storages that are fabricated at both sides of the channel (see Fig. S7 of the supplementary material). Then the solutions were automatically installed into the micro-fluid channel by capillarity phenomenon. Note that the actual amounts of the liquid solution that interact with the SRRs were calculated to be about 318 pl. All the samples were repeatedly measured 10 times for reliability. From our preliminary investigation, the main reason for determining the resonance depth and the frequency is meta-atom array structures, so it does not have a large influence with minute temperature change. Figure 6(a) shows THz transmittance spectra of the meta-atom, the distilled water, and the commercial mineral water. It is noticed that a remarkable resonance at the frequency around 0.55 THz was observed in the measurement of the empty chip and was shifted to the lower frequency around 0.48 THz with depleting the Q factor after filling with distilled water in the channel. This frequency shift could be due to the increase of capacitance in the gaps of SRRs by replacing air with liquid solution, which has a higher dielectric constant. In the measurements of commercial mineral water, the resonance frequency was slightly shifted to higher frequency with an increase of the mineral amounts. For clarity of this frequency shift, the differentiated THz transmittance spectra with deviation from the resonance frequency of distilled water are also shown in Fig. 6(b). The origin of the frequency shift can be estimated via our preliminary investigation on the relationship between hardness and optical constant of water: the refractive index and the absorption coefficient are decreased with an increase of the hardness in water, and the resonance frequency is shifted to higher frequency with a decrease of the refractive index of the solutions inside the microfluidic channel (see Fig. S10 of the supplementary material). Figure 6(c) shows a plot of the resonance frequency as a function of the mole number of minerals in the mineral water. Here, the mole number was calculated by using the molar mass of CaCO3 (100.087 g/mol: World Health Organization standard) and the volume of the channel being in contact with SRR arrays. It is clearly seen that the resonance frequency was shifted to higher frequency with an increase of the amount of minerals in the mineral water. Figure 6(d) shows a plot of the frequency shift (Δf) as a function of the mole number. Here, Δffmf0, where fm and f0 are the resonance frequency of the mineral water with different hardness and distilled water, respectively. It is found that the minute change of the minerals of 10 mg/l was sensitively detected. This sensitivity corresponds to 31.8 fmol according to the calculation as referred earlier. The similar tendencies were also observed in the measurement done on another day (see Fig. S8 of the supplementary material). This approach can be more sensitive and quantitative than that obtained by a small hole type THz chip.

FIG. 6.

(a) Measured THz transmittance spectra of the water solution with various concentration of minerals and (b) the differentiated spectra of (a). (c) Resonance characteristics and (d) resonance frequency shift as a function of the mole number of mineral water.

FIG. 6.

(a) Measured THz transmittance spectra of the water solution with various concentration of minerals and (b) the differentiated spectra of (a). (c) Resonance characteristics and (d) resonance frequency shift as a function of the mole number of mineral water.

Close modal

In conclusion, we fabricated and demonstrated a NLOC-based THz microfluidic chip with a few arrays of meta-atoms for microanalysis, high-sensitive, and label-free measurements of biological samples. It can detect trace amounts of known materials and very small changes in optical constants due to the chemical/bio reaction processes. The microfluidic chip was evaluated by using distilled water and commercial mineral water with different hardness, and we were able to detect 31.8 fmol of the mineral in a 318 pl solution. This sensitivity is comparable to a standard commercial fluorescence-based systems and it can be improved by further optimization of the structure and the periodic arrangement of meta-atoms such as Fano or toroidal resonant types.43,44 As for the measurement of solutions having larger THz absorption coefficient, shallowing the channel depth or the reflective arrangement system for reducing the THz absorption into water will be effective. We could also confirm the quantitative measurement of trace amount of liquid solutions by comparing a simple small hole type THz chip. Since NLOCs are relatively low-cost, we believe that the chip could be used as compact disposable sensors in medical and biological fields. The THz intensity by using this technique is certainly weaker than that by THz quantum cascade lasers (THz-QCLs), but it is easy to control resonance frequency by a simple structural change and also has a function of positioning excitation laser beam, so it is possible to obtain two-dimensional information such as chemical reaction process that is observed in micro-total analysis systems. If a specific frequency is found to be important by evaluating various samples with this chip, we can contribute to less expensive and simple bio sensing by feeding back with sensing using THz-QCLs. Moreover, this technique can be also utilized for THz lab-on-chips. Observing chemical reactions of compounds with water using multi-fluid channels is also attractive and will help understanding of the dynamics of liquids in THz frequency region, which has potential applications in the chemical industry.

