As mentioned in the previous section, THz waves have fingerprint spectrum, so the terahertz spectroscopy of various biomolecules is characteristic and can be relied upon for the identification of various biomolecules. Additionally, the non-ionizing nature of THz radiation makes it possible to detect biological tissues non-invasively. THz waves are strongly absorbed by polar molecules such as water, while biomolecules have limited absorption, so the THz spectrum can be used to distinguish the water content of biological tissues to assess their status. This, coupled with the penetrating of THz and its relatively high resolution, suggests that terahertz spectroscopy has promising applications in the biomedical field [3].

As early as 2002, Ferguson B et al. used THz-TDS to record terahertz spectroscopy of the four nucleobases [adenine (A), cytosine (C), guanine (G) and thymine (T)] and the corresponding nucleosides that form the deoxyribonucleic acid (DNA) building blocks, as shown in Figure 3A,B [40]. As can be seen, the spectrum of each molecule consists of a series of resonances distributed between 1 and 3.5 THz. An absorption peak is observed at each resonance, and the characteristic change is related to the refractive index. In addition, due to the reduction of the bond length in the system at low temperatures, the position of the bands usually shifts towards high frequencies when cooling. Globus T et al. measured the spectroscopy of DNA samples in liquid phase and could distinguish the spectral patterns for native and denatured DNA [41]. Zhang W et al. measured the THz absorption fingerprints of DNA in microliter DNA solution [17], as shown in Figure 3C,D. Then, they clarified that the fingerprint characteristic comes from DNA locally cross-linked by hydrogen bond. In 2013, Tsurkan MV et al. studied the THz spectra of herring degradation DNA using a femtosecond laser and 2 THz spectroscopy methods based on Backward-Wave Oscillator (BWO) and a high Q cavity frequency synthesizer, and showed that herring DNA has characteristic absorption peaks at 306 GHz, 339 GHz, and 375 GHz, providing a method for more accurate analysis of DNA THz spectra [42]. In 2015, Tang M et al. applied terahertz spectroscopy to the detection of DNA mutations, and their method successfully performed label-free analysis of single-base mutations in DNA molecules [43]. Besides, terahertz spectroscopy can be used to detect amino acids [44–46], proteins [47–49], and other biomolecules with good results.

Carcinogenesis causes chemical and structural abnormalities in biomolecules, one of which is widely known to be abnormal methylation of DNA. In 2016, Cheon H et al. proposed that molecular resonances could be monitored using THz-TDS in aqueous solutions of genomic DNA from cancer cell lines [50]. The quantification of resonance signals led to the identification of cancer cell types with some degree of DNA methylation. It was shown that identical resonance peaks at 1.67 THz were found in genomic DNA from various cancer types (PC3, A431, A549, MCF-7, and SNU-1 cell lines), and the THz resonance peaks for DNA samples after baseline correction are shown in Figure 3E. The magnitude of the resonance peaks for each sample was quantified by the amount of genomic DNA methylation, the degree of DNA methylation for each DNA sample is shown in Figure 3F. This demonstrates that cancer cell DNA also has THz fingerprinting spectra and provides a way to diagnose cancer at the molecular level. Shortly after, Cheon H et al. applied this method to the detection of blood cancers [51], using THz-TDS to observe characteristic resonances of DNA in blood cancers at a frequency of about 1.7 THz, and used this to analyze their methylation levels. Subsequently, they applied THz radiation to the treatment of cancer [52].

Terahertz spectroscopy has been applied to human tissue scanning and cancer diagnosis since around 2000. Woodward RM et al. used the absorption characteristics of THz waves for polar molecules such as water [16], and they distinguished between carcinoma diseased tissue and normal tissue of human basal cell by terahertz spectroscopy. Compared to normal tissue, basal cell carcinoma shows positive THz contrast, while inflammatory and scar tissue shows negative THz contrast, whereby THz images can be further drawn, which can help to better diagnose diseases such as cancer. In 2006, Wallace V further studied this and explained how parameters related to reflected THz pulses provide information about the lateral spread of tumors [53]. In 2009, Ashworth PC et al. used terahertz spectroscopy for the identification of human breast cancer tissue and found cancerous samples with high refractive indices and absorption coefficients in the 0.15–2.0 THz range [54]. In 2014, Hou D et al. used terahertz spectroscopy to differentiate between normal and cancerous gastric tissues, with significantly different shapes and amplitudes between the absorption spectra of the two, and a clustering approach was used to achieve automatic identification of gastric cancer [55].

