Ultrasensitive Flexible Memory Phototransistor with Detectivity of 1.8×1013 Jones for Artificial Visual Nociceptor

Emerging intelligent devices that can simulate an artificial intelligence vision system are of great interest for the development of modern information technology. Nociceptor is a crucial sensory neuron that recognizes harmful inputs and sends pain signals to the central nervous system to avoid injury; however, visual nociceptors, considered to be a key bionic function to protect eyesight based on optoelectronic devices, have yet to be developed. Herein this study, a three‐terminal flexible memory phototransistor (MPT) is first fabricated, which simulates the visual nociceptive behavior by adjusting light stimulation. The CsPbBr3 quantum‐dots (QDs)‐few‐layered black phosphorous nanosheets (FLBP NSs) heterojunction MPT demonstrates high responsivity of 7.2 × 103 AW−1 and high detectivity of 1.8 × 1013 Jones due to the high absorption coefficient of CsPbBr3 QDs materials and a high carrier transport property of FLBP NSs. Moreover, the proposed device can be used to emulate ultraviolet‐stimuli‐induced characteristics of visual nociceptors such as a threshold, no adaption, relaxation, allodynia, and hyperalgesia. It provides a new avenue for the realization of next‐generation neural‐integrated devices via its visual pain sense‐perception abilities.

(NSs). Based on that, we first implemented a three-terminal flexible memory phototransistor (MPT) that shows nociceptive behavior with modulation of optical stimuli. The MPT utilizes the characteristics of heterojunction to facilitate the separation and generation of electron-hole pairs, ensuring its charge storage function and improved photodetection ability. Specifically, the holes created in the photoactive CsPbBr 3 QDs-FLBP NSs layer are injected into semiconductor layer channels to increase the carrier concentration. For such a flexible MPT, a high responsivity of 7.2 Â 10 3 AW À1 and detectivity of 1.8 Â 10 13 Jones can be achieved. Moreover, this MPT can be used to emulate characteristics of photonic nociceptor including threshold, no adaption, and relaxation along with allodynia and hyperalgesia in response to light-induced stimuli. Therefore, light-induced visual nociceptive behavior in our MPT has a high response rate (0.05 s), which reduces the heat created during operation and allows for high integration.

Results and Discussion
CsPbBr 3 QDs and FLBP NSs were synthesized according to the previously reported method (see the experimental part in the supporting information for details). First, we prepared CsPbBr 3 QDs (thermal injection method) and FLBP NSs (probe ultrasonic method) separately, then mixed them in a certain proportion, and finally sonicated the mixture to realize the self-assembly of CsPbBr 3 QDs on the surface of FLBP NSs (Figure 1a). The challenge of solution preparation is to choose a suitable universal solvent for both CsPbBr 3 QDs and FLBP NSs to ensure selfassembly. We prepared FLBP in n-methyl-2-pyrrolidone (NMP) solvent and then performed multiple centrifugal exchanges between NMP and toluene to ensure that no NMP solvent remains to degrade CsPbBr 3 QDs. The prepared CsPbBr 3 QDs were added to the FLBP solution at a certain ratio, and the solution was allowed to stand ultrasonically at room temperature to form the self-assembly of CsPbBr 3 QDs on FLBP NSs. Figure 1b shows time-resolved photoluminescence (TRPL) spectra of CsPbBr 3 QDs and FLBP NSs heterostructure. The average lifetime of CsPbBr 3 QDs decreased from 38.5 to 25.7 ns after combining FLBP NSs into composite. As a result, the photoluminescence (PL) life span of CsPbBr 3 QDs-FLBP NSs decreases, implying that fluorescence quenching occurs. In addition, the ultraviolet (UV) absorption spectra of CsPbBr 3 QDs, FLBP NS, and CsPbBr 3 QDs-FLBP NSs solution are given in Figure 1c. The absorption spectrum of pristine FLBP NSs shows a wide absorption band ranging from the UV to near-infrared (NIR) region while the absorption spectrum of CsPbBr 3 QDs shows its absorption band edge at 485 nm. It is worth noting that the CsPbBr 3 QDs-FLBP NSs absorption spectrum retains the inherent behavior of CsPbBr 3 QDs with an additional tail in the range of 600-1100 nm, which is indicative of FLBP NSs presence. Furthermore, we also measured the UV absorption spectrum of pentacene, which has strong absorption bands at 585 and 670 nm ( Figure S1, Supporting Information). Figure 1d presents the steady PL spectra of CsPbBr 3 QDs-FLBP NSs in toluene solution. Under excitation at various wavelengths (330-400 nm), CsPbBr 3 QDs-FLBP NSs exhibit the strongest emission in the blue and green light region at λex ¼ 300 nm. X-ray diffraction (XRD) was employed to determine the crystal structure of