An automated ai-based electroluminescence testing equipment for microleds to test for luminosity and color accuracy

Abstract

The present disclosure provides an automated electroluminescence testing equipment of LED substrates system in which an automated electroluminescence testing equipment inspection module is used to determine quality of incoming processing material, and an automated electroluminescence testing equipment process module is used to process the processing material, an automated electroluminescence testing equipment post-process module is used to inspect the resultant process material for acceptability, and an automated electroluminescence testing equipment historical module is used to analyze the post process results of the most recent process, against all historical data to find enhanced process changes in the automated electroluminescence testing equipment process module, and an automated electroluminescence testing equipment integration module to update the automated electroluminescence testing equipment process module based upon the results of the automated electroluminescence testing equipment historical module.

Description

[0001] AN AUTOMATED AI-BASED ELECTROLUMINESCENCE

[0002] TESTING EQUIPMENT FOR MICROLEDS TO TEST FOR LUMINOSITY AND COLOR ACCURACY

[0003] BACKGROUND AND FIELD OF THE INVENTION

[0004] [1] The present disclosure is generally related to an automated Al-based electroluminescence testing equipment for micro LEDs to test for luminosity and color accuracy

[0005] [2] Currently, the sensitivity and resolution limitations of photoluminescence testing pose a critical challenge in detecting defects in micro LEDs before electroluminescence testing. Defects detected only during EL testing or final product use lead to increased costs and compromised product quality. Addressing this challenge requires enhancing defect detection sensitivity and resolution in PL testing methodologies to ensure thorough defect identification at earlier stages of micro LED production. Also, sample preparation and configuration of testing parameters during electroluminescence testing in micro LED manufacturing present significant challenges. Contamination risks during sample handling and complexities in configuring testing parameters tailored to diverse micro LED configurations impede the seamless progression of micro LEDs through the testing phase. Achieving uniform electrical application during electroluminescence induction further impacts the accuracy and reliability of testing results. Lastly, postelectroluminescence testing stages in micro LED manufacturing encounter substantial challenges in defect detection and repair processes. The demand for exceedingly high precision in defect detection due to the micro-scale nature of LEDs poses significant technical hurdles. Moreover, repair processes such as ultraviolet irradiation and laser welding introduce complexities and risks of further damage to neighboring good dies, hindering the achievement of optimal yield improvement. Thus, there is a need in the prior art for an automated Al-based electroluminescence testing equipment for micro LEDs to test for luminosity and color accuracy. SUMMARY

[0006] [3] The present invention relates to an automated electroluminescence testing equipment method for micro LED chips system, comprising, an automated electroluminescence testing equipment inspection module that measures at least one parameter from a first group of parameters and calculates an output of at least one parameter from a second group of parameters to help control the automated electroluminescence testing equipment processing module, an automated electroluminescence testing equipment processing module which inputs the at least one parameter from said second group to control the electroluminescence testing equipment using at least one parameter from a third group of parameters, an automated electroluminescence testing equipment post-processing module which measures at least one parameter from a fourth group of parameters that determines how well the parameters of said first and second group were running, an automated electroluminescence testing equipment historical module which calculates a plurality of group parameters based upon parameters of the first group, second group, third group, and fourth group; and an automated electroluminescence testing equipment integration module to update the automated electroluminescence testing equipment processing module based upon the results of the automated electroluminescence testing equipment historical module.

[0007] DESCRIPTIONS OF THE DRAWINGS

[0008] [4] FIG. 1 : Illustrates an automated Al-based electroluminescence testing equipment for micro LEDs to test for luminosity and color accuracy, according to an embodiment.

[0009] [5] FIG. 2: Illustrates a Base Module, according to an embodiment.

[0010] [6] FIG. 3 : Illustrates an ET Inspection Module, according to an embodiment.

[0011] [7] FIG. 4: Illustrates an ET Process Module, according to an embodiment.

[0012] [8] FIG. 5: Illustrates an ET Post-Process Module, according to an embodiment.

