A metasurface is a 2D artificial metamaterial composed of periodic sub-wavelength geometric unit cells [1–7]. Its coupling with electromagnetic waves generates a unique resonant response with far-ranging applications to modern photonic devices such as modulators [8–10], absorbers [11,12], lenses [13], and sensors [14–17]. Recently, the plasmonic analog of electromagnetically induced transparency (EIT) has been generated in sub-wavelength geometric structures [18,19]. The absorbance linewidth is limited only by Drude damping, and the narrow EIT-like features render it ideal for biosensors [20]. Attempts have been made to realize ultra-sensitivity with EIT-like metasurface-based biosensors. However, related reports are rare owing to difficulties in changing the dielectric environment. A metasurface with graphene and metal halide perovskites may solve this problem. In such a device, graphene and metal halide perovskites work together to produce a resonant response that is ultra-sensitive to change in the dielectric environment under an external stimulus.

Graphene is a two-dimensional atomic system [21,22]. It allows modification of the Fermi level (${\mathit{E}}_F$) and the corresponding charge carrier density through biological, electrical, and optical methods [8,23,24]. Chemical vapor deposition (CVD) is a conventional method for producing graphene [25]. However, it is inevitably affected by impurities, defects, and disorders. CVD-graphene is usually p-doped, so the initial $E_F$ slightly deviates from the Dirac point and is in the valence band. In other words, the initial $E_F$ is quite close to the Dirac point [26]. The $E_F$ requires extremely weak external stimuli to move from the valence band to the Dirac point [27], which is suitable for an ultra-sensitive biosensor. Patterned graphene devices have also been widely researched for use in modern advanced equipment [28]. Patterning is the key to manufacturing graphene devices. Patterning graphene on the micro/nanoscale may greatly improve the performance of optical devices, which benefit from low-loss plasma characteristics [29]. Therefore, the patterning of graphene is an important factor that affects device performance [30]. Therefore, CVD-patterned graphene is ideal for producing an ultra-sensitive Dirac-point-based biosensor and tunable optics devices.

Metal halide perovskites are up-to-date optoelectronic materials with the crystal formula ${\mathrm{ABX}}_3[\mathrm{A}={\mathrm{CH}}_3{\mathrm{NH}}_3(\mathrm{MA}),\mathrm{CH}{({\mathrm{NH}}_2)}_2(\mathrm{FA})$, Cs; B = Pb, Sn; X = Cl, Br, I] [31–35]. In recent years, perovskites have shown great advantages in optics devices. Examples include photodetectors, solar cells, light-emitting diodes, and optical modulators [36–38]. The excellent optoelectronic properties of metal halide perovskites have received great attention from the optics community. Most importantly, a metal halide perovskite has a high and adjustable relative permittivity that exerts a greater influence on the phase of a terahertz wave passing through it, which plays an active role in phase-based sensors.

In this work, a hybrid dual-optoelectronic-material terahertz metasurface (HDOMM) is used for a Dirac-point-based biosensor. The HDOMM contains an EIT-like metasurface and a sandwich complex composed of graphene, polyimide (PI), and a metal halide perovskite. The HDOMM is used as a platform for ultra-sensitive and qualitative detection of sericin. Using the changes in amplitude and phase, HDOMM can detect sericin with a detection limit of 780 pg/mL. The internal mechanism for ultra-sensitive sensing includes the following two aspects. First, the sericin drives the $E_F$ of graphene to move to the Dirac point, causing a dramatic change in the dielectric environment. This is the main factor. Second, a metal halide perovskite boosts the positive effect on the phase. This plays a secondary role. These two factors are used to achieve ultra-sensitive label-free sensing with the HDOMM.

As shown in Fig. 1(a), the manufacturing process begins with the preparation of an EIT-like metasurface. The unit cell consists of a cut-wire resonator (CW) and two split-oval-ring resonators (SORRs). These resonators were engineered through conventional photolithography [Fig. 1a(i)]. Then, an ${\text{MAPbI}}_3$ film was spin-coated on the metasurface [Fig. 1a(ii)]. The ${\text{MAPbI}}_3$ wrapped the CWs and SORRs with good uniformity. Next, the ${\text{MAPbI}}_3$ was covered with a 1.5-μm-thick PI film [Fig. 1a(iii)]. PI was used to completely isolate ${\text{MAPbI}}_3$ from the ambient conditions, conferring stability to the ${\text{MAPbI}}_3$. After that, trilayer graphene was transferred onto the PI film [Fig. 1a(iv)].

