Hole-only & electron-only devices
The chemical structure of BFDO-4F is shown in Fig. 1a. The BFDO-4F incorporated BHJ layer is composed by PCE-10 as donor, COTIC-4F as acceptor, and BFDO-4F as the third component introducing trap states. The absorption spectra and energy level of the active layer materials are depicted in Fig. 1b and 1c. To investigate the basic working mechanism of BFDO-4F incorporated OPDs, electron-only and hole-only devices containing 1 wt% BFDO-4F were initially fabricated and characterized. The device architectures, along with dark current density–voltage (J–V) curve and EQE, are presented in the Supplementary Fig. 1. The electron-only device performs a EQE below 100%. Conversely, the hole-only device demonstrated the PM-effect, with EQE exceeding 1000% at both − 1 and 1 V bias. The single-carrier devices were further employed to assess the carrier mobility of the ternary blends using the space-charge-limited current (SCLC) method. As shown in Supplementary Fig. 2, the blends exhibit much lower electron mobility compared to hole mobility, with values of 8.75×10− 4 cm² V⁻¹ s⁻¹ for holes and only 3.62×10− 8 cm² V⁻¹ s⁻¹ for electrons, indicating strong electron trapping by BFDO-4F. The trapping effect could possibly result from the low unoccupied molecular orbital (LUMO) level of BFDO-4F, which could serve as electron traps23. Subsequently, to achieve dual-mode characteristic, OPDs are fabricated with the n-i-p architecture of ITO/ZnO/BHJ:BFDO-4F (1–10 wt%)/MoOx/Ag (Fig. 1d).
PM-mode at forward bias
Figure 2 illustrates the working mechanism of the dual-mode OPDs under various conditions. As shown in Fig. 2b, BFDO-4F introduces electron traps, enabling the capture of the photogenerated electrons at the interface between the active layer and the blocking layers (either ZnO or MoOx), leading to varying degrees of carrier accumulation and energy level bending under different conditions29. Under illumination and forward bias, photogenerated electrons are captured while photogenerated holes are blocked at the ZnO/BHJ interface. The consequent energy bending allows external holes to be injected from the anode through the MoOx layer into the active layer. At the very beginning of injection, holes are accumulated at ZnO/BHJ interface, and the resulted energy level bending eventually facilitates an easy tunnelling of the holes to the cathode. Consequently, the OPDs exhibit a strong PM-effect with substantially enhanced light current under low forward bias. Figure 3a illustrates the EQE of OPDs with varied concentrations of BFDO-4F (1–10 wt%) under 1 V forward bias. The EQE decreases from 1612–857% (at wavelength of 820 nm) with increasing concentration of BFDO-4F. This decrease can be attributed to the increased number of trapped electrons introduced by higher BFDO-4F concentrations, leading to increased recombination of injected holes and trapped electrons at the ZnO/BHJ interface30,31.
PV-mode at zero or small reverse bias
When zero or small reverse bias is applied, the device functions as a PV-mode OPD. As depicted in Fig. 2c, the trap-induced energy level bending is insufficient for the injected holes (or electrons) to overcome the energy barrier between ZnO (or MoOx)/BHJ. Figure 3b illustrates the BFDO-4F concentration-dependent EQE. As the concentration increases, the EQE decreases from 12–4% (at a wavelength of 820 nm) due to increased impurities/traps hindering the extraction of photogenerated charge carriers 30.
PM-mode at high reverse bias
As depicted in Fig. 2d, when the applied bias surpasses the blocking layer's threshold, carrier accumulation and energy level bending are intensified, enabling external holes to be injected, thereby switching the device back to PM-mode. Supplementary Fig. 3 illustrates that under a reverse bias of -15 V, the EQE at 820 nm significantly increases from 59–460% as the concentration of BFDO-4F increases from 1 wt% to 10 wt%. This enhancement results from the improved electron trapping capability and increased energy level bending, both attributed to the higher concentration of BFDO-4F introducing more trap states. Although PM can be achieved under both forward and reverse biases, the shapes of the EQE spectra are different, as shown in Supplementary Fig. 4a. This discrepancy likely arises from variations in trapped electron distributions near the ITO or Ag electrodes, causing different extents of recombination under different spectral ranges12,32–34.