[0013] [9] FIG. 6: Illustrates an ET Historical Module, according to an embodiment.

[0014]

[0010] FIG. 7: Illustrates an ET Integration Module, according to an embodiment.

[0015]

[0011] FIG. 8: Illustrates an ET Inspection Database, according to an embodiment.

[0016]

[0012] FIG. 9: Illustrates an ET Process Database, according to an embodiment.

[0017]

[0013] FIG. 10: Illustrates an ET Post-Process Database, according to an embodiment.

[0018] DETAILED DESCRIPTION

[0019]

[0014] Embodiments of the present disclosure will be described more fully hereinafter with reference to the accompanying drawings in which like numerals represent like elements throughout the several figures, and in which example embodiments are shown. Embodiments of the claims may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. The examples set forth herein are non-limiting examples and are merely examples among other possible examples.

[0020]

[0015] FIG. 1 illustrates a system for an automated Al-based electroluminescence testing equipment for micro LEDs to test for luminosity and color accuracy. This system comprises of a 3rd party electroluminescence testing Al network 102 which integrates pre-quality inspection, processing, post-processing, historical analysis, and continuous improvement in an electroluminescence testing system. The ET inspection module 110 measures essential parameters and calculates outputs to guide the subsequent electroluminescence testing process. The ET process module 112 acts as the core controller, utilizing inputs from the ET inspection module 110 to govern the electroluminescence testing process to ensure optimal exposure dose, alignment accuracy, development time, luminous intensity, etc. The ET post-process module 114 conducts inspections to assess the execution of parameters and identify potential deviations. The ET historical module 116 analyzes data collected across various stages, allowing for continuous learning and pattern recognition. The ET integration module 118 then employs historical insights to update and optimize the process, enabling adaptive improvements over time.

[0021]

[0016] Further, embodiments may include a communication interface 104, which may be a hardware or software component that enables communication between the 3rd party electroluminescence testing Al network 102 and the electroluminescence testing pre-process 128, electroluminescence testing process 134, electroluminescence testing post-process 150. The communication interface 104 may include a set of protocols, rules, and standards that define how information is transmitted and received between the devices. The communication interface 104 may be a physical connector, wireless network, or software application and may include components such as drivers, software libraries, and firmware that may be used to control and manage the communication process. In some embodiments, the communication interface 104 may be compatible with USB, Bluetooth, or Wi-Fi. The communication interface 104 may communicate with a network. Examples of networks may include but are not limited to, the Internet, a cloud network, a Wireless Fidelity (Wi-Fi) network, a Wireless Local Area Network (WLAN), a Local Area Network (LAN), a telephone line (POTS), Long Term Evolution (LTE), and / or a Metropolitan Area Network (MAN).

[0022]

[0017] Further, embodiments may include a memory 106, which may include suitable logic, circuitry, and / or interfaces that may be configured to store a machine code and / or a computer program with at least one code section executable by a processor. Examples of implementation of the memory 106 may include, but are not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Hard Disk Drive (HDD), and / or a Secure Digital (SD) card.

[0023]

[0018] Further, embodiments may include a base module 108, which initiates the ET inspection module 110, the ET process module 112, the ET post-process module 114, the ET historical module 116, and the ET integration module 118.

[0024]

[0019] Further, embodiments may include an ET inspection module 110, which begins by being initiated by the base module 108. The ET inspection module 110 connects to the electroluminescence testing pre-process 130. The ET inspection module 110 collects the data from the electroluminescence testing pre-process 130. The ET inspection module 110 performs the inspection algorithm. The ET inspection module 110 sends the determined process data from the inspection algorithm to the ET process module 112. The ET inspection module 110 stores the data in the ET inspection database 120. The ET inspection module 110 returns to the base module 108.

