Surface-Modiﬁed Compounds Improve the Detection Sensitivity of Terahertz Metasurface Biosensors

: Some trace elements in the human body, such as proteins and metabolites, are closely related to human health. Changes in the concentration of proteins can usually be used as indicators for the diagnosis of diseases. On the other side, some metabolites such as lactic acid and uric acid are closely related to human motor function. A large part of cancer markers are proteins, and their concentration changes in human serum and body ﬂuid are usually closely related to cancer diagnosis, staging and prognosis. It has always been the goal of researchers to efﬁciently and sensitively detect the relevant trace elements in human body. Due to the characteristics of low energy, coherence and ﬁngerprint, terahertz (THz) waves have been widely used in the detection of substances in recent years. Metasurface sensor is a new type of sensor with unique electromagnetic characteristics based on artiﬁcial design. The emergence of THz metasurface sensors provides a new technical means for the detection of trace elements proteins. Based on the above background, the research objects in this paper are three common protein cancer markers: carcinoembryonic antigen (CEA), neuron-speciﬁc enolase (NSE) and squamous epithelial cell carcinoma antigen (SCC), in order to study the detection performance of THz metasurface sensor and its improvement effect after modiﬁcation. In the experiment, the cancer marker is dropped onto the surface of sensor, and the spectrum is detected by THz time-domain system (TDS), and the frequency shift is calculated by comparing with the resonance frequency of the blank sensor. Then the experiment was repeated by changing the concentration and class of the marker, and the sensitivity of sensors was calculated by ﬁtting the relationship between the frequency shift and the concentration of markers. After that, two compounds (halloysite nanotube and Tungsten disulﬁde) were used to modify the metasurface sensor, and compared with the unmodiﬁed metasurface sensor. The experimental results show that for the detection of three markers, the metasurface sensor modiﬁed with HNT has signiﬁcantly improved the detection sensitivity compared with the sensor modiﬁed with WS 2 and blank sensor. This provides a new means for THz metasurface sensors in the detection of biomarkers, which is suitable for biomedicine, human health monitoring and other ﬁelds.