CsPbBr 3 QDs, FLBP NSs, and CsPbBr 3 QDs-FLBP NSs, as illustrated in Figure 1e. The CsPbBr 3 QDs-FLBP NSs have the diffraction peaks of both FLBP NSs and CsPbBr 3 QDs, which indicate the dense distribution of QDs on FLBP NSs. Moreover, the X-ray photoelectron spectroscopy (XPS) spectrum characterized the elemental composition of the CsPbBr 3 QDs-FLBP NSs (Figure 1f ). For the CsPbBr 3 QDs-FLBP NSs sample, peaks corresponding to elements of P2p, O1s, C1s, Cs3d, Pb4f, and Br3d can be observed. Figure 1g shows the high-resolution transmission electron microscopy (HR-TEM) image of CsPbBr 3 QDs-FLBP NSs film where many cubic CsPbBr 3 QDs dispersed on the surface of FLBP NSs without obvious agglomeration can be observed. According to the statistics of size distribution obtained from the TEM image, CsPbBr 3 QDs exhibits an average size of %10 nm. The typical lattice of CsPbBr 3 QDs, with spacing about 0.292 nm, which corresponds to the (200) plane of the CsPbBr 3 QDs, confirms the high crystallinity of CsPbBr 3 QDs. In general, the TEM energy-dispersive X-ray (TEM-EDX) elemental mapping image shows the uniform distribution of Cs, Pb, Br, and P, further confirming the successful growth of CsPbBr 3 QDs on thin FLBP NSs.
The CsPbBr 3 QDs-FLBP NSs MPT device was constructed on atomic layer deposition (ALD)-deposited Al 2 O 3 (30 nm)/Si substrate. The typical three-dimensional structure and crosssectional scanning electron microscope (SEM) of CsPbBr 3 QDs-FLBP NSs MPT are illustrated in Figure 2a,b, respectively. In our case, the CsPbBr 3 QDs-FLBP NSs composite film was utilized as both the floating gate and light-sensing layer for memory phototransistor, which interfaces with the pentacene semiconductor layer (30 nm), and 30 nm Au electrodes were served as source and drain electrodes. The SEM and atomic force microscope (AFM) image of self-assembled heterojunction fabricated with different concentrations of CsPbBr 3 QDs (0.5, 1.0, 1.5, and 3 mg mL À1 ) is presented in Figure S3 and S4, Supporting Information, from which it can be seen that the distribution of CsPbBr 3 QDs on FLBP NSs becomes denser with the increase of concentration. The transfer characteristics of the devices based on the heterostructure with different concentrations of CsPbBr 3 QDs are described in Figure S6, Supporting Information. As the CsPbBr 3 QDs concentration increases, the memory window changes from 0.4 to 0.5, 0.6, and 1.2 V. When the concentration of CsPbBr 3 QDs is 1.5 mg mL À1 , the largest memory window can be obtained, so we choose it as the optimal concentration. We further compared the performance of the CsPbBr 3 QDs-based and CsPbBr 3 QDs-FLBP NSs-based MPT, respectively. Figure 2c,d illustrates the optical programming after the irradiation of a 365 nm laser with an intensity of 0.069 mW cm À2 and electronic erasing operation under the À5 V pulse of the two devices. As shown in Figure 2d, the memory window of CsPbBr 3 QDs-FLBP NSs MPT increases significantly compared with CsPbBr 3 QDs MPT. In this structure, the light induces electron transfer from the photoexcited CsPbBr 3 QDs to FLBP NSs. Then the injected electrons will be stabilized in FLBP NSs owing to their large surface area and high carrier mobility, resulting in the accumulation of electrons on FLBP NSs and the formation of the internal electric field. The larger threshold voltage (V th ) shift of CsPbBr 3 QDs-FLBP NSs MPT suggests that the injection of electrons in the FLBP NSs can be adjusted by UV light, and when the voltage of À5 V is used to erase, the transfer curve can be reset to its initial state, indicating that the electrons generated by the light can escape from the FLBP NSs under electrical field. The relationship between the light intensity, current level (I ON /I OFF ), and V th is shown in Figure 2e, showing the statistical switching uniformity and stability. In the 1000-cycle test, the spatial uniformity of the photon-mediated resistance switch was further verified (Figure 2h). The optically programmed state of the CsPbBr 3 QDs-FLBP NSs MPT shows spatial uniformity with device-to-device variability of 0.458. The electrically erased state shows spatial uniformity with inter-device variability of 0.524. Additionally, Figure S7, Supporting Information depicts the output curve of devices under dark and UV illumination. Under dark conditions, the characteristics of p-type transistors can be verified by linearity and saturation regions of current. The current of MPT increases significantly under light irridation (0.163 mW cm À2 ), indicating the increased hole density in the semiconductor channel. Under low V DS , the linear output characteristics indicate that an ohmic contact is achieved at the pentacene-Au electrode interface ( Figure S7a, Supporting Information).
