Wavelength-sensitive CMOS-like optoelectronic inverter circuits based on solution-processable perovskite nanocrystals/organic semiconductor blends

Abstract

The integration of solution-processed perovskites and organic semiconductors (OSCs) offers a promising approach for low-cost and flexible optoelectronic devices owing to the conductivity of OSCs and remarkable photoelectric features of perovskites. However, challenges remain in achieving multi-wavelength recognition and seamless circuit integration. Here we report two polymer–perovskite pairs: CsPbBr₃/PDVT-10 hybrid films exhibiting good photosensitivity and typical synaptic behavior under light at wavelengths below 520 nm, while this range was extended to 800 nm in CsPbI₃/P(NDI2OD-T2) hybrid films due to the staggered heterojunction structure formed between them. Furthermore, an organic complementary inverter with a gain of 28 and a noise margin of 66% was fabricated using these two specific OSCs/perovskite pairs. The output curve of the inverter circuit was shifted towards V_IN = 0 V under red light and towards V_IN = V_DD under blue light illumination revealing a large voltage difference of 21 V, over 1/3 of V_DD. Finally, an optoelectronic synapse with voltage output was demonstrated, showing a clear pathway for integration into neuromorphic circuits. This work demonstrates a dual-wavelength sensing inverter circuit with strong wavelength discrimination capability with high potential for use in photodetectors, optoelectronic synapses and photo-logic circuits.

---

Introduction

Organic semiconductors (OSCs) are a class of materials that have been widely used in organic light-emitting diodes,^1–3 organic field-effect transistors^4–7 and organic photodetectors^8,9 owing to their numerous interesting features. Firstly, OSCs exhibit semiconductor properties by utilizing π-conjugated molecules as the basic structural unit 10 and their charge-transport performances have been shown to reach the level of amorphous silicon.^10–12 Secondly, they are highly versatile and tuneable since a range of photoelectric properties can be achieved by introducing different functional groups.^13,14 Thirdly, they can be employed for detection in UV, visible and near-infrared ranges, owing to bandgap values commonly ranging from 1.5 to 3.0 eV.¹⁵ Fourthly, OSCs include organic semiconducting polymers consisting of a series of semi-flexible molecular chains of varying lengths connected by weak van der Waals interactions. This allows for solution processability and excellent film-forming capabilities that have significant advantages for large-area, low-cost and flexible applications.¹⁶

However, due to the low photosensitivity of OSCs, obtaining optoelectronic devices based on pure OSCs with exceptional photoresponse remains challenging.¹⁷ Therefore, a commonly used approach involves the integration of diverse materials to form heterojunction structures for achieving high-performance and multifunctional devices. Until now, various materials including low-dimensional materials,¹⁸ metal oxides,¹⁹ carbon-based materials²⁰ and perovskites²¹ have been used as additives to promote multifunctional applications. Among them, all-inorganic perovskite nanocrystals (IPNCs) have shown strong potential owing to their tuneable bandgap,²² long exciton diffusion length²³ and significant optical absorption coefficient.²⁴ Meanwhile, OSCs and IPNCs have common properties such as solution processability which makes possible facile fabrication of optoelectronic devices with good performance using spin-coating and printing techniques. More importantly, their energy levels are comparable indicating that they can form heterojunctions effectively.²⁵ Therefore, the combination of OSCs with IPNCs in optoelectronic devices offers a promising platform for substantially enhancing photoresponse performance. These heterojunctions composed of OSCs and IPNCs have been widely used in the past to fabricate photodetectors,^26–28 optoelectronic synapses,^29–32 solar cells,^33,34 and photomemories.^35–37 Their utilisation has proven a strong potential for applications in the field of neuromorphic computing.^38–40 Most of these optoelectronic devices are two-terminal devices, leading to reduced fabrication complexity. However, in three-terminal photodetectors converting an optical signal into a change in source–drain current similarly to a photodiode, the added gate electrode can be exploited as a second mechanism of output current-modification. These devices benefit greatly from the fact that illumination can result in a shift in threshold voltage, leading to a large change in current at specific gate voltages. In this configuration, the dark off-state current becomes a critical parameter for both light and gate voltage inputs, with very low off-state currents (<1 nA) resulting in ultra-high photosensitivities. Several studies have reported perovskite-organic semiconductor heterojunction-based photodetectors/phototransistors, but they differ from our work in terms of structure, heterojunction type, and fabrication. Zhang et al.^41,42 constructed edge-contact heterojunctions via vacuum thermal evaporation of CsPbBr₃/CsPbCl₃ with small molecules (pentacene, C8-BTBT) for 2-terminal photodetectors. Yen-Hung Lin et al.⁴³ and Chen et al.²⁷ separately fabricated layer-stacked Type-I heterojunctions, using FA_0.83Cs_0.17PbI_2.7Br_0.3/C8-BTBT and CsPbBr₃ quantum dots/organic polymers, respectively, with photocarrier transport relying on interfacial diffusion/accumulation or stacked layer configurations.

