A light sensor circuit, methods for operating the light sensor circuit and systems comprising light sensor circuits

Abstract

A light sensor circuit, methods for operating the light sensor circuit and systems comprising light sensor circuits are described The light sensor circuit comprises: a perovskite solar cell configured to generate a photovoltage in response to a light signal and thereby charge the light sensor circuit; and a thyristor, connected in parallel with the perovskite solar cell, and configured to discharge the light sensor circuit when the photovoltage reaches a threshold voltage, thereby generating a spike signal, wherein the perovskite solar cell is further configured to stop charging the light sensor circuit in response to an intensity of the light signal being above a threshold intensity, and to resume charging in response to the intensity of the light signal falling below the threshold intensity.

Description

[0001] A LIGHT SENSOR CIRCUIT, METHODS FOR OPERATING THE LIGHT SENSOR CIRCUIT AND SYSTEMS COMPRISING LIGHT SENSOR CIRCUITS

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to light sensing devices. In particular, the present disclosure relates to a light sensor circuit comprising a perovskite solar cell, methods for operating such a light sensor circuit, and a sensor system comprising multiple light sensor circuits.

[0004] BACKGROUND

[0005] Highly efficient control systems are crucial enablers for energy-sustainable building management systems relevant in offices, dormitories and healthcare facilities. Regulating biologically critical elements such as light, temperature and chemical environment are necessary to maintain proper living standards. Furthermore, monitoring human events (e.g. occupancy, trip-and-fall) and robotic helpers (e.g. navigation, inventory tracking) in the future can require huge data collection, storage and processing. Therefore, typical camera solutions designed to count occupancy and suggest responses are simply unsustainable and unscalable for practical deployment.

[0006] An emerging technique is to utilize sensors embedded in the physical spaces such as walls, furniture and ceilings; each could monitor changes in light, temperature or chemical environment. Not only are the sensors less power-hungry, the quality of data also improves with greater depth and location specificity. The challenges in implementing that are powering and sensitivity.

[0007] Machine learning algorithms are the ‘brains’ while Internet-of-Things (loT) sensors are the ‘sensory organs’ making sense of the physical world. The sensors are required to collect, process and transmit information (e.g. light, pressure, chemical); all requiring power from batteries or wired from main supply. Currently, to remove the need for wired power, nickel-metal hydride (NiMH) and lithium-ion (Li-ion) batteries are used in commercial and industrial edge applications. However, the continuous power consumption of loT devices requires battery life monitoring and even predictive replacement. Self-powered sensors are therefore highly deployable and desirable for longevity and cost-effectiveness. Photovoltaic cells have been previously explored as self-powered, light-sensitive devices for shadow detection, gesture recognition and human occupancy monitoring. However, photovoltaic cells are designed for solar power harvesting and not necessarily optimal for shadow detection in an indoor environment. This is especially critical as indoor lightings are designed to be diffused in nature and moving objects do not easily cast ideal shadows (high coverage by light sources and angles of entry). The faint shadows are approximated to be less than 0.5% of the ambient lighting. That would translate to 500 nanoamperes of photocurrent or less than 25 millivolts of opencircuit voltage. Coupled with intrinsic noises, compact loT devices are unable to reliably distinguish such low signals or obtain meaningful information from it. As a result, most of the earlier demonstrations are confined to highly parallel light sources or required direct coverage.

[0008] It is therefore desirable to provide a light sensor circuit that is self-powered, sensitive to minute variations in light, and capable of encoding information into spike signals. Methods for operating such a light sensor circuit, and a sensor system comprising multiple light sensor circuits are also provided. Further, other desirable features and characteristics will become apparent from the subsequent detailed description and the appended claims, taken in conjunction with the accompanying drawings and this background of the disclosure.

[0009] SUMMARY OF THE INVENTION

[0010] According to a first aspect of the present disclosure, a light sensor circuit is provided. The light sensor circuit comprises: a perovskite solar cell configured to generate a photovoltage in response to a light signal and thereby charge the light sensor circuit; and a thyristor, connected in parallel with the perovskite solar cell, and configured to discharge the light sensor circuit when the photovoltage reaches a threshold voltage, thereby generating a spike signal, wherein the perovskite solar cell is further configured to stop charging the light sensor circuit in response to an intensity of the light signal being above a threshold intensity, and to resume charging in response to the intensity of the light signal falling below the threshold intensity.

[0011] In an embodiment, the perovskite solar cell comprises a structure including a Fluorinedoped tin oxide (FTO) substrate, a compact blocking layer, a mesoporous TiO2 layer, a ZrO2 spacer layer and a carbon counter electrode layer. In an embodiment, the perovskite solar cell comprises a halide-excess perovskite layer.

[0012] In an embodiment, the perovskite solar cell comprises a film material comprising methylammonium (MA), formamidinium (FA), lead (Pb), bromide (Br), and iodide (I).

[0013] In an embodiment, the perovskite solar cell is configured to exhibit three loop impedance responses, including a response unrelated to ionic effects, a response corresponding to ionic charging and discharging, and a response corresponding to ion-mediated recombination.

[0014] In an embodiment, a frequency of the spike signal generated in response to the intensity of the light signal falling below the threshold intensity corresponds to a magnitude of a change in the intensity of the light signal.

[0015] In an embodiment, the light sensor circuit is configured to exhibit an adaptive behavior in response to a light signal with constant intensity above the threshold intensity, wherein a frequency of the spike signal decreases over time and the spike signal eventually ceases.

[0016] The present invention is shown to be second-order in computational capability. Besides encoding light intensity into spike frequency, it shows adaptation and inhibition behaviors, both of which are useful for detecting events and magnitude of changes. The perovskite composition and device configuration of the light sensor circuit described herein enable the detection of small changes in light intensity, such as those caused by faint shadows in diffuse indoor lighting conditions, with a signal- to-noise ratio significantly higher than conventional photovoltaic detectors. This high signal-to-noise ratio facilitates accurate detection of subtle light variations without the need for high-gain amplification or complex filtering circuitry. Furthermore, the present invention achieves broad dynamic ranges with different variations of perovskite composition (tortoise-type perovskite neuron: ~3x, 25Hz to 75Hz; elephant type perovskite neuron: >35x, ~0 to 35Hz; manatee-type perovskite neuron: ~8x, 0.8Hz to 0.1 Hz), enhancing sensitivity and compatibility with diverse sensing environments and Spiking Neural Network (SNN) architectures.

[0017] Furthermore, the light sensor is self-powered as the thyristor functions as a switch and derives its operating energy from the perovskite solar cell. The switching operation consumes only a small amount of energy and does not materially affect the performance of the light sensor circuit, provided that the perovskite solar cell has a sufficient active area. This confers high scalability and operational efficiency to the present invention.