Introduction
Proteins are a class of macromolecules with important biological functions, including replicating genomic information, regulating transcription, signaling, providing structure, catalyzing reactions, and transporting molecules [1,2]. When proteins are misfolded, truncated, or mutated, as well as overexpressed or underexpressed proteins, they are often associated with many diseases. Therefore, proteins are usually used as biomarkers, which are closely related to the physiological state of the human body, and have different metabolic, structural and regulatory functions in different diseases [3,4]. By tracking the concentration of protein, it can reveal the health or disease of the human body. For example, cancer-related proteins include CEA, NSE, and SCC, etc. [5][6][7], and the changes in the concentration of these protein in human serum can assist in the analysis of cancer. In addition, some research have introduced the use of cancer stem cell (CSC) biomarkers for prognostic analysis, among which 8 types CSC biomarkers CD133, CD44, CD90, ALDH1A1, EPCAM, SOX2, SOX9, and LGR5 have significant prognostic value in digestive system tumors (including pancreas, colon, liver, stomach, and esophagus) [8]. Therefore, the detection of protein in human body is of great significance for monitoring human health status, recovery after exercise, disease diagnosis and prognosis analysis. At present, there are many detection methods applied to protein detection, including enzyme-linked immunosorbent assay(ELISA) [9], chemiluminescence Immunoassay(CLIA) [10], electrochemical method [11], biological mass spectrometry [12], etc. However, these methods are often limited by sensitivity, operational difficulty, and high costs. Therefore, there is an urgent need to develop fast, convenient, and highly sensitive protein detection methods.
THz wave is between infrared and microwave, which has the following characteristics. The electron energy is very low and will not cause harmful ionization to biological samples. The detection of THz wave belongs to coherence detection, and physical parameters such as refractive index and absorption coefficient can be obtained through calculation. Most importantly, the vibrational and rotational energy levels of many biomolecules and organic molecules are in the THz frequency band [13][14][15]. Therefore, THz technology has broad application prospects in food safety, biomedicine, human health and other fields [16][17][18]. Metassurface is a kind of artificially designed periodic structure, which have unique electromagnetic characteristics such as negative refractive index and negative conductivity. Its resonance frequency can be drastically changed by the change of local electric field on the surface. Based on these characteristics, sensors for sensing detection can be designed [19][20][21]. However, THz metasurface sensor is a kind of sensor with special electromagnetic characteristics designed by humans in the THz range [22,23]. However, because the change of dielectric environment caused by the trace amounts of biological samples is not obvious, the direct use of THz metasurface sensor detection cannot produce better detection results. Therefore, a new method to improve the detection sensitivity of metasurface sensors need is to explored.
In recent years, metasurface structures working in the terahertz frequency band are a hot research direction, and researchers have carried out a lot of work on them. Harry Miyosi Silalahi et al. [24] proposed a folded metamaterial sensor with high aspect ratio nano profile SRRs. The resonant transmittance of the sensor is far lower than that of ordinary Metamaterial, reaching −49 dB. The Q factor pf sensor reaches 37.0. The sensitivity of the sensor to the refractive index is up to 647 GHz/RIU. In the same year, Zahra Rahimian Omam et al. [25] revealed the unprecedented optical phonon response of CaMg(CO 3 ) 2 (dolomite) thin films in planar ultra-narrow band mid-infrared (MIR) thermal emitter designs. The embedded vanadium dioxide (VO 2 ) phase-change material (PCM) is capable of exciting a hybrid Fano resonance with dynamic spectral tuning. Yibo Pan et al. [26] proposed a dual band multifunctional encoded metasurface that controls the transmission and reflection of the entire space by changing the frequency and direction of the incident wave. proposed a dual band multifunctional encoded metasurface that controls the transmission and reflection of the entire space by changing the frequency and direction of the incident wave. By arranging specific encoding sequences on different frequencies, the functions of reducing RCS, beam splitting, and OAM generation are achieved. Shui Liu et al. [27] proposed the metasurface based on liquid crystal programmable has the functions of beamforming and beam steering, and realizes the dynamic wave control. More elaborate control of meta-unitand quasi-2-bit modulation are realized. Feng Zhao et al. [28] achieved a hyper-dispersive metalens in the range of 2.40 to 2.61 THz by simultaneously controlling the frequency-dependent phase, group delay, and group delay dispersion of the superlattice lens. Its dispersion was 1.76 times larger than that of conventional diffractive hyper-dispersive metalens. Wei Fan Chiang et al. [29] successfully deposited terahertz Metamaterial on LCE films by using the thermal evaporation and high-efficiency cooling system of a programmable electronic shutter. Achieved a large switching contrast of 277 when modulating the transmission ratio.
In order to verify the detection effect of the double-opening ring THz metasurface sensor and the performance improvement effect after modified, three common cancer markers CEA, NSE and SCC were selected as the research objects. First, the unmodified sensor was detection different concentrations of the three markers, and the sensitivity of the sensor was obtained by calculating the change in the resonance frequency of the sensor and establishing the relationship between the concentration of the marker and the frequency shift. Second, the surface modification work was carried out, and WS 2 and HNT were modification to metasurface sensor, respectively. After that, using the modified sensor detection the cancer markers mentioned above. Finally, the detection result of modified sensor was compared with the unmodified sensor, to explore the improvement of the detection ability by different compounds. According to the analysis of experimental results, we found that compared the sensor modified with WS2 and blank sensor, the sensor modified with HNT has the highest sensitivity for the above three cancer markers. This provides a new scheme for THz metasurface sensors in the detection of trace biomarkers, which is suitable for biomedicine, human health monitoring and other fields.