Subsequently, the graphene was patterned into a fishing net structure with round holes whose diameters are 100 μm [Fig. 1a(v, vi)]. Figure 1(b) shows optical microscope images of the bare EIT-like metasurface, the EIT-like metasurface with graphene and ${\text{MAPbI}}_3$, and the EIT-like metasurface with patterned graphene and ${\text{MAPbI}}_3$. Finally, the HDOMM was used as a Dirac-point-based biosensor [Fig. 1a(vi)]. The dimensions of the unit cell and the electric and magnetic boundary conditions are shown in Fig. 1(c). Figure 1(d) displays the Raman spectrum of graphene excited by a 514-nm laser. The D peak has disappeared. The 2D peak is located at $2680{\mathrm{cm}}^{-1}$, and the G peak is at $1600{\mathrm{cm}}^{-1}$. The intensity of the 2D peak is lower than that of the G peak. In addition, the full width at half-maximum of the 2D peak is $\sim 55{\mathrm{cm}}^{-1}$. These results show that the graphene has high quality. Before the X-ray diffraction (XRD) measurement, an ${\text{MAPbI}}_3$ film was deposited on a silicon substrate. Figure 1(e) shows the XRD spectrum. The diffraction peaks are at 14.34°, 28.72°, and 32.16°, representing the (110), (220), and (310) crystal planes, respectively. Figure 1(e) shows no characteristic peaks associated with ${\mathrm{PbI}}_2$ and other redundant phases. The inset of Fig. 1(e) is the top-view SEM image of the ${\text{MAPbI}}_3$ film. The ${\text{MAPbI}}_3$ film has large grains and no pinholes, demonstrating a smooth, uniform morphology.

Figure 2(a) shows the experiment and simulation transmission spectra of the bare EIT-like metasurface, which presents a classic W-line shape. The experiment results agree relatively well with the simulation results in Fig. 2(a). However, there are still some differences due to the inevitable errors in processing and measuring. Figure 2(c) shows the simulated electric field distributions at 0.65 THz. For the CW, the electric field demonstrates a dipole resonant mode acting as a bright mode. Because the CW couples the incident THz waves, the electric field of the CW is consistently oriented in the polarization direction of the THz waves [39,40], whereas the SORRs decouple from the incident THz waves, serving as a dark mode. However, when the incident THz waves excite the CW, the dark mode couples with the bright mode. Figure 2(c) shows that the electric field of two SORRs is opposite that of the CW. This indicates a coupling with the near field of the CW. Because the CW and SORRs keep comparable excitation strength and a $\pi$ phase difference, there is destructive interference between the bright and dark modes, which causes a narrow transparency window [40,41]. The coupled harmonic oscillator model is applied to describe the destructive interference between the bright and dark modes. The interference can be analytically described by the coupled differential equations [39–41] $$\begin{gathered} {\ddot{x}}_1+{\gamma }_1{\dot{x}}_1+{\omega }_0^2{x}_1+\kappa x_2=E, \\ {\ddot{x}}_2+{\gamma }_2{\dot{x}}_2+{({\omega }_0+\delta )}^2{x}_2+\kappa x_1=0, \end{gathered}$$(1)where $x_1$ and $x_2$ are the resonant amplitudes, and ${\gamma }_1$ and ${\gamma }_2$ the losses of bright and dark modes, respectively; $\delta$ is the detuning of the resonant frequency of the dark-mode oscillator from the bright-mode one; and $\kappa$ is the coupling coefficient between the two oscillators. Solving Eq. (1) with the approximation $\omega -{\omega }_0\ll {\omega }_0$ yields the susceptibility $\chi$: $$\chi ={\chi }_r+i{\chi }_i\propto \frac{(\omega -{\omega }_0-\delta )+i\frac{{\gamma }_2}{2}}{(\omega -{\omega }_0+i\frac{{\gamma }_1}{2})(\omega -{\omega }_0-\delta +i\frac{{\gamma }_2}{2})-\frac{{\kappa }^2}{4}}.$$(2)The energy loss is always proportional to the imaginary part of $\chi$; thus, the transmission can be written as $$T\propto 1-{\chi }_i=1-g{\chi }_i,$$(3)where ${\chi }_i$ is the imaginary part of the susceptibility, which is proportional to the energy loss. The transmission $T$ is calculated from $T=1-g{\chi }_i$, in which $g$ is the geometric parameter describing the strength of the coupling of the bright mode with the incident electric field $\mathit{E}$. During coupling, the absorbance linewidth is limited only by Drude damping. That is to say, any change in the transparency window is due only to the external dielectric environment owing to the suppression of radiative losses [20]. In addition, Fig. 2(d) shows circulating current distributions along SORRs, representing low radiation loss. Therefore, the excellent features render the EIT-like metasurface ideal for biosensors.

