Terahertz biomedical imaging using a terahertz emitter based on InAs nanowires

Abstract

We demonstrate the potential of using InAs nanowires (NWs) on Si(111) substrates for terahertz (THz) imaging in biomedical diagnostics. The cost-effectiveness and high precision of InAs NWs are discussed in comparison with those of InAs substrates. Using THz time-domain spectroscopy, the peak-to-peak current signals of catalyst-free InAs NWs and the substrate were measured at 3.37 nA and 1.61 nA, respectively. The THz amplitude of the InAs NWs increased by 47.7% compared to InAs wafers, with a fill factor of 0.034. At frequencies of up to 0.25 THz, the THz amplitude of InAs NWs was higher than that of bulk InAs. The resolution of THz images of a slice of pork belly and rat brain tumour tissue acquired using the InAs NWs was similar to that obtained with the InAs substrates. This suggests that InAs NW-based THz devices on Si substrates, which allow for large-area and cost-effective fabrication, could be useful for biomedical diagnostics.

---

Introduction

Brain tumour imaging, in particular, represents a promising application for terahertz (THz) imaging technology, which has the potential to overcome the limitations of conventional medical imaging techniques and provide groundbreaking benefits in clinical practice.^1–5 Glioblastoma, the most common primary brain tumour of the central nervous system, grows invasively and has unclear margins between the neoplastic and normal regions.⁶ Conventional medical imaging techniques, such as neuronavigation, intraoperative magnetic resonance imaging (ioMRI), positron emission tomography (PET), pyroelectric cameras, and microbolometers, have been employed over the past decade to delineate tumour boundaries during surgery.^7,8 However, these methods face challenges such as being time-consuming, costly, exposing patients to radiation, and having low sensitivity at low frequencies, which collectively hinder real-time delineation of tumour boundaries in surgery and pose a fundamental limitation.^9,10 Previously, we reported that THz reflectometry imaging (TRI) could not only detect low-grade and grade III tumours with an intact blood–brain barrier but also delineate the tumour margin, which is rarely observed by white-light microscopy, in human fresh tissues.¹¹ Canon Inc. reported a method to delineate tumour regions in rat brain tissue using the complex refractive index and principal component analysis in the THz region.¹² Prior THz imaging techniques were implemented using photoconductive antenna emitters fabricated on III–V compound semiconductors such as GaAs or InGaAs wafers grown at low temperatures, which were not price-competitive and could not be fabricated easily. For the practical application of THz techniques in brain tumour surgery, it is essential to develop either handheld miniature probes or large-area THz imaging systems, coupled with cost-effective generation devices. InAs can be a suitable material for a THz emitter due to its high absorption coefficient, high saturation velocity, short carrier lifetime, and appropriate narrow energy bandgap. In addition, the unique features of a nanowire (NW) structure, such as the high aspect ratio, extensive surface area, and growth on hetero-material substrates, are expected to play a key role in the development of very efficient semiconductor emitters in the THz spectral range.^13–20 Considering these points, InAs NW-based devices fabricated on Si could be a promising approach. In our previous study, we reported that Si-based InAs NWs fabricated using molecular beam epitaxy (MBE) could effectively serve as THz generation devices.^21,22 However, MBE processes remain expensive, posing economic limitations for high-throughput production or large-area growth. Metal–organic chemical vapor deposition (MOCVD) offers a potential solution to these challenges such as high-throughput production and large-area growth. Si-based InAs NWs grown using MOCVD could be fabricated on substrates with diameters of up to 10 inches (25 cm), providing a cost-effective alternative to InAs substrates. These NWs could realize large-area sources for THz medical imaging as well as micron-scale THz probes utilizing a single NW for THz endoscopy. Additionally, they can be seamlessly integrated into Si-based platforms for a wide range of applications.²³

In this paper, we showed the feasibility of THz biomedical diagnosis using THz time-domain spectroscopic imaging with a generator prepared with Si-based InAs NWs grown by MOCVD. The THz pulse-emission performances of InAs NWs were compared with those of bulk InAs wafers using THz emission spectroscopy as a cost-effective medium with high precision for biomedical applications. Using the THz imaging system based on InAs NWs, THz images of a slice of pork belly and paraffin-embedded brain tumours were successfully obtained and analyzed for peak signals at each frequency. The THz images of the brain tumours were compared with the images stained with hematoxylin and eosin (H&E) to distinguish cancerous tissue from non-cancerous tissue.

