Integrating intrinsic charge-trapping sites in an insulated MOF nanoparticle-based dielectric layer for effective photo/synaptic transistors

Abstract

The design of charge-trapping sites is the basic and critical factor to explore organic photomultiplier phototransistors and synaptic transistors. There are few examples of charge trapping achieved directly by constructing dielectric materials, as the traditional view holds that their wide band gaps make charge trapping and release difficult. This work employs the liquid phase epitaxy method to construct a two-dimensional confined dielectric MOF-199 island-like film with a 20 nm diameter as the dielectric layer of bulk heterojunction-based photo/synaptic transistors. Diverse energy level structures of metal complex nanostructures integrated intrinsic charge-trapping sites with multi-channel electron and energy exchange to the active layer. The multi-channel exchange combines the response advantages of machines and the human brain, while the intrinsic charge trapping avoids interface charge quenching. It demonstrates simultaneous high photocurrent and good long-term plasticity, a high phototransistor performance of photo-responsivity R = 650.1 A W⁻¹/specific detectivity Jones, and efficient photonic synaptic transistors with maximum paired-pulse facilitation index of 142% and single-pulse remaining ratio of 65%. All these results point to the organic phototransistors for emulating the functions of biological synapses, which indicate their potential as the building blocks of bionic electronic circuit systems.

---

New concepts

Charge/energy transfer pathways between semiconductors and insulators.

Insulators serve as excellent media for charge storage; however, due to their wide bandgap, establishing charge/energy transfer pathways between insulators and semiconductors proves challenging, limiting their widespread application as charge trapping materials.

Metal–organic frameworks (MOFs), with their diverse energy level structures, offer viable pathways for charge energy storage, thereby enabling the construction of high-performance photo/synaptic transistors. In the case study of MOF199 presented in this paper, the conjugated structure of the ligand facilitates charge exchange with the semiconductor, while metal ion d–d transitions associated with the energy exchange, providing a stable and controllable charge trapping pathway.

1. Introduction

Organic photodetectors have small 1/f noise, which allows them to operate at low frequencies, and have a very positive role in the construction of low-power-consumption and bionic electronic circuits.^1,2 Organic phototransistors (OPTs), which have an additional gate voltage compared with photodiodes and photoconductors, exhibit higher photosensitivity for photomultiplier-type devices.^3–5 With the introduction of narrow band gap materials, the charge release and response times in near-infrared OPTs reach millisecond level,^6–8 comparable to the recognition time of the human retina, highlighting their potential as building blocks for bionic electronic circuit systems, such as synaptic transistors. Synaptic transistors, through their integration of biomimetic mechanisms, low-power designs, multimodal signal processing, and process compatibility, exhibit significant advantages over conventional bio-inspired devices in energy efficiency, integration density, functional complexity, and bio-integration capabilities.^2,9,10

The design of charge-trapping sites is the basic and critical factor to explore photomultiplier phototransistors and synaptic transistors, for the photocurrent (I_ph) is proportional to the threshold voltage shift and the total number of trapped charges.^11–13 Charge-trapping engineering normally includes charge separation interface, charge-trapping type, energy level matching and transition, etc. Charge trapping typically occurs near interfaces such as electrode interfaces, donor/acceptor interfaces in active layers, and dielectric layer interfaces.^6,14,15 Besides the introduction of acceptors into the active layer to supply effective charge trapping, additional charge trapping in dielectric layers has been reported as an effective method for realizing a high ratio of charge storage.^16,17 For synaptic transistors, charge-trapping sites with different charge storage stability are very important for the control of short-term plasticity (STP) and long-term plasticity (LTP).^18,19 For the charge-trapping sites (floating gates) in the dielectric layer, semiconductor or metal nanoparticles are mainly selected, but they do not work well in organic active layer systems.^6,20,21 There are few examples of charge trapping achieved directly by constructing dielectric materials, because the traditional view is that the band gap of dielectric materials is very wide for difficulty of the trap and release of charge.

A metal–organic framework (MOF) is an organic–inorganic hybrid material composed of metal ions (or clusters) and organic ligands. Due to high crystallinity, stability, and functional diversity, MOFs have had many applications in the field of optoelectronic devices in recent years.^22–26 Some classic MOF systems have good insulation and a low dielectric constant, which can provide transport channels with a less polarizable environment that reduces energetic disorder.^27–30 In particular, the MOF structure, combined with ionic states, may provide a fixed charge-trapping site and diverse energy level structures, which are very suitable for the regulation of charge trapping.^31–33 It has been reported that MOF-based nanostructures can play a role as charge-trapping sites for photo/synaptic transistors and memory.^34,35 However, there are few reports on insulated MOF-based charge-trapping dielectrics for organic transistor performance enhancement.

In this paper, MOF-199 has been chosen to prepare dielectric thin films with MOF nanoparticles (MOF-NPs) by the liquid phase epitaxy (LPE) layer-by-layer approach as dielectric modification layers for OPTs and organic synaptic transistors (OSTs) (Fig. 1). The effect of the charge trapping action on the performance has been investigated. Compared with control devices with SiO₂ and PMMA dielectric layers, the role of MOF-NPs has been well explored, and appropriate cycles of MOF-NPs can induce stable charge trapping, leading to significant photocurrent enhancement and LTP. All these results point to the OPTs as emulating the functions of biological synapses, which indicate their potential as building blocks of bionic electronic circuit systems.

Fig. 1 (a) Device structure of OPTs and OSTs and schematic diagram of the parts of the device corresponding to the biological synaptic structure. (b) Schematic of a human synapse and a machine memristor. (c) The structures of PDPPBTT and PC₆₁BM. (d) The liquid phase epitaxy (LPE) layer-by-layer impregnation method and the structure of the MOF-NPs layer.

2. Results and discussion

2.1. Device architecture design

The device has a bottom-gate, top-contact, field-effect transistor configuration, as follows: Si/SiO₂/(MOF-NPs)/(PMMA)/PDPPBTT:PC₆₁BM/Ag–Ag electrodes (Fig. 1a). Typically, neurotransmitters released by presynaptic neurons in response to external stimuli attach to acceptors on the postsynaptic membrane, which is closely related to the movement of photogenerated carriers, and our device structure will be able to mimic the important functions of biological synapses. The device also supports direct long-term charge storage pathways. By integrating these two distinct operational modes, it combines photodetection and memory functions, thereby harnessing the complementary strengths of both human brain processing and machine operation (Fig. 1b).

PDPPBTT and PC₆₁BM heterojunctions were chosen as the active layers (Fig. 1c). PDPPBTT is widely used in photovoltaic devices due to its high hole mobility.⁴ Phenyl-C61-butyric acid methyl ester (PC₆₁BM) is a classic electron acceptor material with good stability.^6,7,36 Previous studies have shown that high-performance near-infrared photodetectors can be obtained when the donor/acceptor ratio is optimized to 5 : 1.^6,7 Here, we simultaneously introduce PC₆₁BM and MOF-NPs layer with different charge-trapping effects into the device, realizing photo/synaptic transistors with two different device operation modes.