Then, we investigated the transient response of the CsPbBr 3 QDs-FLBP NSs MPT under UV light illumination (0.0126 mW cm À2 ) for 50 ms (shown in Figure 2f ). The photocurrent of CsPbBr 3 QDs MPT rises from 30 up to 82 nA while www.advancedsciencenews.com www.advintellsyst.com the photocurrent of CsPbBr 3 QDs-FLBP NSs MPT reaches 108 nA within 50 ms. The high speed of light-induced charge trapping is originated from the photovoltaic and photogating effect of heterojunction, ensuring the ultra-fast reading and writing by optically adjusting the interface energy level of the heterojunction. In addition, the readout current of CsPbBr 3 QDs-FLBP NSs MPT and CsPbBr 3 QDs MPT following optical pulses treatment (0.069 mW cm À2 ) is compared in Figure 2g. The optically programmed current of CsPbBr 3 QDs-FLBP NSs MPT could be kept in a highly conductive state, while the current of CsPbBr 3 QDs MPT decreased sharply, indicating that the heterojunction could retain the trapped charge carrier within the Schottky barrier.
In artificial visual systems, the transition from short-term memory (STM) to long-term memory (LTM) based on continuous light perception is a necessary condition for the realization of advanced imaging functions. First, by changing the intensity and width of the light stimulus, the biological function of this CsPbBr 3 QDs-FLBP NSs MPT transition from STM to LTM was successfully explained. The simulation of the enhancement of short-term and long-term forms by optical stimulation with intensity ranging from 0.041, 0.069, 0.126 to 0.163 mW cm À2 as illustrated in Figure 3a. When a low-intensity (0.041 mW cm À2 , duration: 5 s) pulse is applied, the pentacene channel cannot maintain the triggered high-conductivity state and gradually decays back to its initial state, which is similar to the STM mechanism of memories. In comparison, when the input pulse intensity increases to 0.069, 0.126, and 0.163 mW cm À2 , the optical CsPbBr 3 QDs-FLBP NSs MPT also performs a permanent transition to a high-conductivity state, simulating LTM behavior. This is due to a deeper photo-generated electron capture event resulting in a long-term increase in channel conductance. Similarly, Figure 3b describes STM and LTM characteristics are realized by varying the light stimulus www.advancedsciencenews.com www.advintellsyst.com length. The frequency of stimulation with memorizing devices can be proposed to further understand the optical transition behavior. As illustrated in Figure 3c, ten pulses with the frequency varying from 1 to 20 Hz (or the pulse interval is changed from 1 to 0.05 s) were applied to CsPbBr 3 QDs-FLBP NSs MPT. When the frequency is set to 1 Hz, the current does not increase substantially, but when the frequency is raised to 20 Hz, the current increases by more than 30 times, the earlier results thus strongly indicate that the optical sensing and memory abilities can be greatly enhanced in our CsPbBr 3 QDs-FLBP NSs MPT device as the light frequency increases. The photoactive response is related to the charge-trapping behavior in the heterojunction under optical modulation, which is verified by monitoring the electrostatic force, surface potential, and carrier-trapping behavior via the Kelvin probe force microscopy (KPFM) (Figure 3d). The average surface potential gradually changes from 215.5, 465.5, 568.8, 725.3 to 895.6 mV as the light intensity increases from 0, 0.041, 0.069, 0.126 to 0.163 mW cm À2 , which is identical to the light-induced electron accumulation in FLBP NSs (Figure 3e). The light-induced electron capture capability of the CsPbBr 3 QDs-FLBP NSs heterojunction film was also investigated using an AFM in conjunction with a microscale electron injection process and subsequent surface potential calculations. As shown in Figure 3f, the conduction band (valence band) of the FLBP and CsPbBr 3 QDs is À4.58 eV (À4.67 eV) and À3.3 eV (À5.7 eV), respectively; hence, the hybrid junction is expected to be type I with a band offset of 1.28 eV. The large band offset induces the separation of the photogenerated electron-hole pairs in CsPbBr 3 QDs and the transportation of the electrons to FLBP NSs. Additionally, the photo-induced electron transfer from CsPbBr 3 QDs to FLBP NSs was further confirmed by the KPFM measurements in Figure 3g. Surprisingly, when the applied negative bias rises, the surface potential gradually decreases under dark conditions. In contrast, light irradiation slows down the decay speed of the CsPbBr 3 QDs-FLBP NSs heterojunction surface. Light treatment increases the overall surface potential of the injected and non-injected areas, implying that light improves the FLBP NSs electron-trapping capability.