Results and discussion

Catalyst-free InAs nanowires (NWs) were grown on p-type Si(111) substrates using metal–organic chemical vapor deposition (MOCVD, AIXTRON Inc.) with a horizontal reactor. The Si(111) substrates were chemically cleaned using a standard wet-etching process involving acetone, methanol, isopropyl alcohol, and deionised water (DI water). In preparation for the self-assembled growth of InAs NWs, the p-type Si(111) substrate was etched with HF to remove the native oxide, rinsed in DI water for 5 seconds, and then dried with a nitrogen gun. The substrate was subsequently immersed in a poly-L-lysine (PLL) solution (Sigma-Aldrich Inc.) for 3 minutes, rinsed in DI water for 10 seconds, and dried with the nitrogen gun. The positively charged PLL layer attracts the negatively charged arsenic, which facilitates the subsequent reaction for InAs NW growth. The Si substrate was then immediately loaded into the MOCVD chamber, where the InAs NWs were grown at 570 °C for 3 hours. Fig. 1a and b present cross-sectional images of the catalyst-free InAs NWs on the Si(111) substrate, acquired using field-emission scanning electron microscopy (FE-SEM, Hitachi S-4800). Longitudinally grown InAs nanowires experience bending due to strain and merge with neighbouring nanowires. Fig. 1c shows the height distribution of the InAs NWs, with an average height of 25 ± 5 μm and an average diameter of 220 ± 30 nm. The standard deviation of the height of the InAs NWs was calculated to be 3.04 μm. The spatial density of the NWs was estimated to be 1 × 10⁸ cm⁻². The internal structural properties of the InAs NWs were analysed using Cs-corrected field emission transmission electron microscopy (Cs-TEM, Jem-arm200f). Fig. 1d displays a Cs-TEM image and the electron diffraction pattern of the InAs NWs, which indicates the predominant crystal structure to be zinc blende with stacking faults. Catalyst-free III–V compound semiconductor nanowires typically have zinc blende or wurtzite crystal structures with stacking faults. In the Volmer–Weber growth mode, because the interactions between adatoms are stronger than those between the adatom and the substrate surface, InAs manifests as clusters or islands on the Si substrates. Although the zinc blende structure of bulk InAs is stable, nanowires often exhibit the wurtzite structure because of their lower surface energy, which destabilises the bonding energy and gives rise to the occurrence of stacking faults.

Fig. 1 Cross-sectional FE-SEM of InAs NWs: (a) side view (inset scale bar: 500 nm) and (b) tilted view. (c) Height distribution of the InAs NWs and (d) Cs-TEM image and diffraction pattern of InAs NWs.