2.2. Preparation and characterization of MOF-NPs film

We used the liquid phase epitaxy (LPE) method to prepare a low-dielectric-constant MOF-NPs film, as shown in Fig. 1d and Fig. S1. Possible microscopic growth mechanisms are summarized in Fig. 2d. By precisely regulating the impregnation time, solution concentration, and annealing time, high-quality growth of MOF-NP films can be achieved. Formation of MOF-NPs periodic structures is achieved by self-assembly of metal ions and organic ligands. Hydroxyl-functionalized Si/SiO₂ wafers were first immersed in a metal ion solution to allow chemical adsorption of Cu²⁺, then the excess unabsorbed Cu²⁺ was washed with solvent, and immersed in an organic ligand solution to allow the coordination reaction between Cu²⁺ and COO⁻ in 1,3,5-benzenetricarboxylic acid (H₃BTC). Finally, the unreacted material was rinsed off to obtain a 1-cycle MOF-NPs film.

Fig. 2 Structural and morphological analysis of MOF-NPs films. (a and c) XRD patterns and Raman spectra of MOF-NPs films with different LPE dipping cycles, respectively. (b) Size distribution of MOF-NPs at different LPE cycles. The inset is an AFM surface image of a 3-cycle MOF-NPs film. (d) Schematic of LPE preparation of MOF-NPs film on a molecular level.

By varying the number of LPE cycles, we prepared MOF-NP films with different thicknesses on Si/SiO₂ substrates. The interaction of H₃BTC with Cu²⁺, depending on the concentration of Cu²⁺ adsorbed on the substrate, may be via two different modes: when the Cu²⁺ concentration is large for type I mode, and when the Cu²⁺ concentration is small for type II mode. In the classical MOF structure, these two structures grow alternately and form a porous structure due to the variation in copper ion concentration. In the case of multilayer films grown by LPE, we believe that types I and II coexist in the growth mode. A very important phenomenon that supports this view is that there is no obvious difference in performance between odd and even cycles. As 3 cycles give the optimal performance, in this study, we mainly compare the changes of 1-, 3-, and 5-cycle performance. In addition, we spin-coated a PMMA film on the surface of the MOF-NPs film to further modify the surface defects of the MOF-NPs film.

The morphology of the MOF-NPs film was characterized by atomic force microscopy (AFM), X-ray diffraction (XRD) and scanning electron microscopy (SEM). Through a uniform multi-point nucleation process, a nano-island structure of MOF with a maximum diameter of 20 ± 2.7 (0.52) nm is formed on the Si/SiO₂ wafer (Fig. 2a and b and Fig. S2–S4), and the average size of the particles increases slightly with the increase of the number of cycles. It can be observed that the average sizes of the particles are 18.6 nm, 19.1 nm and 21.3 nm for 1, 3 and 5 cycles, respectively. Although the average size of the particles increases slightly with the increase of the number of cycles, the overall range of the change of the particle size is very limited, which is only 2.7 nm. The obtained MOF-NPs are much smaller in size than in previous reports^37,38 and are uniformly distributed, which is conducive to the reversible penetration of electrons. This phenomenon can be attributed to the two-dimensional self-limiting domain effect. Under 2D self-limiting domain conditions, the growth of nanoparticles is constrained by the interface and substrate, resulting in the particles only expanding in specific planes during the growth process. This restriction effectively inhibits the excessive growth and aggregation of particles.^38–41 The surface of the film is very uniform and flat, with a slight increase in surface roughness as the number of LPE cycles increases (Fig. S2). After loading the polymer PMMA on its surface, the surface roughness is further reduced. Nevertheless, single-layer molecular adsorption, even chemical adsorption, is rather challenging. LPE needs to involve in situ chemical reactions and cyclic growth on the interface. Considering the reduction of activation energy at the interface and the reversibility of coordination reactions, the formation of two-dimensional confined nanostructures would be a reasonable outcome. It has been reported that semiconducting MOF structures can form 2D nanosheets for charge trapping,^34,35 while this work relies on utilizing smaller conjugated structures to coordinate with copper ions.

This modification transforms the originally formed nanosheet structure at the interface into a nanoparticle structure. Additionally, the MOF structure undergoes a transition from a semiconductor to a dielectric structure, which is more conducive to the stable storage and regulation of charges.

The MOF film was characterized by XRD measurements as shown in Fig. 2a. The XRD data of the MOF film showed a weak peak corresponding to (222), which was identical to the highest peak in XRD of powder.^42,43 The 2θ value of (222) is 11.8°, and the value of d is 0.74 nm calculated by Bragg's equation, which is basically consistent with the thickness of one ideal cycle. However, we believe that in the case of a relatively small number of LPE cycles, this does not represent the formation of a good layered structure, but is mainly caused by the formation of MOF nanoparticles, i.e., a certain degree of non-two-dimensional domain-limited growth has occurred.

We further analyzed the chemical structure of MOF by Raman scattering spectroscopy, as shown in Fig. 2c. The peaks at 1007 cm⁻¹ correspond to the C=C stretching vibration of the benzene ring, and the peaks at 827 cm⁻¹ and 745 cm⁻¹ correspond to the O–C–O bending and coupling vibrations. All peaks appearing below 600 cm⁻¹ are mainly related to interactions between inorganic cores. The peaks located at 173 cm⁻¹ and 273 cm⁻¹ are attributed to Cu–Cu stretching and O–Cu–O bending vibrations, respectively. It is worth noting that the O–Cu–O band position for Cu(OAC)₂ shifted from 322 cm⁻¹ to 273 cm⁻¹, indicating that the Cu ion and BTC are coordinated to form a MOF structure.^44,45 In addition, the double peaks at 1613 cm⁻¹ and 1543 cm⁻¹ can be referred to the absorption peaks of C=O in the residual COOH in the partially coordinated BTC. With the coordination of COO⁻ with Cu²⁺, the bond energy of C=O decreases. In addition, when the domain-limiting effect increases, the red-shift intensity of the absorption peak becomes weaker. The band at 1613 cm⁻¹ corresponds to monocarboxylate coordination (C=O I), and that at 1543 cm⁻¹ corresponds to bicarboxylate coordination (C=O II), where the remaining two carboxylic acids form intramolecular hydrogen bonds. In the Raman spectra for different LPE cycles, the relative intensities of the two peaks are different, which can provide more detailed information about the film growth and chemical doping. In 1-cycle MOF, the peak of C=O II was relatively strong, indicating that the coordination with Cu²⁺ on the substrate mainly occurred via type II mode, which was due to the small density of Cu²⁺ adsorbed on the substrate. In the subsequent 3- and 5-cycle MOFs, the relative strength of the C=O II peaks decreased significantly and then stabilized, indicating that both type I and type II growth modes occurred during further growth, with their ratio remaining essentially constant. In addition, a new characteristic peak at 229 cm⁻¹ appeared after the combination of MOF and PMMA, which was due to the formation of a Cu–O bond between COOCH₃ in PMMA and Cu²⁺ in MOF. For samples obtained using different numbers of cycles, this peak was observed to exist to varying degrees. This may result not only from the direct interface modification with PMMA but also from ligand exchange with BTC.

