Flexible photodetectors (FPDs) with excellent photoelectric performance and mechanical flexibility have attracted intensive interest for diverse applications including wearable health monitoring1, implantable optoelectronics2, and artificial vision systems3–5, etc. Metal halide perovskites (MHPs) are promising candidates for next-generation FPDs owing to their excellent photoelectric properties, low-temperature solution-processed fabrication, facile integration with complementary metal-oxide-semiconductor techniques, and compatibility with flexible substrates3,6−10. Recently, MHPs-based FPDs have achieved excellent photoelectric performance including high-gain, ultrasensitive, wavelength-selective, and ultralong linear dynamic range, comparable to conventional vacuum-processed counterparts10–12. However, the high defect density (high dark current) and notorious ion migration (unstable dark current) in polycrystalline MHPs13–15 hamper their photodetection performance, such as operational stability, signal-to-noise ratio (SNR), and limit of detection (LoD), which are essential for long-term wearable health monitoring and image sensor applications.
Amounts of efforts have been devoted to controlling the defect density and prohibiting ion migration of MHPs, such as composition engineering16, additive engineering17,18, interfacial engineering19,20, and dimension regulation21–23. Notably, increasing the resistivity of perovskite has been proven a practical approach to obtaining low dark current16,24,25. Nevertheless, it is challenging to endow MHPs with high resistivity comparable to that of commercial low-noise silicon photodetectors (PDs)26. Meanwhile, in conventional photoconductive-type PDs, dark and photo-induced current are conducted in the same path (Fig. 1a). High resistivity will, in turn, deteriorate photoexcited current and sensitivity9,27,28. Additionally, owing to the electric potential difference between the signal and ground electrodes (Fig. 1c), ion migration will inevitably occur in the conventional photoconductive-type PDs13,29.
Herein, we propose a general and effective electrical field modulation strategy to reduce the dark current and suppress the ion migration in the photoconductive-type MHP-based PDs. Compared with the conventional photoconductive-type PDs device structure, a control electrode (CE) is introduced to apply the control voltage (Fig. 1b). Under the electrical field modulation, all the dark current emitted from the ground electrode can be attracted and collected by the control electrode, while the signal electrode can obtain photocurrent without dark current. In addition, the applied control voltage can change the electric field distribution between the signal and ground electrodes (Fig. 1d). Because the potential distribution around the signal electrode is uniform, the ion migration (driven by the electric field) in the area near the signal electrode can be effectively suppressed.
We first theoretically illustrate the process of dark current reduction and ion migration suppression under the electrical field modulation strategy. Additionally, this strategy was proved effective in various scenarios, including different perovskite compositions (MAPbI3, MAPbBr3 and FA0.92Cs0.04MA0.04PbI3), working voltages (0.1, 0.2, and 0.5 V), and thickness of materials (500 nm, 1 µm, 2 µm, and 1 mm). As a result, the dark current of solution-processed MHPs-based FPDs with CE reduced over 1000 times from ~ 5 nA to ~ 5 pA under electrical field modulation, while maintaining high photocurrent (Fig. 1e,f). In addition, the ion migration near the signal electrode was effectively suppressed and the FPDs showed an excellent operational stability with low signal drift (~ 4.2 × 10− 4 pA per second) and ultralow dark current drift (~ 1.3 ×10− 5 pA per second) after 8000 s operation. By virtue of their high performance, the MHPs-based FPDs successfully work as a wearable photoplethysmography (PPG) sensor to detect the blood pulse signals with high-fidelity under low incident light density (2 mW/cm2, 800 nm). Finally, the MHPs-based photodetector with CE was monolithically integrated with a 16 × 16 thin-film transistor (TFT) based flexible active-matrix backplane, capable of imaging weak light distribution with high contrast. This work provides a general approach to achieving high-performance MHPs-based FPDs for wearable and image sensor applications.