Fig. 2a and b illustrate the behaviour of carriers when a pulsed laser is irradiated onto the InAs substrate and nanowires. The THz wave radiation incident on the samples was measured using THz time-domain spectroscopy (THz TDS). This was accomplished using a mode-locked Ti:sapphire laser (Spectra-Physics) with a pulse duration of 80 fs and a central wavelength of 800 nm. The beam was 1 mm in diameter and the incidence angle was 45°. The number of InAs NWs excited by the laser beam was calculated to be approximately 1.09 × 10⁶. The THz signals were measured using a low-temperature grown GaAs receiver with a dipole antenna under a laser power of 11 mW. The TDS current signals of the InAs NWs were converted into frequency spectra by fast Fourier transform (FFT). Fig. 2c shows the THz current signals detected with a fast delay stage for the InAs NWs and a commercial undoped InAs substrate (Wafer Tech). The peak-to-peak current signals (PPCSs) for the InAs NWs (3.37 nA) were 47.7% more intense than those of the substrate (1.61 nA). The fill factor of the InAs NWs was calculated to be 0.034 compared to that of the InAs substrate. The laser area received by the InAs substrate was 8.64 × 10⁻³ cm², whereas if all NWs were to receive the laser, the area would be 6.54 × 10⁻² cm², which is about 7.57 times larger. Fig. 2d shows the THz spectra of the InAs NWs and substrate, which were generated by subjecting the TDS current signals to FFT. The inset is the expanded plot at the frequency range from 0 to 0.4 THz. The intensities of the signals received from the InAs NWs were as much as 0.3 THz higher than those from the InAs substrate. This stronger signal is attributed to the confinement of electrons and holes to one dimension, resulting in an efficient photo–Dember effect. However, due to the lower electron mobility of InAs nanowires (about 1000 cm² V⁻¹ s⁻¹) compared to the undoped InAs substrate (about 20 000 cm² V⁻¹ s⁻¹), the signal strength beyond 0.3 THz was lower for the nanowires. The cutoff frequency for both samples was approximately 3 THz, with the nanowires having a slightly shorter cutoff compared to the substrate. The noise levels for both the InAs nanowires and the substrate were found to be nearly identical. Applications such as biomedical analysis and imaging systems require more reliable and reproducible characteristic curves for THz wave radiation in the ∼0.3 THz frequency range. Previous reports indicated that THz devices with InAs nanowires have a narrower emission range compared to bulk InAs. Nevertheless, the THz spectrum of InAs nanowires is adequate for biomedical analysis, such as in vivo and ex vivo diagnoses around ∼0.3 THz, with a relatively larger effective contact area compared to bulk InAs. Additionally, the nanowires have a larger contact area with biomedical specimens owing to the additional side-wall contact, which increases the reliability of the information. From an economic perspective, Si-based InAs nanowires offer significant advantages over bulk InAs for commercial applications. The use of Si-based InAs nanowires in THz devices can provide single-use indicators for biomedical specimens.

Fig. 2 Schematic diagram of the photo-Dember effect in (a) the InAs substrate and (b) InAs NWs. (c) THz current signal of InAs NWs and the InAs substrate and (d) their THz spectra, obtained by applying FFT to the current signals, with the inset showing the expanded plot in the frequency range from 0 to 0.4 THz.

We evaluated the THz generation efficiency of InAs NWs and InAs substrates by comparing the extent to which their THz generation depends on the input laser power. Fig. 3a and b compare the excitation dependence of THz PPCSs obtained using InAs NWs and a commercial undoped InAs substrate by varying the laser power from 1 mW to 180 mW. The signals generated by both the InAs NWs and InAs substrates increase proportionally with the input laser power. Fig. 3c presents the THz PPCSs as a function of the excitation power. At a laser power of 1 mW, the output power of the InAs NWs is approximately 22% lower than that of the substrate because of the difference in the number of carriers generated. At a power of 180 mW, excessive carrier generation on the substrate hampers the efficient movement of electrons. In contrast, the output power of the InAs NWs, with their confined electron movement, is up to 40% higher compared to that of the substrate. This efficiency is attributed to the 1D electron movement in the NWs. Analysis of the THz frequency domain using FFT (Fig. 3d and e) revealed a cutoff frequency of 5 THz for the InAs NWs at input power levels above 150 mW, whereas the InAs substrate reached the same cutoff frequency at input power levels above 30 mW. The InAs NWs maintain their signal intensity up to ∼0.5 THz, after which the signal intensity declines due to the lower electron mobility caused by stacking faults and because electron movement is limited to the length of the NWs of 20–30 μm.

Fig. 3 THz current signals and frequency domains from the FFT of (a), (d) InAs NWs and (b) and (e) InAs substrate, and (c) PPCS and (f) amplitude at 0.25 THz with excitation laser power dependence.

The InAs substrate produces strong signals up to 2 THz based on its higher charge mobility of 20 000 cm² V⁻¹ s⁻¹ and greater number of generated charges, whereas at frequencies higher than this the signal intensity decreases. At 0.25 THz, the amplitude of the InAs NWs signals is higher than that of the InAs substrate at low input power. Fig. 3f shows the amplitude as a function of the input power at 0.25 THz. Despite the lower number of generated charges and relatively lower charge mobility, the efficient 1D movement in the InAs NWs increases the signal amplitude. However, once the input power exceeds 60 mW, the difference in the generated charge becomes significant, and lowers the signal amplitude compared to that of the InAs substrate. Even at 180 mW, the InAs NWs still maintain nearly 86% of the amplitude of the InAs substrate, despite the PPCS being 40% lower at 0.25 THz. These results indicate that, although the InAs NWs are less efficient at higher frequencies due to their lower electron mobility, they remain a good alternative to InAs substrates at lower frequencies.

