Terahertz circular polarization sensing for protein denaturation based on a twisted dual-layer metasurface

: Protein denaturation has very important research value in nutrition, biomedicine, and the food industry, which is caused by the changes in the molecular structure of the protein. Since the collective vibrational and torsional modes of protein molecules are within the terahertz (THz) frequency range, THz spectroscopy can characterize the protein denaturation with several advantages of non-contact, label-free, real-time, and non-destructive. Therefore, we proposed a reflective THz time-domain polarization spectroscopy sensing method, and use a flexible twisted dual-layer metasurface film as a sensor to realize the thermal denaturation sensing, concentration sensing, and types identification of protein aqueous solutions. The experiment tested three proteins (bovine serum albumin, whey protein, and ovalbumin), and the results show that: for the thermal denaturation sensing, its detection sensitivity can reach 6.30 dB/% and the detection accuracy is 0.77%; for the concentration sensing, the detection sensitivity and detection accuracy reach 52.9 dB · mL/g and 3.6 · 10 − 5 g/mL, respectively; in addition, different protein types can be distinguished by the difference of the circular polarization spectra.


Introduction
Protein is the basic organic matter of cells and plays an important role in life activities. Protein denaturation refers to the change of internal structure and properties of protein molecules under the influence of physical or chemical factors [1,2]. There are many reasons for protein denaturation, including physical factors such as heating [3][4][5][6], pressure [7], dehydration [8], vibration [9], ultraviolet radiation [10], etc., and chemical factors such as heavy metal salts, strong acids or bases, organic solvent [11,12], etc. The study of protein denaturation has important applications in food processing, sterilization, protein purification, etc. It is of great significance in the fields of nutrition, biomedicine, and the food industry [13][14][15].
Biomolecules such as proteins exhibit collective vibrational and torsional modes in the Terahertz (THz) frequency range [16,17], Therefore, THz spectroscopy can perform molecularlevel characterization of proteins. THz wave is a section of the electromagnetic band between microwave and infrared waves that have not yet been fully developed, usually referring to 0.1-10 THz. Because of the non-invasive and non-ionizing characteristics, it has a wide range of applications in sensing detection, especially biosensing. Research on protein spectroscopy in the THz band has been extensively carried out [18]. Markelz's group tested the THz spectrum of lyophilized powder bovine serum albumin (BSA) pressed pellets in the broadband range (0.06∼2.00 THz), and demonstrated that the collective vibration mode related to the protein molecular conformation is optically active in the THz range [19,20]. Paciaroni et al. studied the vibrational collective dynamics of a dry perdeuterated maltose-binding protein in the THz region [21]. Sun et al. use THz spectroscopy to study the binding of hemagglutinin protein and broadly neutralizing monoclonal antibodies in the liquid environment [22].
Because denaturation can change the structure of protein molecules, the THz characteristic spectrum of protein will change with the denaturation. Yoneyama et al. measured the thermal denaturation of BSA protein held in a membrane device with THz spectroscopy, results suggest that the THz transmittance of thermal denatured BSA sample is higher than that of nativeconformation sample [23]. Luong et al. investigated the thermal denaturation of human serum albumin (HSA) using THz spectroscopy in an aqueous buffer solution, the results proved that the absorption coefficient and refractive index of HSA change with temperature [24]. These tests proved the feasibility of THz protein detection and denaturation detection, but there are still some vital problems such as low signal-to-noise ratio and poor sensitivity, especially in the liquid environment with water absorption, which greatly limits the development of THz sensing.
Since the THz optical response of natural biological samples is generally very weak, the signal-to-noise ratio will be very low when the sample is directly sensed. There are many ways to enhance the THz response of biological samples, such as adding gold nanoparticles [25], nanoantenna enhanced [26], using microfluidic sensors [27], etc. The most common method is to use metasurface structures as sensors [28][29][30]. Metasurfaces are artificial electromagnetic structures consisting of orderly arranged elements, by designing their structure, the amplitude, phase, and polarization of waves can be manipulated artificially in the space and time domain [31]. Utilizing the surface plasmon polaritons and resonance enhancement effect of the metasurface as a sensor, the sensing performance can be effectively improved. Cheng et al. used planar Jerusalem cross metamaterial absorber, realize the detection of different viruses [32]. Li et al. used a metal metasurface structure as a sensor to measure the thermal-denaturation temperature of insulin (a protein hormone) [33]. If a double-layer or multi-layer metasurface is used as the sensor, and the structure has geometric chirality, the metasurface will have different transmission or reflection for left-hand circular polarization (LCP) and right-hand circular polarization (RCP), that is, it has a strong response to the polarization of the electromagnetic wave. This kind of metasurface structure has great benefits for polarization sensing [34,35].
In this work, we utilize the reflective THz time-domain polarization spectroscopy (THz-TDPS) sensing system for protein sensing in the liquid environment. Through this system, the complete polarization states of the sensing signals can be obtained to realize polarization sensing. The reflective system is used to avoid strong water absorption of THz. In addition, a flexible twisted dual-layer metasurface structure with geometric chirality is used as the sensor to enhance the polarization response of the protein sample. The circular polarization reflective spectra (RCP) are used to characterize the polarization sensing. Through this experimental setup and sensing method, the sensing experiments were carried out on the three protein aqueous solutions, that is the BSA, whey protein (WP), and ovalbumin (OVA), the thermal denaturation sensing, concentration sensing, and types identification were realized.