2.3. Optical and electrical properties

The MOF-199 structure exhibits excellent optical and electrical properties, enabling the establishment of efficient charge and energy transfer channels between the bulk heterojunction and MOF-199 (Fig. 3a and b). This can be mainly attributed to the following aspects: 1. The mutual conversion between Cu²⁺ and Cu⁺ within the complex has minimal impact on the coordination structure, ensuring relatively stable structural integrity during the charge gain and loss processes. 2. The LUMO energy level of the BTC ligand (−4.14 eV) shows a good matching relationship with that of the donor and acceptor in the bulk heterojunction, facilitating effective charge transfer. 3. The energy levels of carboxylic acids/carboxylate are far from that in the active layer, and possess high photoexcitation energy, resulting in minimal interaction with the active layer. 4. The bulk heterojunction has strong absorption in the near-infrared region and exhibits a wide spectral response range (300–900 nm, Fig. 3d). The absorption of the MOF-NPs layer was also in the 600–850 nm range, corresponding to the d–d spin transition of the Cu²⁺ center (Fig. 3c), with the MOF-NPs layer forming a charged state under near-infrared irradiation or via energy transfer.^46,47 Therefore, the spatial distance between the MOF layer and the active layer plays a crucial role in charge-trapping behavior, which we further investigated by introducing a PMMA interlayer for comparison.

Fig. 3 (a and b) Device diagrams under different charge-trapping effects. (c and d) Optical absorption spectra of PDPPBTT:PC₆₁BM films and MOF-NPs layer. (e–h) Field-effect properties of different dielectric layer devices. (e and f) The transfer curves of different dielectric layer devices. (g) The relationship between D_it value and µ with the number of MOF cycles. (h) The relationship between N_t value and V_on with the number of MOF cycles, 0 cycle being a single SiO₂ or PMMA modified layer device. The 0-cycle devices in the MOF and MOF/PMMA series devices, respectively, represent bare SiO₂ control and the PMMA control.

The dielectric layer has a crucial influence on the field-effect performance of the device. The primary function of the dielectric parameters for a −60 V drain voltage layer is to provide electrical isolation between the gate and the organic semiconductor layer while allowing the gate voltage to modulate the carrier concentration in the semiconductor layer through electric field effects, thereby controlling the conductive state of the device. We analyzed the effect of the MOF film on the field-effect performance of the device. The transfer characteristic curves of different devices are shown in Fig. 3e and f, and the related parameters are summarized in Table 1.

Table 1 Field-effect properties of devices with different dielectric layers

Dielectric layer	K	µ (×10^−2 cm^2 V^−1 s^−1) (forward)	µ (×10^−2 cm^2 V^−1 s^−1) (reverse)	Von (V) (forward)	Von (V) (reverse)	Ioff (A)
SiO2	1.29 × 10^−5	0.12	0.91	28.5	29	6.59 × 10^−10
PMMA	5.07 × 10^−5	1.94	1.82	0.5	0	1.36 × 10^−11
1-cycle MOF	4.56 × 10^−5	1.54	2.20	18	3.5	2.78 × 10^−12
3-cycle MOF	5.36 × 10^−5	2.16	3.35	6.5	−3	8.68 × 10^−12
5-cycle MOF	6.17 × 10^−5	2.95	4.50	20	11.5	1.61 × 10^−11
1-cycle MOF/PMMA	5.43 × 10^−5	2.17	2.06	2	0	1.17 × 10^−11
3-cycle MOF/PMMA	5.17 × 10^−5	2.26	1.99	5.5	3	1.28 × 10^−11
5-cycle MOF/PMMA	4.81 × 10^−5	2.11	2.45	15.5	7.5	1.89 × 10^−11

---

Due to the ultrathin and discontinuous nature of the MOF-NPs layer, the thickness for 1, 3 and 5 cycles being 3 nm, 10 nm and 17 nm, respectively, from UV-visible absorption spectroscopy (Fig. S5) calibrated with a profilometer, it plays a very small role in total capacitance with the underlying 300 nm SiO₂. However, the most pronounced changes were observed in transconductance, subthreshold swing, and threshold voltage, effectively reducing polarization and enabling intrinsic charge storage.

As the MOF thickness increases, the value of K gradually increases, and µ correspondingly increases. After the addition of the PMMA layer, the mobility of the PMMA control device is significantly increased, but the difference between different MOF cycles is not significant.

A risk for dielectrics made with ions is that there can sometimes be free ions that contribute to an ionic mechanism for dielectric polarization that artificially inflates the measured mobility values. The forward and reverse transfer curves of all devices, as shown in Fig. S6, and the corresponding mobilities and V_on were calculated and are summarized in Table 1. The V_on shift and the hysteresis between forward/reverse transfer curves are not as large as on bare SiO₂ after the modification of MOF or MOF/PMMA, where the mobilities extracted from the reverse transfer curves were consistently higher. For devices after 1 and 3 LPE cycles, V_on remains close to zero, whereas devices after 5 LPE cycles exhibit a significant shift. This indicates that some charge trapping has already occurred in the 5-cycle devices under dark conditions, while the 1- and 3-cycle devices remain essentially charge-neutral. The observed variations thus mainly originate from the reversible coordination behavior. Cu⁺ and Cu²⁺ share similar coordination environments. Such structural similarity during electron exchange ensures high stability of the framework, making it a favorable site for charge trapping.

The turn-on voltage characterizes the charge-trapping sites of the whole device. The overall charge-trapping number N_t has a strong relationship with V_on (Fig. 3h), reflecting the total number of charge-trapping sites in the system.^48–50 As the device structure differences used in the experiment are all from the dielectric layer, we assume that the difference in experimental parameters is caused by the differences in the dielectric layer. When unmodified bare SiO₂ is used as the dielectric layer, the device has a large turn-on voltage V_on (28.5 V), which is mainly due to an increase in the high density of defects in the SiO₂ dielectric layer.^34,51 At 1 cycle, the MOF layer partially passivates SiO₂ surface defects, resulting in a reduced (from 28.5 V to 18 V) but still positive V_on; at 3 cycles, optimal passivation is achieved while introducing a moderate density of intrinsic trapping sites, yielding the minimum V_on of 6.5 V; at 5 cycles, additional defects within the thicker MOF layer increase overall charge trapping, leading to a positive V_on shift. The 3-cycle structure consistently delivers the best performance, as evident from both the morphological (Fig. S3) and electrical characterizations as well as the overall device performance. We attribute this behavior to the incomplete passivation of the SiO₂ surface defects in the single-cycle case, whereas three cycles achieve optimal passivation. As the number of cycles further increases, the MOF thickness grows accordingly, which introduces more internal defects within the MOF itself, leading to the observed performance trend. Furthermore, in Fig. 4, the best performance of the phototransistor was also observed at 3 cycles, for both the MOF and MOF/PMMA devices.