[0018] Unlike conventional photodetectors that output continuous analog signals, which requires downstream analog-to-digital conversion and computational filtering, the present light sensor circuit generates discrete spike signals that inherently encode both the occurrence and timing of changes in incident light signal. This event-driven approach simplifies downstream processing, enables direct interfacing with digital processing units or microcontrollers, and eliminates the need for complex signal conditioning.

[0019] As a result, the architecture reduces power consumption, processing latency, and hardware complexity, making it particularly advantageous for self-powered, low- energy Internet-of-Things (loT) applications.

[0020] According to a second aspect of the present disclosure, a method of operating a light sensor circuit comprising a perovskite solar cell and a thyristor connected in parallel with the perovskite solar cell is provided. The method comprises: generating a photovoltage using the perovskite solar cell in response to a light signal, thereby charging the light sensor circuit; halting the charging of the light sensor circuit when an intensity of the light signal being above a threshold intensity; resuming the charging of the light sensor circuit when the intensity of the light signal falls below the threshold intensity; and discharging the light sensor circuit using the thyristor when the photovoltage reaches a threshold voltage, thereby generating a spike signal.

[0021] In an embodiment, the perovskite solar cell comprises a structure including a Fluorinedoped tin oxide (FTO) substrate, a compact blocking layer, a mesoporous TiOa layer, a ZrO2 spacer layer and a carbon counter electrode layer.

[0022] In an embodiment, the perovskite solar cell comprises a halide-excess perovskite layer.

[0023] In an embodiment, the perovskite solar cell comprises a film material comprising methylammonium (MA), formamidinium (FA), lead (Pb), bromide (Br), and iodide (I).

[0024] In an embodiment, a frequency of the spike signal generated in response to the intensity of the light signal falling below the threshold intensity corresponds to a magnitude of a change in the intensity of the light signal. In an embodiment, a frequency of the spike signal generated in response to a light signal of constant intensity above the threshold intensity decreases over time and the spike signal eventually ceases.

[0025] In an embodiment, the method further comprises determining a relative proximity of an object to the light sensor circuit based on a frequency of the spike signal.

[0026] In an embodiment, the method further comprises controlling a robotics element based on a frequency of the spike signal.

[0027] According to a third aspect of the present disclosure, a system comprising a processor and two or more light sensor circuits according to the first aspect of the present disclosure is provided. The processor is configured to: obtain temporal information of a plurality of spike signals, from the light sensor circuits, in response to changes in a light signal caused by a motion of an object; and determine an indication of the motion of the object based on the temporal information of the spike signals.

[0028] In an embodiment, the light sensor circuits are arranged in an array of light sensor circuits.

[0029] In an embodiment, the array of light sensor circuits comprises at least four light sensor circuits and the processor is further configured to determine the indication of the motion of the object that comprises an azimuth or a rotation.

[0030] In an embodiment, the temporal information comprises a time difference of the plurality of spike signals from different light sensor circuits, and the indication of the motion of the object comprises an indication of a direction of movement.

[0031] In an embodiment, the object is a hand making a gesture and the processor is further configured to determine a class of gesture based on the indication of the motion of the object.

[0032] BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In the following, embodiments of the present invention will be described as non-limiting examples with reference to the accompanying drawings in which:

[0034] Figure 1 illustrates a comparison between a biological neuron and a halide perovskite based artificial neuron; Figure 2 illustrates a structure of a light sensor circuit comprising a perovskite solar cell and a thyristor, according to an embodiment of the present invention;

[0035] Figure 3 is a diagram of l-V (current-voltage) characteristics of a thyristor, according to an embodiment of the present invention;

[0036] Figures 4A to 4G illustrate an electrical equivalent circuit of the light sensor circuit of Figure 2 and spiking behavior of the light sensor circuit according to an embodiment; Figure 4A shows the electrical equivalent circuit; Figure 4B shows the spiking profile under low light intensity (5 mW / cm2); Figure 4C shows the spiking profile under intermediate light intensity (7.5 mW / cm2); Figure 4D shows the spiking profile under high light intensity (20 mW / cm2); Figure 4E is a diagram illustrating the relationship between spike signal frequency and the input light intensity for the light sensor circuit; Figure 4F illustrates the spiking profile of the light sensor circuit when a meta-stable state is perturbed by a 10% reduction in light intensity; Figure 4G is a diagram illustrating the relationship between a frequency of spike signals and input current when an electrical signal is applied to the light sensor circuit.

[0037] Figure 5 shows a flowchart of a method of operating a light sensor circuit, according to an embodiment;

[0038] Figures 6A to 6C relate to characteristics of a tortoise-type perovskite neuron according to an embodiment; Figure 6A illustrates an exemplary device configurations; Figure 6B illustrates impedance responses; Figure 6C illustrates impedance responses and intensity-dependent frequency responses;

[0039] Figures 7A to 7C relate to characteristics of an elephant-type perovskite neuron according to an embodiment; Figure 7A illustrates an exemplary device configurations; Figure 7B illustrates impedance responses; Figure 70 illustrates impedance responses and intensity-dependent frequency responses;

[0040] Figures 8A to 80 relate to characteristics of a manatee-type perovskite neuron according to an embodiment; Figure 8A illustrates an exemplary device configurations; Figure 8B illustrates impedance responses; Figure 8C illustrates impedance responses and intensity-dependent frequency responses;

[0041] Figures 9A and 9B illustrate a comparison of spike signal generation in response to a shadow stimulus, as observed in a conventional silicon solar cell and in a perovskite neuron device according to an embodiment; Figure 9A shows the experimental setup; Figure 9B shows the output signals from the silicon solar cell and the perovskite neuron device;

[0042] Figures 10A and 10B illustrate gesture recognition and motion-direction encoding using a system comprising two perovskite neuron devices, according to an embodiment, where Figure 10A shows the experimental set up and Figure 10B shows the implementation of the motion-direction encoding in controlling a robotic arm; and

[0043] Figures 11 A and 11 B illustrate an application of a system comprising a plurality of light sensor circuits in accordance with an embodiment; Figure 11 A shows a configuration of the system; Figure 11 B illustrates a confusion matrix demonstrating gesture classification accuracy achieved using the system.

[0044] DETAILED DESCRIPTION

[0045] The present disclosure relates to a light sensor circuit, associated methods of operation, and a scalable sensing system comprising multiple units of light sensor circuit that collectively enables more complex signal processing and sensing tasks. The light sensor circuit comprises a perovskite solar cell connected in parallel with a thyristor, forming a compact neuromorphic sensor capable of converting variations in incident light into discrete electrical spike signals.