In 2014, Meng K et al. found that paraffin-embedded gliomas had higher refractive index, absorption coefficient and dielectric constant than normal brain tissue under the terahertz spectroscopy [56]. In 2019, Gavdush AA et al. observed differences in THz spectra of intact tissue and gliomas of grades I to IV, reflecting the potential of terahertz spectroscopy in the intraoperative label-free diagnosis of human brain gliomas. The experimental setup is shown in Figure 4A [57]. In 2021, they further proposed physical models describing the THz permittivity of healthy and pathological brain tissue, on the basis of the Double Debye (DD) and Double Overdamped (DO) oscillator models [19]. The DO model and summation rule were applied ex vivo to estimate water content in intact tissues and gliomas, and the observed results are in good agreement with previously reported data demonstrating that water is a major endogenous label for brain tumors in the THz range. Besides, terahertz spectroscopy is also used in the examination of human colon tissue [58, 59], oral tissue [60], pigmented skin nevi [61], liver tissue [62] and related cancers.

Considering the differences between cancer patients, THz imaging technologies were employed to aid the diagnosis of cancer. For example, Bowman TC et al. in 2015 used THz pulse imaging to image and analyze heterogeneous breast cancer tissue [63]. They proposed a THz imaging method capable of distinguishing heterogeneous regions of breast tumors, and the results were compared and validated with standard histopathology images. In addition, THz imaging has also been applied to the diagnosis of colon cancer [64, 65], gastric cancer [66], skin cancer [67], liver cancer [30], etc. With the popularity of artificial intelligence in recent years, artificial intelligence techniques such as machine learning have also been used to diagnose cancer by combining with terahertz spectroscopy and imaging techniques [68–70]. Recently, Liu W et al. devised an automatic recognition strategy for THz pulse signals in breast invasive ductal carcinoma (IDC) based on wavelet entropy feature extraction and machine learning classifier [69]. Using principal component analysis method and machine learning classifier to automatically classify THz signals from breast IDC samples, the accuracy, sensitivity, and specificity of breast IDC recognition can reach 92.85%, 89.66%, and 96.67%, respectively.

In addition to diagnosing cancer, terahertz spectroscopy can also be used for the diagnosis of diseases such as diabetes. As early as the beginning of this century, researchers have used terahertz spectroscopy in dental care [71–74], for example in the detection of dental caries, where there is a higher attenuation of THz radiation in decayed enamel than that in healthy enamel. In 2010, Sy S et al. investigated in detail the correlation between THz properties of liver tissue, water content, structural changes, and cirrhosis, showing significant differences between the THz properties of normal and cirrhotic tissues [75]. Recently, Lykina AA et al. used terahertz spectroscopy to study plasma samples from diabetic and non-diabetic patients hoping to complete the assay nondestructively [76, 77]. The results showed that the normalized refractive index of particles from diabetic samples in the range of 0.2–1.4 THz increased by an average of 9–12% compared to particles from non-diabetic samples, and their terahertz spectroscopy were also spatially separated in principal component space. Detection of corneal edema can prevent corneal disease from reaching its end stage, and Ke L et al. assessed the degree of corneal edema by means of high-sensitivity THz broadband spectroscopy to detect corneal components at different depths. The process of THz signals acquired from the corneal surface and corneal subsurface is illustrated in Figure 4B [78].

