Self-driven event-driven vision sensor based on MISM structure and preparation method thereof

Abstract

This invention belongs to the field of optoelectronic sensor technology, specifically a self-driven event-driven visual sensor based on a MISM structure and its fabrication method. The visual sensor of this invention is composed of a bottom substrate, a lower electrode layer, an organic semiconductor active layer, a ferroelectric polymer insulating layer, and a top electrode layer, stacked vertically from bottom to top. It utilizes the photoelectric and pyroelectric synergistic physical mechanism of the device's bottom layer to directly achieve a highly efficient bipolar pulse current response to dynamic light intensity changes in a zero-bias state, thereby completing hardware-level extraction of dynamic features of moving objects at the sensing source. This invention simplifies the hardware circuit structure, significantly improves the pixel fill factor and spatial resolution, achieves zero-bias, fully self-driven ultra-low power sensing, and is suitable for the hardware deployment of ultra-low power intelligent detection devices. It simplifies the dynamic signal extraction process and provides a high-quality, high-signal-to-noise-ratio sparse hardware event stream for subsequent signal reading and processing circuits.

Description

Technical Field

[0001] This invention belongs to the field of optoelectronic sensor technology, specifically relating to an event-driven visual sensor and its fabrication method, which can be used in neuromorphic visual perception systems. Background Technology

[0002] As intelligent vision systems continue to advance towards low power consumption, high real-time performance, and edge deployment, machine vision, as a core technology of edge artificial intelligence, plays a crucial role in fields such as autonomous driving, drone navigation, and industrial inspection. However, most existing industrial and consumer-grade machine vision systems follow the traditional "perception first, computation later" model. The frame-based vision sensors they employ (such as CCD or CMOS image sensors) are limited by a fixed sampling frequency when facing highly dynamic scenes. Regardless of whether the scene changes, the entire pixel array is sampled and integrated at a fixed frame rate. This discrete time-slicing mechanism inevitably generates massive amounts of repetitive and static background redundant data, which not only occupies a huge amount of communication bandwidth and exacerbates the computational burden and energy consumption pressure of the back-end processing unit, but is also severely constrained by the overall system bandwidth limitations, greatly limiting the real-time response capability of edge terminal sensing nodes.

[0003] Event-driven visual sensors, as a novel type of optoelectronic sensor, have demonstrated immense potential in neuromorphic perception and low-power computing since their introduction to simulate the asynchronous sensing characteristics of the biological retina. They not only eliminate the constraints of fixed frame rates inherent in traditional frame-based sampling, but their internal pixel units can also independently and autonomously monitor changes in light intensity. Only when a change in light intensity exceeds a preset threshold will they asynchronously and instantly output an asynchronous event stream containing spatial coordinates, microsecond-level timestamps, and information reflecting the polarity of the light intensity change direction. Through this unstructured and sparse data acquisition method, event-driven sensors directly filter out a large amount of static background redundancy at the sensing source, significantly reducing data transmission requirements, thereby lowering power consumption and increasing processing speed. These characteristics of low power consumption, low response latency, and high dynamic range are particularly suitable for spiking neural networks and other intelligent computing scenarios. The high efficiency of event-driven sensors can effectively support the large demands for parallel sensing and data dimensionality reduction, simulating the behavior of retinal bipolar cells. This provides the hardware foundation for developing "sensing-computing integrated" neuromorphic computing, making it a promising candidate for significant roles in artificial vision, intelligent edge perception, and other fields.

[0004] In practical applications and large-scale integration, existing event-driven sensor technologies still suffer from significant device structural complexity. To implement complex retinal logic such as differential amplification and dual-threshold comparison at the hardware level, traditional silicon-based event-driven sensors typically integrate dozens of transistors and several large capacitors within a single pixel. While this approach achieves event extraction, it also introduces complexity in circuit design, high manufacturing costs, and a sharp decrease in spatial resolution and fill factor. To simplify pixel circuitry, although novel vertical structures can be used to directly respond to transient light intensity changes at the device's physical layer, the lack of an efficient matching physical event encoding and network mapping strategy results in high information entropy when mapping physical signals to the backend spiking neural network architecture, hindering the full realization of the low-power potential of sparse sensing at the physical layer. Therefore, achieving efficient and robust bipolar pulse response to transient light intensity changes using novel semiconductor structures without complex silicon-based transistor circuitry has become crucial for improving the practicality of event-driven sensors in edge intelligent sensing applications. Summary of the Invention

[0005] The purpose of this invention is to propose a self-driven event-driven vision sensor based on a MISM structure and its fabrication method, so as to solve the technical problems of excessive number of transistors and capacitors in pixels, complex circuit structure, and low spatial resolution and fill factor in existing silicon-based event-driven vision sensors.