[0046] To overcome the low-signal detection limitation of photovoltaic cell sensor, inspiration is drawn from biological sensory neurons. These neurons, fundamental components of the nervous system, perform key functions: receiving input signals (via dendrites), integrating inputs (in the cell body), generating action potentials (in the axon), and encoding information through unique spiking patterns. An action potential or spike is an electrical impulse that can be propagated efficiently along the long neuronal axon. The impulse nature of the action potential spike is also critical in inter-neuron communication where the timing between firing can be synchronized to execute complex responses and in the coding of information where the spiking frequency and pattern can encode specific information.

[0047] By leveraging the optoelectronic properties of halide perovskite materials, such as high absorption coefficients (105 / cm), superior external quantum efficiencies, and photoresponsivity (mA / W) surpassing that of conventional silicon-based materials, the circuit enables highly efficient light sensing with a signal-to-noise ratio improvement exceeding 10 times. This architecture preserves analog input information by encoding it into the frequency of generated spike signals, while also inherently producing both positive and negative signal polarities, eliminating the need for additional computational elements to process or differentiate signal polarity. This event-driven, low-power operation makes the circuit highly suitable for real-time sensing applications such as motion detection, ambient light monitoring, and robotics control.

[0048] Figure 1 illustrates a comparison between a biological neuron and a halide perovskite based artificial neuron, herein referred to as a perovskite neuron. Biological morphology of biological neurons 101 and crystal structure of the perovskite neurons 104 are shown. The ionic-electrical current responses in the perovskite neuron are shown in 105 and the dual-channel ionic current responses in the biological neuron are shown in 102. The comparison demonstrates analogous behaviors in current generation. In the biological neuron, the activity of fast sodium (Na+) channels and slow potassium (K+) channels results in a characteristic action potential. Similarly, the perovskite neuron exhibits ion migration and charge carrier dynamics that emulate this dual-channel behavior, with distinct fast and slow ionic-electronic transport mechanisms contributing to the overall signal. The comparison includes a depiction of carrier transport and mixed ionic-electronic conduction within the perovskite structure shown in 106 and a depiction of fast sodium channels and slow potassium channels with selective ion transport within biological neurons shown in 103. Since the electronic and ionic currents in halide perovskite materials mimic the fast-slow dynamics of the sodium and potassium channels, respectively, in a biological neuron, by engineering these dual ionic-electronic currents, various neuronal processes, such as spike generation, adaptation, and information encoding, can be emulated within halide perovskite based systems.

[0049] In various embodiments of the present disclosure, a perovskite neuron may be implemented as a sensor circuit comprising an electrical component that comprises perovskite material. The circuit leverages the intrinsic properties of perovskite neurons and is configured to produce specific spike generation patterns. The chemical composition and material properties of the perovskite may be tuned to control the spiking behavior, enabling tailored neuromorphic responses based on the desired application. In biological neurons, the sensory inputs are processed by controlling the various major ionic currents (potassium and sodium) and modulate the neuronal membrane voltage, as shown in 103. When the membrane voltage exceeds a threshold, a spike is formed. When the signals are not sufficient to trigger a spike, the information is gradually lost. When a spike is triggered, the neuron goes into a refractory state that inhibits any form of spike generation. Similar to a biological neuron, a perovskite neuron is configured to process input signals, such as light signals or electrical signals, and convert them into spike trains that carry temporal information, including precise spike timings and distinct spike patterns in response to the input signals. Analog information may be encoded in spike trains, where spikes serve as a form of digital communication. Each spike in the spike trains represents a logical '1 ' and the resting state represents 'O'. Digital communication is generally preferred over analog due to the intrinsic noise introduced during signal transmission and readout. This advantage is particularly critical for Internet-of-Things (loT) devices, which operate under strict energy and size constraints while requiring reliable, high-quality analog data acquisition.

[0050] In one embodiment, halide perovskites, a class of material systems exhibiting mixed carrier transport, are used to emulate spiking behavior of the biological neurons. Halide perovskites are mixed ionic-electronic conductors characterized by strongly coupled ionic and electronic transport mechanisms. This coupling can be engineered to tune frequency-dependent responses. In particular, the combination of a slow rise in electronic transport and a fast fall in ionic transport can be utilized to reproduce the oscillatory behavior of biological neurons, as illustrated in Figure 1 .

[0051] Figure 2 illustrates a structure of a light sensor circuit 200 comprising a perovskite solar cell 201 and a thyristor 202 according to an embodiment.

[0052] The perovskite solar cell 201 is configured to generate a photovoltage in response to a light signal and thereby charge the light sensor circuit 200. The photovoltage accumulates and charges the circuit 200 over time.

[0053] The thyristor 202, connected in parallel with the perovskite solar cell 201 , is configured to discharge the light sensor circuit 200 when the photovoltage reaches a threshold voltage, thereby generating a spike signal. The thyristor 202 is configured to remain in a high-impedance (non-conductive) state while the photovoltage remains below a predefined threshold. When the photovoltage reaches or exceeds this threshold voltage, the thyristor 202 switches to a low-impedance (conductive) state, causing the accumulated charge to discharge rapidly through the circuit. This discharge event produces a sharp, discrete spike signal, converting light intensity variation into timedomain digital output in the form of spike signals.

[0054] The perovskite solar cell 201 is further configured to stop charging the light sensor circuit 200 in response to an intensity of the light signal being above a threshold intensity, and to resume charging in response to the intensity of the light signal falling below the threshold intensity. When the perovskite solar cell 201 halts further charging of the circuit, it prevents or inhibits spike signal generation. When the perovskite solar cell 201 resumes charging, this enables the circuit to reinitiate spike signal generation.

[0055] In one embodiment, the perovskite solar cell 201 (also referred to as a printed perovskite photovoltaic cell) may be fabricated using the following process. Fluorinedoped tin oxide (FTO) substrates are first etched and sequentially cleaned using Decon soap solution, deionized water, and ethanol, with 30 minutes of sonication for each step. The cleaning process concludes with a 15-minute UV ozone treatment at 100 °C. A compact blocking layer precursor is then screen-printed onto the FTO substrate and sintered at 500 °C for 40 minutes. The compact blocking layer precursor used may be Greatcell Solar BL. Subsequently, a mesoporous TiO2 (titanium dioxide) layer is deposited using TiOs paste diluted with terpineol in a 1 :1.4 ratio, which is sonicated and stirred prior to use. The TiO2 paste used may be the 30 NR-D Titania Paste. This layer is also sintered at 500 °C for 40 minutes. Following this, a ZrO2 (zirconium dioxide) spacer layer is printed using Solaronix ZrO2 paste and sintered under the same conditions. Lastly, the carbon counter electrode layer is screen-printed using Wonder Solar carbon paste and sintered at 400 °C for 30 minutes. In the embodiment described herein, the perovskite solar cell 201 comprises a structure including a fluorine-doped tin oxide (FTO) substrate, a compact blocking layer, a mesoporous TiO2 layer, a ZrO2 spacer layer, and a carbon counter electrode layer.