To evaluate the detection performance of CsPbBr 3 QDs-FLBP NSs MPT, the transfer characteristics and photoresponse under different light irradiance were measured. Figure 4a depicts the typical transfer characteristics of the MPT under different light intensities where V GS was swept from 0 to À5 V and V DS was fixed at À1 V. The I DS and V th increase significantly as light intensity increases. When the light intensity increases from 0, 0.041, 0.069, 0.126 to 0.163 mW cm À2 , the memory window increases from 1.4,1.6, 1.9, 2.2 to 2.4 V, indicating that the photoactive layer can generate free charge carriers effectively. The photocurrent as a function of the incident light intensity and gate voltage are plotted in Figure 4b. Figure 4c shows the relationship between the photosensitivity (I photo /I dark , P) and the light intensity of the CsPbBr 3 QDs-FLBP NSs MPT to varied V G . It is worth noting that since the carriers of the CsPbBr 3 QDs-FLBP NSs heterojunction are modulated by the incident light intensity and the gate voltage, the I photo /I dark is higher in the OFF state than that in ON state, and the maximum photosensitivity to each light intensity is observed when the gate voltage is À1 V. The photocurrent increases as the incident light intensity increases to reach a maximum of 10 3 under 0.163 mW cm À2 , indicating excellent photosensitivity of our MPT. Phototransistor also has two major variables: photoresponsivity (R) and photodetectivity (D*). R is defined as R ¼ I ph /(P inc S), indicating how efficiently the phototransistor responds to light, where I ph is the photocurrent, P inc is the incident light intensity, and S is the device's channel area. D* denotes the capability of the phototransistor to detect weak optical signals and can be calculated using the following expression: D* ¼ (A 1/2 R)/(2qI d ) 1/2 , where q is the elementary charge and I d is the dark current. The representative R and D* values as functions of irradiation and V G are presented in Figure 4d,e. When V G is À5 V and the optical power is 0.041 mW cm À2 , R reaches the maximum value of 7.25 Â 10 3 AW À1 , which is much larger than the recently reported heterostructure-based photodetectors as shown in Table 1. D* has a similar tendency to I photo /I dark , reaching a peak when the voltage applied to the gate is À1 V. We can realize that the D* peak of our MPT under irradiation is above 10 13 Jones, and the maximum D* reaches 4.2 Â 10 13 Jones at V GS is À1 V. Furthermore, changing the threshold voltage provides a practical approach for characterizing the photodetector's photoresponse. Voltage responsivity (Rv) can be also utilized to evaluate the photoresponse of a phototransistor and is defined as follows: Rv ¼ ΔV th /PA, where ΔV th is the shift threshold voltage after the light irradiation. Figure 4f illustrated the threshold voltage variation as a function of light intensity. The absolute value of the threshold voltage shift increases with increasing light intensity, but it tends to saturate under high intensity. As shown in Figure 4g-i, the uniformity of the R, D*, and Rv characteristics were examined by measuring 30 devices. CsPbBr 3 QDs-FLBP NSs MPT exhibits both higher photocurrent and photosensitivity compared with conventional optoelectronic and hybrid optoelectronic devices. In addition to the optical power perception behavior illustrated earlier, the wavelength-dependent effect of the CsPbBr 3 QDs-FLBP NSs MPT improves the vision system's intrinsic light sensitivity and color selectivity, which improves the device's perception accuracy. The stability of our devices was investigated under different wavelength and low-temperature conditions over time as shown in Figure S13, Supporting Information. When the wavelength is decreased from 660 to 365 nm, the device's switching current ratio rises from 2.8 Â 10 3 to 5.6 Â 10 3 . Apart from its high-performance photodetection, our MPT also exhibits good environmental stability. We tested the transfer characteristic curves (light intensity: 0.069 mW cm À2 , duration: 1 s) and electrical erasing operations of the CsPbBr 3 QDs-FLBP NSs MPT device in the atmospheric environment for at least one week. The optoelectronic devices show that the original memory window (1.2 V) will be preserved during this period ( Figure S14, Supporting Information), indicating good stability.