[0006] The self-driven event-driven vision sensor based on the MISM structure provided by this invention eliminates the need for complex transistor pixel circuits. It utilizes only the underlying photoelectric and pyroelectric synergistic physical mechanism to directly achieve a highly efficient bipolar pulse current response to dynamic light intensity changes in a zero-bias state, thereby completing hardware-level extraction of dynamic features of moving objects at the sensing source. Here, MISM is an abbreviation for Metal-Insulator-Semiconductor-Metal.

[0007] The self-driven event-driven vision sensor based on the MISM structure provided by this invention employs a layered heterogeneous stacked structure in its hardware layer. The specific component names, spatial relationships, and connections are described below: The sensor is composed of a bottom substrate, a lower electrode layer, an organic semiconductor active layer, a ferroelectric polymer insulating layer, and a top electrode layer, stacked vertically from bottom to top. Specifically, the lower electrode layer is deposited on the surface of the bottom substrate; the organic semiconductor active layer is deposited on top of the lower electrode layer, forming a built-in electric field due to the difference between the work function and the Fermi level; the polymer insulating layer covers the surface of the organic semiconductor active layer, with the surfaces of the two layers in close contact to form a physical heterogeneous interface; and the top electrode layer is deposited on the surface of the ferroelectric polymer insulating layer. The lower electrode layer, organic semiconductor active layer, ferroelectric polymer insulating layer, and top electrode layer together in the vertical direction constitute a two-terminal self-driven physical device. The two electrodes are directly connected to external signal output or subsequent signal acquisition circuits, operating in a zero-bias voltage state.

[0008] Preferably, the bottom substrate is a rigid insulating substrate (such as quartz glass); the transparent conductive thin film layer serving as the lower electrode layer is an indium tin oxide (ITO) thin film with a thickness of 90-110 nm; the material serving as the organic semiconductor active layer is a copper phthalocyanine (CuPc) thin film with a thickness of 40-60 nm; the material serving as the polymer insulating layer is a polyvinylidene fluoride (PVDF) thin film with a thickness of 40-150 nm; and the metal thin film layer serving as the top electrode layer is an aluminum (Al) electrode with a thickness of 80-100 nm. All layers are in close contact through physical heterogeneous interfaces, forming an externally stacked, self-driven sensing device consisting of ITO / CuPc / PVDF / Al layers at both ends.

[0009] The specific working (action) process and signal transmission relationship of the visual sensor under dynamic lighting stimulation are as follows:

[0010] When an external moving object causes a transient change in the intensity of the incident pulsed light, the sensor operates in a zero-bias state.

[0011] At the instant the incident light is turned on (ON moment), the organic semiconductor active layer absorbs photons and generates excitons. The excitons generate an internal electric field (E) within the device. biUnder the influence of the active layer, the charge carriers rapidly dissociate into free carriers. Due to the DC blocking characteristics of the polymer insulating layer, these free carriers cannot cross the insulating layer. Therefore, driven by the built-in electric field, they migrate directionally and accumulate at the physical interface between the active layer and the insulating layer, causing a sudden change in the electric displacement vector on both sides of the insulating layer. This results in a sharp positive transient current pulse being self-driven between the two electrodes of the external circuit. As the interface charge density continues to increase, the induced polarization electric field gradually strengthens and eventually cancels the built-in electric field inside the device, causing the current in the external circuit to gradually decay back to zero under continuous illumination.

[0012] At the transient moment when the incident light is turned off (the instant of OFF), due to the loss of the source of photogenerated charge carrier replenishment, the charge accumulated at the physical interface is released and flows back in the reverse direction through the external loop, thereby exciting a transient reverse current spike of opposite polarity between the two electrodes. In this way, the sensor, without external power supply or any external transistor pixel circuitry, utilizes the self-driven mechanism of interface charge dynamics of a single physical device to directly convert the dynamically changing transient optical signal into a bipolar transient electrical signal flow with clearly defined positive and negative polarities at the hardware physical level.

[0013] The specific steps for fabricating a self-driven, event-driven vision sensor provided by this invention are as follows:

[0014] (1) First, the surface of the quartz glass substrate with the ITO lower electrode layer is cleaned. It is then immersed in acetone, isopropanol and deionized water for 15-30 minutes respectively for weak ultrasonic cleaning. Then, the excess liquid on the surface is blown off with a high-purity nitrogen gun and dried on a hot plate for later use.