Two photodetector 3-terminal devices (one n- and one p-type transistor) could be employed to make a complementary inverter with the ability to convert optical signals into a change in output voltage. These two devices could furthermore employ different nanomaterials to achieve a different voltage output response depending on the incident wavelength. However, previously studied organic/perovskite phototransistors mostly focused on CsPbBr₃ and p-type OSCs heterojunctions; although they have demonstrated excellent performance, efforts with CsPbI₃⁴⁴ and n-type OSC devices⁴⁵ are lacking even though they have the potential to be employed in complementary inverter logic circuits together with p-type transistors to achieve colour detection. Previous studies on hybrid complementary inverter-based photoelectric devices include a report by Liu et al. who realized a photodetector based on an inverter circuit with amplified voltage–output using MoS₂/pentacene as the channel materials,⁴⁶ however at the cost of a very shifted inversion point of the circuit. Kim et al. designed a photoinverter circuit composed of a ZnO/QD TFT and load resistance to achieve modulation only under NIR light.⁴⁷ Sungyoung et al. fabricated photodetector inverter circuits based on F16CuPC/pentacene achieved a significant shift in threshold voltage,⁴⁸ but was only sensitive to a single wavelength. Gergely et al. achieved a dual color sensing inverter utilizing the distinct absorption spectra of pure OSCs (PTCDI-C5 and pentacene),⁴⁹ and while it was proven to distinguish two wavelengths, the voltage difference between two wavelengths was smaller than 15% of V_DD.

In this study, in order to fabricate inverter circuits with a large shift in inverter characteristics between dark, blue and red illumination states, CsPbBr₃/PDVT-10 and CsPbI₃/P(NDI2OD-T2) were selected as polymer/NC pairs for integration into the p- and n-transistors of the circuit. Perovskite/organic semiconductor heterojunctions were selected for their complementary advantages in addressing the limitations of the single components. Perovskites enable strong, tuneable visible-to-near-infrared absorption with high carrier generation, while organics ensure good film formation. Since both perovskite NCs and the selected polymers are highly soluble in chloroform, homogeneous thin films could be produced to act as channel materials in 3-terminal devices. In this kind of blended film, perovskite NCs act as primary light absorbers and carrier separation initiators, while the semiconducting polymers act as charge carrier transport channels and protect the perovskite nanocrystals. The CsPbBr₃/PDVT-10 hybrid transistors demonstrated good photosensitivity and large threshold-voltage shifts under illumination below 520 nm. Devices based on CsPbI₃/P(NDI2OD-T2) hybrid films featured a range extended to 800 nm due to the staggered heterojunction.

Inverter circuits based on these two pairs could be tuned to obtain a switching voltage equal to V_DD/2. These inverters retained their logic operation capabilities with gains reaching 28, a noise margin of 66% in the dark and a switching voltage decreased under red light irradiation and increased under illumination of blue light. Additionally, these inverters were operated as a synapse with voltage output.

Results and discussion

The IPNCs employed in this study, CsPbBr₃ and CsPbI₃ NCs, were synthesized using the hot-injection method, with the crystal structures presented in Fig. 1a. CsPbBr₃ NC films exhibited an emission peak at 520 nm and the emission peak of CsPbI₃ located at 707 nm (Fig. S1a). In this approach, lead salts in the presence of ligands were dissolved in octadecane and heated to temperatures of 120 °C to form a clear solution. A hot caesium oleate solution was rapidly injected into the reaction mixture, causing a colour change to bright green for CsPbBr₃ and red for CsPbI₃, which indicates the initiation of rapid nucleation and growth of the respective NCs.^50,51 The uniform size distribution was achieved by maintaining strict inert conditions (detailed in the Experimental section of the SI). Transmission electron microscopy (TEM) images (Fig. 1b) demonstrate that both CsPbBr₃ and CsPbI₃ NCs are nearly monodisperse, with average sizes of approximately 13 nm and 15 nm, respectively. The structural characteristics of the NCs were analysed using X-ray diffraction (XRD) (Fig. S2). The XRD pattern for CsPbBr₃ NCs confirmed an orthorhombic crystal structure, while CsPbI₃ NCs exhibited a cubic structure. Both crystal phases are optically active, which is beneficial for enhancing their performance in optoelectronic applications.

Fig. 1 (a) Crystallographic structure and (b) transmission electron microscopy images of CsPbBr₃ and CsPbI₃ NCs. (c) Schematic representation of OSC/perovskite nanocrystal hybrid phototransistors. (d) Atomic force microscopy height plots of the PDVT-10, CsPbBr₃/PDVT-10, P(NDI2OD-T2) and CsPbI₃/P(NDI2OD-T2) hybrid films. Transfer curves of (e) CsPbBr₃/PDVT-10 and (f) CsPbI₃/P(NDI2OD-T2) under illumination with different light densities and wavelengths at V_DS = ±20 V.

