Self‐powered broadband kesterite photodetector with ultrahigh specific detectivity for weak light applications

Kesterite Cu2ZnSn(S,Se)4 (CZTSSe) is a promising candidate for photodetector (PD) applications thanks to its excellent optoelectronic properties. In this work, a green solution‐ processed spin coating and selenization‐processed thermodynamic or kinetic growth of high‐quality narrow bandgap kesterite CZTSSe thin film is developed. A self‐powered CZTSSe/CdS thin‐film PD is then successfully fabricated. Under optimization of light absorber and heterojunction interface, especially tailoring the defect and carrier kinetics, it can achieve broadband response from 300 to 1300 nm, accompanied with a high responsivity of 1.37 A/W, specific detectivity (D*) up to 4.0 × 1014 Jones under 5 nW/cm2, a linear dynamic range (LDR) of 126 dB, and a maximum Ilight/Idark ratio of 1.3 × 108 within the LDR, and ultrafast response speed (rise/decay time of 16 ns/85 ns), representing the leading‐level performance to date, which is superior to those of commercial and well‐researched photodiodes. Additionally, an imaging system with a 905 nm laser is built for weak light response evaluation, and can respond to 718 pW weak light and infrared imaging at a wavelength as low as 5 nW/cm2. It has also been employed for photoplethysmography detection of pulsating signals at both the finger and wrist, presenting obvious arterial blood volume changes, demonstrating great application potential in broadband and weak light photodetection scenarios.


INTRODUCTION
2][3] Silicon PDs are commonly used for visible (VIS) light detection due to their low cost and excellent performance. 4However, these PDs have a restricted intrinsic detection range, a relatively high dark current, and poor specific detectivity for NIR light detection.6][7] The abovementioned NIR PDs are mostly restricted to military and industrial applications due to their high cost and complexity in production, restricting their extensive use among civilians. 8Therefore, the development of new materials enabling PDs with quickly switchable properties between ON/OFF, and high specific detectivities in the VIS and NIR ranges remain one of the key challenges. 9,10esterite compound semiconductors, including Cu 2 ZnSn(S,Se) 4 (CZTSSe), Cu 2 ZnSnS 4 (CZTS), and Cu 2 ZnSnSe 4 (CZTSe), have been progressively allowing new thin-film optoelectronics, particularly in solar cells, 11 solar water splitting photocathodes, 12 and PDs. 13 The extensive applications of kesterite may be attributed to its exceptional optoelectronic properties, including high absorption coefficient, eco-friendly composition, a suitable bandgap, and thermodynamically stable structure. 14hotogenerated carriers in polycrystalline kesterite films have a recombination lifetime of 10 ns or less, indicating that they can be easily extracted by a built-in field or an external electric field. 15Accordingly, kesterite-based PDs are expected to attain excellent photoresponsivity.A CZTS-based photoconductive detector with a responsivity of 13 A/W (at 5 V bias) and a response speed of 130/700 ms in the NIR region has been developed in recent years. 16e-doping for CZTSSe bandgap tuning exhibited responsivity (18 mA/W), detectivity (10 9 Jones), response speed of 82/97 ms, and a broadband response from the VIS to the NIR region, improving the detection level at the NIR region.17 Na-doping can further improve the response speed to 1.06 ms in the NIR region.18 Moreover, nanostructured heterojunction PDs with black-Si/CZTS nanocrystals demonstrated a high responsivity of 0.411 A/W, a detectivity of 10 12 Jones, and a fast response speed of 86.8 μs.19 In 2019, self-powered PDs based on CZTSSe/CdS heterojunctions have been developed, accompanied with a responsivity of 0.39 A/W, a detectivity of 2.04 × 10 11 Jones, and a fast response speed of 240/340 ns in the wavelength range of 300-1100 nm. 13 Despite such efforts, self-powered kesterite-based PDs functioning in the NIR region still have much room for improvement.A photovoltaicpowered PD is closely related to the light absorber layer (e.g., crystallinity, composition, and bulk defect density), PN heterojunction quality (e.g., energy band alignment and interface defect density), and back-contact barrier, which determine the process of carrier generation, excitation, relaxation, transfer, and recombination.7][18][19] Our group recently reported efficient solution-processed CZTS (12.6% efficiency) 20 and CZTSSe (12.66% efficiency) 21 thin-film solar cells, representing state-of-the-art kesterite solar cells.Therefore, understanding the photoresponse enhancement mechanism based on a significantly great photovoltaic effect is imperative for achieving high-performance, self-powered CZTSSe PDs.