[0015] (2) An organic thermal evaporation coating process is used to deposit an organic semiconductor active layer, copper phthalocyanine (CuPc), thin film on the cleaned ITO electrode layer surface. The evaporation source temperature is controlled at 200℃~380℃, and the vacuum degree is better than 5×10⁻⁶. -4 Pa, the deposition rate was controlled at 0.1-0.2 Å / s, and the final film thickness was controlled at 40-60 nm.

[0016] (3) Prepare a PVDF dimethylformamide (DMF) solution with a concentration of 20-40 mg / mL and stir it at 50-60℃ for 1-2 hours to completely dissolve it; use spin coating to spin coat a polymer insulating layer on the surface of CuPc active layer; set the spin coating speed to 2000-4000 rpm and the spin coating time to 50-60 s; after film formation, place the device on a hot plate at 100-120℃ and dry it for 0.5-1 hour to form an insulating layer with a thickness of 40-150 nm;

[0017] (4) Finally, a top electrode layer of metallic aluminum (Al) is deposited on the surface of the PVDF insulating layer 4 using a thermal evaporation process, with a evaporation vacuum degree better than 5×10⁻⁶. -4 Pa, with a vapor deposition thickness of 80-100 nm.

[0018] The layers are in close contact, and the whole structure ultimately forms a vertically stacked self-driven sensing device of ITO / CuPc / PVDF / Al.

[0019] Compared with the prior art, the specific technical effects achieved by the present invention are as follows:

[0020] (1) The hardware circuit structure is greatly simplified, and the pixel fill factor and spatial resolution are significantly improved: This invention uses a dual-end self-driven MISM heterogeneous layered stacking structure to directly simplify the complex analog pixel circuit, which originally required the integration of dozens of transistors and several large-volume capacitors in the pixel of a traditional silicon-based event-driven sensor, into a single physical device with highly integrated functions. The integration of photoelectric detection, differential processing and bipolar event generation is directly realized at the underlying hardware sensing source. This not only eliminates the noise instability caused by the process deviation of analog circuits, but also greatly reduces the physical size of a single pixel, which is conducive to realizing the integration of device arrays with ultra-high spatial resolution.

[0021] (2) Achieved ultra-low power sensing with zero bias and full self-driving: This invention cleverly utilizes the photoelectric / pyroelectric synergistic mechanism at the interface between the CuPc active layer and the PVDF insulating layer, enabling the device to directly output bipolar electrical signals based on the internal charge movement caused by transient changes in light intensity without the need for an external bias voltage. This completely eliminates static leakage current and hardware maintenance power consumption from a physical mechanism perspective, achieving true passive event extraction, making it extremely suitable for the hardware deployment of edge IoT nodes and ultra-low power intelligent detection devices.

[0022] (3) The extraction process of dynamic signals is simplified, and the data redundancy and interface bandwidth of the system are reduced: This invention utilizes the physical effects of the device to directly output bipolar transient pulses with clear positive and negative polarities, so that the motion characteristics of light intensity increase and decrease in dynamic scenes are initially classified at the physical layer. This avoids the redundant process of traditional image sensors outputting full static image data and then performing complex differential calculations at the back end, greatly reducing the transmission bandwidth of the data interface, and providing high-quality, high signal-to-noise ratio sparse hardware event streams for subsequent signal reading and processing circuits. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the layered stacked structure of a self-driven event-driven vision sensor based on the MISM structure.

[0024] Figure 2This is a response current diagram of the self-driven event-driven vision sensor of the present invention under different power pulse light illumination in a zero bias voltage state.

[0025] Figure 3 This is a current response diagram of the self-driven event-driven vision sensor of the present invention under continuously changing light pulse stimulation.

[0026] Figure 4 This is a linear fitting graph of the response current and optical power change value of the self-driven event-driven vision sensor of the present invention under continuously changing light pulse stimulation.

[0027] The labels in the figure are: 1-bottom substrate, 2-lower electrode layer, 3-organic semiconductor active layer, 4-ferroelectric polymer insulating layer, 5-top electrode layer. Detailed Implementation

[0028] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the drawings are in a very simplified form and use non-precise proportions, intended only to facilitate and clearly illustrate the embodiments of the present invention. The structures, proportions, sizes, etc., depicted in the accompanying drawings are only for illustrative purposes to aid those skilled in the art and to enable them to understand and read the content disclosed herein. They are not intended to limit the implementation conditions of the present invention. Any modifications to the structure, changes in proportions, or adjustments to the size, provided that the effects and objectives achieved by the present invention are the same or similar, should still fall within the scope of the technical content disclosed in the present invention.