Materials
Three common cancer markers were used in the experiment: carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), squamous cell carcinoma antigen (SCC), purchased from Zhengzhou Cell ToAntibody & Antigen Biotechnology. Co., Ltd. Zhengzhou, China; Phosphate buffer solution PBS was purchased from Guangzhou Howei Pharmaceutical Technology Co., Ltd. Guangzhou, China; Halloysite nanotubes (HNT) was purchased from Yuan Xin Nano Technology Co., Ltd. Guangzhou, China; Tungsten disulfide (WS 2 ) was purchased from Suzhou Beasley New Material Co., Ltd. Suzhou, China. All samples were not further purified before use. Figure 1a shows the metasurface sensor with double open metal rings, which is composed of three layers. The bottom layer is high-resistance silicon with a dielectric constant of ε = 11.9, and the thickness of this layer is 500 µm. The middle layer is silica with a dielectric constant of ε = 3.75, and the thickness of the middle layer is 500 nm. The topper layer is gold with a conductivity of σ = 4.561 × 10 7 S/m, and the thickness of this layer is 200 nm. In addition, the metal of topper layer is etched into the shape of two open rings. The period of the structure is P = 80 µm, the metal ring width is w = 7 µm, the gap width is g = 8 µm, the length of the outer metal ring is L 1 = 72 µm, and the length of the inner metal ring is L 2 = 32 µm. Figure 1b shows the simulation results of this sensor, which it can be seen that the sensor produces resonance frequency in 0.2923 THz.  Figure 2 shows the surface current distribution and near-field distribution of the sensor at the resonance frequency. It can be seen that when the THz wave is vertically inci-  Figure 2 shows the surface current distribution and near-field distribution of the sensor at the resonance frequency. It can be seen that when the THz wave is vertically incident, the surface current will flow along the direction of the arrow on the gold. The ring can be approximately seen as the inductor L, and the gap can be equivalent to the capacitor C, and its equivalent circuit is shown in Figure 2b. At this time, the resonance frequency can be approximately calculated by Equation (1),  Figure 2 shows the surface current distribution and near-field distribution of the sensor at the resonance frequency. It can be seen that when the THz wave is vertically incident, the surface current will flow along the direction of the arrow on the gold. The ring can be approximately seen as the inductor L, and the gap can be equivalent to the capacitor C, and its equivalent circuit is shown in Figure 2b. At this time, the resonance frequency can be approximately calculated by Equation (1)

Sensitivity of the Sensor
When there are different analytes on the surface of the sensor, it will have a significant impact on the resonance frequency of the sensor. Therefore, it is necessary to explore the effect of refractive index, thickness and other parameters of the analyte on the sensor. The first parameter is the refractive index, the thickness of the analyte is fixed at 20 µm, and then the refractive index of the analyte is from n = 1.0 to n = 2.0, increasing by 0.2 in turn. The simulation results are shown in Figure 3a,b. As the refractive index increases, the resonance frequency is gradually redshifted. The sensitivity of the sensor to the refractive index is defined as S = ∆f/∆n. Therefore, the sensitivity of sensor is S n = 32.3 GHz/RIU. The first parameter is the refractive index, the thickness of the analyte is fixed at 2 and then the refractive index of the analyte is from n = 1.0 to n = 2.0, increasing by turn. The simulation results are shown in Figure 3a,b. As the refractive index incr the resonance frequency is gradually redshifted. The sensitivity of the sensor to the tive index is defined as S = Δf/Δn. Therefore, the sensitivity of sensor is Sn = 32.3 GH The second important parameter is the thickness of the analyte, therefore t sponse of the sensor surface with different thicknesses of analytes was also simulate refractive index was fixed as n = 2 in the simulation, and the thickness was changed h = 5 µm to 25 µm, and increased by 5 µm in turn. The simulation results are sho Figure 3c,d, where the resonance frequency similarly produces a significant redshift thickness increases. When the thickness of the analyte is greater than 25 µm, the freq shift generated is no longer significant. Therefore, when the thickness of the anal the sensor is greater than 25 µm, it will not have a significant impact on the experim results.

Angle Dependence and Polarization Dependence
In addition, the incidence angle of THz waves can also affect the detection p mance of the sensor. Therefore, this article also simulated the effect of different in angles on the resonance frequency of the sensor. This article simulated the situation vertical incidence to 20° incidence, and the results are shown in Figure 4. From Figu it can be seen that the change in incident angle does not have a significant impact resonant frequency of this sensor. This may be due to the symmetrical structure sensor. So, during the experiment, if there is a slight deviation in the placement sensor, it will not affect the detection results. The second important parameter is the thickness of the analyte, therefore the response of the sensor surface with different thicknesses of analytes was also simulated. The refractive index was fixed as n = 2 in the simulation, and the thickness was changed from h = 5 µm to 25 µm, and increased by 5 µm in turn. The simulation results are shown in Figure 3c,d, where the resonance frequency similarly produces a significant redshift as the thickness increases. When the thickness of the analyte is greater than 25 µm, the frequency shift generated is no longer significant. Therefore, when the thickness of the analyte on the sensor is greater than 25 µm, it will not have a significant impact on the experimental results.