To realize ultra-sensitive sensing, the HDOMM was used as a platform for qualitative detection of sericin, which has strong water solubility. For such detection, the patterned graphene is in direct contact with the sericin and performs the main sensing function. The perovskite is not in direct contact with the sericin, boosting the influence on the phase. This plays a secondary role. The change in the dielectric environment is mainly due to the conductivity of the patterned graphene. Therefore, the relationship between the conductivity and transmission spectra of the HDOMM was simulated to clarify the sensing mechanism of the HDOMM. Figure 2(b) shows the simulated transmission spectra under different conductivities. The transparency window gradually shrinks with increasing conductivity. An in-depth understanding of the electromagnetic behavior is obtained from the simulated electric field distributions in Figs. 2(e)–2(h) for the HDOMM under different conductivities. The field concentrations are strong when the conductivity is 0 S/m [Fig. 2(i)]. However, as the conductivity increases from 0 to 2500 S/m, the electric fields gradually decrease. Correspondingly, the transparency window shrinks.

Figures 3(a)–3(c) show the experimental transmission spectra for the HDOMM at sericin concentrations from 780 pg/mL to 1.25 μg/mL. Compared with that of the bare HDOMM, the transparency window with sericin is significantly enhanced. The obvious enhancement of the transparency window facilitates well the qualitative sensing of sericin. To clarify the sensing effect further, it is defined as $\mathrm{\Delta }T/T=[(T_{\text{sericin}}-T_{\mathrm{bare}})/T_{\mathrm{bare}}]\times 100\%$, where $T_{\text{sericin}}$ ($T_{\mathrm{bare}}$) is the transmittance at the transparency window with (without) sericin. From Figs. 3(a)–3(c), $\mathrm{\Delta }T/T$ for sericin at 1.25 μg/mL (highest concentration) is 6.7%, slightly higher than $\mathrm{\Delta }T/T=5.0\%$ at 1.17 ng/mL; however, this is much lower than $\mathrm{\Delta }T/T=21\%$ for sericin at 780 pg/mL (lowest concentration). The conclusion drawn from these results is that the HDOMM can serve as an ultra-sensitive qualitatively detecting biosensor. This sensor was used to successfully detect sericin with a detection limit of 780 pg/mL. In addition, two abnormal THz responses were found: first, the transparency window with sericin is significantly higher than that for the bare HDOMM; second, $\mathrm{\Delta }T/T$ for 1.25 μg/mL (highest concentration) is much lower than $\mathrm{\Delta }T/T$ for 780 pg/mL (lowest concentration). To understand the internal mechanism of the two abnormal THz responses, the coupled harmonic oscillator model is used to fit the experimental transmission spectra for the HDOMM biosensor at sericin concentrations from 780 pg/mL to 1.25 μg/mL.