Solution-processed self-powered CZTSSe/CdS PN heterojunction PDs were fabricated in this study, wherein, CZTSSe light-absorbing layer as a P-type semiconductor and a CdS buffer layer as an N-type semiconductor.These PDs exhibited broadband (300-1300 nm) detection, accompanied with a highest responsivity (R) of 1.37 A/W, linear dynamic range (LDR) of 126 dB, ultrahigh I light /I dark ratio of 1.3 × 10 8 within the LDR, ultrafast response speed with a rise/decay time of 16 ns/85 ns, and a record specific detectivity (D*) of 4.0 × 10 14 Jones for the best-performing VIS and NIR photodiodes.An imaging system was also built with a 905 nm laser, and it could exhibit weak light infrared imaging as low as 5 nW/cm 2 .The outstanding photo-response performance was attributed to the strong built-in field under a highquality crystalline absorber layer, defect passivation, and energy band alignment.These results suggest that the selfpowered kesterite thin-film PDs with ultrahigh specific detectivity have huge application potential in weak light and/or broadband photodetection scenarios.

RESULTS AND DISCUSSION
Figure 1A shows the cross-sectional scanning electron microscopic (SEM) image of the as-fabricated Mo/CZTSSe/CdS/ITO thin-film PD and a schematic of the photo-response measurement.][22] Accordingly, our photovoltaic-powered PDs enabled ultrafast, self-powered, and broadband imaging with ultrahigh specific detectivity for weak light imaging due to their highly crystalline light absorber, distinguished device interfaces, energy band alignment, and bulk and interface defect passivation.For thin-film optoelectronic devices, a light absorber layer with proper thickness and large, compact crystal grains is essential.Herein, the thickness of the CZTSSe absorber layer was regulated to explore CZTSSe thin-film PDs with simultaneous high response speed and responsivity, and denoted as PD-A with 800 nm, PD-B with 1200 nm, and PD-C with 1600 nm.The morphologies of the CZTSSe films and device cross-section were characterized by SEM (Supporting Information Figure S2).Some pinholes can be found on the surface of the PD-A CZTSSe thin film.The PD-B thin film with suitable thickness results in a more compact crystalline feature with an average grain size of approximately 1.5 μm, which is beneficial in inhibiting recombination loss at the grain boundaries (GBs). 23After the thickness increased to 1600 nm, the insufficient reaction resulted in a double-layer structure with fine grains at the bottom, which limited device performance due to the severe charge recombination caused by the numerous GBs that photo-generated carriers must encounter. 21,24The X-ray diffractometer (XRD) pattern presents three major sharp diffraction peaks of (112), (204), and (312), indicating the absence of any observable peaks out of the standard kesterite CZTS (JCPDS Card No. 52-0868) or CZTSe (JCPDS Card No. 26-0575) pattern.When the XRD and Raman analysis are combined, all the peaks that appeared in the spectra can be linked to kesterite CZTSSe, and no other impurity phases are detected, which validated the high crystallinity and purity of the as-prepared kesterite CZTSSe thin films.Figure 1B shows the desirable spike-like energy band alignment in the CZTSSe/CdS heterojunction after optimizing the light absorber layer thickness.Wherein, the conduction band (E C ), valence band (E V ), and Fermi levels (E F ) of CZTSSe were determined regarding the secondary electron cutoffedge and valence band (VB) position, according to the ultraviolet photoelectron spectroscope (UPS) spectra (Figure S3 and Supplementary Note 1).Additionally, the energy level information of CdS can be derived from our previous work. 23Thus, a positive conduction band offset of 0.3 eV for CZTSSe/CdS is within the optimal value range of 0-0.4 eV belonging to a high-quality PN heterojunction, which contributes to the more efficient transfer of the photo-generated charge carriers by a strong built-in field at the depletion region, thus, generating the photo response under optical signal from UV to NIR. 24 Responsivity (R) is an important parameter used to measure the performance of PDs, which represents the ability to convert incident optical signals into electrical signals.This parameter is usually calculated by using the following formula 3 : where I light represents light current, I dark represents dark current, and P in represents the power of incident light.The UV-VIS-NIR wavelength range is covered by the photoresponses of all the CZTSSe PDs in Figure 1C, which span from 300 to 1300 nm.Furthermore, the PD-B PD possesses much superior responsivity compared with those of other PDs, with the highest R value of 615.4 mA/W at a wavelength of 930 nm.To confirm the ability of weak light detection in the NIR range (Figure 1D), with the light intensity increase from 0.72 nW (5 nW/cm 2 ) to 1.4 mW (10 mW/cm 2 ), the weak light signal can be tested as low as 5 nW/cm 2 with the highest responsivity of 1.37 A/W, while the light current (I light ) gradually increases from 0.98 nA to 0.93 mA (at 0 V).To more clearly illustrate this fluctu-ation in photocurrent caused by light power density, the experimental data were fitted using the following power equation 25 : where θ is an exponent that reflects how electron-hole pairs are generated, separated, and recombined, and α is a constant at a certain wavelength.The result shows that the photocurrent has a quasi-linear fluctuation that is closely correlated with light intensity.The fitting ideal factor (θ = 0.97) is quite close to the ideal value of one, indicating efficient production of charge carriers from the absorbed photoflux and little recombination loss at the interface and/or surface under low trap states.Furthermore, the calculated R values maintain at around 0.7 A/W within light intensity from 10 −7 to 10 −3 W, which is reasonable to occur since the charge carriers excited by light will be filled into the trap states to cause photocurrent saturation.The particular LDR can also be derived from the I ph −P curves according to the following formula 3 : where P max and P min represent the highest and lowest incident power, respectively, within a linear current power area.Consequently, PD-B in the self-powered mode can achieve an LDR of 126 dB, implying the excellent light response characteristics across a broad intensity range.
