Tunable Resonant‐Photopyroelectric Detector Using Chalcogenide−Metal−Fluoropolymer Nanograting

Pyroelectric detectors are often broadband and require external filters for wavelength‐specific applications. This paper reports a tunable, narrowband, and lightweight pyroelectric infrared detector built upon a flexible membrane of As2S3−Ag−P(VDF‐TrFE) with subwavelength grating, which is capable of both on‐chip filtering and photopyroelectric energy conversion. The top surface of this hybrid membrane is a corrugated As2S3−Ag film contributing to narrowband light absorption in the near‐infrared (NIR) regime, and the bottom part is a polyvinylidene fluoride‐trifluoroethylene (PVDF‐TrFE) membrane for the conversion of the absorbed light to an electrical signal. Uniquely, applying a bias voltage to the PVDF‐TrFE membrane enables the tuning of the device's absorption and pyroelectric characteristics owing to the piezoelectrically induced mechanical bending. The resonator exhibited a resonant absorption coefficient of 80% and a full‐width‐half‐maximum of 15 nm within the NIR, a responsivity of 1.4 mV mW−1, and an equivalent noise power of 13 µW Hz−1/2 at 1560 nm. By applying a 15‐V bias to the PVDF‐TrFE membrane, the absorption coefficient decreased to 18% due to the change in the grating period and incident angle. The narrowband and tunable features of the As2S3−Ag−P(VDF‐TrFE) pyroelectric detector will benefit a variety of potential applications in sensors, optical spectroscopy, and imaging.


Introduction
Photopyroelectric detectors can absorb electromagnetic radia tion, convert the absorbed energy into thermal energy, and generate a pyroelectric voltage (PEV) across a pyroelectric material. [1] They have been widely used in various applications, fluoridetrifluoroethylene (PVDFTrFE) copolymer, which has a pyroelectric coefficient of approximately 0.025 C m −2 K −1 . [25] Here, the RPED was designed to absorb nearinfrared (NIR) radiation around 1560 nm and generate PEV across the PVDF TrFE membrane. The RPED structure was numerically modeled and experimentally characterized for its absorption signatures, PEV spectra, responsivities, and inputoutput rela tionships. In addition, the RPED structure can be tuned using a bias voltage applied to the PVDFTrFE membrane. The bias voltage induced the membrane bending and thus changed the grating period and incident angle. The tuning capability of RPED's absorbance and its PEV output at a specific wave length was investigated.

Design and Fabrication of the R-PED
The design goal is to integrate the functions of narrowband light absorption, pyroelectric sensing, and piezoelectric tuning into one membrane. Figure 1a illustrates the RPED structure, which is built upon a 15µmthick membrane. The PVDFTrFE membrane was patterned with the subwavelength 1D grating structure. On top of the PVDFTrFE grating, the thin film stack of Ag and As 2 S 3 layers can effectively absorb NIR light at a specific wavelength. The NIR absorption can locally heat up the membrane, change the PVDFTrFE's polarization, and generate a PEV across the Ag contacts on the top and bottom surface of the PVDFTrFE membrane. The As 2 S 3 /Ag can be considered as an asymmetrical metalcladding waveguide that supports the leaky mode resonance. As illustrated in Figure 1b, when the temperature is stabilized, the PEV signal drops back to zero. Without the light exposure, the device starts to cool down and outputs an opposite PEV signal. The response time and PEV amplitude largely depend on the thermal mass and heat capacity of the device. To reduce thermal mass, the RPED was developed using the 15µmthick PVDFTrFE membrane. At the resonant wavelength, the incidence of light can be cou pled into the waveguide via the grating modulation, and the loss of Ag cladding causes a highly efficient light absorption shown in Figure 1c. During the optical heating phase, the PEV signal rises and reaches its maximum value at the equilibrium temperature.
The membrane RPED was fabricated using a solvent assisted imprint lithography process followed by the evapora tion of Ag and As 2 S 3 thin films. The details of the fabrication process are described in the Experimental Section, and the main steps are summarized in Figure S1, Supporting Infor mation. Briefly, the PVDFTrFE raw material was dissolved in dimethylformamide (DMF) and spun onto an Agcoated glass substrate. The sample viscosity and spin speed were controlled to obtain the PVDFTrFE thickness of 15 µm. The coated PVDF TrFE film was then imprinted at 175 °C using the polydimethyl siloxane (PDMS) mold that carried the opposite grating pattern. After the imprint, the 100nm Ag and 300nm As 2 S 3 films were Figure 1. Design, operation, and fabrication of the tunable R-PED. a) Schematic illustration of the R-PED on a free-standing PVDF-TrFE membrane (left). A DC voltage can deform the membrane (right) and tune the R-PED's absorption. b) Resonant PEV signal generation. The modulated optical excitation causes the heating and cooling of the device and generates positive and negative PEV pulses. c) Tunable absorbance of the As 2 S 3 /Ag/PVDF-TrFE grating. The absorption spectra of a flat and bent R-PED are shown as the black and red curves, respectively. d) Photograph of the R-PED (left), AFM image (center), and cross-sectional SEM image (right) of the As 2 S 3 /Ag/PVDF-TrFE grating. deposited using an ebeam evaporator. The stack consisting of Ag/PVDFTrFE/Ag/As 2 S 3 was carefully peeled off from the glass substrate to form the RPED membrane. The photog raphy, atomic force microscope (AFM), and scanning electron microscopy (SEM) images in Figure 1d shows the fabricated RPED with a grating period of 1 µm and depth of 300 nm.