Controlling the content of pesticide residues and harmful substances in food is an important measurement to ensure food safety. This is also an important application of terahertz spectroscopy in the agricultural field. As early as 2010, Hua Y et al. carried out research on the application of terahertz spectroscopy to detect the pesticide residues [79]. They used THz-TDS at frequencies of 0.5–1.5 THz for cyfluthrin n-hexane solutions in the concentration range of 1–20 mg/ml, and proposed THz-TDS plus partial least squares (PLS) and nonlinear regression methods of simple least-squares support vector machine (LS-SVM) is an effective tool for the quantitative analysis of cyfluthrin n-hexane solutions. Then, Hua Y et al. investigated four pesticides and three food powders as well as polyethylene using THz-TDS in the frequency range of 0.5–1.6 THz [20], and used PLS regression for the detection and analysis of imidacloprid in polyethylene and glutinous rice powder at different weight ratios. The results of the study showed that the four pesticides could be distinguished from each other by their unique absorption peaks, as well as from food powders that did not have this feature. In addition, differences in refractive index between pesticides and food powders can be observed. Imidacloprid can be recognized by its absorption fingerprints in mixtures of imidacloprid and polyethylene and imidacloprid and glutinous rice flour, as shown in Figure 5A,B. Through the linear relationship between the weight ratio of imidacloprid and the absorption coefficient, the authors predicted the weight ratio of imidacloprid in the above two mixtures of powders and achieved a relative error of less than 5%.

Subsequently, more and more researchers applied terahertz spectroscopy to detect pesticides, insecticides, and fungicides residues. Wang Q et al. conducted a study on nitrofen [80], and then measured the THz spectrum of thiabendazole at room temperature [81]. Maeng I et al. used terahertz spectroscopy to detect seven different pesticide residues in wheat flour [82]. Baek SH et al. used THz-TDS to detect methomyl (a carbamate insecticide) in food, and the characteristic absorption peaks of methomyl measured at room temperature in the frequency range of 0.1–3 THz were 1, 1.64, and 1.89 THz as shown in Figure 5C [83]. The three absorption peaks of methomyl are linearly proportional to their concentrations in high-density polyethylene. The peak at 1 THz was used as the fingerprint for detecting methomyl in rice and wheat flours. Figure 5D shows the calibration curve for methomyl which showed a regression coefficient >0.957 and a detection limit <3.74%. Accuracy and precision expressed as recovery and relative standard deviation (RSD, %) in interday repeatability were in the ranges 78.0–96.5 and 2.83–4.98%, respectively. Although THz-TDS can be used to rapidly detect methomyl in food, its sensitivity still needs to be improved. Qin B et al. applied THz-TDS combined with fast search and find of density peaks (CFSFDP) clustering to detect carbendazim residues [21]. Zhang H et al. combined terahertz spectroscopy with an improved PLS method for the detection of the harmful additives Auramine O in the medicinal herb Pollen Typhae [84]. Liu W et al. applied machine learning to terahertz spectroscopy to detect aflatoxin B1 in soybean oil, and obtained an accuracy rate of over 90% [85]. Recently, Ma Q et al. combined terahertz spectroscopy with a back-propagation neural network (BPNN) for the quantitative analysis of ternary pesticide mixtures in wheat flour [23], and obtained good results with correlation coefficients of 0.9913, 0.9948, and 0.9923 for the prediction sets, the corresponding root mean square errors of 0.0211%, 0.0176%, and 0.0191%, respectively.

The application of terahertz spectroscopy in the agricultural field is far more than the detection of pesticide residues and harmful substances, but also in the identification of agricultural products [86–88], the detection of soil [89–91] and the detection of genetically modified crops [92–94]. More brand new applications are being explored.

In 2003, Campbell MB et al. proposed that terahertz spectroscopy could be used to non-destructively detect explosives and other weapons of mass destruction [95], and demonstrated that a THz spectral database that includes biological warfare agent (BWA) simulants, chemical warfare agent (CWA) simulants, explosives, pharmaceuticals, dozens of potential hoax materials as well as several types of papers and plastics used in common containers that may hold threat agents. Chen Y et al. in 2004 used Fourier transform infrared spectroscopy (FTIR) and THz-TDS to measure 14 commonly used explosive samples [96], and the far-infrared spectra of the explosives 2, 3-dinitro toluene (2,4-DNT), 4-nitro toluene (4NT), and 2, 6-dinitro toluene (2,6-DNT) measured via FTIR and THz were in good agreement with their structure and vibrational frequencies obtained using density functional theory (DFT) [97, 98]. In 2005, they applied THz diffuse reflectance spectroscopy (DRS) to the detection of explosives [24]. The experimental results show that DRS technology has higher sensitivity and easier sample preparation than transmission spectroscopy. In 2006, they obtained the absorption spectrum of the explosive hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX) from the reflectance spectrum [99], and were able to distinguish RDX from other materials. In 2007, they used 0.1–2.8 THz THz-TDS to study the absorption spectrum of 17 explosives [100]. Most of the explosives show characteristic absorption, which can be used as a basis to build a relevant database. Moreover, they also considered that explosives are usually hidden, so they studied the absorption spectrum of explosives under the cover material to prove that the THz system can be used to detect hidden explosives. The experimental results show that the transmission spectra of RDX and tetramethylene tetranitramine (HMX) samples covered with three different materials - plastic, cotton, and leather at 0.1–3 THz agree well with those of the samples alone, which indicates that the barrier materials are mostly transparent to the THz wave.