Working Mechanism Of The Electrical Field Modulated Fpds
According to the different control voltage values, the working conditions of the electrical field modulated photodetector are divided into four stages when the signal electrode is applied with constant voltage (> 0 V) (Fig. 2a). At the dark condition, when the control voltage is set as 0 V, the dark current is injected into the perovskite film from the ground electrode and control electrode and transported to the signal electrode due to the potential difference (Fig. 2ai). When the applied control voltage increases (> 0 V), the dark current injected from the ground electrode will be partially attracted by the control electrode. Thus, the dark current transported to the signal electrode will be reduced (Fig. 2aii). When the applied control voltage continuously increases, in principle, there is a critical voltage (CV) that enables the dark current—collected by the signal electrode—to be completely reduced to zero (Fig. 2aiii). If the applied control voltage exceeds CV, the dark current will be injected into the perovskite film from the signal electrode and ground electrode and then transported to the control electrode (Fig. 2aiv). As a result, the current signal measured on the signal electrode will be negative. The photoexcited carriers (electrons and holes) transport process in the electrical field modulated photodetector under light illumination is sketched in Supplementary Fig. 1. Although the control electrode will partially attract the photoexcited carriers and weaken the photocurrent collected by the signal electrode when the control voltage is set as CV. The completely reduced dark current will effectively improve the signal-to-noise ratio (SNR) of the signal electrode. To further illustrate the influence of control voltage, we provide the simulation of the potential distribution in perovskite film using COMSOL software (Fig. 2b). The channel length was set as 100 µm and the thickness of the perovskite film was set as 60 µm. When the control voltage is set as CV, the potential distribution near the signal electrode is uniform (Fig. 2biii). Hence, the ion migration near the signal electrode, driven by the potential difference, can be suppressed sufficiently. Although there is still ion migration near the ground electrode, it will not affect the signal measured from the signal electrode. In this work, to obtain a high SNR and stable photocurrent signal, the control voltage should be set as CV.
To further verify the effectiveness of the electrical field modulated method in practical application, the MHPs-based (FA0.92Cs0.04MA0.04PbI3, thickness of 500 nm) FPD with CE was fabricated on the colorless polyimide (CPI) substrate and the fabrication process is shown in Supplementary Fig. 2. When the signal electrode is applied with 0.1 V, the dark current collected by the signal electrode gradually decreases with the increase of control voltage and even reaches negative values (Fig. 2c). As expected, the dark current can be reduced to zero (stage iii) when the control voltage is set as CV. Figure 2d compares the current-time (I-t) curve of the FPDs under the dark condition at a working voltage (signal electrode) of 0.1 V when the CE was applied with 0, 0.05, 0.1 and 0.15 V (stage i to iv). When the working voltage (0.1 V) was applied and the control voltage was set as 0 V, mobile ions would move in the perovskite film (driven by the potential difference) until reaching the equilibrium state (> 2 s), resulting in the high (5 nA) and drifting dark current measured from the signal electrode (stage i)30. When the control voltage was set as 0.1 V (critical voltage), the dark current was reduced from ~ 5 nA to ~ 5 pA and the equilibrium state was reached in only 0.08 s (stage iii), which proves the effectiveness of the electric field modulation strategy for dark current reduction and the ion migration suppression around the signal electrode. In this scenario, the long-term stability of the electric field modulated FPD was also successfully demonstrated (Fig. 2e). After ~ 8000 s continuous light illumination (520 nm, 50 nW/cm2), the photocurrent and dark current of FPD only drifted from 61.75 to 65.11 pA (~ 4.2 × 10− 4 pA per second) and from 4.95 to 5.06 pA (~ 1.3 × 10− 5 pA per second), respectively.