Fig. 4 Photodetection performance parameters at V_on for different dielectric layer devices. (a and b) Transfer characteristics curves of different dielectric layer devices without illumination and in different light intensities. (c) The relationship between the light-to-dark current ratio and ΔV_on with the number of MOF cycles. (d) The relationship between responsivity and specific detectivity with the number of MOF cycles; 0 cycle is a single SiO₂ or PMMA modified layer device, under −60 V source-drain voltage and 0.018 mW cm⁻² illumination at 810 nm. (e–h) Time responses of different dielectric layer devices under the light co-control method at −60 V drain voltage. (i–l) I–T curves of the device under different gate voltage stages in the dark and under light illumination, where stages 1 and 2 correspond to gate voltages of 0 V and 30 V applied for 10 s each, and stage 3 corresponds to a gate voltage of 0 V applied for 30 s.

The density of interface states (D_it) is the number of defects per unit area in the unit energy range, which can directly measure the quality of the dielectric/active layer interface. The SiO₂ control device has high D_it (7.34 × 10¹² eV⁻¹ cm⁻²), indicating a high defect density at the dielectric/active layer interface. After loading the MOF film, the D_it values decreased significantly with the increase of the number of cycles, and the D_it value of the 5-cycle MOF film was 5.81 × 10¹¹ eV⁻¹ cm⁻², which was 12.6 times lower than that of the SiO₂ control device. D_it gradually decreases with increasing cycles, which indicates that the introduction of the MOF ultrathin film can effectively reduce the defect density of the conductive channel. Thus, the overall increase in V_on and N_t values for the 5-cycle MOF film would come from the defects inside the dielectric layer. The effective separation between the charge-trapping site and the conduction channel can reduce the quenching effect of current, which is beneficial to obtain high photocurrent.

The layer of PMMA film on the surface of the MOF-NPs film is spin-coated to modify the dielectric MOF-NPs. For all experimental conditions, V_on decreased after adding PMMA (Table 1), which proved that PMMA had a certain modification effect on the surface defects of MOF. Interestingly, the modifying effect was obvious when the number of MOF cycles was small, but became weaker when the number increased. Therefore, unlike the MOF series devices, V_on of the MOF/PMMA series devices increases with the number of MOF cycles, which is because PMMA only modifies the surface defects of the MOF and does not affect the charge-trapping sites within the MOF layer.

In addition, the absolute value of D_it is overall smaller after adding PMMA (Fig. 3g). The above results are also illustrated by the decrease in the subthreshold swing (Fig. S6). In the 3- and 5-cycle MOF devices, the modifying effect of PMMA on defects, especially on the defects in the channel, is relatively weak. The spatial separation effect plays a dominant role. Moreover, the analytical results of D_it have a good correspondence with µ, which proves that both MOF and PMMA have a better modification effect on the conductive channel and reduce the density of channel defects.

The evolution of I_dark and I_1/f in MOF devices and MOF/PMMA devices is essentially the same. Except for the Si control device, all devices maintained relatively low I_dark and I_1/f, which increased slightly with the increase of MOF cycles (Table 1 and Fig. S7). Because the variation of I_dark and I_1/f is much smaller than that of V_on and ΔV_on, the final optimal performance of the photodetector is mainly determined by V_on and ΔV_on.

2.4. Photodetection performance of the device

The effect of the dielectric modification layer of MOF-NPs thin films on the photodetection performance of the devices was further investigated. The transfer curves in the dark and under 810 nm incident light are compared for different cycle MOF-based devices, as shown in Fig. 4a and b and S8 and S9, and the relevant parameters are displayed in Table 2 and Table S1. The key parameter changes with the number of cycles are presented in Fig. 4c and d. V_on shifted to the right after illumination in all devices, which was attributed to the trapping of electrons. The source-drain current I_d increases significantly when the gate voltage is larger than V_on in the light. The device performance is closely related to the gate voltage (V_g), where the best light-to-dark ratio (I_light/I_dark) is normally achieved at V_on in the dark. In the dark state, V_on is already 28.5 V (Fig. S8a), and when the device is illuminated with a light intensity of 0.018 mW cm⁻², V_on increases to 30 V, with a ΔV_on of only 1.5 V. Such low ΔV_on makes the photocurrent grow less, where I_light/I_dark is only 2.78, which makes it difficult to use as a photodetector. The evolutions of ΔV_on, I_light/I_dark, R and as a function of MOF cycles are very similar between MOF series and MOF/PMMA series devices, where the 3-cycle ones have the best comprehensive performance. In the 3-cycle MOF device, when V_g = 6.5 V, ΔV_on and I_light/I_dark value are at their highest, R and are 417 A W⁻¹ and 5.01 × 10¹⁵ Jones, respectively, which are increased by 1.8 times and 2.3 times compared with PMMA devices. Under irradiation, ΔV_on of the MOF/PMMA device is further increased compared to that of the MOF device. I_light/I_dark, R and values of the 1-cycle and 3-cycle MOF/PMMA devices are significantly higher than those of the PMMA devices. Among them, the 3-cycle MOF/PMMA device has the best performance. When V_g = 5.5 V, R and are 650.1 A W⁻¹ and 6.76 × 10¹⁵ Jones, respectively. Compared with PMMA devices, R and are increased by 2.9 times and 3.1 times, respectively. Besides, the surface roughness of the active layer increases with the number of MOF layers (Fig. S2), where the 5-layer MOF exhibits greater roughness for a poorer morphology of the surface-loaded active layer (Fig. S10). This might be one of the reasons for the poor performance of the 5-cycle devices.

Table 2 Photodetection performance and synaptic transistor parameters of devices

Dielectric layer	R (A W^−1)		tr/tf (ms)	Maximum PPF index (%)	Maximum EPSC (A)	Remaining ratio (%)
Concrete parameters of device under 810 nm illumination.
SiO2	2.1	0.23	1450/45	—	—	—
PMMA	226.6	21.7	25/37	—	—	—
3-cycle MOF	417	50.1	47/60	142	5.88 × 10^−7	87
3-cycle MOF/PMMA	650.1	67.6	28/31	133	5.01 × 10^−7	83

---

To our surprise, ΔV_on does not undergo significant growth after the addition of MOF, which is completely different from what we have previously observed in Ag NPs and ZnO NPs systems. When the initial V_on in the dark is large and the V_on shift becomes relatively small, the photocurrent, I_light/I_dark, and G still improve significantly after the addition of the MOF layer (Table S1). For example, ΔV_on of the 3-cycle MOF device is only 20 V, which is smaller than the 21.5 V of the PMMA control device, but its optical gain is 1.8 times that of the PMMA control device. ΔV_on of the 3-cycle MOF/PMMA device is 24.5 V, which is only 3 V (14%) larger than that of the PMMA control device, but its gain value is nearly three times that of the PMMA control device. This phenomenon is also mainly caused by the use of a dielectric charge-trapping layer. Theoretically, increasing V_on is not the sole route to enhance the photocurrent. In photodetectors, the photocurrent magnitude is governed by both the transconductance (K) and ΔV_on, as expressed in the relationship I_ph = KΔV_on. The incorporation of the MOF effectively enhances transconductance, which is another critical way of improving photocurrent.

