Open Access
Add to BibSonomy
Abstract
We present the 3D printing of microfluidic devices suitable for the THz regime using TOPAS filaments. We assessed the 3D-printed microfluidic devices with THz time-domain spectroscopy and X-ray microscopy. We compared the results of polished and unpolished windows of the 3D printed devices for studying the quality of the 3D printing in the THz and VIS regimes. Spectroscopic measurements were performed of water and lysozyme dissolved in water. We found that the results from the polished and unpolished devices were similar in the THz regime. A detailed calculation of the measurement uncertainty revealed that the dominant component of the measurement uncertainly was the gap thickness. Because of their optical and mechanical properties, we are convinced that 3D printing from TOPAS offers a cost-effective and rapid production of microfluidic devices usable in X-ray, UV, VIS, NIR, and THz regimes.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
THz radiation generally refers to the frequency band spanning 0.2–10 THz. In the past 20 years, THz science and technology have rapidly been growing because of the non-invasive and non-ionizing radiation, and so new materials and devices are required [1]. Moreover, in the THz frequency range, various biomolecules can be effectively recognized and characterized according to their low-frequency biomolecular motions, including vibration and rotation of the molecular skeleton [2]. Using a single device, such as a chip-based microfluidic device, various materials were identified and characterized in the THz regime, namely biomolecules including DNA/RNA, amino acids/peptides, proteins, and carbohydrates to cells and tissues [3]. Microfluidic devices have a wide range of medical applications, including cellular analysis, genomics, proteomics and metabolomics, immunoassays, point of care diagnostics, and organs on chips [4,5]. Additionally, such devices can find applications in inorganic chemistry, such as in liquid-liquid extraction, and synthesis of inorganic material, including semiconductors, nanoparticles, or colloids [6]. Until now, microfluidic devices were mainly produced based on the technology of traditional processes for micro-electromechanical systems (MEMS) such as lithography, micro hot embossing, and microinjection [7]. Advancements in the resolution and speed of 3D printing methods, Fused Deposition Modeling (FDM) and Stereolithography (SLA) helped to simplify the fabrication process of microfluidic devices [4,8]. A simple microfluidic device for THz applications using 3D printing and polypropylene (PP) and polystyrene (PS) was suggested by S. Schewa et al. [9]. Materials for FDM include Cyclic Olefin Copolymer (COC polymer - TOPAS) [10,11]. It has been used for 3D printing of THz lenses [12], and bendable THz fibers [13]. Furthermore, TOPAS is biocompatible [14] and transparent in a wide frequency range [15,16]. Using a microfluidic device based on TOPAS, measurements in the THz and other spectral ranges are possible. Optimization of its 3D printing parameters has already been demonstrated by our group [17].
In this work, we show that 3D printing of TOPAS can offer a cost-effective solution for developing microfluidic devices for use in the X-ray, UV, VIS, and THz regimes. We used two different methods of fabricating microfluidic devices optimized for the THz regime. In the first method, the windows of the devices were polished mechanically after 3D printing, while in the second method, the windows were not polished. The latter method takes up to 4 hours less than the first, and the measurements show that both types of devices can be used in the THz regime. Devices with a gap thickness ranging from ${177}\;\mathrm{\mu}\textrm{m}$ to ${440}\;\mathrm{\mu}\textrm{m}$ have been made. The refractive index and absorption coefficient of the empty device, the device filled with pure water and filled with a protein lysozyme solution, were measured in the range of 0.3 to 1.5 THz and compared to previous work. The measurement uncertainty of the THz optical parameters is estimated and is dominated by the measurement uncertainty of the thickness of the device, which is as low as $dr={12}\;\mathrm{\mu}\textrm{m}$.
2. Methods
A 3D printer RepRap Industrial (Kühling $\&$ Kühling, Kiel, Germany) and a TOPAS filament with a diameter of ${2.85}\;\textrm{mm}$ from Creamelt (Rapperswil-Jona, Switzerland) were used for 3D printing. All devices were designed with Catia v5-3DX (3D Experience). Figure 1(a) shows an overall view of the microfluidic device, while Fig. 1(b) highlights the spacer used to define the gap thickness. Two different types of microfluidic devices were produced. A device that had all its surfaces polished and an unpolished device (Fig. 1(c$\&$d)). The 3D printing parameters for TOPAS were optimized for the 3D printing of microfluidic devices. In the end, three different parameter sets were used. One for the polished device (P); one for the inner part of the top and bottom piece from the unpolished device, which needs to be transparent (UT), and one for the outer part of the top piece from the unpolished device, which needs to be geometrically accurate (UA). The only significant difference between the different parameter sets is the extrusion multiplier which was varied between 1.0 and 1.3 for the three parts, as shown in Table 1. Other important 3D printing parameters, such as nozzle diameter, bed, and extruder temperature, were the same for the 3D printing of all the devices.