Reflective THz-TDPS system
THz spectrum detection is based on the time-domain spectroscopy (TDS) system. The effective spectral range of the THz-TDS system used in our experiment is 0.1∼2.5 THz. To detect the reflective polarization spectra of liquid samples, we modified the traditional transmission TDS system, as shown in Fig. 1(a) and 1(b). First, we add a reflection module to the system, that is, a metal mirror is used to obliquely incident the THz wave at a fixed angle (30°) to the bottom of the sample cell containing the liquid sample, two THz lenses are used for focusing and collimating. The sample cell is made by 3D printing, and the metasurface structure as the cell bottom can play its role as a sensor. Second, we added two rotatable polarizers at the emitting end and detection ports of the system. The first polarizer (P1) at the emitting port is used to ensure the linear polarization (LP) degree of the incident THz waves. By rotating the second polarizer (P2) at the detection port, a pair of orthogonal signals can be detected to restore the complete polarization state of the outgoing wave. In addition, a THz wave plate with an optical axis direction of 45°i s placed behind P1 to generate the chiral THz polarization state. When the THz wave with a chiral polarization state is obliquely incident on the metasurface sensor and sample, a stronger and more complex polarization response will occur. The experimental THz-TDPS setup in this work is shown in Fig. 1(d). The THz pulse is generated by a low temperature-grown GaAs PCA excited by Ti: sapphire femtosecond laser with a duration of 75 fs at 800 nm. An accurate displacement platform is used to generate optical delay and scan the sample's THz signal in the time domain. THz can excite the electro-optical effect in the ZnTe crystal and change its birefringence, so this effect can be used for THz detection [36]. All experiments were carried out at room temperature and 30% humidity. The effective spectrum range of the system is 0.1∼2.5 THz, the signal-to-noise ratio is 10 5 . Besides, the measurement step of the experimental system is 0.02 ps, and the time domain resolution can reach 6.25 GHz. In the time-domain spectra signal processing, we select the 4 ps time window after the signal appears, and filled the remaining data points with "0", and then performed Fourier transform. This is equivalent to low-pass filtering in the frequency domain. The detection accuracy of the circular polarization (CP) spectrum is mainly determined by the polarization detection signal-to-noise ratio of the system. The polarization detection signal-to-noise ratio (R P ) of the system was tested experimentally to be −36 dB.
The flexible Polyimide (PI) film with only 100 µm thickness is used as the substrate as shown in Fig. 1(e), which has excellent heat resistance, chemical stability, mechanical, and dielectric properties. The periodic metal structures are fabricated on both sides of the PI film by conventional photolithography and lift-off processing, of which detailed geometries are shown in Fig. 1(c). The geometry of structural units on the front and back surface is the same, but there is a θ=60°rotation between them. Such a twisted chiral structure breaks the mirror symmetry in the direction of wave propagation, so it has strong polarization conversion and intrinsic chirality.