[0056] In a subsequent step of fabrication, a perovskite solution may be prepared inside a glove box by mixing 1.2 mmol of Pbl2 (lead iodide), 1.2 mmol of MAI (methylammonium iodide), and 0.5 mmol of 5-AVAI (5-Ammonium valeric acid iodide) in 1 mL of GBL (gamma-butyrolactone), followed by stirring at 50 °C until a clear solution was obtained. The prepared solution is then infiltrated into the device structure, using approximately 3.8 pL per device with an active area of 0.7 cm2. After infiltration, the device is allowed to settle before undergoing a two-step thermal treatment to promote crystallization: first, heating at 50 °C for 90 minutes, followed by 60 °C for 60 minutes in an ambient environment. Finally, electrical contacts are applied to the device for testing.

[0057] In one embodiment, the fabrication process of the perovskite solar cell 201 may comprise depositing an electron transport layer, which comprises the following steps. A thin blocking layer of planar titanium dioxide may be spray-pyrolyzed onto a substrate at a temperature of about 450°C, followed by deposition of a mesoporous titanium dioxide layer spin-coated on top of the compact titanium dioxide (c-TiO2) layer at a speed of about 3000 revolutions per minute for a duration of about 30 seconds. The mesoporous titanium dioxide solution may be prepared by combining approximately one part of TiO2paste (18-NRD) with approximately nine parts of ethyl alcohol (Sigma Aldrich). The electron transport layer-coated substrates may be subsequently treated with UV-ozone for approximately 15 minutes, after which they may be transferred into a nitrogen-filled glovebox for deposition of one or more subsequent layers.

[0058] In one embodiment, the fabrication process of the perovskite solar cell 201 may comprise depositing a cesium chloride (CsCI) doped double cation perovskite. The process may comprises, in the case of CsCI-doped FA(MA)Pbl(Br)s system, weighing 1.35M FAI (formamidinium iodide), 1.35M Pbh (lead iodide), 3 mol% MABr (methylammonium bromide), 5 mol% CsCI and 35 mol% of MACI (methylammonium chloride) in 1 mL of 4:1 DMF (N,N-Dimethylformamide, from Sigma Aldrich): DMSO (dimethyl sulfoxide, from Sigma Aldrich) mixed solvent. The CsCI-doped FA(MA)Pbl(Br)s precursor is stirred overnight at 50°C. The FA(MA)Pbl(Br)s and CsCI- doped FA(MA)Pbl(Br)3 perovskite solution is then spin-coated on the meso-TiC substrate at 6000 RPM for 50 seconds. Approximately 200 pL of chlorobenzene is dripped on the substrate 10 seconds after the start of the spin to induce rapid crystallization. The control films are annealed at 150°C for 10 minutes and the CsCI- doped perovskite films are annealed at 150°C for 15 minutes. In one embodiment, the fabrication process of the perovskite solar cell 201 may comprise depositing a hole transport layer (HTL). The process of depositing HTL may use Spiro-OMeTAD (72.3 mg / mL, from LumTech) in CBZ (chlorobenzene) doped with tBP (4-tert-butylpyridine, 28.5 pL, Sigma Aldrich) and Li-TFSI (lithium bis(trifluoromethanesulfonyl)imide, 520 mg / mL) solution in ACN (acetonitrile, 17.5 pL, Sigma Aldrich). Subsequently, Au electrodes are thermally evaporated (~100 nm) on top of the device fabricated.

[0059] In various embodiments of the present disclosure, the properties of the perovskite material used in the perovskite solar cell 201 may be tuned by modifying the solution composition, thereby enabling control over carrier density, crystallinity, and electronic properties of the resulting film. For example, methylammonium bromide (MABr) may be added to improve crystallinity, while CsCI may be introduced to control carrier density and adjust energy levels. These material modifications are supported and verified through comprehensive characterization techniques, including impedance spectroscopy, optical absorption, photoluminescence spectroscopy, X-ray photoelectron spectroscopy (XPS), and ultraviolet photoelectron spectroscopy (UPS).

[0060] In one embodiment, the thyristor 202 is a silicon thyristor. A suitable commercially available component may be the Littelfuse SCR EC103D series, which has been found to perform effectively under the operating conditions of the sensor circuit. This thyristor exhibits a maximum holding current of approximately 12 pA at a voltage of around 0.8 V, providing consistent switching behavior that supports the generation of discrete spike signals in response to accumulated photovoltage. The thyristor 202 is used as a threshold-switching and reset element within the circuit and it may be replaced with a switching device implemented using alternative materials or device architectures, provided that the switching element exhibits the required electrical characteristics to support the intended circuit functionality. Specifically, the switching device should possess: (1 ) an ultra-smooth negative differential resistance (NDR) region without abrupt transitions or discontinuities, and (2) no hysteresis or electrical instability during both forward and reverse voltage sweeps, as illustrated in Figure 3.

[0061] In various embodiments, the perovskite solar cell 201 may be fabricated using screen printing technology, allowing for scalable manufacturing across a variety of substrates. These include rigid substrates such as glass and flexible plastics. The light sensor circuit 200 and the connection of the circuit’s components may be fully fabricated via printing technology.

[0062] In one embodiment, the perovskite solar cell 201 comprises a halide-excess perovskite layer. The layer exhibits intrinsically high hole densities and contributes to specific impedance characteristics that facilitate spike signal generation. The halide perovskite material's properties, including high ionic mobility, allow the bulk material to undergo polarization, forming halide-rich and halide-deficient regions. Furthermore, in such materials, the carrier density, including both ionic and electronic components, has a strong influence on the material’s ionic polarizability, governed by the Poisson equation (V2V = -p / £0), where the electrostatic potential of the system depends on the total charge distribution. Therefore, both the intrinsic doping level of the perovskite and the energy alignment with the carrier transport layers play critical roles in defining device behavior.

[0063] In one embodiment, the perovskite solar cell 201 comprises a film material comprising methylammonium (MA), formamidinium (FA), lead (Pb), bromide (Br), and iodide (I). A representative composition includes MA0.09FAPbBr0.09l3, providing balanced optoelectronic properties and ionic-electronic interactions suitable for neuromorphic sensing applications.

[0064] In one embodiment, the light sensor circuit 200 comprises two or more perovskite solar cells 201 and one thyristor 202. The light sensor circuit 200 is fully functional without the need for any external power source. This self-powered configuration stands in contrast to conventional electrochemical transistor-based circuits, which typically require a dedicated power source and a more complex architecture, such as 1 S3T1 Inverter (one power source, three transistors, and one inverter), 1 S5T (one power source with five or more transistors) or 1 S1T2lnverter (one power source, one transistor, 2 inverters) configurations. The disclosed light sensor circuit 200 offers a highly compact and scalable design, made possible by the dual functionality of the perovskite material as both a photovoltaic energy harvester and a sensing element. In contrast, reliance on external power sources in conventional designs imposes significant limitations on scalability.