Figures 3(d)–3(f) present the theoretical fitted transmission spectra. The fitting results agree relatively well with the measurement. Figures 3(g)–3(i) show the fitting parameters ${\gamma }_1$ and ${\gamma }_2$ as functions of sericin concentration from 780 pg/mL to 1.25 μg/mL. Compared with those of the bare sample, both the radiative damping of the bright mode ${\gamma }_1$ and non-radiative damping of the dark mode ${\gamma }_2$ dramatically decrease under the influence of sericin. This is because introducing sericin changes the dielectric environment in the HDOMM, thus causing the reductions in ${\gamma }_1$ and ${\gamma }_2$. The combined results in Fig. 2(b) suggest that the conductivity of graphene has decreased. Therefore, the transmission amplitude significantly increases at the transparency window. The second abnormal THz response is more clearly explained by calculating $\mathrm{\Delta }{\gamma }_{1/}{\gamma }_1=[({\gamma }_1^{\mathrm{bare}}-{\gamma }_1^{\text{sericin}})/{\gamma }_1^{\mathrm{bare}}]\times 100\%$ and $\mathrm{\Delta }{\gamma }_2/{\gamma }_2=[({\gamma }_2^{\mathrm{bare}}-{\gamma }_2^{\text{sericin}})/{\gamma }_1^{\mathrm{bare}}]\times 100\%$. The bigger $\mathrm{\Delta }{\gamma }_1/{\gamma }_1(\mathrm{\Delta }{\gamma }_2/{\gamma }_2)$ is, the more ${\gamma }_1({\gamma }_2)$ is reduced. The results are shown in Table 1.Table 1.

$\mathrm{\Delta }{\gamma }_{/}\gamma$ for Different Sericin Concentrations

In Table 1, both $\mathrm{\Delta }{\gamma }_1/{\gamma }_1$ and $\mathrm{\Delta }{\gamma }_2/{\gamma }_2$ are maximum when the sericin concentration is at its minimum of 780 pg/mL. Such a sensing response is mainly caused by the unique bio-doping, which makes the initial $E_F$ move to the Dirac point, leading to a sharp drop in the graphene conductivity. Figures 3(j)–3(l) show the sensing mechanisms of the HDOMM biosensor. CVD graphene is usually p-doped owing to the inevitable influence of impurities, defects, and disorder. The initial $E_F$ slightly deviates from the Dirac point and is in the valence band, as shown in Fig. 3(j). The initial $E_F$ is quite close to the Dirac point. As the 780-pg/mL sericin and the patterned graphene covalently bond, the initial $E_F$ shifts from the valence band to the Dirac point, as shown in Fig. 3(k). The carrier density correspondingly decreases in the patterned graphene. Therefore, the conductivity decreases. This results in the observed marked enhancement of the transparency window. However, when the sericin concentration is 1.17 ng/mL, the initial $E_F$ shifts from the Dirac point to the conduction band [see Fig. 3(l)]. The carrier density correspondingly increases, which enhances the conductivity. Therefore, the transparency window of the 1.17-ng/mL sericin is lower than that of the 780-pg/mL sericin. The results and theoretical analysis show that a substantial increase in the amplitude of the transparency window was achieved at the lowest sericin concentration; that is, ultrasensitive detection of sericin was achieved. The internal mechanism leading to this result is mainly that the $E_F$ of graphene has moved to the Dirac point under the drive of the sericin, causing a dramatic change in the dielectric environment of the EIT coupling. Therefore, the conclusion drawn is that the ultra-sensitive sericin sensing is mainly caused by the $E_F$ shifting to the Dirac point. The designed sensor is considered a Dirac-point-based biosensor. The sensitivity of this biosensor has been greatly improved compared with sensors of previous works (Table 2). Here, we are based on the interaction between external molecules and graphene to compare the sensitivity of previous works. Sericin molecules have no benzene ring-like structures and are without $\pi$-electrons. Therefore, it cannot form $\pi -\pi$ stacking with graphene. First, we compare the no $\pi -\pi$ stacking case. In Table 2, we can find that the sensitivity of the Dirac-point-based HDOMM sensor is as much as two orders of magnitude higher than that of detecting fructose in Ref. [43]. In general, if an external molecule has a benzene ring-like structure with $\pi$-electrons, it strongly interacts with the $\pi$-electrons of graphene through $\pi -\pi$ stacking [42]. Therefore, the graphene-based sensor shows higher sensitivity to the external molecules containing $\pi$-electrons [42]. However, the sensitivity of the Dirac-point-based sensor to detecting sericin molecules without $\pi$-electrons is one order of magnitude higher than that of detecting chlorpyrifos methyl molecules that form $\pi -\pi$ stacking with graphene in previous work [15,42]. Furthermore, the sensitivity of the Dirac-point-based HDOMM sensor is as much as two orders of magnitude higher than that of the sensor (Graphene+PI) whose analytes contain $\pi$-electrons in Ref. [43]. In summary, the HDOMM sensor has higher sensing performance.Table 2.