CsPbBr₃ and CsPbI₃ were blended with p-type (PDVT-10) and n-type (P(NDI2OD-T2)) OSCs respectively, then spin-coated on interdigitated bottom-contact bottom-gate configuration to form bulk heterojunction (BHJ) devices (Fig. 1c). Compared with planar heterojunctions (PHJ) which require two depositions, where the second solvent must not dissolve the first material, BHJ can not only simplify the fabrication process of the devices, but also increase the contact area between OSCs and perovskites since the NCs are blended everywhere in the film. The chemical structures of poly[2,5-bis(alkyl)pyrrolo[3,4-c]pyrrole-1,4(2H,5H)-dine-alt-5,5′-di(thiophen-2-yl)-2,2′-(E)-2-(2(thiophen-2yl)vinyl) thiophene] (p-type, PDVT-10) and poly{[N,N′-bis(2-octyldodecyl)-naphthalene-1,4,5,8-bis(dicarboximide)-2,6-diyl]-alt-5,5′-(2,2′-bithiophene)} (n-type, P(NDI2OD-T2)) are shown in the inset of Fig. 1e and f. Additionally, PDVT-10 has very strong absorption in the region of 600–900 nm with an absorption peak located at 450 nm, and P(NDI2OD-T2) has two absorption peaks around 400 nm and 700 nm (Fig. S1b). PDVT-10 and P(NDI2OD-T2) were chosen as they have relatively high mobility, good output characteristics (Fig. S3a and S4a) and are commercially available.

In order to compare the optoelectronic properties before and after incorporation of perovskites, transfer curves of pure OSC-based transistors were acquired (Fig. S3b and S4b) and characterized in the dark and under their specific absorption wavelengths (450 nm and 800 nm for PDVT-10 and P(NDI2OD-T2) transistors respectively) with different light power intensities. In the dark, the transistors showed typical p-type (mobility ≈ 0.3 cm² V⁻¹ s⁻¹, I_on/I_off ≈ 10⁷) and n-type (mobility ≈ 0.03 cm² V⁻¹ s⁻¹, I_on/I_off ≈ 10⁷) behaviours, in the case of PDVT-10 and P(NDI2OD-T2) transistors, respectively. It is worth noting that when the devices containing pure OSCs were under illumination, the transfer curves of the n-type shifted to a positive direction (p-type shifted to negative direction) with the increased power densities, but the value of the off-state current was not affected, indicating photo-generated carriers increased conduction but were not trapped to form a virtual gate electrode.

Pure CsPbBr₃ and CsPbI₃ films featured low charge transport ability and low photosensitivity (Fig. S5) due to the NC-to-NC gaps in the films and the rapid recombination of excitons.³⁹ After adding the IPNCs (the weight ratio of OSCs and IPNCs is 1 : 1), atomic force microscopy (AFM) measurements were performed to probe the surface morphology. The root-mean square (RMS) roughness of pure PDVT-10 films increased from 0.50 to 2.29 nm upon the addition of CsPbBr₃, while the roughness of pure P(NDI2OD-T2) increased from 0.63 to 3.52 nm upon the addition of CsPbI₃ NCs (Fig. 1d). The thicknesses of the CsPbBr₃/PDVT-10 and CsPbI₃/P(NDI2OD-T2) layers are ≈31.2 and ≈28.1 nm, respectively (Fig. S6). The transfer curves of the CsPbBr₃/PDVT-10 and CsPbI3/P(NDI2OD-T2) phototransistors in the dark and under different wavelengths of light irradiation at various light power intensities were acquired (Fig. 1e and f). In the dark, compared to the pure OSC devices, the threshold voltage shifted toward the positive direction for the n-type transistors (negative direction for the p-type transistors) which can be attributed to the low mobility of the IPNCs hindering the transport of charge carriers. To demonstrate the different absorption performances, we used light of two different wavelengths (450 nm and 800 nm) for illumination. In the case of the CsPbBr₃/PDVT-10 based transistor (mobility ≈10⁻² cm² V⁻¹ s⁻¹, I_on/I_off ≈ 10⁷) the device displayed a weak photoelectric response when irradiated at 800 nm. However, under illumination of blue light, the output changed significantly, the drain current increased and the threshold voltage decreased. AFM characterization revealed distinct morphologies and slightly altered RMS values between pure organic semiconductor films and perovskite/organic blended films, verifying the uniform dispersion of perovskite nanocrystals in the organic matrix. Optoelectronic tests revealed a low photoresponse in pure organic devices (attributed to inefficient carrier separation), while blended devices achieved several orders-of-magnitude enhancement in photoresponse, confirming effective separation of photogenerated electron–hole pairs enabled by the Type-II heterojunction. Pure organic semiconductors featured good electrical conductivity but poor photosensitivity, while pure perovskites exhibited low conductivity and photosensitivity. For polymer-perovskite hybrids, electrical performance decreased slightly but photosensitivity improved significantly: this can be attributed to perovskite nanoparticles (uniformly dispersed in the polymer matrix, as confirmed by AFM) acting as efficient light absorbers, and the polymer serving as the main charge transport channel. Their matched energy levels form a heterojunction that enables effective separation of photogenerated electron–hole pairs. The key photoelectric performance parameters of the hybrid films with different compositions under varying light intensities, including responsivity (R), detectivity (D), and photosensitivity (P), were calculated. The formulas used for these calculations and the corresponding curves are presented in the SI (Fig. S7). In the case of phototransistors based on PDVT-10/CsPbBr₃, the photosensitivity reached 10⁵, the responsivity was 10¹ A W⁻¹, and the detectivity was 10¹² Jones. In the case of phototransistors based on P(NDI2OD-T2)/CsPbI₃, the photosensitivity reached 10², the responsivity was 10¹ A W⁻¹, and the detectivity was 10¹⁴ Jones.