Figures-of-merit of the dual-mode OPDs
To evaluate the key performance metrics of the dual-mode OPDs, we analyzed devices incorporating 1 wt% BFDO-4F. Figure 3c presents the EQE spectra of the optimized device under various biases. As the bias shifts from zero to the forward direction, the EQE progressively increases, surpassing 5700% at a forward bias of 2.5 V. Additionally, under reverse bias, the maximum EQE rises from 17–143% as the applied bias increases up to -20 V (Supplementary Fig. 4b). The responsivity (R) can be calculated from the EQE using Eq. (1):
$$\:R\left(\lambda\:\right)=EQE\times\:\frac{q\lambda\:}{hc}\approx\:EQE\times\:\frac{\lambda\:\left(nm\right)}{1239.8}$$
1
where λ is the wavelength of the incident light, q is the elementary charge, h is the Planck constant, and c is the speed of light. The R under varied bias is depicted in Supplementary Fig. 5. Under reverse bias, the R increased from 0.08 A/W (at 0 V) to 0.62 A/W (at -20 V) at the wavelength of 820 nm. When applying forward bias, a maximum R of 23.41 A/W (at 2.5 V) was achieved at the same wavelength. D* is one of the most vital figure-of-merits for photodetectors (PDs), reflecting the sensitivity of the device, which can be calculated using Eq. (2):
$$\:{D}^{\text{*}}=\frac{R\sqrt{A}}{{S}_{n}}$$
2
where A is the device effective area (0.0516 cm2) and Sn represents the source of noise current. Figure 3d shows the measured Sn spectra, with the corresponding D* for different modes presented in Fig. 3e. The dashed lines in Fig. 3d indicate the calculated white noise, comprising shot and thermal noise, for each mode. Under forward bias, the device exhibits significant photogain, reaching a D* of 1.13×10¹² Jones at 1 kHz. In self-powered PV-mode, it also achieves a maximum D* of 1.92×10¹² Jones. According to the f− 3dB in Fig. 3f, we selected Sn at 1 kHz and 40 kHz to calculate D* for PM and PV modes, respectively. Since signal degradation within the f− 3dB range was minimal, we assumed the EQE remained stable. However, due to challenges in achieving a light chopping frequency above 1 kHz, we conducted the EQE measurements using a chopping frequency of 39 Hz.
-3dB cut-off frequency ( f − 3dB ) and response speed
In PV-mode, the OPD exhibits a f− 3dB of approximately 70 kHz at zero bias, comparable to most self-powered OPDs35–37. As the applied bias increases, intensified electron accumulation gradually reduces the f− 3dB, though it remains above 1 kHz in PM-mode (Fig. 3f). The optimized dual-mode OPD demonstrates a rapid response speed (defined as the duration from 10–90% of the maximal photocurrent), with a rise time (tr) of 2.83 µs and a fall time (tf) of 4.43 µs in PV-mode (Fig. 4a). The initial overshoot observed upon illumination onset is attributed to the pyroelectric effect, which can be subject to the applied bias, as shown in Supplementary Fig. 638,39. In PM-mode under a small forward bias, the photoresponse speed decreases due to additional electron accumulation, yet the device still achieve a relatively short response time on the millisecond scale (Fig. 4b)40. Along with the broad spectral response range, high D*, and fast response speed, these devices rank among the best-performing PM-OPDs, especially within the recently reported dual-mode OPDs category (Supplementary Table 1).
Mechanism analysis of BFDO-4F induced traps
To better understand the operational mechanism of dual-mode OPDs, it is essential to examine their photoresponse behavior under different bias conditions. Figure 4c illustrates how the J–V curves rise with increasing concentrations of BFDO-4F, and further details of Jd are shown in Supplementary Fig. 7. This increase can be attributed to changes in the density of trap states, which introduce additional trap levels and result in enhanced thermal excitation and recombination35. As a reflection of trap states and the resulting recombination loss41,42, the nongeminate recombination rate coefficient (krec) was determined from the relationship between the lifetime (τ) and density (n) of photogenerated carriers, which are obtained through transient photovoltage measurement and charge extraction technique, respectively43,44. The relationship can be described by Eqs. 3 and 4:
$$\:\tau\:={\tau\:}_{0}{\left(\frac{{n}_{0}}{n}\right)}^{\lambda\:}$$
3
where τ0 and n0 are constants and λ is the recombination exponent.
$$\:{k}_{rec}=\frac{1}{\left(\lambda\:+1\right)n\tau\:}$$
4
Figure 4 | Investigation of the trap effect in dual-mode OPDs. Transient photoresponse of the 1 wt% BFDO-4F OPD at a 0 V and b 1 V bias under LED illumination. c J–V curves of OPDs with varying BFDO-4F concentrations, measured in the dark and light (AM 1.5G) conditions. d τ–n curves and e krec of the OPDs based on different BFDO-4F concentrations. f Light J–V characteristics of devices with varying BFDO-4F concentrations, measured under AM1.5G spectrum using a sun simulator. g ESR spectra of the involved active layer components. h DR of the 1wt% BFDO-4F device in both PM and PV-mode, under 1050 nm LED illumination.