Charge trap sites in dielectric materials exhibit spatial localization, with their energy distribution decoupled from the semiconductor band structure, thereby conferring quenching resistance, which can help to obtain a stable photomultiplication effect and LTP. Nevertheless, if the gate voltage is applied while turning off the light, the charge trapping by the MOF can be effectively quenched, which makes the charge-trapping process reversible (Fig. 4e–h and Fig. S11 and S12).

The photocurrent of the 1-cycle MOF device increases with illumination time. After the illumination is turned off, the dark current is driven to be stable near 30% of the photocurrent and no longer decreases further. In addition, as the number of MOF cycles increases, the rise time (t_r) value gradually increases, and the fall time (t_f) value cannot even be measured for 3-cycle MOF and 5-cycle MOF, indicating that the MOF can trap charges, and the trapped charges are not easily released. The response speed of the MOF/PMMA devices was significantly faster than that of the MOF devices, indicating that the load of PMMA weakened the charge-trapping effect of MOF. Meanwhile, dielectric charge trapping improves the photocurrent and LTP.

It is more interesting that in the second charge-trap site system of the dielectric, the trapped charge can be erased by introducing a gate voltage, which can achieve fast light detection response and memory erasure. In our previous work, we used an optoelectronic dual control method in semiconductor- and conductor-based charge-trap sites,⁵² which is still applicable to dielectric trapping sites. Adding a synchronous gate voltage at light on and light off can effectively increase charge trapping and release. The photocurrent quickly saturates and stabilizes after illumination, and returns to a lower dark current after the end of illumination. For the 3-cycle MOF/PMMA device, t_r and t_f are as quick as 28/31 ms.

A pulsed-gate voltage measurement was conducted to investigate the charge-trapping behavior of the device. As in Fig. 4i–l, three consecutive testing stages were defined, while the MOF and MOF/PMMA device series were compared in the dark and under light. In stage I, the device was biased at a gate voltage of 0 V for 10 s under the initial condition. Subsequently, in stage II, the gate voltage was switched to +30 V for 10 s to enable electron trapping. Finally, in stage III, the gate voltage was reset to 0 V and maintained for 40 s. The drain current was recorded and compared across the three stages. During stage I, the device exhibited a very small operating current, as only a few electrons could be trapped in the MOF under a 0 V gate bias. The weak electric field in the channel limited hole accumulation, resulting in a low hole current. In stage II, although a positive gate voltage of +30 V was applied, the current remained low. This is because the strong gate field formed in this state is opposite in direction to the hole accumulation in the channel, thus suppressing the hole current. Meanwhile, the positive gate bias led to electron trapping in the MOF, generating an internal electric field that partially compensated the external gate field; however, this compensation was insufficient to induce significant hole accumulation. In contrast, stage III displayed a pronounced increase in current under 0 V gate bias. Unlike stage I, the removal of the external gate field allowed the trapped electrons within the MOF to produce an internal field that facilitated hole accumulation in the channel, leading to a large current.

This distinct behavior strongly confirms the charge-trapping capability of the MOF layer, and the differences in trapping ability across different devices and conditions can help determine whether charges are transferred from the active layer to the MOF or if the MOF itself exhibits some photoactivity. The coexistence and rationality of these two processes would be mainly supported by the energy-level analysis and optical absorption spectra discussed in this work. With respect to the different charge-trapping processes in MOFs, two aspects are particularly noteworthy. First, by comparing the MOF and MOF/PMMA series devices under dark conditions, it can be observed that the current significantly decreases after the addition of the PMMA layer. This demonstrates the presence of a short-range charge-trapping effect between the MOF and the active layer; more importantly, under illumination, the current in stage III exhibits a notable increase compared with that in the dark. Although this could potentially be attributed to the overall current enhancement induced by illumination, the fact that the charge-trapping capabilities of samples with varying thicknesses remain identical in MOF-based devices suggests that the actual charge trapping is far less than the charge-holding capacity of the MOF, indicating the generation of new photoactive trapping sites after illumination.

2.5. Biological synaptic performance

In addition to their photodetection capabilities, our devices can also function as efficient photonic synaptic transistors that emulate the functions of biological synapses. By modulating with optical pulses, we replicate key biological synaptic functions, including excitatory postsynaptic currents (EPSC), paired-pulse facilitation (PPF), and the transition from short-term plasticity (STP) to long-term plasticity (LTP). Furthermore, the photonic synaptic devices are utilized to simulate brain-like behaviors of “learning-forgetting-relearning”. As we mentioned before, this is mainly due to the fact that the MOF-NPs layer has a stable charge trapping and releasing effect, and thus is able to mimic the synaptic structure of the human brain as a synaptic transistor. Charge trapping dynamically affects the amplitude and attenuation of the EPSC and the timing dependence of the PPF by modulating the capture-release balance of carriers, and the continued accumulation of carriers by charge trapping drives the STP to LTP transition. Here, we use the 3-cycle MOF device with a good charge-trapping effect for demonstration.

Biological synapses are essential components for the transmission and regulation of information within the brain, acting as junctions where synaptic neurons are functionally linked.⁵³ We utilized 810 nm light pulses as presynaptic stimuli, the active layer as the pre-synapse, and the MOF-NPs layer as post-synapse. Moreover, the EPSC under different conditions was measured by varying the pulse excitation time (pulse width) and interval time (Fig. 5a). Fig. 5b depicts the typical EPSC response triggered by 810 nm light pulses (68 mW cm⁻², 20 s). Upon application of the light pulse stimulus, the current rises rapidly. After the light pulse is removed, the current gradually decreases. This behavior closely resembles that of a biological synapse.

Fig. 5 Synaptic plasticity and “learning experience” behavior of the device. (a) Schematic of EPSC triggered by different pulse widths and interval times. (b) EPSC triggered by a single light pulse (68 mW cm⁻², 810 nm, 20 s). (c) EPSC triggered by two consecutive light pulses (68 mW cm⁻², 810 nm, 2 s) with an interval of 3 s. (d) PPF index variation as a function of Δt. The light intensity is fixed at 68 mW cm⁻², and the light pulse duration is 2 s. (e and f) Transition of STM to LTM behavior by pulse width (68 mW cm⁻², 810 nm) and pulse intensity. (g and h) Max-EPSC (68 mW cm⁻², 810 nm) and the remaining ratio with pulse intensity and pulse width, respectively.

PPF is a distinctive characteristic of STP that is crucial for the recognition and interpretation of visual, auditory, and other signals within the biological nervous system.⁵⁴ This phenomenon occurs when two consecutive stimuli of the same intensity are presented, resulting in the EPSC induced by the second stimulus being stronger than that generated by the first⁴⁵ as shown in Fig. 5c, we applied two successive light pulses to the device (68 mW cm⁻², 810 nm, 2 s), with a 3 s interval between them. The findings indicate that the EPSC for the second pulse was significantly higher than for the first. This enhancement occurs because the photogenerated holes from the initial light pulse had not completely recombined by the time the second pulse was delivered within a short duration. When the second light pulse was applied, the newly generated photogenerated holes combined with those from the first pulse that had not yet recombined, leading to a larger EPSC during the second pulse and demonstrating typical PPF behavior.

The PPF index was further used to investigate synaptic behavior, and it can be defined by the following equation:

where A₁ and A₂ represent the ΔEPSC values from the first and second light stimulations, respectively.