Figure 1E shows the champion PD-B PD's semilogarithmic I-V curves under 905 nm laser illumination.The result shows a clear rectifying feature with a rectification ratio up to 2574 (Supporting Information Figure S4), which demonstrates a good diode quality for CZTSSe/CdS heterojunction and belongs to the typical behavior of photovoltaic driving PDs.Specifically, the extremely low dark current of 5.9 pA can be found, and the photocurrent reaches 0.98 nA with more than two orders of magnitude under 5 nW/cm 2 weak light illumination by a 905 nm laser, indicating the excellent weak light detection ability.Under a certain light intensity, the photocurrent exhibits a steady state from a given negative bias of −0.5 to 0 V, even to a certain positive bias, which results in a record I light /I dark ratio of 1.3 × 10 8 within the LDR (Supporting Information Figure S5), demonstrating a superior self-powered operation ability via a built-in electric field formed at the CZTSSe/CdS heterojunction.Without any encapsulation, the critical operational stability of these self-powered PDs was assessed by employing a signal generator to regulate the light illumination on the device.After multiple cycles, the light response is still the same, and there are barely any differences between the dark and light currents throughout the whole test period (Supporting Information Figure S6),  demonstrating a dependable operation stability.In addition, our CZTSSe-based PDs continue to display stable performance at a test temperature of 100, -50 • C and with an indoor storage for half a year, paving a bright avenue to tolerate some extreme application environments (Figure 1F).Another important measure that impacts the ability of a PD to track a rapidly varying optical signal is its inherent response speed. 26Figure 1G shows that the rising and decay times of the PD-B are 1.1 and 2 μs, which are close to the modulation limit of the 905 nm laser.Considering the resolution of the modulation frequency or bandwidth and the saturation characteristic of the lateral photovoltage for a continuous-wave laser, a pulsed laser should be utilized.The more realistic and ultra-fast response speeds of the PD-B are 16 and 85 ns (Figure 1H).All the samples are longer than the laser pulse width (herein, 3 ns), demonstrating the consistent response speed of our PDs.Based on the response time of the PD-B, its 3 dB cut-off frequency (f 3dB ) is calculated by the following formula 26 : The 3 dB bandwidth of the PD-B is as high as 4.1 MHz, representing the highest level of the CZTSSe-based PDs.
The key parameters of our PDs are summarized in Table 1 and a comparison to the literature that reported kesterite-based PDs and the representative PbS, perovskite, graphene-based, and commercial PDs.The specific detectivity (D*) depicts the sensitivity of PDs to weak optical signals, determining their ability for light detection, imaging, optical communication, etc.The specific detectivity was calculated according to the following two equations 25 : In Equation ( 5), q represents the unit charge, J dark represents the dark current density, and the  * 1 value calculated by this equation is 4.0 × 10 14 Jones.In Equation (6), S represents the device area, Δf represents the electrical bandwidth (0.5 Hz), I noise represents the noise current, and the D* value calculated by this equation is 3.9 × 10 13 Jones.As shown in Figure 2A and Table 1, our best-performing CZTSSe/CdS PD can achieve high detectivity, high responsivity, ultra-fast response speed, and high "ON/OFF" switching ratio, covering the UV-VIS-NIR wavelength range, which is also superior to those of well-researched and commercial photodiodes.An NIR imaging system (Figure 2B) was built to further prove the feasibility of NIR detection and weak light imaging.Ten devices were used to form a 2 × 5 array, and each device has 4 × 4 pixels.A hollow mask with "SZU" patterns was positioned between the PDs and a 905 nm laser source.The laser moved on a step of 4 mm along the X or Y axis and was controlled by the signal generator to form an exposure-like optical switch.Then, the current signals of the corresponding pixels were collected by a source meter and recorded on the computer.A legible picture of the "SZU" pattern (8 × 20 steps) with a striking color contrast that generated with different light powers (5.32 nW/cm 2 , 10.2 μW/cm 2 , and 12.5 mW/cm 2 ) further confirmed its strong photocurrent mapping capability and NIR imaging potential, even in an extremely weak light illumination (Figure 2C). Figure 2D shows the transient dark current and root-mean-square current at zero bias, which may be used to explore the lowest detectable light intensity that the CZTSSe-based PDs can measure.The root-mean-square current can be used to calculate the electronic noise (I rms ), and the corresponding formula is as follows 3,27 : where I j is a discrete current value, and Ī is the average value of I j .The calculated I rms value is 50 fA.Figure 2E shows the measured current noise of the CZTSSe-based PDs as a function of frequency at 0 V bias.The calculated shot noise, 1/f noise, and G-R noise