Fig. 1. (a) Blender schematic representations of the two channels microfluidic device and (b) the 3D printed spacer. (c) Picture of the unpolished 3D printed microfluidic device; (d) the polished microfluidic device. (e) Blender illustration of the THz-TDS setup. (f) Picture of the holder.
Table 1. 3D printing parameters for the transparent (UT) and accurate (UA) parts of the unpolished device and for the polished (P) device.
For this work, we compared polished and unpolished devices because the surface roughness of the unpolished devices, due to the resolution of the 3D printing (nozzle diameter), can affect the THz transparency of the devices. Very importantly, the polished devices do not show any significant surface roughness (< ${10}\;\mathrm{\mu}\textrm{m}$, due to the Sandpaper P2400 used for the polishing) after the polishing process, while the unpolished devices show roughness comparable to the nozzle diameter (d$_{nozzle}$ = ${0.35}\;\textrm{mm}$, as presented in Table 1). This surface roughness can create scattering and diffraction because it is comparable to the THz wavelength (f = 0.85 THz corresponds to a wavelength of ${0.35}\;\textrm{mm}$ in air equal to the nozzle diameter). To produce the polished device (P), two TOPAS cuboids ${40}\;\textrm{mm}$ ${\times }$ ${40}\;\textrm{mm}$ ${\times }$ ${10}\;\textrm{mm}$ were printed and polished using a grinding machine. The cuboids were wet-grinded, starting from a P180 grid to the P2400 grid. The machine was set to 200 RPM. Force was applied by hand and offset, resulting in a more uniform height. A frame for the microfluidic device was 3D printed to align the cuboid more precisely. The frame and cuboid were fixed to the printer bed using Kapton tape. The printed spacer was ground to the desired height using P500 grid paper. Into the other cuboid, two ${2.5}\;\textrm{mm}$ holes were drilled and threaded using an M3 metal screw to achieve sealing. The spacer was then filled with glue (UHU Plus endfest 300 2 K Epoxidkleber), and the excess glue was removed. After the glue had hardened, the sides were ground on the P320 grid paper to ensure reproducible mounting and thickness measurements. Two PA 6.6 screws were used to seal the two microfluidic channels. Without additional polishing, the production time was ${4}\;\textrm{h}$ shorter in comparison with the ${8}\;\textrm{h}$ that were required for the fabrication of the polished device. The bottom part was printed with a spacer (UT). The top part was printed using two different parameters, for the transparent window (UT) and the accurate outer part with the screw hole (UA). The spacer was ground to the desired height, the hole threaded, and the device was glued. The sides were not grounded.
For the experiments, we purchased ultrapure water for molecular biology (sterile and filtered) and lysozyme from Sigma Aldrich. A 200 mg/ml lysozyme solution was prepared by mixing lysozyme and water.
The THz-TDS measurements, as shown in Fig. 1(e), were performed using a commercial system TeraFlash (TOPTICA Photonics AG, Graefelfing, Germany) with a bandwidth from 0.2 to 6.0 THz. The spot diameter in focus on the sample was ${2}\;\textrm{mm}$. For the THz-TDS measurements, the microfluidic device was mounted in an in-house 3D printed holder (Fig. 1(f)). The microfluidic device was filled with a single-use pipette with a ${2.5}\;\textrm{mm}$ opening. It, therefore, made a good seal with the hole of the microfluidic device. For emptying the device, pressurized air was used. The liquids were characterized in the restricted frequency range between 0.3 and 1.5 THz because of the high absorption coefficient of liquid water and the absorption lines of water vapor at higher frequencies.