Angle Dependence and Polarization Dependence
In addition, the incidence angle of THz waves can also affect the detection performance of the sensor. Therefore, this article also simulated the effect of different incident angles on the resonance frequency of the sensor. This article simulated the situation from vertical incidence to 20 • incidence, and the results are shown in Figure 4. From Figure 4b, it can be seen that the change in incident angle does not have a significant impact on the resonant frequency of this sensor. This may be due to the symmetrical structure of the sensor. So, during the experiment, if there is a slight deviation in the placement of the sensor, it will not affect the detection results.
Polarization response is an important factor that must be considered in analyzing sensor performance. To research the polarization dependence of the absorber, we first set the incident angle θ = 0 • , then change the azimuthal angle (i.e., ϕ) from 0 • to 45 • . In the simulation, cell boundary conditions will be set in the x and y directions, and open boundary conditions will be set in the z direction. The simulation result is shown in Figure 4c, the relationship of azimuthal angle and the shift account of resonance frequency is shown in Figure 4d. It indicates that the polarization response of the sensor to the incident wave is obvious. sensor performance. To research the polarization dependence of the absorber, we first set the incident angle θ = 0•, then change the azimuthal angle (i.e., φ) from 0° to 45°. In the simulation, cell boundary conditions will be set in the x and y directions, and open boundary conditions will be set in the z direction. The simulation result is shown in Figure 4c, the relationship of azimuthal angle and the shift account of resonance frequency is shown in Figure 4d. It indicates that the polarization response of the sensor to the incident wave is obvious.

Fabrication
The fabrication process of the THz metasurface sensor is shown in Figure 5. The first step is shown in Figure 5a, a lay of SiO2 film is deposition on silicon substrate by chemical vapor deposition (CVD). The second step is shown in Figure 5b, spin-coating a layer of photoresist. The third step is shown in Figure 5c, exposure of photoresist with UV exposure system. The fourth step is shown in Figure 5d, developing, fixing, and baking to form structural pattern. The fifth step is shown in Figure 5e, evaporate a 200-nm-thick layer of aluminum on the surface of SiO2. The final step is shown in Figure 5f, stripping the photoresist from the sample by using the liftoff process.

Fabrication
The fabrication process of the THz metasurface sensor is shown in Figure 5. The first step is shown in Figure 5a, a lay of SiO 2 film is deposition on silicon substrate by chemical vapor deposition (CVD). The second step is shown in Figure 5b, spin-coating a layer of photoresist. The third step is shown in Figure 5c, exposure of photoresist with UV exposure system. The fourth step is shown in Figure 5d, developing, fixing, and baking to form structural pattern. The fifth step is shown in Figure 5e, evaporate a 200-nm-thick layer of aluminum on the surface of SiO 2 . The final step is shown in Figure 5f, stripping the photoresist from the sample by using the liftoff process.

THz-TDS
As shown in Figure 7a, the THz time-domain spectroscopy system (THz-TDS) used in the experiment was the CCT-1800 spectrometer produced by the Institute of THz Science and Technology Innovation of ECCOM Ark Group, and its spectral range was 0.06-4 THz. Figure 7b is a schematic diagram of the placement of the sample holder in the testing chamber, where the little opaque square represents the THz metasurface sensor.

Metasurface Sensors Detect Cancer Markers
Before the experiment, PBS was used as buffer to prepare cancer marker of different concentrations.
As a control, the cancer markers were detected using an unmodified metasurface sensor. The detection process was as follows, after the metasurface sensor was washed and dried with purified water, 10 µL of cancer markers was added to sensor, and then incubated for 30 min. After that, the sensor was then fixed in THz-TDS and scanned to obtain the THz time-domain spectrum. Then, change the concentration and class of cancer marker, and repeat the above steps. THz frequency-domain spectrum can be obtained by

THz-TDS
As shown in Figure 7a, the THz time-domain spectroscopy system (THz-TDS) used in the experiment was the CCT-1800 spectrometer produced by the Institute of THz Science and Technology Innovation of ECCOM Ark Group, and its spectral range was 0.06-4 THz. Figure 7b is a schematic diagram of the placement of the sample holder in the testing chamber, where the little opaque square represents the THz metasurface sensor.