are also provided.The calculated D* value decreasing from 4.0 × 10 14 to 3.9 × 10 13 Jones is reasonable based on the above-mentioned equation.After extrapolating photocurrent to the as-calculated I rms of 50 fA, the theoretically detectable minimum light intensity is as low as 38.8 fW at a signal-to-noise ratio of 1, the estimated LDR reaches 211 dB, surpassing the 160 dB LDR of silicon diode 3 (Figure 2F), echoing superior detectivity.Further experiments were also designed to confirm our CZTSSe-based PDs with the extreme weak light detection ability in the NIR region.The multiple attenuators were used to reduce the weak light intensity below the power meter test limit (Thorlabs PM100A test limit: 1 nW/cm 2 ) and connect CZTSSe-based PDs to the electrometer for measurement.In Supporting Information Figure S7, the impressive optical switch and the stable current peak can be found, verifying the extreme weak light detection feasibility in the NIR region.Given these superior features, successful photoplethysmography (PPG) detection of the sphygmic signals can be obtained at the finger and wrist with theobvious arterial blood volume changes shown in Supporting Video.The CZTSSe PDs' spectral response ranges from 300 to 1300 nm, covering hemoglobin's distinctive absorption, 28 making it suitable for heart beat detection.Figure 2G shows the schematic of the heart rate detection system.CZTSSe PDs may convert the transmitted laser light with 635 and 905 nm at 0 V bias from the top surface of the wrist to the bottom in accordance with the blood flow volume in the veins into an electrical signal (Figure 2H).The temporal photocurrent response revealed the distinct pulsing PPG signal due to its ultra-high specific detectivity in weak light.Furthermore, the indoor light worked as an excitation light source for the finger test, and the distinguishable signals before and after running were clearly observed, indicating the accurate real-time health monitoring and light-source-free feature.
The as-measured heart rate results were compared with the corresponding detection results of a smart bracelet worn on another wrist at the same time.The heart rate detection hardware of the smart bracelet is a green LED light coupled with a detector.A video of the real-time working model is also given in Supporting Video.According to the results as shown in Figure 2G, our PDs are comparable to those of the commercial electronic wristband in detection accuracy, and even faster response.
As previously mentioned, the device's noise is influenced by flicker 1/f noise and generation-recombination (G-R) noise in addition to shot noise. 3The shot noise (I shot ), flicker 1/f noise (I 1/f ), and generationrecombination (I G-R ) were calculated by using the following equations to avoid overestimation of the realistic detectivity 3 : where α, β, and C are constant, M is the photocurrent gain, f is the operating frequency, and τ is the average carrier lifetime.Here, we set α and β at 2 and 1, respectively.Therefore, the total noise current (I noise ) can be obtained by using the following expression 3 Figure 3A displays the noise power spectra with different frequencies for PD-B, and the measured total noise derived by performing a fast Fourier transform on the current versus time.The flicker 1/f noise produced by scattering at the device interface and the CZTSSe thin film contributes the majority of noise at a frequency of <10 Hz (Figure 2E).Thus, if we take I noise = I 1/f , then Equation ( 6) for the calculated specific detectivity D* of 3.9 × 10 13 Jones can be obtained by 25 : and the noise equivalent power (NEP) is usually calculated by 25 : When the frequency fixed is between 10 and 1 000 000 Hz, the noise is primarily decided from G-R noise.At a frequency of >1 MHz, shot noise (I noise = I shot ) is predominant, providing a response to Equation ( 5) for the calculated specific detectivity  * 1 of 4.0 × 10 14 Jones.Supporting Information Figure S8 shows the total noise determined NEP as a function of frequency (a) and light intensity (b) at 0 bias.Those NEPs are as low as 1 × 10 −14 W/Hz 0.5 (under 905 nm), resulting in a highly interesting D*, enhancing its NIR region capability to detect weak light.The photovoltaic-based PDs are created by using a diode architecture operated in the reverse direction. 29The Shockley equation is used to describe the total current flowing through the device, ideally when illuminated where I 0 is the reverse saturation current.When I light = 0, I can be equal to I dark .In a real diode, I dark is frequently much higher than I 0 and originated from a trap-assisted recombination due to the existence of traps, pinholes, hunts, or charge-carrier injection from the contacts.The Shockley-Read-Hall statistics provide for the description of I dark , which is expressed as follows: where I shunt = V/R shunt and R shunt is the shunt resistance.Figure 3B shows a good fit with the measured I dark under reverse bias.The resulting reverse saturation current was approximately 7.07 × 10 −8 A. A typical space charge-limited current of the CZTSSe films is measured by regulating the forward voltage and current (Figure 3C).The border points of the linear and quadratic segments can be found.Accordingly, the filling limit voltage (V TFL ) value is 145 mV.According to the following equation 30 : where L is the thickness of CZTSSe, ε 0 signifies the vacuum permittivity (8.85 × 10 −12 F/m), and ε r represents the relative permittivity (8.0 for CZTSSe).The calculated N trap value is 1.40 × 10 14 /cm 3 , which is lower than those of the other reports, demonstrating a longer minority carrier lifetime. 