Fig. 4a shows a photograph of pork belly, which was imaged using THz imaging with both the InAs substrates and InAs NWs (Fig. 4b). The THz imaging system was used in reflection mode, with the pulse laser beam size being 1 mm and the incidence angle set at 75°. Considering the incidence angle and beam size, the number of InAs NWs excited by the laser beam was estimated to be approximately 8.63 × 10⁵. A mode-locked Ti:sapphire laser with a central wavelength of 800 nm and a pulse width of 80 fs was employed to generate and detect THz pulses. The detector was a low-temperature grown GaAs photoconductive antenna (PCA). Both InAs substrates and nanowires were able to clearly differentiate between the fat and lean meat of the pork belly in the THz images. Generally, for pork belly, the refractive index of lean meat is 1.52–1.55, and that of fat is 1.46–1.48. The significant difference in refractive indices is attributable to the higher water content of lean meat (75%) compared to fat (20%). For samples such as pork belly, the components of which have a large difference in their water content, the difference between the InAs substrate and NWs is insignificant. Images based on the PPCS are well resolved owing to the strong THz power resulting from the combination of various frequencies from low to high. One of the main advantages of THz technology is its ability to obtain a dielectric constant and phase information. However, extracting them from the reflection signal requires precise experiments, making THz devices difficult to apply in practical clinical fields. In addition, the signals in the frequency band from 0.2 to 0.6 THz, which contains most of the energy, are significantly important. Analysis of only the 0.5 THz and 0.25 THz bands using FFT revealed that the resolution of the 0.5 THz image was similar to that of the PPCS image. However, the 0.25 THz images were less well resolved because the signal amplitude and wavelength presented problems compared to the PPCS and 0.5 THz images. Despite these problems, the difference between the refractive indices of fat and lean meat enabled a clear distinction to be made. Considering that InAs NWs can be grown on silicon substrates and that these substrates are available in sizes up to 12 inches (300 mm) in diameter, InAs NWs offer a cost advantage and larger area production capability compared to the more expensive InAs substrates, which are not suitable for large-area production. For samples with significant differences in their refractive indices, InAs NWs, being relatively less expensive and exhibiting similar characteristics, can serve as a viable alternative to InAs substrates. These findings provide a pivotal demonstration, visually elucidating the reflection mechanisms in biological tissues based on the refractive index and absorption rates of water and lipids using InAs NWs.

Fig. 4 (a) Photograph of pork belly and (b) THz images using the InAs substrate and InAs NWs of pork belly.

Fig. 5a and b show the brain tissue of a rat with part of a tumour after formalin fixation and paraffin embedding, and the H&E-stained images, respectively. Fig. 5c presents the THz images obtained using the InAs substrates and NWs. The PPCS images allowed us to distinguish the brain tumour (indicated by red areas) which is not visualised in the visible light photographs. Unlike the pork belly, which consists of parts where the water content differs significantly, the difference between normal and abnormal (e.g., cancerous) cells is less pronounced. However, abnormal tissues such as tumours are clearly identified in the THz images. For images acquired in the time domain, all frequencies are mixed, and thus the THz images obtained from InAs substrates and NWs do not differ greatly. At 0.5 THz, the resolution of the images obtained using the InAs NWs is lower. This is because, as is evident from the THz TDS results, InAs substrates maintain an intense signal over a wide frequency range from 0.5 to 2 THz, whereas the signal of the InAs NWs reaches its maximum intensity at 0.25 THz with a rapid decrease thereafter. At 0.25 THz, the resolution is affected by the wavelength and low signal strength, yet the resolution of the THz images obtained with the InAs NWs is relatively superior to those acquired with the InAs substrates. This is due to the more efficient carrier dynamics of the InAs NWs at the pulse laser power used in this imaging system. The results indicate that, although the image quality of InAs NWs is relatively poorer than that of InAs substrates in the frequency range higher than 0.5 THz due to the differences in carrier mobility, the quality of the PPCS images is similar to that of the images obtained at 0.25 THz. These results indicate a significant achievement, showcasing reflection imaging of paraffin-embedded biological tissues using InAs NWs, effectively isolating the influence of water and clearly revealing variations in tissue density and composition.