As can be seen in Fig. 2a, the photosensitivity values measured at each wavelength from 300 nm to 800 nm (interval of 10 nm), show that the CsPbBr₃/PDVT-10 hybrid film exhibits a weak response above 520 nm (P < 5). However, P increases rapidly when the wavelength reaches values below 520 nm, reaching values over 10³ around 350 nm. The CsPbI₃/P(NDI2OD-T2) device (mobility was around 10⁻⁴ cm² V⁻¹ s⁻¹ with I_on/I_off of 10⁵) could see its characteristics strongly affected by illumination at both wavelength ranges. Both the threshold voltage and the source–drain current were changed at the two ranges, due to the photo-gating effect.⁵² This can clearly be observed in Fig. 2b where the CsPbI₃/P(NDI2OD-T2) hybrid film has excellent photosensitivity over 10⁴ across the entire 300–800 nm range, the highest P value reaching 10⁵. The threshold voltage shift values of these two transistors under different wavelengths are presented in Fig. 2c and d.

Fig. 2 Photosensitivity values of 3-terminal devices based on the (a) CsPbBr₃/PDVT-10 and (b) CsPbI₃/P(NDI2OD-T2) heterojunctions as a function of the incident light wavelength. Threshold shift of 3-terminal devices based on (c) CsPbBr₃/PDVT-10 and (d) CsPbI₃/P(NDI2OD-T2) heterojunctions as a function of the incident light wavelength. Illustrations of the light-stimulated charge transfer processes in the (e) CsPbBr₃/PDVT-10 devices under illumination at a wavelength below 520 nm and (f) CsPbI₃/P(NDI2OD-T2) devices under illumination at wavelengths below 800 nm.

The output curves of hybrid phototransistors in the dark and under 450 nm light irradiation are presented in Fig. S8. The output current was linear under low bias, indicating a good ohmic contact between the active layer and the gold electrodes. The photosensitivity of pure OSCs was similarly determined, PDVT-10 features high absorption around 450 nm and above 650 nm (Fig. S3d), and P(NDI2OD-T2) exhibits strong absorption around 390 nm and 700 nm (Fig. S4d). As expected, the photosensitivity increases matched well with the absorption spectra of the polymers, although the values were over two orders of magnitude lower owing to the lack of photogating effect.^53,54

To clarify the differing synaptic behaviours of hybrid thin-film devices at 450 nm and 800 nm, we first characterized the valence band edge (E_av) or HOMO energy levels of key materials via Photoelectron Spectroscopy in Air (PESA, Fig. S9). Perovskites featured values of −5.30 eV and −4.94 eV for CsPbBr₃ and CsPbI₃, respectively, while the organic semiconductors featured HOMO levels of −5.02 eV and −5.77 eV, for PDVT-10 and P(NDI2OD-T2), respectively. Using these values, we constructed simplified energy band diagrams for the two heterojunction systems (CsPbBr₃/PDVT-10 in Fig. 2e, CsPbI₃/P(NDI2OD-T2) in Fig. 2f) and analysed their photoresponse mechanisms. In the case of CsPbBr₃/PDVT-10 Type-I heterojunctions, 450 nm blue light (below CsPbBr₃'s 520 nm absorption edge) excites both materials, and photogenerated carriers transfer from CsPbBr₃ to PDVT-10 to increase hole concentration, boosting conductivity and inducing a strong synaptic response. When irradiated with 800 nm light (above 520 nm), only PDVT-10 is excited (CsPbBr₃ NCs remain unexcited), and its photogenerated carriers undergo rapid recombination, leading to weak synaptic behaviour. In the case of CsPbI₃/P(NDI2OD-T2) Type-II heterojunctions: 450 nm and 800 nm light both excite at least one material (the hybrid film's absorption range extends to the edge of P(NDI2OD-T2)'s absorption), and the Type-II alignment enables efficient photocarrier separation—holes transfer from P(NDI2OD-T2) to CsPbI₃ and are trapped there, forming a virtual gate voltage. This effect explains why the CsPbI₃/P(NDI2OD-T2) films exhibit stronger synaptic responses (and higher photosensitivity, P) than the CsPbBr₃/PDVT-10 blend at the same weight ratio.^55,56 Furthermore, the absorption range of the hybrid film extends to the edge of the absorption range of P(NDI2OD-T2).²⁸

The time-domain characteristics of these phototransistors under two different wavelengths have been studied. The blend of OSC/IPNC film was used as an optoelectronic synapse owing to the energy level misalignment between perovskite NCs and OSCs leading to the capture of electrons or holes under illumination, leading to a gradual decrease in current. Fig. 3a presents a schematic diagram depicting the neural signal transmission process between two biological synapses in a neuron under illumination from UV to NIR. The current of CsPbBr₃/PDVT-10 and CsPbI₃/P(NDI2OD-T2) phototransistors was measured during ten consecutive light pulses of irradiation (450 nm, 90 µW cm⁻²) with a source–drain voltage of −1 V and 1 V, respectively (Fig. 3b and d). The devices featured typical synaptic behavior, with the current rapidly increasing and then slowly decaying after the illumination was stopped. However, when exposed to longer wavelength (800 nm, 102 µW cm⁻²), the synaptic performance of CsPbBr₃/PDVT-10 phototransistors disappeared (Fig. 3c), because this wavelength exceeded the absorption range of CsPbBr₃. In contrast, the phototransistors based on CsPbI₃/P(NDI2OD-T2) showed good synaptic behavior even under illumination at 800 nm (Fig. 3e). Therefore, we developed a synapse that is better suited for application in the human visual system, as it covers the entire wavelength range of visible light.