See supplementary material for more details on the setup and extra information on the THz chip properties.

This work was partially supported by JSPS KAKENHI Grant Nos. JP15K18053 and JP17H01269 and a research granted from The Murata Science Foundation.

1.
M. U.
Kopp
,
A. J.
De Mello
, and
A.
Manz
,
Science
280
,
1046
(
1998
).
2.
A.
Arora
,
G.
Simone
,
G. B.
Salieb-Beugelaar
,
J. T.
Kim
, and
A.
Manz
,
Anal. Chem.
82
,
4830
(
2010
).
3.
C. T.
Culbertson
,
T. G.
Mickleburgh
,
S. A.
Stewart-James
,
K. A.
Sellens
, and
M.
Pressnall
,
Anal. Chem.
86
,
95
(
2014
).
4.
Lab-on-a-Chip Devices and Micro-Total Analysis Systems
, edited by
J.
Castillo-León
and
W. E.
Svendsen
(
Springer
,
2014
).
5.
V.
Faustino
,
S. O.
Catarino
,
R.
Lima
, and
G.
Minas
,
J. Biomech.
49
,
2280
(
2016
).
6.
V.
Doln
k
ı,
J. Biochem. Biophys. Methods
41
,
103
(
1999
).
7.
Z.
Zhu
,
J.
Lu
, and
S.
Liu
,
Anal. Chim. Acta
709
,
21
(
2012
).
8.
F.
Liu
,
X.
Liu
,
S.
Ng
, and
H.
Chan
,
Sens. Actuators, B
113
,
234
(
2006
).
9.
R.
Slavík
and
J.
Homola
,
Sens. Actuators, B
123
,
10
(
2007
).
10.
M.
Tonouchi
,
Nat. Photonics
1
,
97
(
2007
).
11.
M.
Nagel
,
M.
Först
, and
H.
Kurz
,
J. Phys.: Condens. Matter
18
,
S601
(
2006
).
12.
Terahertz Biomedical Science and Technology
, edited by
J.
Son
(
CRC Press
,
2014
).
13.
U.
Heugen
,
G.
Schwaab
,
E.
Bründermann
,
M.
Heyden
,
X.
Yu
,
D. M.
Leitner
, and
M.
Havenith
,
Proc. Natl. Acad. Sci. U. S. A.
103
,
12301
(
2006
).
14.
L.
Thrane
,
R.
Jacobsen
,
P.
Jepsen
, and
S.
Keiding
,
Chem. Phys. Lett.
240
,
330
(
1995
).
15.
H.
Hirori
,
K.
Yamashita
,
M.
Nagai
, and
K.
Tanaka
,
Jpn. J. Appl. Phys.
43
,
L1287
(
2004
).
16.
T.
Arikawa
,
M.
Nagai
, and
K.
Tanaka
,
Chem. Phys. Lett.
457
,
12
(
2008
).
17.
B.
Fischer
,
M.
Walther
, and
P.
Jepsen
,
Phys. Med. Biol.
47
,
3807
(
2002
).
18.
A.
Soltani
,
H.
Neshasteh
,
A.
Mataji-Kojouri
,
N.
Born
,
E.
Castro-Camus
,
M.
Shahabadi
, and
M.
Koch
,
Appl. Phys. Lett.
108
,
191105
(
2016
).
19.
M.
Nagel
,
P.
Bolivar
,
M.
Brucherseifer
,
H.
Kurz
,
A.
Bosserhoff
, and
R.
Büttner
,
Appl. Phys. Lett.
80
,
154
(
2002
).
20.
S.
Ranjana
,
P.
Bhatt
,
H.
Surdi
,
B.
Sangala
,
M.
Satyanarayan
,
G.
Umesh
, and
S.
Prabhu
,
Biomed. Phys. Eng. Express
1
,
025003
(
2015
).
21.
L.
Xie
,
W.
Gao
,
J.
Shu
,
Y.
Ying
, and
J.
Kono
,
Sci. Rep.
5
,
8671
(
2015
).
22.
H.
Murakami
,
K.
Serita
,
Y.
Maekawa
,
S.
Fujiwara
,
E.
Matsuda
,
S.
Kim
,
I.
Kawayama
, and
M.
Tonouchi
,
J. Phys. D: Appl. Phys.
47
,
374007
(
2014
).
23.
H.
Murakami
,
S.
Fujiwara
,
I.
Kawayama