To further validate the variation of the density of trap states, the light J–V curves are depicted in Fig. 4f, and the corresponding parameters are included in Supplementary Table 2. By increasing the concentration of BFDO-4F, distinct S-kinks are observed and become more pronounced. Tress et al. have demonstrated that such phenomenon is attributed to charge transport problems45, and charge accumulation due to poor charge extraction properties might be one major reason30. According to the local enlarged image, increasing the concentration of BFDO-4F results in a decreasing trend in open-circuit voltage (VOC), fill factor (FF) and short-circuit current density (JSC). The decline in JSC and FF is suspected to result from an increase of shallow traps due to the dispersive charge transport caused by trapping and detrapping processes. The lower FF is also believed to be a proof of elevated non-geminate recombination46, consistent with our results in Fig. 4e. The drop in VOC indicates the presence of deeper traps responsible for increased defect recombination and subsequent non-radiative loss30,47. To furhter investigate the relationship between BFDO-4F and the trap states, electron spin resonance (ESR) spectroscopy was carried out for the three component used in acirve layer. As shown in Fig. 4g, the significant ESR signal of BFDO-4F indicates strong radical characteristic48, while COTIC-4F exhibits clearly no signal. This result suggests that BFDO-4F could have radical species, exhibiting similar features to quinoidal polymers like PCE-1049. Radicals have been proved to behave as traps for mobile charge carriers in the previous works50–52.
Dynamic range ( DR ) reflecting blocking effect
In addition to the trap effect induced by BFDO-4F, the effective blocking within the devices is another crucial factor in explaining the dual-mode mechanism. Although the evolution of krec demonstrates the influence of introduced traps, the device performance can vary with different applied bias. To gain a deeper understanding of these effects, the DR of the optimized devices, containing 1 wt% BFDO-4F, were measured. The DR for OPDs can be calculated by the following equation:
$$\:DR=20\times\:\text{l}\text{o}\text{g}\frac{{P}_{\text{m}\text{a}\text{x}}}{{P}_{\text{m}\text{i}\text{n}}}$$
5
where Pmax and Pmin represent the maximum and minimum illumination power densities derived from the linear range, respectively. As depicted in Fig. 4h, in PV-mode without applying bias, the device exhibits a DR of 122 dB with a slope of 0.89. The sub-linear slope results from increased bimolecular recombination, caused by trap states introduced by BFDO-4F. These trap states hinder the transport and extraction of photogenerated charge carriers. Under a 1 V forward bias, the DR decreases further to 83 dB, with the slope declining to 0.59. The relationship between light current density (Jph) and light power (P) is described by \(\:{J}_{ph}\propto\:{P}^{\alpha\:}\). The exponent α, derived from the sub-linear slope of the DR fitting line, is used to estimate the level of bimolecular recombination13,53,54. The significantly reduced slope under forward bias indicates more severe recombination, with a smaller slope suggesting more efficient competition between recombination and the extraction of photogenerated carriers. This phenomenon is primarily due to the blocking effect at the ZnO/BHJ interface. Although a higher concentration of BFDO-4F enhances tunneling and energy level bending, the injected holes are still impeded by the ZnO, leading to charge accumulation under forward bias. Conversely, as the reverse bias increased to -20 V, the dual-mode OPD was switched back to PM-mode, thus the DR sharply dropped to 45 dB but the slope only slightly decreased to 0.79 (Supplementary Fig. 8). The slight change in the slope indicates that holes face less obstruction when injected from ZnO into the BHJ layer, and the small energy barrier between BHJ and MoOx layer facilitates their effective collection by the Ag electrode. This results in an increasing (decreasing) trend of EQE versus the BFDO-4F concentration in PM-mode under reverse (forward) bias.