30,31The series resistance may be ignored at sufficiently low bias voltages, and I shunt shall be considered for the sake of simplicity.I shunt can be suppressed by the bulk and interface defect passivation or energy band alignment heterojunctions. 29When the R shunt is sufficiently high, the I light /I 0 ratio relates to the V OC of the CZTSSe thin-film solar cell.The relevant formula is as follows 29 : Based on the trend shown in Figure 3D, the experimental V OC and the I light /I dark will be far above or close to the intrinsic limit (within the experimental error), but not much below.When the reverse voltage is low enough to assess I dark , it can be equal to I 0 , and the measured V OC under 905 nm laser illumination and simulated sunlight are also near the limit.The outstanding photo-response performance of the CZTSSe thin-film PD is also explained by a strong built-in field due to the high quality of the CZTSSe/CdS heterojunction.The typical 1/C 2 -V curves for the devices are plotted in accordance with the capacitance-voltage (C-V) measurements to further examine the junction quality (Figure 3E).Using linear fitting and extrapolating to the x-axis, the heterojunctiondependent built-in voltage (V bi ) can be calculated.Accordingly, the PD-B possesses a higher V bi (498 mV), which is beneficial to the increase of the V OC and reduction of I 0 , which can unquestionably encourage the separation of the electron-hole pairs produced by the photons to boost the photocurrent.Furthermore, capacitance profiling data can provide the spatial defect density of interface defects, free carriers, and bulk defects based on the C-V scans.Drivelevel capacitance profiling (DLCP) measurement is used to more precisely assess the free carriers and bulk defects in order to more clearly detect the interfacial defects. 32The following equations can be used to depict the C-V determined doping density (N C-V ) and DLCP measured doping density (N DLCP ) versus the profile depth x 23 : where S denotes the device area, N D signifies the doping density of CdS, ε 0 , ε r,p , and ε r,n represent the permittivity of free space and the relative permittivity of CZTSSe and CdS, respectively; C 0 and C 1 are two quadratic fitting parameters derived from the C-V curves.The discrepancy between N C-V and N DLCP equals to the interface defect density (N i ).Accordingly, the estimated N i value of PD-B decreased to 1.47 × 10 15 /cm 3 , indicating a superior CZTSSe/CdS heterojunction interface, accelerating the carrier transport across the interface and reducing the carrier recombination at the interface (Figure 3F).Moreover, PD-B exhibits a larger depletion width (W d ) of approximately 244 nm (equals to the profile depth × value at V = 0 V), which is predictable to a certain extent because W d positively varies with V bi 33 : The PD-B with a wider W d can accelerate the minority carrier transport, which is essential for decreasing the response speed.As the thickness of CdS is about 50-80 nm, which is fully depleted in this situation, thus, the whole depletion region includes the CdS buffer layer and part of the CZTSSe absorber layer that is located at spacecharge-region (SCR).Under photovoltaic measurement conditions, the junction-dependent electrical behaviors of the CZTSSe thin film PDs are shown in Figure 3G-I.The key performance was computed using a single exponential diode equation 24 : where J SC , J 0 , and J represent the short-circuit current density, reverse saturation current density, and current density, respectively.V stands for voltage, G for shunt conductance, and R for series resistance, and A is the diode ideality factor.The linearly fitted G value for PD-B is the lowest (2.0 mS/cm 2 ).The plots of dV/dJ are in relation to (J + J SC ) −1 , from which R and A can be obtained by using a slope of AkT/q and the y-axis intercept.The linearly fitted R value for PD-B is the lowest (1.04 Ω cm 2 ).The calculated A value of 1.89 confirms a reduced interface and SCR recombination for PD-B, which is consistent with an improved diode rectification characteristic following the optimization of absorber layer thickness.The reverse saturation current density J 0 is obtained from the plots of ln (J + J SC -GV) against V-JR, and the matching intercept yields a J 0 of 6.2 × 10 −3 mA/cm 2 for the PD-A, 6.0 × 10 −4 mA/cm 2 for the PD-B, and 1.9 × 10 −3 mA/cm 2 for the PD-C, respectively.The lower J 0 also confirmed a decrease in nonradiative recombination in the CZTSSe PDs.Supporting Information Table S1 provides a summary of these characteristics related with the interface or heterojunction.