Comparison with Previous Works^a

Figures 4(a)–4(c) show the experimental phase spectra for the HDOMM at sericin concentrations from 780 pg/mL to 1.25 μg/mL. The reference point of the phase spectra for the HDOMM with sericin is the corresponding phase in dry air. In Figs. 4(a)–4(c), the phase spectra of the HDOMM with sericin are significantly different from those for the bare HDOMM. The obvious change in the phase spectra also facilitates the qualitative sensing of sericin. To clarify the sensing effect further, it is defined as $\mathrm{\Delta }P/P=|(P_{\text{sericin}}-P_{\mathrm{bare}})/P_{\mathrm{bare}}|$, where $P_{\text{sericin}}$ (${\mathit{P}}_{\mathrm{bare}}$) is the phase at the transparency window with (without) sericin. From Figs. 4(a)–4(c), $\mathrm{\Delta }P/P$ for sericin at 1.25 μg/mL (highest concentration) is 0.6, slightly lower than $\mathrm{\Delta }P/P=2.6$ at 1.17 ng/mL; however, this is much lower than $\mathrm{\Delta }P/P=10$ for sericin at 780 pg/mL (lowest concentration). The conclusion drawn from these results is that the HDOMM can serve as an ultra-sensitive qualitatively detecting biosensor. This sensor was used to successfully detect sericin with a detection limit of 780 pg/mL. As is the case for the phase-based sensing effect, and the phase spectra undergo the maximum change at the minimum sericin concentration of 780 pg/mL. The internal mechanism is also mainly that the $E_F$ of graphene has moved to the Dirac point, causing a dramatic change in the dielectric environment. Figure 4(d) shows the mechanism through which perovskite facilitates phase-based sensing. When terahertz waves pass through graphene, ${\phi }_0$ changes to ${\phi }_0+\mathrm{\Delta }\phi$ owing to the biological doping in graphene, where ${\phi }_0$ is the initial phase of the terahertz wave, and $\mathrm{\Delta }\phi$ is the change in phase caused by the change in conductivity. After the terahertz waves pass through perovskite, ${\phi }_0+\mathrm{\Delta }\phi$ changes to $({\phi }_0+\mathrm{\Delta }\phi )+(\omega /\mathit{c})nl$, where $\omega$ is the frequency, $c$ is the speed of light, $n$ is the refractive index of perovskite, and $l$ is the thickness of the perovskite. Because perovskite has a relatively high refractive index, Figs. 4(a)–4(c) show that the phase of terahertz waves [$({\phi }_0+\mathrm{\Delta }\phi )+(\omega /\mathit{c})nl$] greatly changes with the sericin concentration. Therefore, perovskite plays an important role in phase-based sensing.

Ultra-sensitive and qualitative detection of sericin has been demonstrated using a new patterned graphene–PI–perovskite integrated terahertz metasurface as a platform for Dirac-point-based biosensing. The changes in amplitude and phase were used to successfully detect sericin with a detection limit of 780 pg/mL. The sensitivity of this sensor is as much as one order of magnitude higher than that of sensors in published works. The internal mechanism of ultra-sensitive biosensing was explained via simulation, coupled-harmonic-oscillator model fitting, and theoretical analysis of the changes in the graphene $E_F$. It was found that as 780-pg/mL sericin and the patterned graphene covalently bond, the initial $E_F$ shifts from the valence band to the Dirac point. Correspondingly, the carrier density decreases in the patterned graphene, which decreases the conductivity. This results in marked enhancement of the transparency window. Thus, ultra-sensitive and qualitative detection of sericin was achieved. This work could lead to a wide range of ultra-sensitive and qualitative detection of biomolecules based on the Dirac-point-based biosensor.