Fig. 5 (a) Images of part of a tumour and (b) the surrounding brain tissues of a rat. (c) THz images using the InAs substrate and InAs NWs.

Conclusions

We demonstrated a cost-effective approach towards THz emission using InAs NWs, making them a viable alternative to InAs substrates in certain THz applications. InAs NWs were grown on inexpensive Si(111) substrates using MOCVD. The deviations in the length and width of InAs NWs were 25 ± 5 μm and 220 ± 30 nm, respectively, and the density of the InAs NWs was 1 × 10⁸ cm⁻² with a standard deviation of 3.04 μm for the length. Evaluation of the THz generation characteristics using THz TDS, despite a fill factor of 0.034, revealed that the PPCSs of InAs NWs were approximately 47.7% of those of InAs substrates due to efficient charge transport. In imaging experiments, when measuring objects with significant differences in their refractive indices, such as the parts of pork belly, both the InAs NWs and substrates clearly distinguished the boundaries between lean meat and fat. Additionally, with respect to the diagnosis of rat brain cancer, despite the relatively small difference in the water content between cancerous and non-cancerous tissue compared to that between fat and lean meat, the PPCS performance was similar. Although the signal generation decreases rapidly above 0.5 THz for InAs NWs, they can be grown on 12-inch silicon wafers and could serve as a cost-effective alternative to InAs substrates for large-area THz generation, especially for targets around 0.25 THz.

Author contributions

S. J. O. and J. S. K. designed the research project and supervised the experiments. D. W. P., Y. B. J., I. C., J. H., J. C. S., J. C. C., E. S. L., S. N. and J. S. performed the experiments and analysed the data. D. W. P., Y. B. J., I. C., S. J. O. and J. S. K. wrote the paper, which was discussed by all the authors.

Data availability

Data that support the findings of this study are available from the corresponding author (Jin Soo Kim) upon reasonable request.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the Institute of Information & Communications Technology Planning & Evaluation (IITP) grant funded by the Korean government (MSIT) (RS-2022-II221044), the Nano & Material Technology Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (RS-2024-00411969), and the Global-Learning & Academic Research Institution for Master's·& PhD students, and Postdocs (LAMP) Program of the NRF grant funded by the Ministry of Education (No. RS-2024-00443714).