After production, the gap thickness of the microfluidic devices was measured using a Microfocus X-ray microscope (X-TEK SYSTEMS LTD., Tring, UK). The sample was moved to the position where one of the inner window planes was in the X-ray vertical beam path (Fig. 2(a)). Then, the sample was moved until the other inner window plane was in the X-ray vertical beam path. The translation distance was thus equal to the gap thickness of the device (Fig. 2(b)). In these pictures, the bottom halves are in the vertical beam path [18]. The procedure was repeated for different orientations of the device, and the uncertainty of the gap measurement was estimated to $u_{\text {gap}}={12}\;\mathrm{\mu}\textrm{m}$. Because of the surface roughness, it was not possible to have an accurate thickness measurement for the unpolished devices.
Fig. 2. (a) Schematic X-ray measurement. The gap walls are aligned with the perpendicular ray (red line) by moving the sample. (b) X-ray measurement result for a polished device with thickness approximately 1 mm.
3. Results and discussion
For the characterization of our devices, we measured the optical properties of the empty THz devices first, see Fig. 3, which served as the reference measurement. From the transmission data, the refractive index and the absorption coefficient of TOPAS were calculated using a parameter extraction algorithm [19]. Afterward, the devices were filled with water, and the optical parameters of water were extracted from the transmission change relative to the empty device.
Fig. 3. Optical parameters of empty devices. Reproducibility due to remounting (#1, #3, #5) and drift over time for a mounted device (2 min, 6 min, 10 min). (a) Refractive index of the polished device. (b) Absorption of the polished device. c) Refractive index of the unpolished device (d) Absorption of the unpolished device.
All the results presented in this work are in the range from ${0.3}\;\textrm{THz}$ to ${1.5}\;\textrm{THz}$. We report the values of the refractive index and the absorption of our devices at the frequency f = ${0.5}\;\textrm{THz}$ due to the high signal-to-noise ratio. For this reason, the figures show insets focused in the range between ${0.45}\;\textrm{THz}$ and ${0.55}\;\textrm{THz}$.
To determine the reproducibility of the THz-TDS setup, including the mounting of the holder, the polished and unpolished devices were measured empty and filled with water five times. The empty samples were unmounted and then remounted to the holder, which took roughly two minutes. Therefore, to assess possible drifts in the apparatus, another five measurements were made every two minutes with the device left untouched. The results are shown in Figs. 3 and 4. The measured refractive index was slightly lower than expected for a TOPAS substrate. At a frequency $f={0.5}\;\textrm{THz}$ the refractive index was $n_{\text {microfluidic}}$ = 1.524, and for the bare substrate it is $n_{\text {TOPAS}}$=1.530. The difference is presumably because of the air inclusions caused by 3D printing as has been previously described [17]. The refractive index of the unpolished device was lower than the one of the polished device (Figs. 3(a&c)). Again, this can be explained by the different 3D printing parameters for the unpolished device. The data show that remounting the device in the setup had a minor effect on the results. Moreover, as can be seen in Figs. 3(b&d), the polishing reduces the absorption of the devices in the THz regime, showing absorption $\alpha <{0.5}\;\textrm{cm}^{-1}$ at a frequency of $f={0.5}\;\textrm{THz}$. In Fig. 3(b), the oscillations can be assigned to interference that occurred due to the air gap.
Fig. 4. Optical parameters of water. (a) Reproducibility of refractive index of the water-filled unpolished device. (b) Reproducibility of absorption of the water-filled unpolished device. (c) Reproducibility of refractive index of the water-filled polished device. (d) Reproducibility of absorption of the water-filled polished device for the ${0.254}\;\textrm{mm}$ thickness.
Then, the polished and unpolished devices filled with water were measured. To exclude the influence of possible bubble formation for the water-filled device, the device was unmounted, emptied, refilled, and then remounted. This process took roughly eight minutes. The reference measurement (empty device) was made with eight minutes delay (Fig. 4((a)-(d))). Because of TOPAS’s transparency in the VIS spectrum, the device could also be easily checked for air bubbles which occurred only rarely. The refilling process had a negligible effect on the results. The results were reproducible and gave a refractive index for water $n_{\text {water}}$ = 2.28 at frequency $f={0.5}\;\textrm{THz}$ and absorption $\alpha <{150}\;\textrm{cm}^{-1}$.