[0065] Figures 4A to 4G illustrate the electrical equivalent circuit 410 of the light sensor circuit 200 and spiking behavior of the light sensor circuit according to an embodiment. Figure 4A illustrates an electrical equivalent circuit of a light sensor circuit 200, comprising a perovskite solar cell 201 (or alternatively referred to as a perovskite memdiode), connected in parallel with a thyristor 202, which, in one embodiment, is a silicon-based thyristor device. Upon exposure to a light stimulus, the perovskite solar cell 201 generates a photocurrent 411 . The equivalent circuit further models the three characteristic temporal responses of the perovskite solar cell.

[0066] The intrinsic properties of the perovskite material, such dielectric and geometric properties, are represented by a small capacitor 412 (approximately 20 nF), which accounts for the high-frequency impedance response, typically in the range of 10 kHz to 1 MHz. Ion-related properties, such as ionic polarization and associated charging / discharging dynamics, are represented by a resistor-capacitor (RC) network 413. The RC network exhibits a characteristic time constant in the range of 10 milliseconds and corresponds to mid-frequency responses between 10 Hz and 10 kHz. Electron-related properties, including ion-mediated recombination currents, can be represented using a resistor-inductor (RL) low-pass component 414, characterized by a time constant (L / R) in the range of 1 second, and responsible for low-frequency behavior in the range of 10 Hz to 0.01 Hz.

[0067] When continuous light is applied to the circuit, the photovoltage across the perovskite solar cell 201 gradually increases, analogous to the membrane potential in a biological neuron. Upon reaching a threshold voltage, the thyristor 202 acts as a threshold-and- reset mechanism, triggering a rapid discharge of the circuit. The dynamics of charging and discharging are directly influenced by the impedance characteristics of the circuit, which consequently determine the spike generation behavior.

[0068] In one embodiment, the perovskite solar cell 201 is configured to exhibit three loop impedance responses: (1 ) a response unrelated to ionic effects, (2) a response associated with ionic charging and discharging, and (3) a response corresponding to ion-mediated recombination processes.

[0069] Figures 4B to 4D illustrate the spiking behavior of the light sensor circuit 200 according to an embodiment under varying light stimulus intensities. Figure 4B shows the spiking profile under low light intensity (5 mW / cm2), Figure 4C under intermediate light intensity (7.5 mW / cm2), and Figure 4D under high light intensity (20 mW / cm2). In the case of high light intensity shown in Figure 4D, the circuit 200 exhibits adaptive spiking behavior, characterized by a gradual decrease in spiking frequency over time. This adaptation behavior enables the neuron to transition into a meta-stable state, which may be exploited for downstream processing or event detection.

[0070] In one embodiment, the light sensor circuit 200 is configured to exhibit an adaptive behavior in response to a light signal with constant intensity above the threshold intensity, wherein a frequency of the spike signal decreases over time and the spike signal eventually ceases. The adaptive behavior is attributed to the fast-rise, slow-fall dynamics of the photovoltage in the perovskite solar cell. Under high-intensity illumination, the photovoltage rises rapidly, initiating a burst of high-frequency spiking (as shown in Figure 4D). However, as the exposure continues, the photovoltage decreases, resulting in a progressive reduction in spike frequency, demonstrating an adaptation to sustained stimulation.

[0071] Figure 4E illustrates the relationship between spike signal frequency and the input light intensity for the light sensor circuit 200 according to an embodiment, wherein the spike signal frequency increases proportionally with increasing light intensity. The light sensor circuit 200 exhibits a characteristic tortoise-shaped impedance response 450. The fast-rise, slow-fall adaptive behavior is a consequence of the tortoise-type impedance response. Specifically, the lower inductive loop leads to the fast-rise and upper capacitive loop leads to the slow-fall of the photovoltage. This demonstrates that the spike signal frequency of the perovskite-based sensor circuit corresponds to input light intensity, thereby enabling intensity-encoded temporal signal processing.

[0072] Figure 4F illustrates the spiking profile of the light sensor circuit 200 according to an embodiment when the meta-stable state is perturbed by a 10% reduction in light intensity. This perturbation induces the neuron into an oscillatory spiking state, wherein the frequency of spike signal generation is proportional to the magnitude of the change in light intensity. This behavior demonstrates an inhibition spike response, which is a biorealistic function observed in the perovskite neuron.

[0073] In one embodiment, the perovskite solar cell 201 first adapts to a high light stimulus, during which the circuit enters a stable, non-spiking state. Upon the sudden application of a minor shadowing event (i.e., a 10% reduction in ambient illumination), the circuit 200 is triggered to generate a spike. The sensitivity to slight, sudden changes in illumination intensity enables the circuit to serve as a highly effective event detector, particularly useful in detecting low-level, transient perturbations in the visual or environmental signal field.

[0074] In one embodiment, a frequency of the spike signal generated in response to the intensity of the light signal falling below the threshold intensity corresponds to a magnitude of a change in the intensity of the light signal. This behavior enables the light sensor circuit 200 to detect varying magnitude of changes in light intensity, thereby allowing it to infer contextual information such as the proximity or movement of an object relative to the sensor.

[0075] Figure 4G illustrates the relationship between the frequency of spike signals and the input current when an electrical signal is applied to the light sensor circuit according to an embodiment. In addition to light signal, the circuit 200 may receive electrical inputs derived from other types of sensors, such as pressure sensors and chemical sensors, thereby extending its functionality across multiple sensing modalities.

[0076] In one embodiment, the light sensor circuit 200 comprises a perovskite solar cell 201 configured to receive a light stimulus to generate an increasing photovoltage, and a silicon thyristor device 202 configured to discharge the circuit to generate a decreasing photovoltage when the increasing photovoltage reaches a threshold, wherein a rate of increase and decrease of the photovoltage is dependent on the intensity of the light stimulus. The spiking response as a result of increasing and decreasing photovoltage, including the rate and firing pattern of generated spike signals, can be modulated by the choice of intrinsic doping in the perovskite layer and interfacial energy level of the electron / hole transport layers in the perovskite solar cell 201 . In one embodiment, the perovskite solar cell 201 exhibits a tortoise-type impedance response without the use of external additives. The impedance behavior arises from the intrinsic properties of the perovskite layer, which is characterized by a high hole density resulting from its intrinsic doping. This high carrier concentration contributes to the observed frequencydependent impedance and supports the generation of spiking behavior in response to light signal.