Numerical Design of the R-PED
To design an RPED that can effectively absorb the telecom wavelength around 1.55 µm, the rigorous coupledwave anal ysis (RCWA) was used to model the As 2 S 3 /Ag/PT grating. The details of the RCWA simulation model are given in the Experimental Section. Figure 2 shows the calculated absorp tion spectra and near field distributions for the RPED with the grating period, grating depth, Ag thickness, and As 2 S 3 thickness of Λ = 1 µm, d = 300 nm, t Ag = 100 nm, and t As2S3 = 300 nm, respectively. For the transverse electric (TE) mode, whose electric field is polarized along the xaxis, the absorption peak resides at the wavelength of λ TE = 1560 nm and incidence angle of θ i = 0° with a fullwidth halfmaximum (FWHM) of 15 nm, as shown in Figure 2a. The resonant feature is sensitive to the incidence angle. The increase of θ i resulted in the splitting and shifting of the resonant peaks towards red and blue spectral regions. The electric field distributions at given coupling conditions were calculated and plotted in Figure 2b. It can be seen that the near field of the TE mode (|E y | 2 ) is confined in the As 2 S 3 layer and significantly enhanced compared with the intensity of the incidence field. When the incidence light is coupled into a guidedmode resonance mode, the material loss of the Ag cladding causes the resonant absorp tion and effectively converts photon energy into heat. [22] In contrast, the right panel of Figure 2b shows the near field dis tribution at λ = 1560 nm and the incidence angle of θ i = 10°. Without resonant absorption, the local field intensity is weak, and lighttoheat conversion capability is limited. For the trans verse magnetic (TM) modes, the absorption resonances locate around λ TM = 1750 nm, as shown in Figure 2c. Figure 2d shows the near field distributions (|E x | 2 + |E z | 2 ) for two TM modes at λ TM = 1750 nm and θ i = 0° and 10°, respectively. Compared with the TE modes, the electric fields of the TM resonances mainly reside at the As 2 S 3 Ag interface and exhibit a larger FWHM of 100 nm since the TM modes are associated with the surface plasmon resonance. Because the TE resonances exhibit an absorption resonance with narrower linewidth, the TE modes were chosen to develop the RPED.
To estimate how the absorbed optical energy can raise the RPED's temperature, we modeled the heat transfer process using a finite element method (FEM) simulation. The details of the thermodynamic FEM simulation are described in the Experimental Section. Figure 3a showed the temperature Adv. Optical Mater. 2021, 9, 2101147 Figure 2. Numerical design of the resonant absorption grating. a) Simulated absorption spectra of the As 2 S 3 /Ag/PVDF-TrFE grating with a TE polarized incidence and θ i = 0°, 5°, and 10°, respectively. b) Calculated near field distributions for the TE modes with the absorption resonance when λ i = 1560 nm and θ i = 0° (left panel) and without the resonance when λ i = 1560 nm and θ i = 10° (right panel). The electric fields were plotted within one period of the grating. c) Absorption spectra of the TM mode resonances with θ i = 0°, 5°, and 10°, respectively. d) The local field distributions are associated with the TM modes and without the resonance effect in the left and right panels, respectively. Scale bar: 300 nm. distribution around the RPED membrane when the device was heated using a NIR laser beam (λ = 1560 nm and P = 7.26 mW) at its center. According to the result of prior electromagnetic simulation, we assumed the absorption coefficient of 80% and placed a 5.8 mW at the 1mmdiameter heating spot. Under the ambient temperature of 20 °C, the NIR absorption can increase the membrane to 22.5 °C for the 15µmthick mem brane. Figure 3b plots the dynamic temperature responses measured at the center of the membrane when the membrane was heated and subsequently cooled by shutting off the light source. The amount of temperature change and response time is mainly determined by the membrane thickness. As shown in Figure 3b, three membranes with a thickness of 15, 20, and 25 µm were simulated. For the 15µmthick membrane, the temperature can reach the equilibrium temperature of 22.5 °C within 32 ms.