Except for the heterojunction interface, carrier generation and transport efficiencies can also be impacted by the CZTSSe bulk characteristics.To minimize the detrimental nonradiative recombination and improve the PD performance, it is crucial to comprehend and manage the features of bulk defects, especially deep-level defects.To learn more about the density and distribution of defects in our CZTSSe PDs, admittance spectroscopy was implemented.Figure 4A-C shows the capacitance-frequency (C-f) spectra of CZTSSe PDs measured at various temperatures from 110 to 330 K with an increment of 10 K.Meanwhile, the low-frequency capacitance indicates the response brought on by the deep traps and free carrier, and the high-frequency capacitance response is mostly brought on by a free carrier density.PD-B's capacitance showed lesser volatility in the low frequency range compared with PD-A and -C, indicating a lower defect density in its absorber layer.One capacitance step for all the devices was distinguishable with varying frequencies, which can be ascribed to the defect level.The Arrhenius plot identifies the inflection point for the admittance spectrum.The angular frequency point ω at the peak of the ωdC/dω plot is where its frequency value, ω 0 , is obtained.The slopes of the Arrhenius plot were linearly fitted by using the equation to determine the defect activation energies (E a ) 21 : where E a is the defect activation energy, which indicates the average energetic depth of the defect in relation to the VB maximum or conduction band minimum; ν 0 is the attempt-to-escape frequency; and ω 0 is the inflection point frequency.Supporting Information Figure S9 demonstrates that the predicted trap energy levels for PD-A, -B, and -C are 138.7,106.0, and 170.9 meV, respectively.According to first-principal calculations, the activation energies of Cu Zn 's deep acceptor level are between 100 and 200 meV. 33Thus, they all can be classified as Cu Zn .The PD-B has a lower E a , which means a higher hole emissiv-ity and a lower recombination center concentration.The Kimerling model was used to fit a Gaussian distribution to each defect type based on the following equations 21 : where ω is the angular frequency, V d is the built-in potential of the p-n junction, and N t (E(ω)) is the defect density.Figure 4D-F shows that the N t of Cu Zn (N t = 1.55 × 10 16 /cm 3 ) in PD-B is the lowest, contributing to the low free carrier density, which is consistent with the reduced charge density observed in the C-V and DLCP measurements.The lower and shallower acceptor defects might successfully limit nonradiative recombination by decreasing the absorber carrier density and lengthening the interface hole lifetime. 21The lowered N t and E a of the acceptor defects after the thickness optimization may dramatically boost carrier collection efficiency and significantly contribute to the rise in V OC by decreasing the nonradiative interface recombination. Figure 4G shows the external quantum efficiency (EQE) and corresponding integrated J SC of the devices.The PD-B demonstrates a stronger photo response than the other devices over the whole spectral region.Plotting (E × ln(1 − )) 2 versus energy (Figure 4H) yields an estimate of the CZTSSe absorber layers' bandgap, which is shown to have lowered from 1.06 to 1.02 eV after the thickness optimization.Specifically, the decreased nonradiative recombination within the SCR increases the EQE of PD-B in the wavelength range between 550 and 1000 nm, 30 while the enhanced photo response in the wavelength range greater than 1000 nm should naturally result from the narrower bandgap.The Urbach energy (E u ) is used to evaluate the band-tail states induced by defects, impurities, and disorders in the photo-absorber materials. 33The lowest E u of 24.27 meV in the PD-B (Figure 4I) indicates the reduction of defect states with the optimal thickness and increase of depletion width due to the promoted crystalline growth, defect passivation, and "spike-like" energy band alignment.
The microstructural variation by high-resolution transmission electron microscopy (HRTEM) and Kelvin probe force microscopy (KPFM) was investigated to elucidate this efficient CZTSSe/CdS heterojunction nature.A smooth contact, void-free, and well-adherent CZTSSe/CdS interface can be observed at the heterojunction in the PD-B (Figure 5A), which is beneficial to suppress carrier recombination and current leakage.transmission electron microscopic image with Gaussian blur treatment for CdS.The lattice fringes with interplanar d-spacing of 0.336, 0.338, and 0.343 nm are consistent with the CdS (033, 001, and 303) plane.In addition, Figure 5D, e also shows the CZTSSe HAADF-STM picture obtained inside the bulk absorber layer region.The lattice fringes with interplanar d-spacing of 0.168, 0.198, and 0.320 nm are consistent with the CZTSSe (220, 312, and 112) plane, echoing the aforementioned XRD results.Especially, the main (112) plane of CZTSSe matches well with exposed crystal plane of CdS, such benign lattice-matched characters can further passivate the interfacial defects, which is expected to suppress carrier recombination and facilitate carrier transport, contributing to the superior response speed and responsivity in PDs.Supporting Information Figure S10 shows the lattice models of CdS and CZTSSe.According to the TEM-coupled elemental distributions (EDS) results (Supporting Information Figure S11), the marked elements present a reasonably compositional distribution, and elements Cu, Zn, Sn, S, and Se present good uniformity throughout the whole CZTSSe absorber layer.Cd and S also show stability in the CdS buffer layer, according to the TEM analysis of the smooth contact in the CZTSSe/CdS heterojunction (Figure 5F).We investigated the band bending at the GBs and the lateral electrostatic potential fluctuation between the grains of the CZTSSe absorber layer in PD-B by using a combination of atomic force microscopy and KPFM (Figure 5G-I).The surface topographies match well with those shown in the SEM images, confirming full coverage and highly crystalline.The topography and potential line profiling show synchronous variation, presenting a higher surface potential at the GBs compared with the grain internals, and the contact potential difference (CPD) fluctuation is estimated to be approximately 53.1 mV.This downward band could attract the electrons and repel the holes, thereby promoting the separation of electron and hole (Figure 5J).The rapid separation of electron and hole provides excellent response speed.The corresponding band bending is more across GBs and potentially favored for minority charge collection (i.e., electrons) 34 and support the observations of greater built-in field, resulting in superior response speed and responsivity in our CZTSSe thin-film PDs.