The difference between the polished and unpolished device in the refractive index and the absorption is minor and is probably caused by the gap thickness measurement uncertainty or the thickness variability caused by the inner surface roughness. The polished and unpolished devices had a gap thickness of $d$ = ${0.254}\;\textrm{mm}$ and $d$ = ${0.310}\;\textrm{mm}$, respectively. Again, in Figs. 4((a)-(d)) insets are shown in the frequency range from ${0.45}\;\textrm{THz}$ to ${0.55}\;\textrm{THz}$ for a more detailed representation of the data.
It is crucial, though, to assess the measurement uncertainty of the optical parameters (refractive index and absorption coefficient) while we use the 3D-printed microfluidic devices. The refractive index was extracted from a comparison of the transmission function of the filled microfluidic cell with the empty one, $H$exp, with the theoretical expression [20].
Multiple reflections caused by the microfluidic cell were suppressed by applying a Gaussian windowing function to the THz-TDS time signal, which is therefore ignored in Eq. (1). The complex refractive index $\tilde {n}(\omega )$ was numerically evaluated by equating the modulus and argument of the experimental transmission function with Eq. (1).
While these equations for $n$ and $\kappa$ can be numerically evaluated for each frequency $\omega$ individually, the results are prone to artifacts caused by the measurement uncertainty, especially close to the water vapor absorption lines. Taking the logarithm of Eq. (2), $\kappa$ is obtained from the slope:
For estimating the measurement uncertainty, we assume $\tilde {n}$ to be independent of $\omega$. Then, the attenuation coefficient $\kappa$ and its relative uncertainty are obtained from the slope of $\ln \left (\left \vert H_{\text {exp}}\right \vert \right )$, Fig. 5(a), and error propagation, respectively.
Fig. 5. (a) Transmission $ln\left (\left \vert H_{\text {exp}}\right \vert \right )$. (b) Phase of $\left \vert H_{\text {exp}}\right \vert$. (c) Relative uncertainty of $ln\left (\left \vert H_{\text {exp}}\right \vert \right )$. (d) Relative uncertainty of the phase. (e) Refractive index with uncertainty band. (f) Absorption with uncertainty band.
The measurement uncertainty of $\left \vert H_{\text {exp}}\right \vert$ was estimated by the standard uncertainty of five measurements of the empty and filled microfluidic cell. Its relative uncertainty was less than $5\%$ for frequencies below 1.5 THz (Fig. 5(c)). It turns out that the measurement uncertainty of the gap thickness is the dominant contribution.
The uncertainty of the slope $b$ of a function $y(x)$ is calculated from the (local) linear fit to the function. Using the local center of gravity coordinate system, the slope and its variance are given by [21].
where the weights $p_i$ are estimated by $1/u^2(y_i)$. Applying these relations to the slope of $\ln \left \vert H_{\text {exp}}\right \vert$ we obtain the first estimation of the attenuation coefficient and its uncertainty (see Figs. 5(a) & (b)).Similarly, from Eq. (3), the refractive index $n$ and its relative variance are obtained from the slope of $\arg \left (H_{\text {exp}}\right )$ and error propagation, respectively:
Again, the measurement uncertainty of the gap thickness is dominant.
In order to corroborate the results, polished microfluidic devices with different gap thicknesses, namely ${0.177}\;\textrm{mm}$ ${0.250}\;\textrm{mm}$ and ${0.440}\;\textrm{mm}$ were produced and compared as shown in Figs. 6(a& b). The gap thickness is defined by the spacer thickness (Fig. 1(a& b)). Various gap thicknesses were used to show the reproducibility of the measurements and, most importantly, to show that the 3D printed microfluidic devices can be used for the measurement of materials with high absorption, which requires a very small gap thickness. These results are reproducible, and in agreement with the water measurements reported by Afsar et al., [22]. The lower signal-to-noise ratio of the thickest device is due to the high water absorption. Finally, we assessed the accuracy of the devices for measuring lysozyme in solution compared with pure water. The measurements of these samples were conducted in a polished device with a gap thickness of ${0.254}\;\textrm{mm}$. The refractive index and the absorption are in agreement with the values reported by Aoki et al. which are also shown in Fig. 6(c)&(d) [23]. The difference between the measurement of lysozyme and the literature values can be attributed to the measurement uncertainty. As can be seen, the refractive index and the absorption of lysozyme solutions in Fig. 6(a)) and (b), are lower than the water values, which can be explained by the reduced water concentration in the solution with lysozyme.