Optical Characterization of the NIR R-PED
The resonant absorption features of the RPED strongly depend on the coupling conditions, such as the wavelength and incident angle of the excitation light. Only when the coupling conditions are met, the device can absorb the excitation effec tively. To characterize the absorption signatures, the optical dis persion diagram of the device was measured using the optical setup illustrated in Figure 4a. The sample was illuminated using a TEpolarized broadband light, and the reflectance (R(λ)) was analyzed using a fibercoupled spectrometer. Because the Ag film's thickness exceeded its skin effect depth of 4.5 nm at λ = 1560 nm, there was no transmission through the device, and T(λ) = 0. The absorbance of the device was calculated using A(λ) = 1−R(λ). Figure 4b compares the measured and simulated absorption optical dispersion diagrams of the RPED struc ture shown in Figure 1. The incident angle and wavelength ranged from 0° to 10° and 1450 to 1700 nm, respectively. For the TEmodes, two resonant bands can be seen across the NIR range. For the normal incidence at θ i = 0°, the device's absorp tion peak is located at λ r = 1560 nm with the FWHM of 16 nm ( Figure 4c). By tuning θ i from 0° to 5°, the absorption resonance split into two resonances and moved to λ r = 1580 and 1550 nm, respectively. The resonance modes can be shifted more by increasing the θ i to 10°, as shown in Figure 4c. The upper and lower absorption bands showed the dispersion slopes of 4 and −2 nm/°, respectively.

Resonant Pyroelectricity
The PEV output of the device depends on the intensity, polari zation, wavelength, and angle of incidence of the incoming light. We measured the PEV output signals using an oscil loscope when the RPED device was illuminated by a tunable laser. The laser beam was collimated and polarized to excite the TE modes. Figure 5a plots the device output as a function of the time when the laser's intensity was modulated using a 50ms period square wave. The peak PEV was V max = 10 mV with the laser's wavelength, power, and incidence angle of λ = 1560 nm, P = 7.26 mW, and θ i = 0°, respectively. At the red sections shown in Figure 5a, the laser light was on, and the device was heated. The device's PEV output reached its maximum value within 10 ms, representing a temperature rise of approximately 2.4 °C. After the temperature stabilized at equilibrium, the PEV signal dropped back to zero in 10 ms. When the laser was turned off, the device's temperature started to decrease to the ambient temperature during the cooling phase (the blue sec tions in Figure 5a). The temperature drop resulted in an oppo site PEV with a V min = −7 mV and peak to peak width of 0.1 s. The resonanceenhanced PEV phenomena on the PEV output were investigated for heating and cooling phases, respectively.
The spectral responsivities of the device were character ized at the incidence angle of θ i = 0°, 5°, and 10°, as shown in Figure 5b,c. At each angle of incidence, the laser wavelength was tuned from 1520 to 1620 nm with an increment of 5 nm.    was illuminated by a TEpolarized and intensity-modulated laser beam at λ i = 1560 nm. b,c) Measured peak and dip PEV signals versus wavelength when the laser was turned on and off, respectively. The laser power was kept at P = 7.26 mW, and the emission wavelength was scanned from 1520 to 1620 nm. The R-PED's spectral response was measured at θ i = 0°, 5°, and 10°, respectively. d) Input and output relationship when the device was excited at the resonance condition of λ r = 1560 nm and θ i = 0° (black) and of the resonance condition of λ r = 1560 nm and θ i = 10°. which was five times stronger than the offresonance PEV of 0.25 mV mW −1 at λ = 1620 nm (Figure 5b). The increase of inci dent angle to θ i = 5° resulted in two peaks at λ r1 = 1540 nm and λ r2 = 1570 nm in the response curve. At θ i = 10°, one of the PEV peaks shifted to 1590 nm, and the other one moved out of the laser's tuning range. The PEV spectra agree well with the RPED's absorbance spectra shown in Figure 4c. During cooling phases, the polarity of the PEV signal was inverted, and the negative resonant PEV peaks can be seen in Figure 5c. Figure 5d compares the PEV inputoutput relationships for the onresonance (θ i = 0° and λ i = 1560 nm) and offresonance (θ i = 10° and λ i = 1560 nm) cases, respectively. The laser emis sion power scanned from 2.5 to 7.25 mW, and the peak PEV signals were plotted. For both cases, the PEV output varied lin early versus the excitation laser power. The onresonance input output relationship showed a slope of 1.36 mV mW −1 , which is 45 times higher than the offresonance slope of 0.03 mV mW −1 . Since the Johnson-Nyquist noise was the dominant noise source, the noise spectral density (NSD) is 4 B NSD Rk T = , where T is the room temperature of 20 °C, R = 470 Ω is the device resistance, and k B is the Boltzmann's constant. The NSD of the RPED sensor was 6 µV Hz −1/2 at room tempera ture. The noise equivalent power (NEP) can be calculated using the NSD divided by the voltage responsivity. The NEP spectra of the device at different angles of incidence are plotted in Figure S4, Supporting Information. The minimal NEP value of 13 µW Hz −1/2 was found at the resonance of λ r = 1560 nm and θ i = 0 °.