In addition, to determine the generality of the electrical field modulated strategy, we also fabricated MHPs-based FPD with different perovskite compositions (e.g., MAPbI3 and MAPbBr3) and perovskite film thickness (500 nm, 1 µm, 2 µm, and 1 mm). It was demonstrated that the electrical field modulated strategy is applicable to different perovskite materials and different working voltages (0.1, 0.2, 0.5, and 10 V) and the dark current can be reduced to nearly zero when the control voltage was set as CV (Supplementary Figs. 3–8). When the thickness of perovskite is 500 nm, 1 µm, and 2 µm, the value of CV is equal to the working voltage (Supplementary Figs. 5 and 6). When the thickness of perovskite reaches 1mm, a larger voltage (CV = 35 V) is needed to attract the dark current, so that the dark current of the signal electrode (working voltage = 10 V) can be reduced to zero. The simulation of potential distribution in perovskite film of the corresponding thicknesses (500 nm and 1 mm) are shown in Supplementary Figs. 7 and 8.
Optical properties of the flexible photodetector.
To systematically investigate the photoelectric properties of the electrical field modulated FPD, visible light with different wavelength (405, 520, 600, and 800 nm) were used to illuminate the FPD, as described in Fig. 3a. Compared with the composition of pure FAPbI3 and MAPbI3 perovskite, a more stable cesium-containing triple cation perovskite (FA0.92Cs0.04MA0.04PbI3, thickness of 500 nm) is used as photo-sensing material in this work, which has been proven to be more stable and less affected by fluctuating surrounding variables31. The optical properties of the perovskite are investigated by ultraviolet-visible (UV-Vis) spectrophotometer and photoluminescence (PL) spectrometers (Fig. 3b). The PL spectrum of the perovskite film measured with the excitation of a 540 nm laser exhibits a peak at 803 nm (blue trace), which is consistent with the bandgap of perovskite (1.55 eV) 32. And the UV-Vis absorption spectrum of the perovskite (Fig. 1b, orange trace) exhibits an apparent absorption in the visible light region, rendering it feasible for visible light detection. The working voltage was fixed at 0.1 V, unless otherwise specified, to collect the photoexcited signal of the FPD.
To further evaluate the improvement of the FPD photo-sensing performance under CV, the responsivity (R) and specific detectivity (D*) of the FPD were extracted using the following equations9:
$$R=\frac{{{I_{photo}} - {I_{dark}}}}{{{P_{in}}A}}$$
1
$$D{\text{*}}=\frac{{R\sqrt {AB} }}{{{I_{noise}}}}$$
2
Where \({I}_{photo}\) is photocurrent, \({I}_{dark}\) is dark current, \({P}_{in}\) is incident light power density, \(B\) is the bandwidth, \({I}_{noise}\) is the measured total noise, and \(A\) is active area. The wavelength-dependent responsivity and detectivity curves with the light intensity fixed at 1 µW/cm2 are shown in Fig. 3c. The spectral profile of responsivity peaks at 520 nm, reaching 0.55 (control voltage = 0 V) and 0.37 A W− 1 (control voltage = 0.1 V), respectively. Owing to the attractive effect of the control electrode on the photoexcited carriers, the values of R are lower when the control voltage is 0.1 V. Meanwhile, the significant reduction of the dark current lead to an increase in the D* from 8.3 × 109 (control voltage = 0 V) to 5.2 × 1012 (control voltage = 0.1 V) Jones at 520 nm, indicating a great potential for detecting weak light. And the dependence of R and D* on the light intensity at different wavelengths (405, 520, 600, and 800 nm) are shown in Supplementary Fig. 9. When the control voltage was set at 0.1 V and incident light intensity was 1 nW/cm2, the highest detectivity of FPD reached 1.1 × 1014 Jones at 520 nm, outperforming previously reported FPDs7,33−35. The values of R and D* decrease with the growing light intensity, which is attributed to the increased recombination of photoexcited carriers under high light intensity36. To avoid performance overestimation, we probed the noise power spectrum of the FPD (Fig. 3d). When the control voltage was set as CV, the measured noise decreased from 4.4 × 10− 24 to 5.1 × 10− 30 A2 Hz− 1 at a frequency of 1 Hz (working voltage = 0.1 V). Figure 3e shows the photocurrent of the FPD under light illumination of various intensities (wavelength of 520 nm), indicating the linear relationship between the incident light density and photocurrent.