Protein sample preparation
We choose three common proteins, BSA, WP, and OVA for experimental testing, all three proteins are from Shanghai Macklin Biochemical Co., Ltd. All proteins are dissolved in distilled water. The protein solution is denatured by heating in a water bath, the heating time is 2 mins at different temperatures. Figure 1(f) shows the appearance changes of BSA solution at different heat-treatment temperatures. It can be seen that when the temperature is higher than 80°C, the protein solution changes significantly, and the viscosity of the solution becomes higher, changing from the light-yellow transparent liquid to a white colloid.
In the following sensing experiments, to avoid experimental errors, we prepared 3 sets of the same samples, and each set of samples was measured 5 times under the same conditions. A total of 15 sets of experimental data were obtained. The experimental results are obtained by removing the results with too large errors and averaging the remaining data.

Functions of the wave plate and metasurface sensor
Before analyzing the functions of wave plate and metasurface sensor, we need to understand the definition and calculation method of the CP spectra. In the sensing experiment, by rotating P2, we get two THz time-domain signals in two perpendicular directions of ±45°. By Fourier transform, the amplitude spectrum A ±45°( ω) and the phase spectrum σ ±45°( ω) can be obtained, and the reflective spectra for LCP and RCP (R LCP and R RCP ) can be calculated by In addition, we defined the polarization ellipticity (PE) to describe the polarization state of the outgoing wave, i.e., the difference between LCP and RCP, the PE spectrum in dB can be calculated by the following formula The experimental results are also verified by numerical simulation. All the simulations are performed using the finite-difference time-domain (FDTD) method with the commercial software FDTD Solution. Since the THz waves are incident obliquely, the plane wave source is set to BFAST type in the software. Correspondingly, the boundary condition is set to the Bloch type in the direction where the unit structure is periodically arranged. Figure 2 shows the circular polarization transmission characteristics of the wave plate and the twisted metasurface both in experiment and simulation. From Fig. 2(a), the polarization state of the LP incident light has been changed significantly due to the polarization conversion of the wave plate. It can be seen from Fig. 2(b) that the PE spectrum through the wave plate reaches the extremum near 0.36 THz and 1.15 THz, indicating that the transmitted waves are circularly polarized at these two frequencies (LCP state at 0.36 THz and RCP state at 1.15 THz). While at 0.78 THz, the PE value is close to 0, so the output wave at this frequency is still in the LP state. Therefore, the results of Fig. 2 show that the sensor and its system make the polarization state of the emitted light field undergo strong polarization transformation and have chirality. The CP light field (i.e. chiral light field) will interact with the chiral metasurface sensor and protein molecules to produce a stronger polarization response, which can effectively improve the polarization sensing performance of the system.  Using water as a sample can test the sensing performance of the sensor structure. It can be seen that the R LCP spectrum has a strong resonance near 0.30 and 1.40 THz, while the resonance of R RCP is around 1.40 THz. Figure 2(d) shows the PE spectra for water sensing. The PE spectrum shows that the sensing signal is LCP state near 0.10 THz and RCP state near 0.30 THz, while there is no significant polarization change at high frequency (around 1.40 THz). Comparing the CP and PE spectra with and without the sensor structure, it can be seen that the polarization spectra are changed significantly, which proves that the twisted dual-layer metasurface structure has a strong polarization response. The above experiment and simulation results are all in good agreement. Figure 3(a) shows the schematic diagram of the twisted dual-layer metasurface sensor. It can be seen that the THz wave is incident on the metasurface sensor at an angle of 30°. Since the total thickness of the metasurface is only 100 µm, this scale is in the same order of magnitude as the incident THz wavelength, the Fabry-Perot effect between the two metallic layer structures can be ignored, and the metasurface sensor needs to be analyzed as a whole structure. Figure 3(b) shows the simulative field distribution near the metasurface in the Y-Z section at 1.4 THz when the sample is the air and the water, respectively. It can be seen that due to the existence of the metallic structure, the field localized enhancement occurs on the surfaces and inside of the sensor. In addition, the change of the sample on the surface of the sensor will lead to the change of field distribution, thereby affecting the intensity and polarization state of the reflected wave. The twisted structure has the intrinsic chirality, and when the electromagnetic field is incident obliquely, it also has the external chirality. Therefore, Fig. 3(c)∼3(e) show that the field distributions of up-surface, middle-layer, and down-surface are rotating with the orientation difference of the twisted metallic units. Therefore, strong polarization and chiral responses can be exhibited in this sensor.