Fig. 3 (a) Schematic illustrations of the transmission process of neural signals through biological synapses in neurons. Synaptic behaviour of the phototransistors under 10 light pulses, pulse width and pulse interval of 1 s. CsPbBr₃/PDVT-10 devices under (b) 450 nm and (c) 800 nm illumination and CsPbI₃/P(NDI2OD-T2) devices under (d) 450 nm and (e) 800 nm illumination.

Complementary inverters, as the basic component of an integrated circuit, are often characterized based on the voltage transfer characteristic (V_TC) curve, which shows the output voltage (V_OUT) as a function of the input voltage (V_IN). An important performance parameter that can be extracted from the V_TC curve is the switching voltage (V_SW), also called the midpoint voltage which is defined as the point where V_IN = V_OUT. The V_SW of an ideal inverter is located at V_IN = V_DD/2 which is influenced by the threshold voltage and mobility of n- and p-transistors.⁵⁷ However, different OSCs have different threshold voltage and mobility values. Hence, the on-state current of the two devices can be matched by changing the W/L of each channel, making it challenging to devise a universal circuit architecture. However, in the present case it was possible to tune the perovskite to OSC ratio to modulate the threshold voltage and mobility of the transistors. Therefore, we investigated the influence of IPNCs/OSCs weight ratios (10 : 1, 5 : 1, 1 : 1, and 1 : 10), each measured on four identical devices. We found that ΔV_Th (ΔV_Th = V_Th,Dark – V_Th,Light) and P values were increased with the increase of perovskite in the blend, while the maximum current decreased with the increased perovskite proportion (Fig. S10), which shows that that the threshold shift and current can be adjusted through tuning the weight ratio between IPNCs and OSCs. Furthermore, the threshold voltage shift under different wavelengths affected the position of V_SW, offering a novel approach to the detection of different wavelengths depending on the positive or negative value of ΔV_SW (ΔV_SW = V_SW,Dark – V_SW,Light).

Due to the different bandgaps of CsPbBr₃ and CsPbI₃, the materials feature different absorption regions, which we exploited in different channels of a complementary-like inverter circuit (Fig. 4a, the optical microscopy image of the inverter can be found in Fig. S11). Usually, p-type OSCs have higher mobility and current values than n-type OSCs, owing to the limited selection of electron withdrawing building blocks available for ensuring stable electron transport in n-type OSCs.^58,59 Therefore, in order to get comparable current in two channels, 50% P(NDI2OD-T2) +50%CsPbI₃ was used as the active layer of the n-type channel and in the p-type channel, and the weight ratio of PDVT-10 and CsPbBr₃ was 1 : 10 (the transfer curves under different wavelengths are shown in Fig. S12). In the dark, the V_TC curve featured a standard shape with a gain of 28 and the V_SW located at the point of V_DD/2 (Fig. 4b and c). The noise margin (NM) was calculated by dividing the maximum side length of the square fitted to the butterfly curves by V_DD/2, the NM reached 66%, demonstrating excellent noise tolerance (Fig. S13). Under illumination of red light (650–750 nm), only the threshold voltage of the n-channel is negatively shifted, so this transistor will reach the “on state” prematurely while the p-channel transistor is only weakly affected. This leads to the V_SW shifting towards V_IN = 0 V. In the case of illumination of the inverter with blue light (400–500 nm), although both transistors are excited, the p-channel has a stronger threshold voltage shift than the n-channel under blue light irradiation (∼34 V compared to ∼20 V, Fig. S14), leading to a shift of V_SW in the V_IN = V_DD direction. As can be seen in Fig. 4d, under red and blue light illumination, the maximum voltage difference can reach 21 V (over 1/3 of V_DD) highlighting a novel approach for the discriminate detection of these two wavelengths. In terms of logic operation, when the circuit is exposed to a shorter wavelength, the output voltage of the logic value “0” region is higher than under dark conditions because the p-type transistor is not “off” completely, there is still a relatively large photo-current flowing. Similarly, under longer wavelength irradiation, the logic value “0” is closer to zero due to the stronger current in the n-channel. The results show this hybrid inverter can be used for colour detection without filters indicating great potential in future highly integrated circuits, while maintaining strong voltage gain and switching voltage difference, outperforming most reported optoelectronic inverter devices (Table S1, SI).

Fig. 4 (a) Schematic representation of the colour-sensing inverter circuit, (b) voltage transfer characteristics and (c) voltage gain of the inverter in the dark and under illumination at different wavelengths. (d) Shift in switching voltage ΔV_SW under different incident light wavelengths.