LoD is another important performance metric of PDs, which plays an vital role in various applications such as biomedical imaging, environmental monitoring, and communications37–39. To estimate the LoD of the PFD, we defined LoD as the light intensity which can produce a signal greater than three times noise level, meaning the signal-to-noise ratio (SNR) is 340. And the SNR was extracted as:
$${\text{SNR}}=\frac{{\left( {{{\bar {I}}_{photo}} - {{\bar {I}}_{dark}}} \right)}}{{\sqrt {\frac{1}{N}\sum\limits_{i}^{n} {{{\left( {{I_i} - {{\bar {I}}_{photo}}} \right)}^2}} } }}$$
3
Where \({I}_{i}\) is the measured photocurrent, \({\stackrel{-}{I}}_{photo}\) is the average measured photocurrent, and \({\stackrel{-}{I}}_{dark}\) is the average measured dark current. We derived the SNR of FPD under different light intensities (wavelength = 520 nm), as shown in Fig. 3f. The extracted LoD of the FPD is 6.5 pW/cm2 at 0.1 V control voltage, which is 460 times lower than that of 3 nW/cm2 at 0 V control voltage. Benefiting from the the low dark current and high signal current characteristics, the FPD showed prominent optical switching properties (SNR = 5.4) under 520 nm weak light illumination (30 pW/cm2), as shown in Supplementary Fig. 10. Similarly, the dependence of SNR on light intensity at different wavelengths (405, 600 and 800 nm) is shown in Supplementary Fig. 11. When the control voltage is 0.1 V, the LoD decreases from 29.5, 5, and 15.4 nW/cm2 to 52, 37, and 66 pW/cm2 at 405, 600, and 800 nm wavelengths, respectively, which proves that the applied control voltage could effectively improve the weak light detection performance of MHPs-based FPDs (Supplementary Fig. 12).
Furthermore, we investigated the operational stability of the FPD under the weak light of 520 nm wavelength (0.2 Hz, 50 nW/cm2). In the test duration of 600 s with 120 light ON/OFF cycles, the dark current drifted from 6.9 to 5.1 nA (~ 3 pA per second) when the control voltage is 0 V, as shown in Fig. 3g. Especially, during the first 50 seconds of the test, the dark current drifted at a rate of ~ 10 pA per second (from 6.9 to 6.4 nA), which leads to the low SNR of the signal, as shown in the inset of Fig. 3g. In contrast, the dark current under 0.1 V control voltage modulation, reveals almost no drift, indicating ion migration was effectively suppressed. The response speed of FPDs was measured by chopper modulated 532 nm light illumination (1 mW/cm2) under different working voltages (Supplementary Fig. 13). With the increase of working voltage, the rise time of FPDs gradually decreases, because the increased working voltage reduces the carrier transit time. In addition, the photocurrent also increased with the increase of working voltage, as shown in Supplementary Fig. 14.
Application of the FPD for wearable bio-signal detection.
To demonstrate the practical application for wearable electronics, we carried out a PPG test based on the FPD. The basic working mechanism of a transmission mode PPG sensor is shown in Fig. 4a. When a light beam with a wavelength of 800 nm is projected onto the skin surface of the fingertip, the light beam will be transmitted to the FPD by transmission, while a part of the light is absorbed, reflected, and scattered by human tissues. During the test process, the light intensity detected by FPD will fluctuate with the changes in the volume of blood vessels caused by heartbeat. Specifically, when the heart is contracting, the increased blood volume in vessels will reduce the light intensity (be absorbed by blood) that can be detected by the FPD. And during diastole, on the contrary, more light can be transmitted to the FPD. Hence, the heart rate (HR) can be extracted from the FPD signals to evaluate the cardiopulmonary function of humans41,42.