THz-TDS
As shown in Figure 7a, the THz time-domain spectroscopy system (THz-TDS) used in the experiment was the CCT-1800 spectrometer produced by the Institute of THz Science and Technology Innovation of ECCOM Ark Group, and its spectral range was 0.06-4 THz. Figure 7b is a schematic diagram of the placement of the sample holder in the testing chamber, where the little opaque square represents the THz metasurface sensor.

Metasurface Sensors Detect Cancer Markers
Before the experiment, PBS was used as buffer to prepare cancer marker of different concentrations.
As a control, the cancer markers were detected using an unmodified metasurface sensor. The detection process was as follows, after the metasurface sensor was washed and dried with purified water, 10 µL of cancer markers was added to sensor, and then incubated for 30 min. After that, the sensor was then fixed in THz-TDS and scanned to obtain the THz time-domain spectrum. Then, change the concentration and class of cancer marker, and repeat the above steps. THz frequency-domain spectrum can be obtained by

Metasurface Sensors Detect Cancer Markers
Before the experiment, PBS was used as buffer to prepare cancer marker of different concentrations.
As a control, the cancer markers were detected using an unmodified metasurface sensor. The detection process was as follows, after the metasurface sensor was washed and dried with purified water, 10 µL of cancer markers was added to sensor, and then incubated for 30 min. After that, the sensor was then fixed in THz-TDS and scanned to obtain the THz time-domain spectrum. Then, change the concentration and class of cancer marker, and repeat the above steps. THz frequency-domain spectrum can be obtained by Fast Fourier transform (FFT) of THz time-domain spectrum. Analyze the experimental results through frequency-domain spectrum. The experimental results are shown in Figure 8.
concentration can be defined as S = ΔF/ΔC, which is the change in frequency shift c by each unit of concentration. The sensitivity of the unmodified sensor to SCC was 0.5429 GHz/lg(pg·mL −1 ).
For CEA, Figure 8c,d show the experiment results of the unmodified sensor to The relationship between the concentration of CEA and the frequency shift 0.5714×CCEA + 3.6714, R 2 = 0.8342. Therefore, the sensitivity of the sensor to CEA was 0.5714 GHz/lg(pg·mL −1 ). The relationship between the concentration of NSE and the frequency shift 0.5321×CNSE + 3.8214, R 2 = 0.8244. Therefore, the sensitivity of the sensor to NSE is 0.5321 GHz/lg(pg·mL −1 ).
The experimental results indicate that when the THz metasurface sensor is no ified, its detection sensitivity for the above three biomarkers is not high, all of wh Among them, Figure 8a is the frequency-domain spectrum of SCC and Figure 8b is the frequency shift of the unmodified sensor caused by changing the SCC concentration. After linear fitting, the relationship between SCC concentration and frequency shift can be obtained as follows: F = 0.5429 × C SCC + 4.1, R 2 = 0.8660, where F is the frequency shift, C is the concentration of SCC after taking the logarithm. The sensitivity of the sensor to the concentration can be defined as S = ∆F/∆C, which is the change in frequency shift caused by each unit of concentration. The sensitivity of the unmodified sensor to SCC was S SCC = 0.5429 GHz/lg(pg·mL −1 ).
For CEA, Figure 8c,d show the experiment results of the unmodified sensor to CEA. The relationship between the concentration of CEA and the frequency shift is Appl. Sci. 2023, 13, 8818 9 of 14 F = 0.5714 × C CEA + 3.6714, R 2 = 0.8342. Therefore, the sensitivity of the sensor to CEA was S CEA = 0.5714 GHz/lg(pg·mL −1 ).
For NSE, Figure 8e,f show the experiment results of the unmodified sensor to NSE. The relationship between the concentration of NSE and the frequency shift is F = 0.5321 × C NSE + 3.8214, R 2 = 0.8244. Therefore, the sensitivity of the sensor to NSE is S NSE = 0.5321 GHz/lg(pg·mL −1 ).
The experimental results indicate that when the THz metasurface sensor is not modified, its detection sensitivity for the above three biomarkers is not high, all of which are less than 0.6 GHz/lg(pg·mL −1 ). Moreover, the linear relationship between the concentration and the frequency shift was not obvious, and the R 2 was less than 0.9. Therefore, it is necessary to modify the surface of the sensor to improve detection sensitivity.