Fig. 5d illustrates the relationship between the pulse interval (3 s to 60 s) and the PPF index. As the interval increased, the PPF exhibited an exponential decay. This decay occurs because with a longer interval between the two light pulses, most of the photogenerated holes recombine before the arrival of the second pulse, resulting in a decrease in current. Consequently, the PPF index diminishes as the intervals between light stimulations increase. This decay trend is able to be accurately modeled by the following double-exponential function, similar to the behavior observed in biological synapses:

where Δt represents the time interval between light pulses, C₀, C₁, and C₂ correspond to the initial facilitation magnitudes, and τ₁ and τ₂ represent the characteristic relaxation time for rapid and slow decay phases. In addition, the PPF index of 3-cycle MOF/PMMA is lower than that of 3-cycle MOF, indicating that the addition of PMMA weakened the charge-trapping effect of MOF, which is consistent with previous conclusions. Besides, the experimental results of the synaptic transistor might be the best evidence that free ions are not present and the MOF-NPs thin film has a good quality. It is generally believed that the electrochemical principle would result in a PPF increase of more than 200%, but in this study, the PPF value was around 140%.

Memory can be categorized into short-term memory (STM) and long-term memory (LTM) based on the duration of information retention. The memory capacity of the human brain is related to the intensity of learning, and the transition from STP to LTP at biological synapses can be achieved through repeated learning and consolidation. To further emulate the synaptic plasticity observed in biological synapses, we adjusted the light pulse width and intensity to facilitate the transition from STP to LTP. As illustrated in Fig. 5e, the EPSC significantly increased with longer pulse widths, suggesting that the enhancement in memory level takes more time to return to its initial state. This reflects a transition from STP to LTP, closely resembling the memory processes in the human brain. Additionally, the intensity of light pulses exhibited similar effects on the transition from STP to LTP at biological synapses to those of the pulse width (Fig. 5f). Thus, by manipulating the duration and intensity of light pulses, we can effectively regulate synaptic memory levels, demonstrating the potential of our device to mimic the learning and memory capabilities of the human brain.

In addition, we compared the Max-EPSC and remaining ratio of 3-cycle MOF and 3-cycle MOF/PMMA under different pulse intensities and pulse widths. As shown in Fig. 5g and h, with an increase of pulse intensity and pulse width, the Max-EPSC of different devices is gradually enhanced, indicating that the stronger the optical stimulation, then the larger the EPSC value, which is similar to the response behavior of biological synapses. Moreover, the Max-EPSC and remaining ratio of the 3-cycle MOF are higher than those of the 3-cycle MOF/PMMA (Fig. S13), indicating that the MOF-NPs have higher memory performance, and the addition of PMMA leads to a decrease in the performance. This is also mainly due to the fact that the addition of PMMA attenuates the charge-trapping effect of the MOF.

Moreover, the repeatable plasticity of biological synapses allows the human brain to engage in a learn-forget-relearn cycle when acquiring new knowledge. The relearning process requires less time to recall and to replicate this “learning experience” behavior, we applied 51 continuous light pulses to stimulate the device (Fig. S14a). The EPSC continued to rise during the ongoing light pulse stimulation. Following this stimulation, the EPSC began to decrease, reflecting the forgetting that occurs during the learning process (Fig. S14b). To recover the forgotten information, light pulses were reintroduced to the device. Remarkably, only 35 light pulses were necessary to restore the previous learning level, significantly fewer than what was required during the initial learning phase (Fig. S14c).

Based on the application of harnessing the complementary strengths of both human brain processing and machine operation, we consider the following parameters as the core parameters of the research system: the device's perception capability (D*), response time, bionic characteristics (PPF), and energy efficiency (power). These parameters are not independent of each other; instead, they have a mutually restrictive relationship. For example, to meet the detectivity requirement, the photocurrent needs to be as large as possible, which will result in relatively high power consumption.

For devices with near-infrared response, the trade-off between D* and response time has been relatively well addressed.⁷ The device D* reported in this article is relatively high among those found in organic photodetectors (Table S2). From the perspective of trade-off between D* and power consumption, MOF devices are more suitable for the research of photoelectric neural synapses compared to MOF/PMMA devices. According to the calculation in Fig. 5c, for a 3-layer MOF device, with V set to −0.1 V, I set to 0.55µA, and Δt set to 2 s, the value obtained by E_spike calculation is 0.1 µJ. Compared with previous reports, although the EPSC is relatively large, the power consumption has reached the requirements of the synaptic device (Tables S3 and S4).

3. Conclusions

This work focuses on insulators' wide bandgap hindering charge transfer with semiconductors, limiting their use as charge-trapping materials. MOF-199 island-like film was chosen as a dielectric layer. In dielectric MOFs, charge-trapping sites are not doping or defect structures within the system but are instead based on the charge trapping and release inherent to the dielectric material itself. This approach enables stable and controllable charge storage and release, a feat that is unimaginable with charge-trapping sites in semiconductors and conductors. Its diverse energy levels enable intrinsic charge trapping, establishing stable charge/energy transfer pathways between semiconductor and insulator. This leads to high-performance photo/synaptic transistors, showing potential for bionic electronic circuits.

4. Materials and methods

4.1. Chemicals and materials

Copper(II) acetate (Cu(OAC)₂·H₂O) and 1,3,5-benzenetricarboxylic acid (H₃BTC) were purchased from Aladdin Reagent Co. Ltd. Ethanol and acetone were obtained from Guangzhou Chemical Reagent Co. Ltd. Chloroform, butyl acetate, and PMMA (ε_r ≈ 3.5, M_w ≈ 350 000) were purchased from Aldrich. PC₆₁BM was purchased from 1-Material. PDPPBTT was purchased from Luminescence Technology Corp. All of these commercially available chemicals were used without further purification.

4.2. Preparation of the MOF-NPs ultrathin film

An ultrasonic cleaner was used for the silicon wafer in turn with deionized water, acetone, deionized water, isopropanol, and deionized water. Then, the SiO₂/Si substrate was pretreated with H₂SO₄ and H₂O₂ (H₂SO₄/H₂O₂ = 7 : 3) at 80 °C for 30 min to achieve surface hydroxylation modification. The hydroxy-functionalized substrate was a good candidate for the preparation of MOF thin film. There were three containers for different solutions (1.5 mM Cu(OAc)₂, 1 mM H₃BTC and pure ethanol) as shown in Fig. S1. The functionalized SiO₂/Si substrate was immersed in each container in turn and cleaned with ethanol. By changing the number of cycles, MOF ultrathin films with different thicknesses can be prepared.

4.3. Device preparation

The active layer solution of PDPPBTT : PC₆₁BM (D/A ratio 5 : 1, 7 mg mL⁻¹ of PDPPBTT in chloroform) was used to prepare devices. In order to completely dissolve the polymer, the active layer solution was heated and stirred on a hot plate at 30 °C for 24 hours. Then, the semiconducting polymers PDPPBTT and PC₆₁BM were deposited on the bare SiO₂/Si and MOF/SiO₂/Si substrates by spin-coating the polymer solution at 2500 rpm for 40 s. Moreover, we used PMMA as a dielectric layer to spin-coat the SiO₂/Si substrate and MOF/SiO₂/Si substrate to prepare devices. The PMMA (15 mg mL⁻¹, 2000 rpm) in butyl acetate was spin-coated on silicon wafers and crosslinked at 220 °C for 30 min. Then, Ag film (100 nm) was deposited for use as electrodes under vacuum.