Morphology analysis and GIWAXS characteristics
Grazing incidence wide angle X-ray scattering (GIWAXS) was used to examine the crystalline packing in blend films. The scattering images and the corresponding averaged I–Q curves in both the in-plane (IP) and out-of-plane (OOP) directions are shown in Fig. 5a and Fig. 5b. The GIWAXS signals reflect the crystalline structures of both PCE-10 and the acceptors. By fitting the lamellar reflections at ~ 0.33 Å−1 with the Scherrer Eq. 55, it was observed that adding BFDO-4F increases the crystal coherence lengths (CCLs), from 4.80 nm with 1% added to 5.23 nm with 10% added (Supplementary Table 3), indicating a reduction in packing dislocations within the crystallites56,57. These improvements in crystallization probably lead to the formation of distinct crystallite domains at the phase-separated scale, as revealed by the dark regions in transmission electron microscopy (TEM) images of the blend film with 5% and 10% BFDO-4F, depicted in Fig. 5c. Meanwhile, the atomic force microscopy (AFM) images, as shown in Supplementary Fig. 9, also exhibit slightly increased root-mean-square (RMS) roughness with higher ratio of BFDO-4F, indicating enhanced crystallinity35. However, these crystallite domains do not form an interconnected network across the thin film, which aids exciton dissociation but fails to significantly enhance macroscopic charge transport. From a morphological standpoint, the addition of BFDO-4F introduces charge transport traps, consistent with the electrical properties discussed earlier.
Application demonstration of dual-mode OPDs
An on-chip self-powered module: The dual-mode functionality of the OPDs showcases their potential for highly integrated and miniaturized applications. To explore this potential, we developed a multifunctional on-chip module capable of operating as a self-powered PM-OPD. As depicted in Fig. 6a, the on-chip module consists of two components: a PV section that functions as a power supply and a PM section that acts as a PD. These components are connected in series within a circuit housed in a 3D-printed outer shell (Fig. 6b). When illuminated, the PV section generates a photo-induced voltage that provides a forward bias to the PM section, enabling substantial photogain without the need for external bias or amplifiers (Fig. 6c). As illustrated in Fig. 6d, the self-powered device produces a signal intensity approximately five times greater than that of a reference device under zero bias, with the photogenerated voltage controlled at around 1 V. Although there is a slight reduction in PM compared to the EQE in Fig. 3c—likely due to voltage losses in the series circuit—the on-chip self-powered module demonstrates a promising strategy for achieving multifunctional and highly integrated OPD.
Photocurrent amplification in Bio-sensing
The PM OPD shows significant potential for detecting weak light signals, making it particularly well-suited for photoplethysmography (PPG) measurements. In this study, the dual-mode OPD was employed for PPG measurements, as illustrated in Supplementary Fig. 10. The measurements were performed in PM-mode, with a commercial silicon (Si) PD serving as a comparison under a faint light illumination with an irradiance of 9.45 µW/cm2. As shown in Fig. 6e, the PM-mode measurement produces a signal 21 times stronger than that of the Si PD. This amplified signal, with comparable clarity (Supplementary Fig. 10c), highlights the PM-OPD's ability to detect and enhance small variations in light intensity. This capability makes it an excellent candidate for medical applications like PPG and oximetry, where precise signal detection is crucial.
Flexible device for wearable electronics
The device architecture was also adapted to a flexible PI substrate, as illustrated in Fig. 6f. The blend films demonstrated good compatibility with the PI substrate, maintaining a significant PM effect. While the photoresponse experienced some reduction due to PI's intrinsic absorption properties, the device still achieved an EQE exceeding 1000% under relatively low bias conditions. This performance underscores the device's potential for integration into flexible and wearable electronics, where maintaining high sensitivity and efficiency in non-rigid form factors is crucial. The ability to retain a substantial PM effect on a flexible substrate paves the way for its application in next-generation wearable sensors, medical monitoring devices, and other portable technologies that require high-performance, lightweight, and adaptable PDs.
Universality of BFDO-4F induced dual-mode OPDs
To evaluate the broader applicability of the dual-mode OPD approach, we incorporated 1 wt% BFDO-4F into an active layer consisting of PCE-10 and IEICO-4F, chosen for their comparable energy levels (Supplementary Fig. 11a). The resulting device's EQE is presented in Supplementary Fig. 11b, alongside its corresponding dark J–V curve in Supplementary Fig. 11c. Remarkably, the maximum EQE increased from 22% at zero bias to approximately 600% under a forward bias of 1 V. Additionally, the dual-mode OPD exhibited fast response speed characteristics, with rise and fall times of 5.99 µs and 8.57 µs at zero bias, respectively, and 2.58 ms and 0.35 ms at 1 V forward bias (Supplementary Fig. 12). These results confirm that the proposed mechanism and fabrication process are not only effective but also versatile, capable of being successfully applied across different BHJ active layers. This adaptability underscores the potential of BFDO-4F induced dual-mode OPDs in various applications, making them suitable for a wide range of advanced optoelectronic devices.