[0077] Figure 5 shows a flowchart of a method 500 of operating a light sensor circuit 200 comprising a perovskite solar cell 201 and a thyristor 202 connected in parallel with the perovskite solar cell 201 , according to an embodiment. This method is applicable to the light sensor circuit 200 as shown in Figure 2 and other embodiments of the light sensor circuit 200 described herein.

[0078] In various embodiments mentioned above, the perovskite solar cell 201 may comprise a structure including a Fluorine-doped tin oxide (FTO) substrate, a compact blocking layer, a mesoporous TiC layer, a ZrC>2 spacer layer and a carbon counter electrode layer. The perovskite solar cell 201 may also comprise a halide-excess perovskite layer. The perovskite solar cell 201 may comprise a film material comprising methylammonium (MA), formamidinium (FA), lead (Pb), bromide (Br), and iodide (I).

[0079] The method 500 comprises the following steps.

[0080] In step 502, the perovskite solar cell 201 generates a photovoltage in response to a light signal, thereby charging the light sensor circuit 200. The perovskite material, when exposed to photons, absorbs light and produces charge carriers (electrons and holes). These carriers are separated and transported across the internal junctions of the device, resulting in the buildup of an electrical potential (photovoltage) across the terminals of the perovskite cell 201 . The generated photovoltage contributes to charging the light sensor circuit over time. The charging rate is dependent on the intensity and duration of the light signal as well as the intrinsic electrical and ionic properties of the perovskite layer.

[0081] In step 504, the perovskite solar cell 201 halts the charging of the light sensor circuit 200 when an intensity of the light signal being above a threshold intensity. In this state, the accumulation of additional photovoltage is suspended and the circuit enters a stable state.

[0082] In step 506, the perovskite solar cell 201 resumes the charging of the light sensor circuit when the intensity of the light signal falls below the threshold intensity. This enables the circuit to detect low-level changes in light intensity during the stable state.

[0083] In step 508, the thyristor 202 discharges the light sensor circuit when the photovoltage reaches a threshold voltage, thereby generating a spike signal.

[0084] In one embodiment, a frequency of the spike signal generated in response to the intensity of the light signal falling below the threshold intensity corresponds to a magnitude of a change in the intensity of the light signal. This enables the circuit 200 to detect and quantify small or abrupt changes in environmental illumination. In one embodiment, a frequency of the spike signal generated in response to a light signal of constant intensity above the threshold intensity decreases over time and the spike signal eventually ceases. This behavior results from the adaptive properties of the perovskite material, which mimics biological neuron adaptation by reducing sensitivity to sustained stimuli.

[0085] In one embodiment, the method 500 further comprises a step of determining a relative proximity of an object to the light sensor circuit based on a frequency of the spike signal. For example, a transient decrease in intensity caused by a passing object may lead to a burst of spikes, the frequency of which is proportional to the shadowing magnitude, thus enabling proximity sensing.

[0086] In one embodiment, the method 500 further comprises a step of controlling a robotics element based on a frequency of the spike signal. The temporal information of the spike signals may be mapped to motion control commands, enabling light sensing for real-time robotic control or interaction.

[0087] The embodiments illustrated in Figures 4A to 4G relate to perovskite neuron having a tortoise-type impedance response. Figures 6A to 6C further illustrate an exemplary device configurations, perovskite film compositions, impedance responses and intensity-dependent frequency response corresponding to tortoise-type perovskite neurons. The tortoise-type perovskite neurons exhibit material characteristics including an electron offset greater than the hole offset, large hole densities, and electron-limited Shockley-Read-Hall (SRH) recombination.

[0088] As shown in Figure 6A, in one embodiment of a tortoise-type perovskite neuron, the perovskite solar cell comprises a pristine MA009FAPbBr009l3 film 613, which possesses high electron densities. The presence of high carrier densities introduces an additional mid-frequency temporal region in the impedance spectrum, manifested as a second top semicircle, which is associated with ionic polarization within the bulk perovskite film. In contrast to suppressed-carrier-density implementations, the tortoise-type perovskite neuron maintains a large intrinsic hole density in the halide- excess perovskite composition. Other structural layers of the tortoise-type perovskite neuron may include a gold (Au) electrode layer 611 , a Spiro-OMeTAD hole transport layer 612, a PEDOT:PSS layer 614, and a fluorine-doped tin oxide (FTO) substrate layer 615. As shown in Figure 6B, the impedance response of such a tortoise-type perovskite neuron typically includes three distinct loop features: (1 ) The first capacitive loop 621 , appearing at high frequencies, is unrelated to ionic effects and is common across all perovskite cell configurations. (2) The second loop 622, observed in the mid-frequency range, corresponds to ionic charging and discharging processes in the perovskite bulk. (3) The third response 623, observed at low frequencies, is attributed to ion-mediated recombination dynamics, and appears as a bottom semicircle in the impedance plot. The tortoise-type perovskite neuron is configured to exhibit adaptive behavior, as described in various embodiments of the present disclosure, due to its characteristic impedance response. As shown in Figure 6C, tortoise-type neuron exhibits a distinct sublinear response at low light intensities 631 and an adaptive response at high light intensities 632. This behavior is characteristic of a Class 2 neuron, indicating enhanced computational capabilities.

[0089] Figures 7A to 7C illustrate exemplary device configurations, perovskite film compositions, impedance responses and intensity-dependent frequency response corresponding to elephant-type perovskite neurons. The elephant-type perovskite neurons exhibit material characteristics including an electron offset greater than the hole offset, small electron and hole densities, and electron-limited SRH recombination.

[0090] As shown in Figure 7A, the elephant-type neuron is implemented using a perovskite film wherein the carrier densities are experimentally suppressed through cesium chloride (CsCI) doping. As a result of the reduced carrier concentration, the ionic polarizability within the perovskite film is significantly diminished, resulting in the absence of a distinct mid-frequency impedance response. The elephant-type perovskite neuron may include a gold (Au) electrode layer 71 1 , a Spiro-OMeTAD hole transport layer 712, a 5% CsCI doped MA0.09FAPbBr0.09l3 layer 713, a PEDOT SS layer 714, and a fluorine-doped tin oxide (FTO) substrate layer 715.

[0091] As shown in Figure 7B, the impedance response of the elephant-type neuron exhibits only two temporal regions: (1) a high-frequency capacitive response 721 , attributed to the intrinsic dielectric and geometric properties of the perovskite material, and (2) a low-frequency inductive response 722, associated with ion-controlled modulation of recombination currents. In this configuration, the elephant-type neuron presents only one ion-related temporal region, specifically the ion-mediated recombination process represented by the low-frequency semicircle in the impedance spectrum 722. Figure 7C shows that the intensity-dependent frequency response of the elephant-type perovskite neuron is linear.