Fig. 6. Optical parameters for water and lysozyme. (a) The refractive index of water from polished devices of different gap thickness (${0.177}\;\textrm{mm}$, ${0.250}\;\textrm{mm}$, and ${0.440}\;\textrm{mm}$) compared with the literature [22]. (b) The absorption coefficient of water from polished devices of different gap thicknesses. (c) The refractive index of lysozyme solution compared with water and literature values for lysozyme. (d) The absorption coefficient of the lysozyme solution in comparison with water and literature values for lysozyme.
4. Conclusion
We have shown that simple THz microfluidic devices can be easily designed and 3D printed with TOPAS polymer in less than four hours. We have also confirmed that polishing the devices does not affect the reproducibility of our results in the THz regime but can improve the transparency of the device in the VIS regime. The polished devices can find applications for the characterization of various proteins simultaneously in the visible range and the THz regime. Moreover, various gap thicknesses of microfluidic devices were fabricated, and the results of the different thicknesses were consistent with the measurement uncertainty. We have shown that the dominant parameter that affects the measurement uncertainty is the gap thickness of the device. Finally, we used the devices for the measurement of lysozyme, and the results are in agreement with previous works. We are convinced that the 3D printing of TOPAS is a cost-effective, fast and efficient production method also for more complex THz microfluidic devices. 3D printed microfluidic devices with TOPAS will contribute to the rapid characterization of various biological samples from THz to X-ray.
Funding
Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (CRSK-2_190426).
Acknowledgements
The authors would like to thank Peter Nirmalraj for his assistance in the process of writing the paper, and apprentices Nick Caceres, Sofie Gnannt, and Claudio Sommer for their work on 3D printing.
This project was financially supported by the project 3D printed terahertz waveguide spectrometer (3DTera-WaSp) under SNSF grant number$\:$ CRSK_2_190426.
Disclosures
The authors declare no conflicts of interest.
Data availability
All the experimental data that are presented in the figures are available from authors after request.
References
1. S. S. Dhillon, M. S. Vitiello, E. H. Linfield, et al., “The 2017 terahertz science and technology roadmap,” J. Phys. D: Appl. Phys. 50(4), 043001 (2017). [CrossRef]
2. C. Weisenstein, A. K. Wigger, M. Richter, R. Sczech, A. K. Bosserhoff, and P. H. Bolívar, “THz detection of biomolecules in aqueous environments-status and perspectives for analysis under physiological conditions and clinical use,” J. Infrared, Millimeter, Terahertz Waves 42(6), 607–646 (2021). [CrossRef]
3. X. Yang, X. Zhao, K. Yang, Y. Liu, Y. Liu, W. Fu, and Y. Luo, “Biomedical applications of terahertz spectroscopy and imaging,” Trends Biotechnol. 34(10), 810–824 (2016). [CrossRef]
4. A. K. Au, W. Huynh, L. F. Horowitz, and A. Folch, “3D-printed microfluidics,” Angew. Chem. Int. Ed. 55(12), 3862–3881 (2016). [CrossRef]
5. C. G. van Dijk, M. M. Brandt, N. Poulis, J. Anten, M. van der Moolen, L. Kramer, E. F. Homburg, L. Louzao-Martinez, J. Pei, M. M. Krebber, B. W. van Balkom, P. de Graaf, D. J. Duncker, M. C. Verhaar, R. Luttge, and C. Cheng, “A new microfluidic model that allows monitoring of complex vascular structures and cell interactions in a 3D biological matrix,” Lab Chip 20(10), 1827–1844 (2020). [CrossRef]
6. A. Abou-Hassan, O. Sandre, and V. Cabuil, “Microfluidics in inorganic chemistry,” Angewandte Chemie - Int. Ed. 49(36), 6268–6286 (2010). [CrossRef]