The CsPbBr 3 QDs-FLBP NSs MPT shows great promise in a variety of critical applications such as flexible wearable electronics. As demonstrated in Figure 5a, the flexible MPT is fabricated on a polyethylene terephthalate (PET) substrate, and the manufacturing process is detailed in Figure S16, Supporting Information. Figure 5b shows the basic electrical characteristics of the flexible MPT under various bending conditions using UV programming. Even after 500 cycles, the memory window and ON/OFF current ratio remain intact, demonstrating the device's exceptional versatility. The strain is extracted by www.advancedsciencenews.com www.advintellsyst.com where d is the thickness of the PET film, and r ¼ 10 mm. When the strain is less than 1% (r > 5 mm), the device's photocurrent degrades by less than 5%. Following that, we explored the effect of under-strain states on the photodetection capability of the CsPbBr 3 QDs-FLBP NSs MPT as illustrated in Figure 5c-e, and Figure S17, Supporting Information. There is almost no change in its photosensitivity, while the photoresponsivity and photodetectivity have a slightly decreasing trend under the compression/extension strain state. As can be seen, the depression is less than 8% under the condition of 1% strain, demonstrating the film's exceptional photodetection durability. Through the optical memory and photodetection characteristics of the CsPbBr 3 QDs-FLBP NSs MPT, we have confirmed that the e-h pair generated by the photocurrent is filling the trapped state, thereby producing relatively smooth conduction. Simultaneously, it is observed that the current of the device is close to saturation, indicating that it can be utilized to design light-triggered visual nociceptors. When neurons at free nerve terminals are stimulated, they generate an electrical signal and transmit it to the nociceptors. The visual nociceptors compare the signal's amplitude to its threshold to determine whether or not to generate an action potential and transmit it to the brain via the central nervous system (Figure 6a). It not only exhibits the threshold switching and relaxation dynamics observed in normal neurons in response to noxious stimuli, but also exhibits its characteristic properties such as "no adaptation," "allodynia," and "hyperalgesia" in response to repeated stimuli and excessively intense stimuli, respectively. [9][10][11] The trigger of biological nociceptors is dependent on duration, intensity, and the number of stimuli. Therefore, we replicate external stimulation of our nociceptors by using light pulses with varying amplitudes, pulse widths, and pulse counts. We set the threshold of the CsPbBr 3 QDs-FLBP NSs MPT to 60 nA. The MPT cannot be switched on until the light intensity reached 0.069 mW cm À2 under single light pulse with width of 0.5 s. When the light intensity increases further to 0.163 mW cm À2 , the output current increases significantly (Figure 6c). Similarly, under a fixed light intensity (0.069 mW cm À2 ), we observed that sufficiently long pulse width (0.5 s) is necessary to reach the threshold value for our MPT (Figure 6d), and a longer pulse width results in a larger output current. To further validate the device's threshold properties, the light response was tested under self-biased conditions by varying the pulse intensity from 0 to 0.069 mW cm À2 , as shown in Figure 6f. Interestingly, the device did not exhibit a significant photoresponse at intensities less than 0.006 mW cm À2 , but the photocurrent progressively increases and reaches saturation when exposed to continuous light pulses (intensities larger than 0.02 mW cm À2 ). After attaining saturation, even if additional pulses are applied, the photocurrent remains constant, indicating that the device's reaction is constant when the same stimulus is applied repeatedly. It is worth noting that once a balance is struck between the generation and capture of light-generated e-h pairs, the photocurrent flowing through the device remains constant, even after repeated illumination www.advancedsciencenews.com www.advintellsyst.com pulses. This phenomenon is similar to the nonadaptive characteristics of pain receptors, which is essential for the human body or humanoid robots to protect themselves from repeated harmful stimuli. Apart from the threshold and nonadaption