Detection of Cancer Markers after WS 2 Modification
The first compound used for modification is WS 2 , which is a gray fine crystal or powder with metallic luster, belongs to the hexagonal crystal family, and has semiconductivity and diamagnetism. The concentration of WS 2 dispersions at 10 mg/mL was prepared and sonicated to fully disperse. After that, 10 µL of WS 2 dispersion was added to the sensor surface and left for 30 min. The microscope image (20×) of the sensor has modified by WS 2 is shown in Figure 9a, from which, it can be seen that many black dots of WS 2 cover the surface of sensor. The Figure 9b shown the frequency shift caused by WS 2 , after calculated, the Q-factor of sensor is 21.56, and the Q-factor of WS 2 -modified sensor is 11.5. And then, the experiments procedure was similar to that in Section 3.1. The experimental results are shown in Figure 10.

FOR PEER REVIEW
10 of 16 less than 0.6 GHz/lg(pg·mL −1 ). Moreover, the linear relationship between the concentration and the frequency shift was not obvious, and the R 2 was less than 0.9. Therefore, it is necessary to modify the surface of the sensor to improve detection sensitivity.

Detection of Cancer Markers after WS2 Modification
The first compound used for modification is WS2, which is a gray fine crystal or powder with metallic luster, belongs to the hexagonal crystal family, and has semiconductivity and diamagnetism. The concentration of WS2 dispersions at 10 mg/mL was prepared and sonicated to fully disperse. After that, 10 µL of WS2 dispersion was added to the sensor surface and left for 30 min. The microscope image (20×) of the sensor has modified by WS2 is shown in Figure 9a, from which, it can be seen that many black dots of WS2 cover the surface of sensor. The Figure 9b shown the frequency shift caused by WS2, after calculated, the Q-factor of sensor is 21.56, and the Q-factor of WS2-modified sensor is 11.5. And then, the experiments procedure was similar to that in Section 3.1. The experimental results are shown in Figure 10.  For SCC, Figure 10a,b show the experiment results of WS 2 modified sensor to detect SCC. The relationship between the concentration of SCC and the frequency shift is F = 0.85 × C SCC + 3.2, R 2 = 0.9912, and the sensitivity S SCC = 0.85 GHz/lg(pg·mL −1 ). For CEA, Figure 10c,d show the experiment results of CEA detection by WS 2 modified sensor. The relationship between CEA concentration and frequency shift is: F = 0.775 × C CEA + 2.65, R 2 = 0.9428, and the sensitivity S CEA = 0.775 GHz/lg(pg·mL −1 ). Figure 10e,f show the experiment results of NSE detection by WS 2 modified sensor. The relationship between the concentration of NSE and the frequency shift is F = 0.7286 × C NSE + 2.9286, R 2 = 0.9868, and the sensitivity S NSE = 0.7286 GHz/lg(pg·mL −1 ).
The experimental results indicate that, compared with the unmodified sensor, the sensitivity of WS 2 modified sensor to these three markers were significantly improved. At the some time, the relationship between the shift of resonance frequency and the concentration of marker is more obvious. The experimental results indicate that, compared with the unmodified sens sensitivity of WS2 modified sensor to these three markers were significantly improv the some time, the relationship between the shift of resonance frequency and the c tration of marker is more obvious.