[0092] Figures 8A to 8C illustrate exemplary device configurations, perovskite film compositions, impedance responses and intensity-dependent frequency response corresponding to manatee-type perovskite neurons. The manatee-type perovskite neurons exhibit material characteristics including an electron offset greater than the hole offset, large electron densities, and electron-limited SRH recombination.

[0093] Figure 8A illustrates a manatee-type perovskite neuron characterized by large carrier densities, wherein the carrier transport is influenced by a hole-limiting energy barrier. The manatee-type perovskite neuron may comprise a carbon electrode layer 811 , a MAPbh perovskite layer 812, a scaffold layer 813 comprising a compact TIO2layer, mesoporous TiOz layer, and mesoporous ZrO2layer, and a fluorine-doped tin oxide (FTO) substrate layer 814.

[0094] Figure 8B illustrates the manatee-type perovskite neuron exhibiting a capacitive response 821 and a dual inductive response 822, 823 arising from ion-controlled modulation of two distinct carrier types. The impedance profile exhibits two ion-related temporal regions, namely ionic charging / discharging and ion-mediated recombination, represented as two bottom semicircles in the Nyquist plot.

[0095] Figure 8C shows the corresponding intensity-dependent frequency response of the manatee-type perovskite neuron. The manatee-type neuron exhibits a strong inductive dynamic that results in an inverted frequency response to light intensity, which can be used for shadow detection.

[0096] Figures 9A and 9B illustrate a comparison of spike signal generation in response to a shadow stimulus, as observed in a conventional silicon solar cell and in a perovskite neuron device (i . e. , a light sensor circuit comprising a perovskite solar cell and thyristor as described herein), according to an embodiment of the present disclosure. A faint shadow is cast over both the silicon solar cell and the perovskite neuron device. Figure 9A shows the experimental setup and Figure 9B shows the output signals from the silicon solar cell and the perovskite neuron device. Due to the diffuse and low-intensity characteristics of indoor lighting environments (typically ~700 lux), conventional solar cell-based gesture detection systems require very close proximity between the gesture and the cell to produce a measurable signal. At greater distances, only a faint and indistinguishable shadow is cast, translating to an approximate signal of 25 mV in silicon solar cells. This value is near the lower limit of detectability, particularly when considering typical system noise (e.g., ~10 mV introduced by wiring).

[0097] In contrast, the perovskite neuron device demonstrates a significantly higher signal- to-noise ratio, converting subtle shadow-induced variations into large, clean spike signals of approximately 350 mV, which are easily discernable and processable. As a result, the device maintains the analog properties of the shadow input while enabling digital transmission of information. This demonstrates the superior responsiveness and spike generation capability of the perovskite neuron under low-contrast light variations.

[0098] Furthermore, the frequency of spike generation in the perovskite neuron device increases proportionally with the strength of the shadow stimulus. This property enables the light sensor circuit to not only detect changes in light intensity but also encode the magnitude of change as a temporal spiking frequency.

[0099] In one embodiment, the output spike signals may be transmitted to control a robotic element. For example, the proximity of a user's hand casting a shadow can directly influence the speed of movement of a robotic arm. A closer hand results in a darker shadow, which increases the spike frequency, thereby increasing the actuation speed of the robotic component.

[0100] Figures 10A and 10B illustrate gesture recognition and motion-direction encoding using a system comprising two perovskite neuron devices, according to an embodiment. The system leverages the timing differential between spike signals generated by each neuron to determine the direction of motion of an object, such as a hand gesture.

[0101] As shown in Figure 10A, an experimental setup includes two spatially separated perovskite neuron devices configured to receive light stimuli. A moving object, such as a hand, casts a dynamic shadow over the two neurons in sequence, with a temporal difference of approximately 50 milliseconds between their respective spike responses, depending on the direction of movement.

[0102] In one embodiment, a left-to-right hand motion causes the left neuron to generate a spike prior to the right neuron. This temporal pattern is interpreted as a negative signal. Conversely, a right-to-left motion causes the right neuron to spike before the left neuron, producing a positive signal. These polarity-encoded spike signals can then be used to relay directional control commands to a robotic system.

[0103] Figure 10B demonstrates the implementation of this directional encoding in controlling a robotic arm, wherein the arm is configured to move right upon detecting a positive spike and left upon detecting a negative spike. The system thus enables intuitive gesture-based control of robotic elements through simple and direct neuromorphic encoding.

[0104] The high signal contrast among the three key states: (i) Rest, (ii) Activity, and (iii) Directional response (positive or negative) enables reliable signal interpretation with significantly enhanced accuracy. The directional gesture recognition method exhibits: high signal contrast, with over 10x improvement compared to conventional systems; low latency, due to minimal computational overhead as spike computation is inherently performed by the perovskite neuron hardware itself; and low power operation, leveraging the self-powered nature of the device.

[0105] The present disclosure further provides a system comprising a processor and two or more light sensor circuits according to various embodiments described herein, wherein the processor is configured to: obtain temporal information of a plurality of spike signals, from the light sensor circuits, in response to changes in a light signal caused by a motion of an object; and determine an indication of the motion of the object based on the temporal information of the spike signals.

[0106] In one embodiment, the temporal information comprises a time difference of the plurality of spike signals from different light sensor circuits, and the indication of the motion of the object comprises an indication of a direction of movement.

[0107] In one embodiment of the system comprising a processor and two or more light sensor circuits, the light sensor circuits may be arranged in an array of light sensor circuits. The processor may aggregate spike signals from multiple light sensor circuits in the array and, based on the spatial arrangement of the light sensor circuits within the array, interpret the aggregated signals to determine the direction and rotational movement of the object.

[0108] Figures 11 A and 11 B illustrate an application of a system comprising a plurality of perovskite-enabled saddle-node neurons, which are embodiments of the light sensor circuit described herein, the neurons being arranged in a 3x3 array, in accordance with an embodiment of the present disclosure.

[0109] In the embodiment shown in Figure 11 A, the system comprises a 3x3 array (9 pixels) of neurons 1111 embedded with a multilayer perceptron (MLP) neural network and connected to a processor, such as an Arduino GIGA R1 WiFi microcontroller, for motion detection 1112. Each neuron is configured to respond selectively to perturbations in ambient light.

[0110] When a shadow or transient light variation (an "event") is cast onto a given pixel, the corresponding neuron generates a train of spike signals. In the absence of such stimulus, the neuron remains in a resting state. This dynamic behavior enables the detection of transient events while maintaining low power consumption in the absence of activity. The saddle-node oscillator configuration of the neuron provides an inhibition firing mechanism that facilitates high temporal precision in the initiation of event detection.