Tuning of Resonant Absorption
As a piezoelectric material, the PVDFTrFE membrane pro duces mechanical stress when a bias voltage is applied across it. The strain mismatch between the membrane and Ag/As 2 S 3 thin films can bend the membrane, as illustrated in Figure 6a. The controlled membrane deformation can be employed to tune the resonance absorption characteristics by increasing the grating period and the angle of incidence, which is similar to the piezoelectric photonic crystal resonators. [26] Figure 6b compares the FEM simulation results of membrane displace ments when the bias voltage of 0, 5, 10, and 15 V were applied. The membrane was suspended by fixing two edges along the yaxis, and the membrane edges along the xaxis were free to move. The amount of membrane displacement along the zaxis was present by the color scale. Figure 6c   by measuring the laser beam deflection. The details of the deflection angle measurement are given in the supplementary materials ( Figure S2, Supporting Information). The reflection angle θ = sin −1 (0.5L'/d), where d represents the membrane dis placement along the zaxis, was calculated by simulation. Then the L', which is the length of the membrane after bending, can be calculated, as shown in Figure 6c. Figure 6d compares the measured absorption spectra of TE modes with the bias voltage ranging from 0 to 20 V. Based on results, we calcu lated the change of the grating period, which is proportional to the membrane length along the yaxis: ΔΛ = (L'−L)/L, where L is the length of the membrane, the period changing of the grating is about 0.8 nm V −1 . Without a zero bias, the absorp tion resonance is located at λ r = 1560 nm with an FWHM of 16 nm. Owing to the voltageinduced grating period and radius changes, the resonance mode split and shifted in the way given by the dispersion relationship (Figure 4c). With the bias voltage of 20 V, the absorption peak moved to 1597 and 1540 nm. For the resonance modes that shifted towards the red and blue directions, the voltage tuning sensitivities are 1.85 and −2 nm V −1 , respectively.

Tuning of R-PED's PEV Output
The capability of controlling the device's absorption enabled the development of a tunable RPED. Figure 6e shows the absorption spectra of the tunable PVDF sensor when the bias voltage (from 0 to 15 V) was applied to the device, and Figure 6f presents the PEV outputs versus bias voltage for heating and cooling processes. To demonstrate the tunable pyroelectric effect, we applied the DC bias voltage to bend the membrane and, in the meantime, measured its PEV output using the oscilloscope. The device was illuminated using the NIR laser (λ = 1560 nm, θ i = 0°, and P = 7.26 mW), whose intensity was modulated at 10 Hz. Figure 6e plots the RPED's dynamic output. Without the bias voltage, the device can absorb the laser light efficiently and generate strong PEV output of 10.2 mV. When the bias increased from 0 to 15 V, the device absorption was reduced, and the PEV output consequently decreased from 10.2 to 2.9 mV. At 20 V, the device absorption was only 18% at 1560 nm due to the resonance shifting, and the PEV signal was too weak to be distinguished from its noise floor.