As shown in Fig. 5a, the photoinverter circuit architecture can be constructed using IPNCs/OSCs which exhibit synaptic behaviour with voltage output, thereby overcoming the challenges associated with detecting nA currents. Using the same architecture, the synaptic behaviour of this inverter was characterized at V_IN = 60 V. In this case, the p-transistor is in the cut-off region with V_GS = 0 V and the n-transistor is in linear region with V_GS = 60 V, so the output voltage is close to 0. When this device under illuminated of 450 nm light pulses, current in both channels will be increase simultaneously so the output voltage will increase firstly due to the “pull-up” effect of the p-transistor. After that, the current in the p-transistor decayed slowly, with a good synaptic behaviour under low gate bias (V_GS = 0 V) while the n-transistor, under a large V_GS, induces rapid current decay,^60,61 so the output voltage will decrease slowly owing to the slow current decay in the p-channel as shown in Fig. 5b–e with a different number and duration of pulses.

Fig. 5 (a) Illustration and schematic diagram of the photoinverter circuit for simulating the learning-forgetting behaviours. Dynamic voltage response of the inverter at V_IN = 60 V under 450 nm illumination for (b) double pulses, 0.5 s, (c) double pulses, 1 s (d) single pulse, 1.5 s and (e) single pulse, 2 s.

While in a conventional neuromorphic device outputting a current value for each pulse, the PPF (paired-pulse facilitation) Index values are calculated as PPF Index (%) = (ΔI/I₀) × 100, in our case, the PPF Index (%, V_OUT) is estimated as (ΔV_OUT/V_OUT,0) × 100. It should be noted that this adjusted index cannot be compared directly to the literature. In a conventional neuromorphic device outputting a current value, the off state before any pulse is a precise current value over 0, which scales with the input voltage. Once the device is submitted to a pulse, the newly outputted value also scales with the input voltage. However, in our case of an inverter circuit, V_OUT is constrained between the bounds between 0 and V_DD, and hence the value of the PPF index is largely dependent on the boundary of V_DD, while its precision is strongly dependent on the V_OUT ≈ 0 value. Our calculated values for two identical pulses with inter-pulse interval of 0.5 s and 1 s (Fig. 5b and c) indicate a similar PPF Index of 65%. Increasing the pulse durations to 1 s (inter-pulse interval of 0.5) a PPF Index of 85% can be observed. Further increasing the pulse durations to 2 s allows reaching values of PPF Index over 300%. These findings can be explained by the operation of the inverter circuits and in particular its V_IN–V_OUT curves as seen in Fig. 4b and c: a pulse induces a modest change in the value at which the inversion occurs. As V_IN,inversion is shifted closer to V_IN, the effect of the pulse becomes stronger, until V_IN,inversion reaches V_IN, at which point the effect of each pulse will become weaker. Exploiting this effect offers a novel approach for neuromorphic computing systems and reduces the circuit complexity to detect ultralow current.

Conclusions

In summary, we conducted an in-depth study on phototransistors based on IPNC/OSC hybrid films, including electrical characteristics, optical properties and synaptic behaviours, then applied them into different channels of the inverter to realize red/blue light recognition and optoelectronic synapse. CsPbBr₃/PDVT-10 based phototransistors exhibited good photosensitivity and synaptic behaviour below 520 nm owing to the type-I straddling gap which cannot induce the photogating effect. However, the CsPbI₃/P(NDI2OD-T2) based phototransistor was able to retain synaptic behaviour when irradiated under UV to NIR with higher photosensitivity due to the type-II heterojunction strategy. Furthermore, a standard inverter with a gain of 28 and a noise margin of 66% was fabricated using these two polymer-IPNC pairs by adjusting the weight ratio, then it was applied as a red/blue voltage-output photodetector owing to the reverse motion of switching voltage under red/blue light irradiation, the shift of switching voltage between red and blue light reached 21 V (>1/3V_DD), and this device exhibited the best performance among similar reported devices. The device had the ability to distinguish multiple wavelengths integrated into a fully operational circuit, eliminating the need for separate red and blue photodetectors and reducing circuit complexity significantly. Finally, we have demonstrated voltage-output synaptic behaviour, including EPSC and PPF, revealing the potential of our device for neuromorphic electronics without requiring complex low-current detection circuits.

Conflicts of interest

There are no conflicts to declare.

Data availability

The authors state that the data will be made available upon reasonable request.

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d5tc02341f.

Acknowledgements

This research was supported by the National Natural Science Foundation of China (NSFC Grant 22050410280), the Chengdu Science and Technology Bureau (Grant 2022YFWZ0004), and UESTC grants. T. L. would like to acknowledge support from the NSFC grant Research Fund for International Young Scientists (Grant No. 52050410342) and the National Department of Science and Education of China (Grant No. QNJ2021167003).