[0111] In one embodiment, the array of nine light sensor circuits (i.e., nine pixels) is interfaced with a commercial microcontroller configured to detect, encode, and classify the generated spike signals. The proximity of an object causing a shadow affects the frequency of the spike signals, enabling estimation of azimuth or rotational movement. To obtain azimuth or rotational movement information from the spike signals, a two- dimensional array of light sensor circuits of any size may be employed, with a minimum array size of 2x2. Larger arrays provide additional data points, thereby improving the accuracy and robustness of motion detection performed by the processor. Therefore, in one embodiment, the array comprises at least four light sensor circuits, and the processor is configured to determine an indication of motion of the object that includes at least one of an azimuth or a rotational movement.

[0112] The generated spike signals exhibit a high signal-to-noise ratio and temporal sparsity, which supports efficient and accurate recognition by low-power embedded processors. Further, the processor may include wireless connectivity for remote transmission of data.

[0113] In one embodiment, the processor is configured to receive signals from a plurality of light sensor circuits, the signals being processed to determine the motion of a detected object. When the object is a hand performing a gesture, the processor may be further configured to determine a class of the gesture based on the detected motion of the object using a neural network model. The neural network model may be trained to classify types of gestures, such as left or right swipes, based on the motion of the object, including azimuth, rotation, or direction of movement.

[0114] In the embodiment illustrated in Figures 11 A and 1 1 B, the spike signals from the 3x3 light sensor array are processed using a lightweight one-dimensional convolutional neural network (1 D CNN), occupying approximately 25.8 kB of RAM and 454.6 kB of flash memory. The network is deployed on-device within the Arduino GIGA R1 WiFi microcontroller, enabling real-time gesture recognition without external computation.

[0115] Figure 11 B illustrates a confusion matrix 1120 demonstrating gesture classification accuracy achieved using the system according to an embodiment. In one embodiment, the trained model may be configured to recognize ten gesture classes, including eight unidirectional swipes and two rotational swipes. The resulting confusion matrix demonstrates classification accuracy exceeding 99%, sufficient for most Internet-of- Things (loT) and human-machine interface applications. The performance also indicates that the network has sufficient capacity to support a greater number of gesture classes if required.

[0116] Whilst the foregoing description has described exemplary embodiments, it will be understood by those skilled in the art that many variations of the embodiments can be made within the scope of the present invention.

Claims

CLAIMS1 . A light sensor circuit, comprising: a perovskite solar cell configured to generate a photovoltage in response to a light signal and thereby charge the light sensor circuit; and a thyristor, connected in parallel with the perovskite solar cell, and configured to discharge the light sensor circuit when the photovoltage reaches a threshold voltage, thereby generating a spike signal, wherein the perovskite solar cell is further configured to stop charging the light sensor circuit in response to an intensity of the light signal being above a threshold intensity, and to resume charging in response to the intensity of the light signal falling below the threshold intensity.

2. The light sensor circuit of claim 1 , wherein the perovskite solar cell comprises a structure including a Fluorine-doped tin oxide (FTO) substrate, a compact blocking layer, a mesoporous TiC>2 layer, a ZrC>2 spacer layer and a carbon counter electrode layer.

3. The light sensor circuit of claim 2, wherein the perovskite solar cell comprises a halide-excess perovskite layer.

4. The light sensor circuit of any one of the preceding claims, wherein the perovskite solar cell comprises a film material comprising methylammonium (MA), formamidinium (FA), lead (Pb), bromide (Br), and iodide (I).

5. The light sensor circuit of any one of the preceding claims, wherein the perovskite solar cell is configured to exhibit three loop impedance responses, including a response unrelated to ionic effects, a response corresponding to ionic charging and discharging, and a response corresponding to ion-mediated recombination.

6. The light sensor circuit of any one of the preceding claims, wherein a frequency of the spike signal generated in response to the intensity of the light signal falling below the threshold intensity corresponds to a magnitude of a change in the intensity of the light signal.

7. The light sensor circuit of any one of the preceding claims, wherein the light sensor circuit is configured to exhibit an adaptive behavior in response to a light signal with constant intensity above the threshold intensity, wherein a frequency of the spike signal decreases over time and the spike signal eventually ceases.

8. A method of operating a light sensor circuit comprising a perovskite solar cell and a thyristor connected in parallel with the perovskite solar cell, the method comprising: generating a photovoltage using the perovskite solar cell in response to a light signal, thereby charging the light sensor circuit; halting the charging of the light sensor circuit when an intensity of the light signal being above a threshold intensity; resuming the charging of the light sensor circuit when the intensity of the light signal falls below the threshold intensity; and discharging the light sensor circuit using the thyristor when the photovoltage reaches a threshold voltage, thereby generating a spike signal.

9. The method of claim 8, wherein the perovskite solar cell comprises a structure including a Fluorine-doped tin oxide (FTO) substrate, a compact blocking layer, a mesoporous TiC layer, a ZrO2 spacer layer and a carbon counter electrode layer.

10. The method of claim 9, wherein the perovskite solar cell comprises a halide- excess perovskite layer.11 . The method of any one of the claims 8 to 10, wherein the perovskite solar cell comprises a film material comprising methylammonium (MA), formamidinium (FA), lead (Pb), bromide (Br), and iodide (I).

12. The method of any one of the claims 8 to 11 , wherein a frequency of the spike signal generated in response to the intensity of the light signal falling below the threshold intensity corresponds to a magnitude of a change in the intensity of the light signal.

13. The method of any one of the claims 8 to 12, wherein a frequency of the spike signal generated in response to a light signal of constant intensity above the threshold intensity decreases over time and the spike signal eventually ceases.

14. The method of any one of the claims 8 to 13, further comprising determining a relative proximity of an object to the light sensor circuit based on a frequency of the spike signal.

15. The method of any one of the claims 8 to 14, further comprising controlling a robotics element based on a frequency of the spike signal.

16. A system comprising a processor and two or more light sensor circuits according to claim 1 , wherein the processor is configured to: obtain temporal information of a plurality of spike signals, from the light sensor circuits, in response to changes in a light signal caused by a motion of an object; and determine an indication of the motion of the object based on the temporal information of the spike signals.

17. The system of claim 16, wherein the light sensor circuits are arranged in an array of light sensor circuits.

18. The system of claim 17, wherein the array of light sensor circuits comprises at least four light sensor circuits and the processor is further configured to determine the indication of the motion of the object that comprises an azimuth or a rotation.

19. The system of any one of claims 16 to 18, wherein the temporal information comprises a time difference of the plurality of spike signals from different light sensor circuits, and the indication of the motion of the object comprises an indication of a direction of movement.

20. The system of any one of claims 16 to 19, wherein the object is a hand making a gesture and the processor is further configured to determine a class of gesture based on the indication of the motion of the object.

Images