Conclusion
The narrowband pyroelectric sensor demonstrated in this work seamlessly integrated the resonant grating absorber and the pyroelectric material to detect the change of NIR excitation. The Agclad As 2 S 3 grating was able to absorb radiation around 1560 nm with a bandwidth of 15 nm and effectively convert the absorbed optical energy into heat. The lightinduced tem perature variations generated PEV signals from the PVDFTrFE membrane. The results showed that the PEV signal is sensitive to the coupling of infrared radiation and the absorption reso nance modes. With the resonant absorption at 1560 nm, the device showed the responsivity of 1.4 mVmW −1 , which was fifty times higher than the device responsivity without incurring any resonance mode. Furthermore, we studied the tuning capability of the resonant absorber by applying a bias voltage across the PVDFTrFE membrane. Because the bias voltage can induce internal mechanical stress and deform the membrane, the reso nant absorption can be tuned within the range of 50 nm and sensitivity of 1.85 nm V −1 . The tunable resonant absorption was utilized for the development of the tunable RPEDs.
In future work, the device performances, such as the response time, bandwidth, and sensitivity, will be improved from the following three aspects. First, materials with a higher pyroelectric coefficient will be adopted in the RPEDs to replace the PVDFTrFE substrate. The resonant grating structure can be fabricated on thin films with a high pyroelectric coefficient, such as barium titanate, aluminum nitride, cesium nitrate, and lithium tantalite, to enhance PEV output. Second, by opti mizing the grating structures, the bandwidth and response time of the RPED can be further reduced. Last, an array of RPEDs with different spectral bands can be integrated on one chip to detect multiple wavelengths simultaneously for multi band sensing applications.

Experimental Section
Fabrication Process: The hot embossing approach, shown in Figure S1, Supporting Information, was used to fabricate the subwavelength grating structure in the PVDF-TrFE membrane. [27] The 1-µm-period grating was replicated from a holographic grating film (#40-267, Edmund Optics) to the PDMS mold. Before the embossing process, the PVDF-TrFE film was spun onto a silver-coated glass substrate. The 100-nm silver layer was evaporated using an electron beam evaporator (BJD-1800, Temescal) to facilitate the release of the PVDF-TrFE membrane. The PVDF-TrFE powder was dissolved in DMF at 75 °C to 15%(w/v%). The dissolved PVDF-TrFE was spin-coated on the silver-coated glass slide at a spinning speed of 4000 rpm for 45 s. After the spin coating, the sample was baked at 160 °C for 4 h to completely evaporate DMF. Then, the film was embossed using the PDMS mold at the glass transition temperature of PVDF (≈ 175 °C) on a hotplate under a pressure of 80 kPa for 20 s. After being cooled down to room temperature, the PDMS mold was peeled away to leave the grating pattern on top of the PVDF-TrFE film. Following the molding process, the Ag and As 2 S 3 thin films were deposited using the evaporator. The thickness and refractive index of the Ag and As 2 S 3 layers were measured using a spectroscopic ellipsometer (J.A. Woollam Co., Inc.). To release the membrane, the stack of Ag/PVDF-TrFE/ As 2 S 3 /Ag films were carefully peeled off from the glass substrate and suspended on a 5 mm × 5 mm frame for tests.
Numerical Modeling: The RCWA simulation model consisted of a single period of the 1D grating pattern along the x-axis. The refractive index (n(λ)) and extinction coefficient (k(λ)) of Ag and As 2 S 3 thin films were interpolated using the results obtained from the ellipsometry measurement as shown in Figure S3, Supporting Information. The incidence light was linearly polarized along the x-axis and y-axis for the TM and TE modes, respectively. The model output reflectance (R(λ)) and transmittance (T(λ)) were in the wavelength range of 1400 to 1700 nm. The absorption spectra were calculated using A(λ) = 1−R(λ)−T(λ). At the resonance wavelength, the near field distribution was plotted using the total electric field of E( , ) | | | | | | 2 2 2 x z E E E x y z = + + . The temperature distribution profile across the PVDF device was modeled using finite element analysis (COMSOL Multiphysics 5.3). The simulation domain contained the membrane stack of PVDF, silver, and As 2 S 3 layers with the thickness of 15 µm, 100 nm, and 300 nm, respectively. The simulation domain was discretized using 4-noded tetrahedral meshes and was truncated using the open boundary on all sides of the membrane. The open boundary meant the membrane was