References

1. Z. Li, Y. Cheng, H. Luo, F. Chen and X. Li, Dual-band tunable terahertz perfect absorber based on all-dielectric InSb resonator structure for sensing application, J. Alloys Compd., 2022, 925, 166617.
2. C. Hou, Z. Xu, W. Qiu, R. Wu, Y. Wang, Q. Xu, X. Y. Liu and W. Guo, A biodegradable and stretchable protein–based sensor as artificial electronic skin for human motion detection, Small, 2019, 15, 1805084.
3. M. Wan, J. J. Healy and J. T. Sheridan, Terahertz phase imaging and biomedical applications, Opt. Laser Technol., 2020, 122, 105859.
4. A. Gong, Y. Qiu, X. Chen, Z. Zhao, L. Xia and Y. Shao, Biomedical applications of terahertz technology, Appl. Spectrosc. Rev., 2020, 55, 418–438.
5. G. Lee, J. Lee, Q.-H. Park and M. Seo, Frontiers in terahertz imaging applications beyond absorption cross-section and diffraction limits, ACS Photonics, 2022, 9, 1500–1512.
6. Q. T. Ostrom, L. Bauchet, F. G. Davis, I. Deltour, J. L. Fisher, C. E. Langer, M. Pekmezci, J. A. Schwartzbaum, M. C. Turner and K. M. Walsh, The epidemiology of glioma in adults: a “state of the science” review, Neurooncology, 2014, 16, 896–913.
7. Y. T. Lo, H. Lee, C. Shui, N. Lamba, R. Korde, S. Devi, S. Chawla, Y. Nam, R. Patel, J. Doucette, A. Bunevicius and R. A. Mekary, Intraoperative magnetic resonance imaging for low-grade and high-grade gliomas: What is the evidence? A meta-analysis, World Neurosurgery, 2021, 149, 232–243.
8. J. Rong, A. Haider, T. E. Jeppesen, L. Josephson and S. H. Liang, Radiochemistry for positron emission tomography, Nat. Coummun., 2023, 14, 3257.
9. A. Banerjee, C. Chakraborty and M. Rathi, Medical imaging, artificial intelligence, internet of things, wearable devices in terahertz healthcare technologies, Terahertz biomedical and healthcare technologies, Elsevier, 2020, pp. 145–165.
10. L. Hou, J. Rong, A. Haider, D. Ogasawara, C. Varlow, M. A. Schafroth, L. Mu, J. Gan, H. Xu and C. J. Fowler, Positron emission tomography imaging of the endocannabinoid system: opportunities and challenges in radiotracer development, J. Med. Chem., 2020, 64, 123–149.
11. Y. B. Ji, S. J. Oh, S.-G. Kang, J. Heo, S.-H. Kim, Y. Choi, S. Song, H. Y. Son, S. H. Kim and J. H. Lee, Terahertz reflectometry imaging for low and high grade gliomas, Sci. Rep., 2016, 6, 36040.
12. S. Yamaguchi, Y. Fukushi, O. Kubota, T. Itsuji, T. Ouchi and S. Yamamoto, Brain tumor imaging of rat fresh tissue using terahertz spectroscopy, Sci. Rep., 2016, 6, 30124.
13. J. Shin, S. Noh, J. Lee, J. Oh, M.-Y. Ryu and J. S. Kim, Structural and optical properties of InN nanowires formed on Si(111), Appl. Sci. Converg. Technol., 2022, 31, 141–144.
14. L. Balaghi, S. Shan, I. Fotev, F. Moebus, R. Rana, T. Venanzi, R. Hübner, T. Mikolajick, H. Schneider and M. Helm, High electron mobility in strained GaAs nanowires, Nat. Commun., 2021, 12, 6642.
15. M. Marnauza, M. Tornberg, E. K. Mårtensson, D. Jacobsson and K. A. Dick, In situ observations of size effects in GaAs nanowire growth, Nanoscale Horiz., 2023, 8, 291–296.
16. S. Noh, S. Han, J. Shin, J. Lee, I. Choi, H. M. Oh, M.-Y. Ryu and J. S. Kim, Photoelectrochemical water splitting using GaN nanowires with reverse-mesa structures as photoanode material, Appl. Sci. Convergence Technol., 2022, 31, 51–55.
17. D. W. Park, S. G. Jeon, C.-R. Lee, S. J. Lee, J. Y. Song, J. O. Kim, S. K. Noh, J.-Y. Leem and J. S. Kim, Structural and electrical properties of catalyst-free Si-doped InAs nanowires formed on Si(111), Sci. Rep., 2015, 5, 16652.
18. J. Seaberg, H. Montazerian, M. N. Hossen, R. Bhattacharya, A. Khademhosseini and P. Mukherjee, Hybrid nanosystems for biomedical applications, ACS Nano, 2021, 15, 2099–2142.
19. S. Chen, A. van den Berg and E. T. Carlen, Sensitivity and detection limit analysis of silicon nanowire bio(chemical) sensors, Sens. Actuators, B, 2015, 209, 486–489.
20. A. F. Smith, X. Liu, T. L. Woodard, T. Fu, T. Emrick, J. M. Jiménez, D. R. Lovley and J. Yao, Bioelectronic protein nanowire sensors for ammonia detection, Nano Res., 2020, 13, 1479–1484.
21. K. Peng, D. Jevtics, F. Zhang, S. Sterzl, D. A. Damry, M. U. Rothmann, B. Guilhabert, M. J. Strain, H. H. Tan and L. M. Herz, Three-dimensional cross-nanowire networks recover full terahertz state, Science, 2020, 368, 510–513.
22. A. Arlauskas, J. Treu, K. Saller, I. Beleckaite, G. Koblmüller and A. Krotkus, Strong terahertz emission and its origin from catalyst-free InAs nanowire arrays, Nano Lett., 2014, 14, 1508–1514.
23. J. H. Shin, H. Ryu, Y. B. Ji, S. J. Oh and W. Lee, Anodically induced chemical etching of GaAs wafers for a GaAs nanowire-based flexible terahertz wave emitter, ACS Appl. Mater. Interfaces, 2020, 12, 50703–50712.

---

Footnote

† These authors contributed equally to this work.

---