4.4. Characterization methods

All measurements were performed under ambient conditions. Each device was placed in a vacuum chamber (≈8 h, starting immediately after the device was ready) to completely remove the solvent before measurements. The film thicknesses were measured by a Dektak 150 from Veeco USA. Absorption and transmission measurements were performed with a Shimadzu UV-3100 spectrophotometer. The OPT characterizations were performed under ambient conditions using a Cascade RF1 manual probe station and a semiconductor parameter analyzer (Keithley 2636B). The light illumination was provided by an Opolette 355 LD (low divergence) with a wavelength range of 410–2400 nm (7 ns, 20 Hz), where the light intensity was tested by a laser power meter (Header Ophir NOVA II and probe PD300-UV). AFM images were obtained using a MultiMode 8 (Bruker). SEM images for the morphology of thin films were obtained using a Regulus8100. XRD patterns were collected with an X-ray diffractometer (D8 Advance, Bruker) using Cu Kα radiation with a scanning rate of 0.5° min⁻¹. Raman scattering patterns were measured using an inVia Reflex microscopic laser confocal Raman spectrometer (Renishaw). MOF-NPs particle size distribution was statistically derived using ImageJ software.

Calculation formulae. The device mobility, which is used to characterize semiconductor device features, is calculated according to the following equation:here, L is the channel length, W is the channel width, C_i is the capacitance per unit area of the gate dielectric, I_SD is the source-drain current, and V_G is the gate voltage. The K value is the growth rate of source-drain current with the change of gate voltage.

The performance of a photodetector can be characterized by three important parameters: the responsivity (R), the gain (G), and the specific detectivity , and R, G, and are calculated according to the following equations:

here, I_ph is the photogenerated current, I_light is the photocurrent, I_dark is the dark current, P_in is the optical power incident on the channel, A is the device conductive channel area, h is the Planck constant, and q is the elementary charge.

The formulas for calculating the density of interfacial defects and overall charge-trapping number N_t are as follows:

here, e is the Euler's number, k is the Boltzmann constant, q is the unit charge, T is the absolute temperature, C_i is the capacitance per unit area, and V_on is the turn-on voltage.

Author contributions

C. L.: preparation characterization of MOF-NPs and OPT devices, article writing. D. S.: characterizing the performance of synaptic devices, article writing. R. F.: UV-visible and Raman spectroscopy testing and analysis. C. L: preparation and characterization of MOF-NPs. H. Q.: electrochemical analysis. Q. L.: OPT and OST characterization. Z. X.: analysis of physical and chemical mechanisms, L. L.: the initial idea and article writing.

Conflicts of interest

There are no conflicts to declare.

Data availability

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/d5mh02148k.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (52273178, 52373178, 21975076) and the Natural Science Foundation of Guangdong Province (2023B1212060003).