References

1. K. Yoshida, J. Gong, A. L. Kanibolotsky, P. J. Skabara, G. A. Turnbull and I. D. W. Samuel, Nature, 2023, 621, 746–752.
2. T. Matsushima, F. Bencheikh, T. Komino, M. R. Leyden, A. S. D. Sandanayaka, C. Qin and C. Adachi, Nature, 2019, 572, 502–506.
3. A. Salehi, X. Fu, D. Shin and F. So, Adv. Funct. Mater., 2019, 29, 1808803.
4. M. Li, J. Zheng, X. Wang, R. Yu, Y. Wang, Y. Qiu, X. Cheng, G. Wang, G. Chen, K. Xie and J. Tang, Nat. Commun., 2022, 13, 4912.
5. H. Chen, W. Zhang, M. Li, G. He and X. Guo, Chem. Rev., 2020, 120, 2879–2949.
6. A. Nawaz, L. Merces, L. Ferro, M. M. Sonar and C. C. B. Bufon, Adv. Mater., 2023, 35, 2204804.
7. T. Leydecker, M. Herder, E. Pavlica, G. Bratina, S. Hecht, E. Orgiu and P. Samorì, Nat. Nanotechnol., 2016, 11, 769–775.
8. H. Ren, J. Chen, Y. Li and J. Tang, Adv. Sci., 2021, 8, 2002418.
9. Q. Cui, Y. Hu, C. Zhou, F. Teng, J. Huang, A. Zhugayevych, S. Tretiak, T. Nguyen and G. C. Bazan, Adv. Funct. Mater., 2018, 28, 1702073.
10. S. Allard, M. Forster, B. Souharce, H. Thiem and U. Scherf, Angew. Chem., Int. Ed., 2008, 47, 4070–4098.
11. Y. Zheng, S. Zhang, J. B. H. Tok and Z. Bao, J. Am. Chem. Soc., 2022, 144, 4699–4715.
12. S. Yan, X. W. Zhang, Z. L. Wang, C. H. Xu, Y. B. Shi, Y. F. Deng, Y. G. Han and Y. H. Geng, Chin. J. Polym. Sci., 2023, 41, 824–831.
13. H. Dong, H. Zhu, Q. Meng, X. Gong and W. Hu, Chem. Soc. Rev., 2012, 41, 1754–1808.
14. M. Ashizawa, Y. Zheng, H. Tran and Z. Bao, Prog. Polym. Sci., 2020, 100, 101181.
15. A. Mavridi-Printezi, A. Menichetti, M. Guernelli and M. Montalti, Nanoscale, 2021, 13, 9147–9159.
16. D. Yang and D. Ma, Adv. Opt. Mater., 2019, 7, 1800522.
17. Z. Guo, J. Zhang, X. Liu, L. Wang, L. Xiong and J. Huang, Adv. Funct. Mater., 2023, 33, 2305508.
18. J. Wang, B. Yang, S. Dai, P. Guo, Y. Gao, L. Li, Z. Guo, J. Zhang, J. Zhang and J. Huang, Adv. Mater. Technol., 2023, 8, 2300449.
19. L. Ren, L. Liang, Z. Zhang, Z. Zhang, Q. Xiong, N. Zhao, Y. Yu, R. Scopelliti and P. Gao, RSC Adv., 2021, 11, 3792–3800.
20. J. Ha, D. Akinwande and A. Dodabalapur, Appl. Phys. Lett., 2012, 101, 033309.
21. X. Wang, D. Hao and J. Huang, Sci. China Mater., 2022, 65, 2521–2528.
22. K. Leng, I. Abdelwahab, I. Verzhbitskiy, M. Telychko, L. Chu, W. Fu, X. Chi, N. Guo, Z. Chen, Z. Chen, C. Zhang, Q. H. Xu, J. Lu, M. Chhowalla, G. Eda and K. P. Loh, Nat. Mater., 2018, 17, 908–914.
23. D. Stranks, G. E. Eperon, G. Grancini, C. Menelaou, M. J. P. Alcocer, T. Leijtens, L. M. Herz, A. Petrozza and H. J. Snaith, Science, 2013, 342, 341–344.
24. K. Wang, Y. Hou, B. Poudel, D. Yang, Y. Jiang, M. Kang, K. Wang, C. Wu and S. Priya, Adv. Energy Mater., 2019, 9, 1901753.
25. M. Chen, J. Wang, F. Yin, Z. Du, L. A. Belfiore and J. Tang, J. Mater. Chem. A, 2021, 9, 4505–4527.
26. C. Xie, P. You, Z. Liu, L. Li and F. Yan, Light: Sci. Appl., 2017, 6, e17023.
27. K. Chen, X. Zhang, P. Chen, J. Guo, M. He, Y. Chen, X. Qiu, Y. Liu, H. Chen, Z. Zeng, X. Wang, J. Yuan, W. Ma, L. Liao, T. Nguyen and Y. Hu, Adv. Sci., 2022, 9, 2105856.
28. C. Zou, Y. Xi, C. Huang, E. G. Keeler, T. Feng, S. Zhu, L. D. Pozzo and L. Y. Lin, Adv. Opt. Mater., 2018, 6, 1800324.
29. C. Wang, Q. Sun, G. Peng, Y. Yan, X. Yu, E. Li, R. Yu, C. Gao, X. Zhang, S. Duan, H. Chen, J. Wu and W. Hu, Sci. China Mater., 2022, 65, 3077–3086.