CONCLUSIONS
In summary, an effective solution-processed CZTSSe light absorber layer coupled with a chemical bath deposition (CBD)-CdS layer was prepared to fabricate a CZTSSe/CdS planar heterojunction prototype PD.After optimizing the CZTSSe light absorber layer, the superior CZTSSe with micro-sized grains and benign growth orientation can be obtained, contributing to bulk and interface defect passivation, better energy band alignment with CdS, higher built-in voltage and depletion width.Moreover, the defect-assisted carrier recombination is suppressed, and carrier transport is enhanced.Consequently, our selfpowered CZTSSe thin-film PDs are developing to receive a broadband photo-response (300-1300 nm), a high responsivity (R) of 1.37 A/W, LDR of 126 dB, and I light /I dark ratio exceeding 10 8 within the LDR, and an ultrafast response speed (rise time of 16 ns and decay time of 85 ns), surpassing the previously reported CZTSSe-based self-powered PDs.An ultra-high specific detectivity (D*) of up to 4.0 × 10 14 Jones is also achieved among the well-researched and commercial photodiodes suitable for VIS and NIR regions.
An imaging system was built with a 905 nm laser, and it could achieve 718 pW of weak light detection and infrared imaging as low as 5 nW/cm 2 .Further theoretical calculation shows that our CZTSSe thin-film PDs can detect the NIR light as low as 38.8 fW and LDR of 211 dB.Benefitting from its ultra-high specific detectivity, our CZTSSe thinfilm PDs were also successfully integrated for PPG detection of pulsating signals with obvious arterial blood volume changes, working in transmission mode by 63, 905 nm laser, and by indoor light illumination.These findings suggest that the self-powered kesterite CZTSSe thin-film PDs possess enormous potential for broadband photodetection and weak light detection in widespread application scenarios.were dissolved into 2-methoxyethanol and mixed to obtain a precursor solution.The relevant details can be seen in our previous work. 21,23On Mo-coated soda-lime glass substrates, the CZTSSe precursor films were spin coated.To create films with various thicknesses, spin coating was carried out six to ten times at 3000 rpm for 20 s, followed by a 2 min preheating step at 280 • C. Subsequently, the precursor films were placed into a graphite box and annealed at 555 • C for 15 min in Ar gas.

EXPERIMENTAL SECTION
Fabrication of CZTSSe-based thin film PDs: We used the CBD approach to grow a CdS layer with a thickness of approximately 80 nm on the CZTSSe thin films.A mixture of thiourea (0.75 M, 20 mL), CdSO 4 (0.015 M, 20 mL), ammonium hydroxide (15 M, 22 mL), and deionized water (140 mL) was used to submerge the CZTSSe thin films.The CBD process was conducted for 9 min with constant stirring at 80 • C.Then, the ITO thin films were covered by magnetron sputtering under a suitable O 2 /Ar ratio atmo-sphere (O 2 /Ar≈2%) at a power of 120 W and a pressure of 0.4 Pa.Subsequently, Ag electrodes were thermally vaporized on the ITO thin films as a top contact.Finally, the SLG/Mo/CZTSSe/CdS/ITO/Ag thin-film PDs with different absorbing layer thicknesses were fabricated, and the device was divided into many squares with identical area (0.16 cm 2 ) by mechanical scribing.The active area of the CZTSSe PD device is 0.135 cm 2 .
Characterizations: The photodetection performance of the CZTSSe-based PDs was characterized with a 635 nm power adjustable laser (MRL-III-635L-80 mW) and a 905 nm laser (MDL-III−905-100 mW) and collected by using a computer-controlled source meter (Keithley 2400) in a shielded box.The dynamic photo-response measurements were divided into an optical signal output part and an electrical signal input part.A modulated laser (UTG2025A-driven) and a pulsed laser (pulse width of 3 ns) were used to generate the optical signal, and a high-speed oscilloscope was used to record the electrical signal (DSOX1202A).Using a Keithley 2400 source meter and a Zolix SCS101 system with a monochromator, the wavelength-dependent responsivity was shown.The noise current of the PDs was directly measured with a CHI660e electrochemical workstation.The C-V measurements and DLCP measurements were performed in a dark at room temperature by using a Keithley 4200A-SCS system.The measurements were carried out at an amplitude of 30 mV and a frequency of 10 kHz, with a DC bias voltage from -1 to 0.3 V.The DLCP measurements were conducted by using an AC amplitude ranging from 20 to 140 mV and a DC bias voltage from -0.2 to 0.2 V.A Lakeshore 325 temperature controller was used to acquire admittance characterization.The EQE measurements were carried out by using a Zolix solar cell QE/IPCE measurement system (Solar Cell Scan 100).The morphological and structural characterizations of an ablated thin-film PD were also done using a transmission electron microscope (FEI Titan Cubed Themis G2 300).TEM-coupled energy-dispersive X-ray spectroscopy could be used to determine the chemical compositions and EDS.The surface and cross-sectional morphology were observed using a thermal field emission SEM (Zeiss SUPRA 55).An X-ray diffractometer (XRD, Ultima-iv) with CuK α radiation was used to examine the crystal structure of the CZTSSe thin films.The Raman spectra were obtained by using a Raman spectrometer (Renishaw inVia) with an excitation wavelength of 532 nm.The CZTSSe thin films were analyzed with an X-ray photoelectron spectroscope (ESCALAB 250Xi) (Supporting Information Figure S12).The band-level information of the thin films was further investigated using an UPS (Supporting Information Figure S3).