Considering that the sample may not be a single-component explosive but a mixture in the actual security inspection, based on the spectral data of known pure explosives components, Chen Y et al. proposed the THz spectral uncertainty analysis using micro genetic algorithm [101], by applying intelligent computing to the analysis and optimization of THz spectral combination. In this work, the mixture of benzoic acid, o-toluic acid and p-toluic acid were specifically analyzed. In 2015, Trofimov VA et al. proposed a method for the efficient analysis of explosives by pulsed terahertz spectroscopy using spectral dynamics [102]. In recent years, they have also developed an efficient tool to detect and identify substances in ternary explosive mixtures with similar spectral properties using broadband reflected THz signals [103].

In security applications, terahertz spectroscopy is used not only for the inspection of explosives, but also for the inspection of other contraband such as drugs [104, 105]. The THz images obtained after the analysis and processing of THz spectral information can also be used for the rapid detection of firearms and controlled knives [106–108]. As the application of terahertz spectroscopy technology for security screening technology continues to mature, more and more THz security inspection equipment began to put into practical application [25, 109].

As we have already mentioned, terahertz spectroscopy can be used for non-destructive testing. We have also described its applications in biomedicine, security inspection, etc. This property can also be used for the detection of other substances and materials. In 2008, Jackson JB et al. used terahertz spectroscopy for the nondestructive examination of frescoes [110]. Subsequently, Abraham E et al. applied terahertz spectroscopy to the inspection of art paintings [26], and in 2015, Koch Dandolo CL et al. applied it to the inspection of panel paintings beneath gilded finishes [111].

THz fingerprints are available for pigments, adhesives and substrates of artworks, however, researchers still lack a database of THz spectra of these substances and materials, which makes the application of terahertz spectroscopy for artwork identification still difficult, and many researchers have set out to determine the THz spectra of different pigments. Li CY et al. in 2017 performed terahertz spectroscopy on several traditional Chinese pigments including cinnabar, red lead, oyster shell white, realgar, malachite, orpiment and azurite [112]. In 2019, Kleist EM et al. provided insight into the molecular and intermolecular forces that give rise to spectroscopy absorption features of crystalline copper-containing historical pigments in solid-state density flooding theory simulations and reveal deviations from simple harmonic vibrational behavior that may complicate these spectra, while investigating terahertz spectroscopy of solid azurite, malachite, and verdigris [113]. In the same year, Squires AD et al. analyzed phthalocyanine pigments under terahertz spectroscopy [114], while the identification and differentiation of copper phthalocyanines’ α, β and ε crystal polymorphs is demonstrated. In 2021, Lee JE et al. performed terahertz spectroscopic analysis of vermillion pigments in freestanding and polyethylene mixed forms [115].

The aforementioned applications of terahertz spectroscopy are almost based on typical THz-TDS. Nevertheless, other types of terahertz spectroscopy have also been reported. In 2003, Kiwa T et al. used THz emission spectroscopy to check electrical faults in integrated circuits (IC), and the two-dimensional THz emission images of IC chips could be clearly observed [33]. In 2008, Dressel M et al. investigated the superconducting energy gap of superconductors using terahertz time-resolved spectroscopy [32]. They obtained terahertz spectroscopy of niobium and niobium alloys, magnesium diboride, and some organic and high-temperature superconductors to explore the inherent laws of superconductors.

Furthermore, terahertz spectroscopy can also be applied to building detection [28], energy power [116, 117], pharmaceuticals [118], food science [119], marine engineering [27], and other fields [120, 121].