behavior, relaxation is another important nociceptive characteristic. As shown in Figure 6e, pulse measurements were taken at various intervals (0.5, 1, and 2 s). We activated the device with the first light pulse (0.069 mW cm À2 , 0.5 s) and then applied the second light pulse (0.069 mW cm À2 , 0.5 s) at a different interval. However, when the identical stimulus is administered prior to the complete relaxation of the device, a considerable current response is recorded Figure 6. Visual nociceptive behavior of the CsPbBr 3 QDs-FLBP NSs MPT. a) Working mechanism of the human eye. b) Schematic presentation of the allodynia and hyperalgesia features with increasing stimuli intensity in normal (no injured) and injured conditions. c) By applying optical pulses with 0.5 s width and different pulse intensity from 0.025 to 0.163 mW cm À2 , the device cannot reach threshold current (I th ¼ 60 μA) until the pulse intensity reached 0.069 mW cm À2 . d) A fixed optical stimulus (0.069 mW cm À2 ) with different pulse widths (0.15-1 s) applied to our device leads to a gradual step-wise increase of the excitatory postsynaptic current (EPSC) outputs accordingly. e) The resulting EPSCs using 0.069 mW cm À2 light pulse train followed by 0.069 mW cm À2 pulse train with different time intervals between these two continuous stimulus trains ranging from 0.5 to 2 s. f ) The response of device to continuous multiple pulses with different light intensities (0, 0.006, 0.020, 0.035, 0.041, and 0.069 mW cm À2 , respectively). The maximum output currents at different input voltage amplitudes g) in linear scale and h) in log scale, demonstrating the shift of the ON current toward higher currents (Hyperalgesia) and ON-switching light intensity toward a lower threshold (Allodynia).
www.advancedsciencenews.com www.advintellsyst.com (black and red curves in Figure 6e). Additionally, the other primary properties of nociceptors, allodynia, and hyperalgesia were further examined. When a powerful enough stimulus is provided to the nociceptor, the nociceptor will be in an injured state. After the injury, the nociceptor will exhibit an amplified reaction when the threshold is lowered, referred to as allodynia and hyperalgesia (Figure 6b). In this work, an injured state is induced using high ultraviolet (HUV) and low ultraviolet (LUV) pulses. Prior to HUV illumination, "no injury" nociceptors originally responded poorly to low photocurrent values. Following exposure to HUV, the nociceptors are "injured," the injured nociceptor becomes more sensitive, and the photocurrent as a function of the light intensity is expressed in linear ( Figure 6g) and logarithm scale (Figure 6h). This demonstrates that a lower threshold light intensity is required to activate a device capable of simulating the " allodynia " and "hyperalgesia" properties of nociceptors. As a result of the preceding study, it is clear that our MPT device can emulate the visual nociceptor properties of "threshold", "relaxation", "non-adaptation", "allodynia ", and "hyperalgesia."

Conclusion
We have proposed an artificial visual nociceptor based on a CsPbBr 3 QDs-FLBP NSs memory phototransistor, which realizes all the key functions of the nociceptor in a single device, including threshold, relaxation, no adaption, allodynia, and hyperalgesia. Not only that, the MPT device also has excellent optical memory and optical detection capabilities, which use the mechanism of adding light to promote the separation and generation of heterojunction electron-hole pairs. The new artificial nociceptor can easily be extended from light stimulation to processing other stimuli, such as chemical, mechanical, etc., showing the potential application of integrating optoelectronic circuits and artificial intelligence receptors on the chip.
Preparation of Cs-Oleate: The CsPbBr 3 QDs were effectively synthesized according to the procedure reported by Protesescu et al. [33] The ruthenium oleate precursor was synthesized by loading 0.81 g of Cs 2 CO 3 , 40 mL of ODE, and 2.5 mL of OA to 100 mL of 3-necked flask, and then heating to 120 C for 1 h under a nitrogen atmosphere to ensure complete reaction of Cs 2 CO 3 with OA. The prepared solution was maintained at 160 C to avoid further precipitation. Finally, atransparent solution was observed and it was preheated at 100 C.