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.

F I G U R E 1
Design and performance of kesterite thin film PDs.(A) Schematic diagram of the photo-response measurement and the self-powered PDs structure.(B) The desirable spike-like energy band alignment in the CZTSSe/CdS heterjunction.(C) Demonstration of spectral responsivity at zero bias measured under 300−1300 nm wavelength.(D) Responsivity and linear dynamic range (LDR) of the PD-B.(E) Semilogarithmic current−voltage (J−V) curves of PDs based on a 905 nm laser with different light power.(F) The stability of PDs in some simulated environments.(G) The response time of the PD-B under a 905 nm laser with a light intensity of 10 μW/cm 2 irradiation.(H) The response time of the PD-A, -B, and C under a 1064 nm pulsed laser irradiation (pulse width of 3 ns and pulse intensity of 20 μJ).

TA B L E 1
Comparison of the characteristic parameters of the PDs reported in the previous literature.

F I G U R E 2
Noise analysis and weak light applications of the PDs.(A) Statistics of specific detectivity (D*) based on Equation (5) with shot noise and Equation (6) with 1/f noise and G-R noise in those of well-researched state-of-the-art PDs.(B) Schematic diagram of configuration for NIR imaging.(C) NIR imaging output for "SZU" patterns illuminated under a 905 nm laser.(D) Transient dark current and root-mean-square current of CZTSSe-based PDs at 0 bias.(E) Measured current noise of CZTSSe-based PDs as a function of frequency at 0 V bias, the calculated shot noise, 1/f noise, and G-R noise are also provided.(F) The lowest detectable power is calculated by extrapolating the photocurrent equal to the noise current value.(G) The schematic diagram of heart rate detection at the finger and wrist is presented, the illustration shows the comparison between our detector and the smart bracelet, and the heart rate detection hardware of the smart bracelet is a green LED light and corresponding detector.(H) Detected pulse signals of the CZTSSe-based PDs in different places and under different light sources (laser intensity of 10 μW/cm 2 ).

F I G U R E 3
Electrical characterizations of the PDs.(A) The frequency-dependent fit noise power spectra of PD-B.(B-C) Dark current fitting under reverse bias and forward voltage.(D) I light to I dark was plotted against the V OC on a semi-logarithmic scale.(E) 1/C 2 −V plots.(F) C−V and DLCP profiles.(G) Shunt conductance G analysis.(H) Series resistance R and diode ideality factor A analysis.(I) Reverse saturation current density J 0 analysis.
Figure 5B,C shows an atomic resolution high-angle annular dark field scanning F I G U R E 4 Defects characterizations of the PDs.(A-C) Capacitance-frequency-temperature (C-f-T) spectra of PD-A, PD-B, and PD-C.(D-F) Defect distributions of PD-A, PD-B, and PD-C derived from the admittance spectra.(G) EQE and integrated J SC of the PDs.(H) Bandgap derived from the EQE data of the PDs.(I) Urbach energy derived from the EQE data of the PDs.

F I G U R E 5
TEM and KPFM characterizations of PD-B.(A) Bright-field TEM cross-sectional image of PD-B.(B) HRTEM image corresponding to CdS buffer layer region.(C) HAADF signal intensity profile and the matched atomic configuration of CdS.(D) HRTEM image corresponding to CZTSSe absober layer region.(E) HAADF signal intensity profile and the matched atomic configuration of CZTSSe.(F) EDS elemental line scanning profiles associated to the CZTSSe/CdS heterojunction.(G-I) KPFM scanning surface topography, CPD maps, topography, and potential line scans.(J) Schematic diagrams of the energy band structure and CPD near the GBs for CZTSSe thin films.

A
C K N O W L E D G M E N T S This work was supported by National Natural Science Foundation of China (62074102, 62104156), Guangdong Basic and Applied Basic Research Foundation (2022A1515010979, 2023A1515011256) China, Science and Technology Plan Project of Shenzhen (20220808165025003) China.The authors wish to acknowledge the assistance on (TEM/STEM/FIB) received from the Electron Microscope Center of the Shenzhen University.L. Ding thanks the National Key Research and Development Program of China (2022YFB3803300), Open Research Fund of Songshan Lake Materials Laboratory (2021SLABFK02), and the National Natural Science Foundation of China (21961160720).