Thermal denaturation sensing of three proteins
Next, we utilize the reflective THz-TDPS system and metasurface sensor to detect protein denaturation caused by temperature. First, BSA was studied, the concentration of BSA solution was 0.2 g/mL. We rotate the polarizer P2 to 0°and obtain the LP component of the outgoing signal for LP sensing. Figures 4(a) and 4(b) show the changes of LP spectra with temperatures around 0.25 and 1.45 THz, respectively. It can be seen that although the difference between water and protein can be distinguished by the LP sensing method, the LP spectral difference of BSA solution at different temperatures is not significant. The LP spectral peak difference at 30°C and 90°C is only 0.4 dB at 0.25 THz and 1.5 dB at 1.45 THz, respectively, which are difficult to be effectively distinguished by the traditional LP sensing. Therefore, we rotate the polarizer P2 to ±45°to obtain the complete polarization information of the outgoing signal and use the CP spectra as the new sensing parameters. Figures 4(c) and 4(d) show the changes of LCP and RCP spectra with temperature, respectively. For R LCP , the sensing performance of the characteristic valley near 0.30 THz is better than that near 1.40 THz, so the valley near 0.30 was selected for sensing characterization. For R RCP , the only characteristic valley near 1.45 THz was selected for sensing characterization. Different from LP sensing, for R LCP and R RCP , except for the obvious difference between the water and the BSA solution, the changes of CP spectra with temperature are significant. The peak differences of R LCP and R RCP at 30°C and 90°C are 3.8 dB and 5.3 dB, respectively, much larger than that of LP sensing.
where the E x and E ye represent the components of the electric vector E on the X-and Y-axes, while ∆σ is calculated by ∆σ=σ +45°-σ −45°. It can be seen that the polarization states of the output THz waves of the water and BSA solution are significantly different. At 0.30 THz, the polarization state of the water is more circular than that of the BSA solution, while at 1.45 THz, the polarization state of the water is more linear than that of the BSA solution. Compared to the polarization ellipses of the BSA solution at 30°C and 90°C, the polarization direction rotates clockwise by 10°both at 0.30 THz and 1.45 THz. Figure 5 shows the CP sensing results of WP and OVA solutions. The concentration of WP and OVA solutions is 0.2 g/mL. For WP, no matter for R LCP or R RCP , the peak values of water are the lowest, and the peak values of WP solution both increase with the increase of temperature, which is similar to the changing trend of BSA solution. However, for OVA, the peak values of CP spectra for water are still the lowest regardless of R LCP or R RCP , but the peak values of OVA solution both decrease with the increase of temperature, which is opposite to that of BSA and WP. Moreover, at 0.3 THz, the polarization spectra only change in the characteristic valley intensity, while at 1.45 THz, in addition to the intensity change, there are also slight frequency shifts. This is because the resonant frequency of the higher-order mode in metasurface is more sensitive to the change of surrounding refractive index than that of the lower-order mode.  Figure 6(a) and 6(b) show the relationship between the peak values of CP spectra and heattreatment temperatures. The temperatures corresponding to each curve in the figures represent the complete denaturation temperatures of the three protein solutions, when the heating temperature is higher than the complete denaturation temperature, the CP spectrum will no longer change. The gradient colors on the right side represent the solutions' viscosity increase. It can be seen that the complete denaturation temperatures of BSA, WP, and OVA are about 90°C, 80°C, and 70°C, respectively. Among them, the changing trend of BSA and WP are the same, and the peak value of the CP spectra increases with the increase of temperature, while the changing trend of OVA is the opposite. This is because the molecular changes of OVA caused by thermal denaturation are significantly different from BSA and WP: the molecular structure of BSA and WP becomes loose and easy to precipitate after heating up, while OVA forms a colloid after heating up, and the intermolecular structure is more stable [37,38]. Then, we express the degree of protein denaturation as a percentage, the denaturation percentage at 30°C is 0%, and the complete denaturation is 100%. The vertical axes are the CP spectra peak values differences before and after denaturation, as shown in Fig. 6(c) and 6(d). Therefore, the degree of denaturation of all three proteins can be detected by the THz CP sensing system. It can be seen that the changing trend of OVA is also opposite to that of BSA and WP. In addition, the reflection change of BSA at low frequency (0.30 THz) is greater than WP, while it is less than WP at high frequency (1.45 THz). Therefore, the three protein solutions can be distinguished by the denaturation curves.
To measure the sensing ability of the system for different proteins, we define the sensing sensitivity "S d " according to the change of the peak values of the CP spectra caused by the where R P is the polarization detection signal-to-noise ratio mentioned above. The experimental results of the three protein solutions are summarized as shown in the Table. 1. The sensing sensitivities of CP spectra reach the order of 10 −2 dB/%, and M % reach the order of 1%. The sensing sensitivities of R RCP are generally higher than that of R LCP . Among the three proteins, the sensing sensitivity of WP solution is the highest, reaching 6.30·10 −2 dB/%, and its M % is the lowest, reaching 0.77%.