Synthesis of CsPbBr 3 QDs: Five milliliters of ODE mixed with 0.188 mmol of PbBr 2 was placed in another 3-necked flask, and the temperature was maintained at 120 C for 60 min under vacuum, and 0.5 mL of oleylamine (technical grade, 70%) and 0.5 mL of OA were injected, and the temperature was increased to 140 C under a flow rate of N 2 . Immediately, 0.4 mL of the yttrium oleate precursor was injected into the flask. The solution turned to fluorescent green, and the flask was immersed in an ice/water bath for quenching 5 s after hot injection.
Purification of CsPbBr 3 QDs: The purification of CsPbBr 3 QDs was performed by following the previous report by Swarnkar et al. [34] QDs of assynthesized CsPbBr 3 precipitated in 15 mL of MeOAc and then centrifuged at 9000 rpm for 5 min. In addition, the CsPbBr 3 QDs of each centrifuge tube were wet pellets in 1 mL of redispersed hexane with the same volume of MeOAc and reprecipitated at 9000 rpm for 5 min. QDs were redispersed in 2 mL of anhydrous toluene or hexane.
Purification of FLBP: Using BP crystals as raw materials, low-layer BP was prepared by the probe method in NMP solvent. [35] Preparation of Self-Assembled CsPbBr 3 QDs-FLBP NSs: Through the solvent exchange of toluene, the self-assembly of CsPbBr 3 QDs on FLBP was realized. First, CsPbBr 3 QDs solution of different concentrations (0.5, 1, 1.5, 3 mg mL À1 ) was added, then FLBP with a concentration of 0.5 mg mL À1 was added, and finally it was bathed in ultrasound at room temperature.
Fabrication of Flexible CsPbBr 3 QDs-FLBP NSs Memory Phototransistor: The CsPbBr 3 QDs-FLBP MPT was manufactured with a bottom-gate top-contact structure on a PET/indium tin oxides (ITO)/Al 2 O 3 substrate. Strip ITO was evaporated onto the PET layer as a gate by thermal evaporation. Then 30 nm dielectric layer Al 2 O 3 is deposited on PET/ITO by ALD under 80 C. After that, the CsPbBr 3 QDs-FLBP was spin-coated onto the Al 2 O 3 surface at a speed of 3000 rpm for 40 s, followed by annealing with 100 C for 15 min in a glove box. Next, under 4 Â 10 À4 Pa, p-type 30 nmpentacene semiconductor layer was thermally evaporated (rate: 0.1 Å s À1 ) on the top of the CsPbBr 3 QDs-FLBP charge-trapping layer. Finally, 30 nm Au source and drain electrodes were thermally evaporated on the pentacene film via a shadow mask. And the channel length for the as-prepared memory device was 50 μm.
Characterizations and Measurements: PL spectra were conducted on an Edinburgh Instruments (FLS 920). And the UV-visible spectra were recorded on an UV-visible spectrophotometer (Agilent Cary 60). The prepared CsPbBr 3 QDs-FLBP were characterized by HR-TEM (Tecnai F30). All nanoelectronic measurements were made at room temperature using the Bruker size icon AFM. SCM-PIT (which is a type of electrical force modulation AFM probe) conductive sheet (Bruker; platinum rhodium coating; frequency: 60-100 kHz). Scan, scan rate 10 Â 10 μm 0.8 Hz, respectively. What's more, the top surface distance of all samples was measured at a surface distance of 75 nm (resolution: 256 Â 256 pixels), and the charge-trapping ability of pentacene thin films was studied by AFM technique in combination with contact mode and KPFM mode. There are two steps for each measurement: two-dimensional charge injection and surface potential measurement. First, in the contact mode, the scan rate of 0.8 Hz, the 6 V bias was a technique applied to the injection of conductive electrons (scan area: 1 Â 1 μm; resolution: 256 Â 256 pixels). Injection of the well was achieved by biasing the tip of a 5 V at different scanning zones. Then, the AFM system was converted to KPFM amplitude modulation mode for on-site surface potential measurement. Scan, scan rate 10 Â 10 μm 0.8 Hz, respectively. The top surface distance of all samples was measured at a surface distance of 75 nm (resolution: 256 Â 256 pixels). Finally, the electrical characteristics of all devices were measured using the Keysight B2902A Precision Source/ Measurement Unit and the Keysight 4200-SCS Parameter Analyzer. All device tests were performed at room temperature.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.