[0029] Example:

[0030] The physical device structures at both ends of the self-driven event-driven vision sensor are shown in the attached figure. Figure 1 As shown. The bottom substrate 1 is a quartz glass substrate, the bottom electrode layer 2 is a transparent conductive indium tin oxide (ITO) thin film layer, the organic semiconductor active layer 3 is a copper phthalocyanine (CuPc) thin film layer, the ferroelectric polymer insulating layer 4 is a polyvinylidene fluoride (PVDF) thin film layer, and the top electrode layer 5 is a metallic aluminum (Al) thin film electrode layer. The layers are in close contact to form a vertically oriented layered heterogeneous stacked MISM with self-driven structures at both ends.

[0031] The specific steps for fabricating this self-driven event-driven vision sensor are as follows:

[0032] (1) First, the quartz glass substrate 1 with ITO lower electrode layer 2 is cleaned by immersing it in acetone, isopropanol and deionized water for 30 minutes each. Then, excess liquid on the surface is blown away with a high-purity nitrogen gun and dried on a hot plate for later use.

[0033] (2) An organic thermal evaporation coating process is used to deposit an organic semiconductor active layer 3, copper phthalocyanine (CuPc) film, on the cleaned ITO lower electrode layer 2. The evaporation source temperature is controlled at 200℃~380℃, and the vacuum degree is better than 5×10 -4 Pa, the deposition rate was controlled at 0.2 Å / s, and the final film thickness was controlled at 60 nm.

[0034] (3) Prepare a PVDF dimethylformamide (DMF) solution with a concentration of 40 mg / mL and stir at 60°C for 2 hours to completely dissolve it. A polymer insulating layer 4 is prepared by spin-coating onto the surface of the CuPc active layer 3. The spin-coating speed is set to 2000 rpm and the spin-coating time is 60 s. After film formation, the device is dried on a hot plate at 120°C for 1 hour to form an insulating layer with a thickness of 150 nm.

[0035] (4) Finally, a top electrode layer 5 of metallic aluminum (Al) is deposited on the surface of the PVDF insulating layer 4 using a thermal evaporation process, with a evaporation vacuum degree better than 5×10⁻⁶. -4 Pa, with a vapor deposition thickness of 100 nm. The layers are in close contact, ultimately forming a vertically stacked self-driven sensing device of ITO / CuPc / PVDF / Al.

[0036] The electrical characteristics of the self-driven event-driven vision sensor fabricated in this embodiment were tested under conditions of no external power supply (zero bias voltage operation). The two electrodes of the device (lower electrode layer 2 and top electrode layer 5) were directly connected to the test port of a semiconductor parameter analyzer for signal acquisition. (See attached diagram) Figure 2 As shown, under zero bias voltage conditions, different power wavelengths of 520nm pulsed light are applied, demonstrating the current response characteristics of the self-driven event-driven vision sensor. Under zero bias voltage conditions, the optical power density of the incident dynamic pulsed light is gradually adjusted, and the measured transient displacement current pulses are shown in the attached figure. Figure 3 As shown, this demonstrates the superior ability of a self-driven event-driven vision sensor to continuously and dynamically track fluctuating light intensity in complex dynamic scenes. Through... Figure 3 The results show that as the dynamic optical power density increases, there is a good correlation between the amplitude of the transient current pulse output by the device and the change in optical power, and the fitting linearity is good, as shown in the attached figure. Figure 4 As shown, an efficient physical event encoding and optical signal-electrical signal mapping strategy for a self-driven event-driven vision sensor is implemented.

Claims

1. A self-driven event-driven vision sensor based on MISM structure, characterized in that, By utilizing only the photoelectric and pyroelectric synergistic physical mechanism at the device's underlying level, a highly efficient bipolar pulse current response to dynamic light intensity changes can be directly achieved in a zero-bias state, thereby completing the hardware-level extraction of dynamic features of moving objects at the sensing source. Its hardware layer adopts a layered heterogeneous stacked structure, specifically composed of a bottom substrate, a bottom electrode layer, an organic semiconductor active layer, a ferroelectric polymer insulating layer, and a top electrode layer, which are vertically stacked and connected from bottom to top; wherein: The lower electrode layer is deposited on the surface of the bottom substrate; the organic semiconductor active layer is deposited on the lower electrode layer, and a built-in electric field is formed due to the difference between the work function and the Fermi level; the polymer insulating layer covers the surface of the organic semiconductor active layer, and the surfaces of the two materials are in close contact with each other to form a physical heterogeneous interface; the top electrode layer is deposited on the surface of the ferroelectric polymer insulating layer. The lower electrode layer, organic semiconductor active layer, ferroelectric polymer insulating layer and top electrode layer together in the vertical direction constitute a two-end self-driven physical device. The two end electrodes are directly connected to the external signal output or subsequent signal acquisition circuit and are in a zero bias voltage working state.