Concentration sensing of BSA before and after thermal denaturation
Next, we study the application of THz CP sensing in the quantitative detection of protein solutions with different concentrations. Figure 7 shows the sensing results of the BSA solution with different concentrations before and after thermal denaturation. It can be seen from Fig. 7(a)∼7(d) that whether for R LCP or R RCP , before or after denaturation, the peak values of water are significantly lower than that of the BSA solution, and the peak values decrease as the concentration of the BSA solution decrease. Figure 7(e) and 7(f) show the changing trends of the CP spectra peak value with the sample concentration. When the BSA concentration is 0, i.e., the detection sample is water, the spectra at 30°C and 90°C are almost unchanged as expected. When the concentration of BSA increases, the peak values of R LCP and R RCP both become larger. Comparing R LCP and R RCP , the variation range of R RCP peak value is greater than that of R LCP . In addition, comparing the BSA solution before and after denaturation at 30°C and 90°C, the results show that the spectral peak of the BSA solution after denaturation has a larger range. Similar to the definition of S d , we define the sensing sensitivity "S c " according to the change of the peak values of the CP spectra caused by the change of sample concentration per unit, calculated by S C = ∆R CP /∆c (unit: dB·mL/g), where ∆RCP is the change of the CP resonance intensity and ∆c is the change of the BSA solution concentration. Moreover, the minimum protein concentration (M C ) that can be detected in this sensing system can be calculated by The experimental results of the sensing sensitivities and M C of the BSA solution are shown in Table 2. It can be seen that the sensing sensitivities for CP resonance intensity are all in the order of 10 dB·mL/g, with a maximum of 52.9 dB·mL/g. The minimum detection concentration of BSA solution is all in the order of 10 −4 ∼10 −5 g/mL, with a minimum value of 3.6×10 −5 g/mL.

Conclusion
In conclusion, the reflective THz-TDPS sensing system and flexible twisted dual-layer metasurface sensor were proposed to realize concentration sensing, thermal denaturation sensing, and types identification of protein aqueous solutions. Through this new sensing method, the liquid sample can be effectively detected by the reflective module, and its complete polarization and phase information can be obtained by the orthogonal polarization detection method. Compared with the traditional LP spectra, the CP sensing sensitivity is improved and more sensing information can be obtained. The double-layer metasurface sensor enhances the polarization response of the sample. The experimental results show that: for the thermal denaturation sensing, its detection sensitivity and detection accuracy reach 6.30 dB/% and 0.77% respectively; for the concentration sensing, the detection sensitivity and detection accuracy reach 52.9 dB·mL/g and 3.6·10 −5 g/mL respectively; in addition, different protein types can be distinguished by the difference of CP spectra. This new THz sensing system and method has great application potential in the biochemical detection of active samples and is expected to become a wide-applicable, high-sensitivity, and non-destructive sensing method.

Disclosures. The authors declare no conflicts of interest.
Data availability. Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.