As shown in Fig. 4b, a light beam (800 nm) emitted from the optical fiber was projected onto the fingertip. And the FPD was stuck on the finger pulp by a transparent adhesive film to detect the light intensity changes. To minimize noise signals from the finger movement and surrounding environments, a flexible cable was used to connect the FPD with the external signal acquisition circuit and the whole measurement was conducted in the dark condition. To investigate the performance improvement of FPD by our proposed method, the PPG tests were conducted with working voltage of 0.1 V under different incident light intensities (72, 4.6, and 2 mW/cm2) when the control voltage was set to 0 V and 0.1 V, respectively (Fig. 4, c-e).
Figure 4c shows the FPD performance under high incident light intensity (72 mW/cm2), and the device in both conditions (with/without control voltage) could successfully obtain high-fidelity PPG signals. When the incident light intensity decreased to 4.6 mW/cm2, significant baseline drift appeared in the signal of FPD (control voltage = 0 V), although the periodic blood pulse signal can be distinguished (Fig. 4di). As the incident light intensity was further reduced to 2 mW/cm2, the blood pulse signal was overshadowed by the baseline drift and cannot be discriminated (Fig. 4ei). In contrast, high-fidelity PPG signals were acquired under the 0.1 V control voltage modulation regardless of the light intensity (Fig. 4dii and 4eii). From the measurement results, the blood pulse frequency was calculated to be 67 beats per minute. The above experimental results successfully demonstrate the capability of the FPD to obtain high SNR PPG signals under low light density through electric field modulation, which shows the great potential of our method in fabricating low-power consumption wearable electronics.
Weak light imaging with active-matrix FPD array.
Finally, we integrated the perovskite film with a 16 × 16 TFT backplane, based on solution-processed In2O3 semiconductors, to construct the active-matrix FPD (AM-FPD) array (with an active area of 2 × 2 cm2). The TFT based active-matrix FPD array provides merits of high spatial resolution, low signal crosstalk, and low-power consumption43,44. Significantly, to improve its photo-sensing performance, a CE was directly deposited on the perovskite film to apply the modulation electric field. The detailed structure and process scheme of the AM-FPD array, fabricated on the CPI substrate, are shown in Fig. 5a and Supplementary Fig. 15. Photograph of the as-fabricated AM-FPD array wound on a glass rod demonstrating the good mechanical flexibility of the device (Fig. 5b). Each pixel in the array consists of an In2O3 switching TFT for pixel addressing and a perovskite photoconductor for photo-sensing (Fig. 5c). And the corresponding equivalent circuit and cross-section diagrams of a single pixel are presented in Fig. 5d and Supplementary Fig. 16, respectively. The statistical distributions of threshold voltage (average Vth=10.85 V) and subthreshold swing (average SS = 0.54 V dec− 1) of the 256 (16 × 16) In2O3 TFTs show a narrow performance distribution, indicating the high-quality imaging capability of the TFT array as a switching backplane (Fig. 5e).
Figure 5f shows the real-time relative drain current changes (∆ID/I0) curves when the In2O3 transistor connected with perovskite film in series at different light intensities (0.5, 1, 2, 5, and 10 µW/cm2). When the control voltage was applied (0.1 V), the real-time current responses showed high stability, low drift, and high SNR value even under weak incident light density, which is critical for high contrast imaging. Still, even under lower working voltage, the specific detectivity of our FPD outperforms documented flexible perovskite PD array, as shown in Fig. 5F and Table S17,35,45−49.
To evaluate the weak light imaging capability of the AM-FPD array, a weak light beam (520 nm, 500 nW/cm2) was illuminated via a shadow mask, with pattern of character ‘U’, onto the sensor array, as schematically shown in Fig. 5h. The detailed measurement equipment and process are described in the Methods. Figure 5i,j present the corresponding relative current changes mappings of AM-FPD without (Fig. 5i) and with (Fig. 5j) applied control voltage. Owing to the large dark current and baseline drift, the patterned shape could not be clearly distinguished by the signal mappings under weak light intensity without control voltage (Fig. 5i). In contrast, high SNR and contrast image of the ‘U’ pattern was obtained when the control voltage was applied with 0.1 V (Fig. 5j).