2. The event-driven vision sensor according to claim 1, characterized in that, in: The underlying substrate is a rigid insulating substrate; The lower electrode layer is an indium tin oxide thin film with a thickness of 90-110 nm. The organic semiconductor active layer is made of copper phthalocyanine thin film with a thickness of 40-60 nm. The polymer insulating layer is made of polyvinylidene fluoride film with a thickness of 40-150 nm. The top electrode layer material is an aluminum thin film with a thickness of 80-100 nm. Each layer is in close contact with the others through physical heterogeneous interfaces, forming a self-driven sensing device with ITO / CuPc / PVDF / Al layers stacked at both ends.

3. The event-driven vision sensor according to claim 2, characterized in that, The specific action process and signal transmission relationship under dynamic light stimulation are as follows: When an external moving object causes a transient change in the intensity of the incident pulse light, the sensor operates in a zero-bias state. At the instant when the incident light is turned on, i.e., the ON instant, the active layer of the organic semiconductor absorbs photons and generates excitons. Under the action of the built-in electric field inside the device, the excitons quickly dissociate into free carriers. Due to the DC blocking characteristics of the polymer insulating layer, the free carriers cannot cross the insulating layer. Therefore, driven by the built-in electric field, they migrate directionally and accumulate at the physical interface between the active layer and the insulating layer, causing a sudden change in the electric displacement vector on both sides of the insulating layer. This results in a sharp positive transient current pulse being self-driven between the two electrodes of the external circuit. As the interface charge density continues to increase, the induced polarization electric field gradually strengthens and eventually cancels the built-in electric field inside the device, causing the current in the external circuit to gradually decay back to zero under continuous illumination. At the instant when the incident light is turned off, the charge accumulated at the physical interface is released due to the loss of the source of photogenerated charge carriers and flows back in the opposite direction through the external loop, thereby exciting a transient reverse current spike of opposite polarity between the two electrodes. In this way, the sensor, without external power supply and without any external transistor pixel circuitry, utilizes the interface charge dynamics self-driving mechanism of a single physical device to directly convert dynamically changing transient optical signals into bipolar transient electrical signal flows with distinct positive and negative polarity characteristics at the hardware physical level.

4. The method for fabricating an event-driven visual sensor as described in any one of claims 1-3, characterized in that, The specific steps are as follows: (1) First, the surface of the quartz glass substrate with the ITO lower electrode layer is cleaned. It is then immersed in acetone, isopropanol and deionized water for 15-30 minutes respectively for weak ultrasonic cleaning. Then, the excess liquid on the surface is blown off with a high-purity nitrogen gun and dried on a hot plate for later use. (2) The organic semiconductor active layer phthalocyanine copper film is deposited on the surface of the cleaned ITO lower electrode layer by using the organic thermal evaporation coating process, the evaporation source temperature is controlled at 200-380°C, the vacuum degree is better than 5x10 -4 Pa, the deposition rate is controlled at 0.1-0.2 Å / s, and the final growth thickness of the film is controlled at 40-60 nm; (3) Prepare a PVDF dimethylformamide solution with a concentration of 20-40 mg / mL and stir it at 50-60℃ for 1-2 hours to completely dissolve it; use spin coating to spin coat a polymer insulating layer on the surface of the copper phthalocyanine active layer; set the spin coating speed to 2000-4000 rpm and the spin coating time to 50-60 s; after film formation, place the device on a hot plate at 100-120℃ and dry it for 0.5-1 hour to form an insulating layer with a thickness of 40-150 nm; (4) Finally, the metal aluminum film of the top electrode layer is deposited on the surface of the PVDF insulating layer by using the thermal evaporation process, the evaporation vacuum degree is better than 5x10 -4 Pa, and the film thickness is 80-100 nm; The layers are in close contact, and the whole structure ultimately forms a vertically stacked self-driven sensing device of ITO / CuPc / PVDF / Al.

Images