Detection of Cancer Markers after HNT Modification
The second compound used for modification is Halloysite nanotubes (HNT), which is hollow tubular aluminum silicate clays. Its Molecular formula is Al 2 (OH) 4 Si 2 O 5 · nH 2 O. The inner diameter of a tubular structure is usually between 10 nm to 40 nm, and the outer diameter is usually between 40-70 nm, with a total length is between 0.2 µm to 2 µm [30].
The concentration of HNT dispersions at 10 mg/mL were prepared and fully dispersed by sonication. After that, 10 µL of HNT dispersion was added to the sensor surface for 30 min. The microscope image (20×) of the sensor has modified by HNT is shown in Figure 11a, from which, it can be seen that many black pins of HNT cover the surface of sensor. The Figure 11b shown the frequency shift caused by HNT, after calculated, the Q-factor of HNT-modified sensor is 13.57. It still less than the blank sensor, but more than WS 2 . And then, the experimental procedure was similar to that in Section 3.1. The experimental results are shown in Figure 12. for 30 min. The microscope image (20×) of the sensor has modified by HNT is shown in Figure 11a, from which, it can be seen that many black pins of HNT cover the surface of sensor. The Figure 11b shown the frequency shift caused by HNT, after calculated, the Qfactor of HNT-modified sensor is 13.57. It still less than the blank sensor, but more than WS2. And then, the experimental procedure was similar to that in Section 3.1. The experimental results are shown in Figure 12.    Figure 12a,b show the experiment results of the HNT modified sensor to detection SCC. The relationship between the concentration of SCC and the frequency shift is F = 1.25 × C SCC + 2.6429, R 2 = 0.9804, and the sensitivity S SCC = 1.25 GHz/lg(pg·mL −1 ). Figure 12c,d show the experiment results of detecting CEA by HNT modified sensor. The relationship between CEA concentration and frequency shift is: F = 1.1786 × C CEA + 2.6429, R 2 = 0.9840, and the sensitivity S CEA = 1.1786 GHz/lg(pg·mL −1 ). Figure 12e,f show the experiment results of NSE detection by HNT modified sensor. The relationship between the concentration of NSE and the frequency shift is F = 1.2643 × C NSE + 2.3, R 2 = 0.9804, and the sensitivity S NSE = 1.2643 GHz/lg(pg·mL −1 ).
The experiment results show that compared to the unmodified sensor and the sensor modified with WS 2 , the sensor modified with HNT has significantly improved sensitivity to these three markers. Meanwhile, the relationship between the change in resonance frequency and the concentration of the marker is more pronounced than that of sensors modified with WS 2 .

Discussion
To clearly discussion the experiment result, there need to compared the results of the above three parts. When there had not modification on the sensor, the detection of CEA, NSE and SCC did not show high sensitivity, only about 0.53-0.57 GHz/lg(pg·mL −1 ), and there was no obvious linear relationship between the change of the concentration and the frequency shift. After the surface of the sensor was modified with WS 2 , the detection effect of the three markers was significantly improved, and the sensitivity reaches 0.72-0.85 GHz/lg(pg·mL −1 ). There was also an obvious linear relationship between the concentration and the frequency shift. Meanwhile, the detection results after the sensor are modified by HNT, the sensitivity reaches 1.17-1.26 GHz/lg(pg·mL −1 ), which has a significant improvement, and also keeps a good linear relationship. So, the sum up, the modification of metasurface sensors using HNT will lead to higher detection sensitivity and linearity of the sensors for cancer markers such as CEA, NSE, and SCC. The reason for this effect may be that HNT has a larger specific surface area than the blank sensor surface and WS 2 , which can provide more attachment points for biomarkers. Meanwhile, HNT also has a better effect on surface local electric field enhancement.

Conclusions
This article uses the THz time-domain system to measure double ring type THz metasurface sensors. At the same time, three commonly used cancer marker (CEA, NSE and SCC) were selected as research objects. First, the unmodified sensor was used to detection cancer marker as a control. Then, HNT and WS 2 were used to modify the element surface sensor, respectively. And then the above cancer marker were detected. The experimental results indicate that the sensor modified with HNT has the highest detection sensitivity between unmodified sensors, WS 2 modified sensors, and HNT modified sensors. At the same time, the linear relationship between the frequency shift of the resonance frequency and the concentration of the marker is also the most obvious. The experimental results provide a new method for detecting trace biomarkers, proteins, and metabolites using THz metasurface sensors. The results of this study have certain reference value for the application of terahertz metasurface sensors in biomedicine and human health.