7. S. M. Scott and Z. Ali, “Fabrication methods for microfluidic devices: an overview,” Micromachines 12(3), 319 (2021). [CrossRef]
8. Y. He, Y. Wu, J. Z. Fu, Q. Gao, and J. J. Qiu, “Developments of 3D printing microfluidics and applications in chemistry and biology: a review,” Electroanalysis 28(8), 1658–1678 (2016). [CrossRef]
9. S. Schewa, M. A. Schroer, T. Zickmantel, Y. H. Song, C. E. Blanchet, A. Y. Gruzinov, G. Katona, D. I. Svergun, and M. Roessle, “A THz transparent 3D printed microfluidic cell for small angle X-ray scattering,” Rev. Sci. Instrum. 91(8), 084101 (2020). [CrossRef]
10. J. Sun and F. Hu, “Three-dimensional printing technologies for terahertz applications: a review,” Int. J. RF Microw. Comput. Eng. 30(1), e21983 (2020). [CrossRef]
11. S. F. Busch, M. Weidenbach, M. Fey, F. Schäfer, T. Probst, and M. Koch, “Optical properties of 3D printable plastics in the THz regime and their application for 3d printed THz optics,” J. Infrared, Millimeter, Terahertz Waves 35(12), 993–997 (2014). [CrossRef]
12. S. F. Busch, M. Weidenbach, J. C. Balzer, and M. Koch, “THz optics 3D printed with TOPAS,” J. Infrared, Millimeter, Terahertz Waves 37(4), 303–307 (2016). [CrossRef]
13. A. J. L. Adam, H. K. Rasmussen, K. Nielsen, O. Bang, P. C. M. Planken, and P. U. Jepsen, “Bendable, low-loss TOPAS fibers for the terahertz frequency range,” Opt. Express 17(10), 8592–8601 (2009). [CrossRef]
14. M. Bernard, E. Jubeli, J. Bakar, L. Tortolano, J. Saunier, and N. Yagoubi, “Biocompatibility assessment of cyclic olefin copolymers: impact of two additives on cytotoxicity, oxidative stress, inflammatory reactions, and hemocompatibility,” J. Biomed. Mater. Res. 105(12), 3333–3349 (2017). [CrossRef]
15. E. J. Wollack, K. L. Denis, A. Barlis, M.-P. Chang, A. S. Kutyrev, K. H. Miller, and P. C. Nagler, “Far-infrared properties of cyclic olefin copolymer,” Opt. Lett. 45(3), 780 (2020). [CrossRef]
16. P. D. Cunningham, N. N. Valdes, F. A. Vallejo, L. M. Hayden, B. Polishak, X.-H. Zhou, J. Luo, A. K.-Y. Jen, J. C. Williams, and R. J. Twieg, “Broadband terahertz characterization of the refractive index and absorption of some important polymeric and organic electro-optic materials,” J. Appl. Phys. 109(4), 043505 (2011). [CrossRef]
17. E. Mavrona, J. Graf, E. Hack, and P. Zolliker, “Optimized 3D printing of THz waveguides with cyclic olefin copolymer,” Opt. Mater. Express 11(8), 2495 (2021). [CrossRef]
18. C. Gras and S. J. Stanley, “Post-irradiation examination of a fuel pin using a microscopic X-ray system: measurement of carbon deposition and pin metrology,” Annals Nucl. Energy 35(5), 829–837 (2008). [CrossRef]
19. E. Mavrona, U. Chodorow, M. E. Barnes, J. Parka, N. Palka, S. Saitzek, J. F. Blach, V. Apostolopoulos, and M. Kaczmarek, “Refractive indices and birefringence of hybrid liquid crystal-nanoparticles composite materials in the terahertz region,” AIP Adv. 5(7), 077143 (2015). [CrossRef]
20. E. Mavrona, “Functionalised liquid crystals for manipulating terahertz radiation,” University of Southampton, Local EPrints ID: 419030, http://eprints.soton.ac.uk/id/eprint/419030 (2016).
21. P. R. Bevington, D. K. Robinson, J. M. Blair, A. J. Mallinckrodt, and S. McKay, “Data reduction and error analysis for the physical sciences,” Comput. Phys. 7(4), 415 (1993). [CrossRef]
22. J. B. Hasted and M. N. Afsar, “Measurements of the optical constants of liquid H2O and D2O between 6 and 450 -cm1,” J. Opt. Soc. Am. 67(7), 902–904 (1977). [CrossRef]
23. K. Aoki, K. Shiraki, and T. Hattori, “Observation of salt effects on hydration water of lysozyme in aqueous solution using terahertz time-domain spectroscopy,” Appl. Phys. Lett. 103(17), 173704 (2013). [CrossRef]