30. D. Hao, J. Zhang, S. Dai, J. Zhang and J. Huang, ACS Appl. Mater. Interfaces, 2020, 12, 39487–39495.
31. J. Liu, Z. Yang, Z. Gong, Z. Shen, Y. Ye, B. Yang, Y. Qiu, B. Ye, L. Xu, T. Guo and S. Xu, ACS Appl. Mater. Interfaces, 2021, 13, 13362–13371.
32. Y. Wang, Z. Lv, J. Chen, Z. Wang, Y. Zhou, L. Zhou, X. Chen and S. Han, Adv. Mater., 2018, 30, 1802883.
33. I. Susic, K. P. Zanoni, A. Paliwal, I. C. Kaya, Z. Hawash, M. Sessolo, E. Moons and H. J. Bolink, Solar RRL, 2022, 6, 2100882.
34. S. Lim, D. H. Lee, H. Choi, Y. Choi, D. G. Lee, S. B. Cho, S. Ko, J. Choi, Y. Kim and T. Park, Nano-Micro Lett., 2022, 14, 204.
35. Q. Li, T. Li, Y. Zhang, Y. Yu, Z. Chen, L. Jin, Y. Li, Y. Yang, H. Zhao, J. Li and J. Yao, Org. Electron., 2020, 77, 105461.
36. E. Ercan, Y. Lin, L. Hsu, C. Chen and W. Chen, Adv. Mater. Technol., 2021, 6, 2100080.
37. H. Yang, Y. Yan, X. Wu, Y. Liu, Q. Chen, G. Zhang, S. Chen, H. Chen and T. A. Guo, J. Mater. Chem. C, 2020, 8, 2861–2869.
38. Y. Li, J. Wang, Q. Yang, H. Zhang, L. Chen and Z. Zhao, Adv. Sci., 2022, 9, 2202123.
39. G. Ding, H. Li, J. Y. Zhao, X. Wu, Y. Zhou and L. Sun, Chem. Rev., 2024, 124, 12738–12843.
40. T. Liu, Z. Yuan, L. Wang, M. Chen, Y. Zhang and H. Xu, Nat. Commun., 2025, 16, 4261.
41. T. Zhang, Y. Chen, Y. Chu, L. Wang, J. Liu and F. Huang, ACS Appl. Electron. Mater., 2022, 4, 2805–2814.
42. Q. Zhang, D. Yang, J. An, S. Li, Y. Zhao and M. Wang, ACS Appl. Electron. Mater., 2025, 7, 1914–1920.
43. H. Lin, W. Huang, P. Pattanasattayavong, J. Li, C. Wang and Z. Chen, Nat. Commun., 2019, 10, 4475.
44. Y. Chen, X. Wu, Y. Chu, J. Zhou, B. Zhou and J. Huang, Nano-Micro Lett., 2018, 10, 57.
45. H. Hong, S. N. Afraj, P. Y. Huang, Y. Z. Yeh, S. H. Tung, M. C. Chen and C. L. Liu, Nanoscale, 2021, 13, 20498–20507.
46. F. Liu, Y. Zhang, J. Wang, Y. Chen, L. Wang, G. Wang, J. Dong and C. Jiang, Nanotechnology, 2021, 32, 015203.
47. J. Kim, S. Park, S. K. Cha, I. K. Han and S. J. Kang, RSC Adv., 2018, 8, 23421–23425.
48. S. Kim, T. Lim, K. Sim, H. Kim, Y. Choi, K. Park and S. Pyo, ACS Appl. Mater. Interfaces, 2011, 3, 1451–1456.
49. G. Tarsoly and S. Pyo, Sens. Actuators, A, 2021, 332, 113178.
50. Z. Ge, S. Wan, M. Moin, Y. Liu, H. Zhang and L. Chen, Adv. Sci., 2025, 12, 2417304.
51. X. Han, S. Wan, L. He, J. Wang, Y. Zhao and M. Zhou, Adv. Sci., 2024, 11, 2401973.
52. J. Shin and H. Yoo, Nanomaterials, 2023, 13, 882.
53. P. Guo, J. Zhang and J. Huang, J. Semicond., 2025, 46, 021405.
54. Z. Wang, Z. Yin, Z. Yang, L. Chen, Y. Zhang and H. Xu, Chem. Eng. J., 2024, 501, 157512.
55. Y. Wu, S. Dai, X. Liu, J. Li, Z. Chen and L. Wang, Adv. Funct. Mater., 2024, 34, 2315175.
56. X. Wang, D. Hao and J. Huang, Sci. China Mater., 2022, 65, 2521–2528.
57. T. Leydecker, Z. M. Wang, F. Torricelli and E. Orgiu, Chem. Soc. Rev., 2020, 49, 7627–7670.
58. E. Quinn, J. Zhu, X. Li, J. Wang and Y. Li, J. Mater. Chem. C, 2017, 5, 8654–8681.
59. M. Liu, Y. Shih, X. Yu, M. Yu, X. Sun, C. Chueh and Z. Li, Adv. Funct. Mater., 2024, 34, 2405171.
60. P. Guo, J. Zhang, D. Liu, R. Wang, L. Li, L. Tian and J. Huang, ACS Appl. Mater. Interfaces, 2023, 15, 51483–51491.
61. J. Liu, Z. Yang, Z. Gong, Z. Shen, Y. Ye, B. Yang, Y. Qiu, B. Ye, L. Xu, T. Guo and S. Xu, ACS Appl. Mater. Interfaces, 2021, 13, 13362–13371.

---