Notes and references

1. C. C. Hung, Y. C. Chiang, Y. C. Lin, Y. C. Chiu and W. C. Chen, Adv. Sci., 2021, 8, 2100742.
2. E. L. Li, C. S. Gao, R. J. Yu, X. M. Wang, L. H. He, Y. Y. Hu, H. J. Chen, H. P. Chen and T. L. Guo, Nat. Commun., 2022, 13, 2898.
3. C. Wang, X. Zhang and W. Hu, Chem. Soc. Rev., 2020, 49, 653–670.
4. K. J. Baeg, M. Binda, D. Natali, M. Caironi and Y. Y. Noh, Adv. Mater., 2013, 25, 4267–4295.
5. H. Yoo, I. S. Lee, S. Jung, S. M. Rho, B. H. Kang and H. J. Kim, Adv. Mater., 2021, 33, 2006091.
6. M. Shou, Q. Zhang, H. Li, S. Xiong, B. Hu, J. Zhou, N. Zheng, Z. Xie, L. Ying and L. Liu, Adv. Opt. Mater., 2021, 9, 2002031.
7. M. Shou, Q. Zhang, S. Xiong, T. Han, J. Zhou, N. Zheng, Z. Xie and L. Liu, ACS Appl. Mater. Interfaces, 2021, 13, 5293–5301.
8. C. R. Liu, S. C. Xiong, D. Sun, Z. Q. Xie and L. L. Liu, J. Mater. Chem. C, 2024, 12, 16100–16107.
9. R. J. Yu, L. H. He, C. S. Gao, X. H. Zhang, E. L. Li, T. L. Guo, W. W. Li and H. P. Chen, Nat. Commun., 2022, 13, 7019.
10. X. T. Zhu, C. S. Gao, Y. W. Ren, X. H. Zhang, E. L. Li, C. Y. Wang, F. X. Yang, J. S. Wu, W. P. Hu and H. P. Chen, Adv. Mater., 2023, 35, 2301468.
11. Z. Zhao, C. Xu, Y. Ma, K. Yang, M. Liu, X. Zhu, Z. Zhou, L. Shen, G. Yuan and F. Zhang, Adv. Funct. Mater., 2022, 32, 2203606.
12. R. M. Nie, Y. Y. Wang and X. Y. Deng, ACS Appl. Mater. Interfaces, 2014, 6, 7032–7037.
13. J. L. Miao and F. J. Zhang, Laser Photonics Rev., 2019, 13, 1800204.
14. J. W. John, A. Mishra, R. Debbarma, I. Verzhbitskiy and K. E. J. Goh, Nanoscale, 2023, 15, 16818–16835.
15. D. Xue, Q. Wang, M. Xie, W. J. Gong, Y. Y. Zhang, Y. Yin, Y. J. Wei, J. Lu, J. D. Zhang, S. Duhm, Z. Wang, L. F. Chi and L. Z. Huang, Adv. Opt. Mater., 2024, 12, 202302091.
16. J. Robertson and R. M. Wallace, Mater. Sci. Eng. R., 2015, 88, 1–41.
17. A. Nawaz and I. A. Hümmelgen, J. Mater. Sci.: Mater. Electron., 2019, 30, 5299–5326.
18. Z. Y. Guo, J. Y. Zhang, X. Liu, L. Wang, L. Z. Xiong and J. Huang, Adv. Funct. Mater., 2023, 33, 202305508.
19. S. J. Chen and J. Huang, ACS Appl. Electron. Mater., 2020, 2, 1815–1825.
20. T. Han, L. L. Liu, M. Y. Wei, C. Wang, X. Y. Wu, Z. Q. Xie and Y. G. Ma, Phys. Chem. Chem. Phys., 2017, 19, 17653–17660.
21. T. Han, Y. Xu, M. Shou, Z. Xie, L. Ying, C. Jiang, H. Wang, S. Ding, G. Jin, Q. Liu, X. Huang and L. Liu, Org. Electron., 2020, 86, 105883.
22. X. H. Ren, G. C. Liao, Z. J. Li, H. Qiao, Y. Zhang, X. Yu, B. Wang, H. Tan, L. Shi, X. Qi and H. Zhang, Coord. Chem. Rev., 2021, 435, 213781.
23. C. C. Chueh, C. I. Chen, Y. A. Su, H. Konnerth, Y. J. Gu, C. W. Kung and K. C. W. Wu, J. Mater. Chem. A, 2019, 7, 17079–17095.
24. V. Stavila, A. A. Talin and M. D. Allendorf, Chem. Soc. Rev., 2014, 43, 5994–6010.
25. Y. S. Zheng, F. Z. Sun, X. Han, J. L. Xu and X. H. Bu, Adv. Opt. Mater., 2020, 8, 2000110.
26. G. O. N. Ndjawa, M. R. Tchalala, O. Shekhah, J. I. Khan, A. E. Mansour, J. Czaban-Józwiak, L. J. Weselinski, H. A. Ahsaine, A. Amassian and M. Eddaoudi, Materials, 2019, 12, 2457.
27. J. Veres, S. D. Ogier, S. W. Leeming, D. C. Cupertino and S. M. Khaffaf, Adv. Funct. Mater., 2003, 13, 199–204.
28. S. H. Lee, Y. Xu, D. Khim, W. T. Park, D. Y. Kim and Y. Y. Noh, ACS Appl. Mater. Interfaces, 2016, 8, 32421–32431.
29. F. S. Kim, D. K. Hwang, B. Kippelen and S. A. Jenekhe, Appl. Phys. Lett., 2011, 99, 173303.
30. J. Lee, J. W. Chung, G. B. Yoon, M. H. Lee, D. H. Kim, J. Park, J. K. Lee and M. S. Kang, ACS Appl. Mater. Interfaces, 2016, 8, 30344–30350.
31. S. Eslava, L. P. Zhang, S. Esconjauregui, J. W. Yang, K. Vanstreels, M. R. Baklanov and E. Saiz, Chem. Mater., 2013, 25, 27–33.
32. W. J. Li, J. Liu, Z. H. Sun, T. F. Liu, J. Lü, S. Y. Gao, C. He, R. Cao and J. H. Luo, Nat. Commun., 2016, 7, 11830.
33. M. Krishtab, I. Stassen, T. Stassin, A. J. Cruz, O. O. Okudur, S. Armini, C. Wilson, S. De Gendt and R. Ameloot, Nat. Commun., 2019, 10, 3729.
34. N. Shi, J. Zhang, Z. Ding, H. Jiang, Y. Yan, D. Gu, W. Li, M. Yi, F. Huang, S. Chen, L. Xie, Y. Ren, Y. Li and W. Huang, Adv. Funct. Mater., 2022, 32, 2110784.
35. G. Ding, B. Yang, K. Zhou, C. Zhang, Y. Wang, J.-Q. Yang, S.-T. Han, Y. Zhai, V. A. L. Roy and Y. Zhou, Adv. Electron. Mater., 2020, 6, 1900978.
36. M. Shou, Q. Zhang, Y. Zhang, X. Hou, J. Zheng, J. Zhou, S. Xiong, N. Zheng, Z. Xie and L. Liu, J. Mater. Chem. C, 2022, 10, 16070–16077.
37. J. Fonseca, L. X. Meng, I. Imaz and D. Maspoch, Chem. Soc. Rev., 2023, 52, 2528–2543.
38. S. Z. Wang, C. M. McGuirk, A. d'Aquino, J. A. Mason and C. A. Mirkin, Adv. Mater., 2018, 30, 1800202.
39. R. Zheng, Z. H. Fu, W. H. Deng, Y. Y. Wen, A. Q. Wu, X. L. Ye and G. Xu, Angew. Chem., Int. Ed., 2022, 61, e202212797.
40. A. L. Semrau, S. Wannapaiboon, S. P. Pujari, P. Vervoorts, B. Albada, H. Zuilhof and R. A. Fischer, Cryst. Growth Des., 2019, 19, 1738–1747.
41. Z. G. Gu, H. Fu, T. Neumann, Z. X. Xu, W. Q. Fu, W. Wenzel, L. Zhang, J. Zhang and C. Wöll, ACS Nano, 2016, 10, 977–983.
42. Z. G. Gu, A. Pfriem, S. Hamsch, H. Breitwieser, J. Wohlgemuth, L. Heinke, H. Gliemann and C. Wöll, Micropor. Mesopor. Mater., 2015, 211, 82–87.
43. C. J. Chen, Y. Qi and Z. P. Qiao, J. Mater. Res., 2012, 27, 2911–2915.
44. M. Todaro, A. Alessi, L. Sciortino, S. Agnello, M. Cannas, F. M. Gelardi and G. Buscarino, J. Spectro., 2016, 2016, 8074297.
45. F. S. Gentile, M. Pannico, M. Causà, G. Mensitieri, G. Di Palma, G. Scherillo and P. Musto, J. Mater. Chem. A, 2020, 8, 10796–10812.
46. X. B. Min, X. Y. Li, J. Zhao, X. X. Hu and W. C. Yang,, Environ. Technol., 2021, 42, 4134–4144.
47. L. Liu, D. Q. Wu, B. Zhao, X. Han, J. Wu, H. W. Hou and Y. T. Fan, Dalton Trans., 2015, 44, 1406–1411.
48. N. Yoshida, E. Waki, M. Arai, K. Yamasaki, J. H. Han, M. Takenaka and S. Takagi, Thin Solid Films, 2014, 557, 237–240.
49. Y. Bargaoui, M. Troudi, N. Sghaier, N. Yacoubi, V. Aimez and A. Souifi, Microelectron. Eng., 2016, 159, 151–154.
50. M. C. Tsai, C. I. Wang, Y. C. Chen, Y. J. Chen, K. S. Li, M. C. Chen and M. J. Chen, Semicond. Sci. Technol., 2018, 33, 035013.
51. T. Han, L. Liu, M. Shou, Z. Xie, L. Ying, C. Jiang, X. Huang, H. Li and Y. Ma, ACS Photonics, 2018, 5, 4650–4659.
52. T. Han, M. H. Shou, L. L. Liu, Z. Q. Xie, L. Ying, C. Z. Jiang, H. Y. Wang, M. Yao, H. M. Deng, G. Jin, J. W. Chen and Y. G. Ma, J. Mater. Chem. C, 2019, 7, 4725–4732.
53. C. Han, X. W. Han, J. Y. Han, M. Y. He, S. L. Peng, C. Y. Zhang, X. C. Liu, J. Gou and J. Wang, Adv. Funct. Mater., 2022, 32, 2113053.
54. S. Y. Feng, J. X. Li, L. Z. Feng, Z. T. Liu, J. C. Wang, C. Cui, O. X. Zhou, L. J. Deng, H. N. Xu, B. Leng, X. Q. Chen, X. Jiang, B. D. Liu and X. L. Zhang, Adv. Mater., 2023, 35, 202308090.

---

Footnote

† Chaoran Liu and Di Sun contributed equally to this work.

---