, and
M.
Tonouchi
,
Photonics Res.
4
,
A9
(
2016
).
24.
T.
Kiwa
,
S.
Oka
,
J.
Kondo
,
I.
Kawayama
,
H.
Yamada
,
M.
Tonouchi
, and
K.
Tsukada
,
Jpn. J. Appl. Phys.
46
,
L1052
(
2007
).
25.
Y.
Sakai
,
I.
Kawayama
,
H.
Nakanishi
, and
M.
Tonouchi
,
APL Photonics
2
,
041304
(
2017
).
26.
K.
Serita
,
S.
Mizuno
,
H.
Murakami
,
I.
Kawayama
,
Y.
Takahashi
,
M.
Yoshimura
,
Y.
Mori
,
J.
Darmo
, and
M.
Tonouchi
,
Opt. Express
20
,
12959
(
2012
).
27.
W.
Withayachumnankul
,
H.
Lin
,
K.
Serita
,
C.
Shah
,
S.
Sriram
,
M.
Bhaskaran
,
M.
Tonouchi
,
C.
Fumeaux
, and
D.
Abbott
,
Opt. Express
20
,
3345
(
2012
).
28.
K.
Serita
,
H.
Murakami
,
I.
Kawayama
,
Y.
Takahashi
,
M.
Yoshimura
,
Y.
Mori
, and
M.
Tonouchi
,
Opt. Eng.
53
,
031205
(
2013
).
29.
C.
Fumeaux
,
H.
Lin
,
K.
Serita
,
W.
Withayachumnankul
,
T.
Kaufmann
,
M.
Tonouchi
, and
D.
Abbott
,
Opt. Express
20
,
18397
(
2012
).
30.
K.
Serite
,
J.
Darmo
,
I.
Kawayama
,
H.
Murakami
, and
M.
Tonouchi
,
J. Infrared, Millimeter, Terahertz Waves
38
,
1107
(
2017
).
31.
R.
Rakshit
,
D.
Sarkar
,
M.
Pal
,
K.
Serite
,
M.
Tonouchi
, and
K.
Mandal
,
J. Phys. D: Appl. Phys.
48
,
245301
(
2015
).
32.
R.
Rakshit
,
M.
Pal
,
K.
Serita
,
A.
Chaudhuri
,
M.
Tonouchi
, and
K.
Mandal
,
AIP Conf. Proc.
1728
,
020084
(
2015
).
33.
R.
Rakshit
,
K.
Serita
,
M.
Tonouchi
, and
K.
Mandal
,
J. Appl. Phys.
120
,
203901
(
2016
).
34.
H.
Chen
,
W.
Padilla
,
J.
Zide
,
A.
Gossard
,
A.
Taylor
, and
R.
Averitt
,
Nature
444
,
597
(
2006
).
35.
H.
Chen
,
W.
Padilla
,
M.
Cich
,
A.
Azad
,
R.
Averitt
, and
A.
Taylor
,
Nat. Photonics
3
,
148
(
2009
).
36.
F. J.
Garcia-Vidal
,
E.
Moreno
,
J. A.
Porto
, and
L.
Martin-Moreno
,
Phys. Rev. Lett.
95
,
103901
(
2005
).
37.
D.
Grischkowsky
,
S.
Keiding
,
M.
Van Exter
, and
C.
Fattinger
,
J. Opt. Soc. Am. B
7
,
2006
(
1990
).
38.
R.
Mendis
,
V.
Astley
,
J.
Liu
, and
D.
Mittleman
,
Appl. Phys. Lett.
95
,
171113
(
2009
).
39.
P.
Jepsen
and
B. M.
Fischer
,
Opt. Lett.
30
,
29
(
2005
).
40.
A.
Rice
,
Y.
Jin
,
F.
Ma
,
X.
Zhang
,
D.
Bliss
,
J.
Larkin
, and
M.
Alexander
,
Appl. Phys. Lett.
64
,
1324
(
1994
).
41.
W.
Withayachumnankul
and
D.
Abbott
,
IEEE Photonics J.
1
,
99
(
2009
).
42.
D.
Chowdhury
,
R.
Singh
,
M.
Reiten
,
J.
Zhou
,
A.
Taylor
, and
J.
O’Hara
,
Opt. Express
19
,
10679
(
2011
).
43.
R.
Singh
,
W.
Cao
,
I.
Al-Naib
,
L.
Cong
,
W.
Withayachumnankul
, and
W.
Zhang
,
Appl. Phys. Lett.
105
,
171101
(
2014
).
44.
M.
Gupta
,
Y. K.
Srivastava
,
M.
Manjappa
, and
R.
Singh
,
Appl. Phys. Lett.
110
,
121108
(
2017
).

Supplementary Material