[0025]

[0020] Further, embodiments may include an ET process module 112, which begins by being initiated by the base module 108. The ET process module 112 receives the process data, such as the data on the micro LED processed wafer, from the ET inspection module 110. The ET process module 112 connects to the electroluminescence testing process 140. The ET process module 112 collects the data from the electroluminescence testing process 140. The ET process module 112 performs the control algorithm. The ET process module 112 adjusts the control parameters of the electroluminescence testing process 140. The ET process module 112 executes the electroluminescence testing process. The ET process module 112 stores the data in the ET process database 122. The ET process module 112 returns to the base module 108.

[0021] Further, embodiments may include an ET post-process module 114, which begins by being initiated by the base module 108. The ET post-process module 114 connects to the electroluminescence testing post-process 150. The ET post-process module 114 collects the postprocess data. The ET post-process module 114 extracts the data from the ET inspection database 120 and ET process database 122. The ET post-process module 114 performs the post-process algorithm. The ET post-process module 114 performs the post-process inspection. The ET postprocess module 114 stores the data in the ET post-process database 124. The ET post-process module 114 returns to the base module 108.

[0026]

[0022] Further, embodiments may include an ET historical module 116, which begins by being initiated by the base module 108. The ET historical module 116 connects to the electroluminescence testing pre-process 130, the electroluminescence testing process 140, and the electroluminescence testing post-process 150. The ET historical module 116 aggregates the data from the various types of processes and stores the data in the ET network database 126. The ET historical module 116 performs the historical machine learning algorithm on the historical data stored in the ET network database 126. The ET historical module 116 sends the process adjustments to the ET inspection module 110, the ET process module 112, and the ET post-process module 114. The ET historical module 116 returns to the base module 108.

[0027]

[0023] Further, embodiments may include an ET integration module 118, which begins by being initiated by the base module 108. The ET integration module 118 performs the integration machine learning algorithm. The ET integration module 118 connects to the electroluminescence testing pre-process Al module 136, the electroluminescence testing process Al module 146, and the electroluminescence testing post-process Al module 156. The ET integration module 118 sends the process adjustments to the electroluminescence testing pre-process Al module 136, the electroluminescence testing process Al module 146, and the electroluminescence testing postprocess Al module 156. The ET integration module 118 returns to the base module 108.

[0028]

[0024] Further, embodiments may include an ET inspection database 120, which provides an example of the results of the inspection algorithm performed in the ET inspection module 110. The ET inspection database 120 may contain the wafer ID, the initial throughput percentage, the initial temperature stability, the adjusted throughput percentage, the temperature stability optimized by the inspection algorithm, the throughput percentage optimized by the inspection algorithm, etc. For example, the ET inspection database 120 provides an example of the inspection algorithm optimizing the temperature to optimize the throughput percentage. The inspection algorithm may achieve a balance between high throughput and the accuracy of defect detection. High throughput is essential for cost-effective manufacturing, but it may compromise the detection accuracy which may allow defective units to proceed to later stages of production and may result in increased costs if defects are detected only at the EL testing phase or beyond. The first example illustrates a "Temperature Stability (degrees C)" column which displays deviations from the optimal temperature condition (0 degrees C), for normal process variations such as temperature that may occur in EL Testing, the positive values indicate a temperature above the optimal and the negative values indicating below optimal. The "Adjusted Thruput for EL Testing (%)" column reflects the throughput adjustment based on the temperature stability, where each degree of deviation adjusts the throughput by approximately -2 percent. This example illustrates how temperature variations can impact the efficiency of the EL testing process, to maintain as close to the optimal temperature as possible to ensure high throughput. The second example reflects how the inspection algorithm may reduce temperature stability deviations and improve throughput results for EL testing. The example data in the inspection database 120 illustrates approximately a 50 percent reduction, or improvement, in temperature stability deviations due to the inspection algorithm optimizations with the improved optimized throughput percentage. For example, the inspection algorithm may reduce temperature stability deviations and potentially improve throughput results for EL testing. The inspection algorithm may predict optimal EL testing parameters, such as temperature settings, that maximize throughput while ensuring high-quality testing outcomes. The inspection algorithm may use a linear regression model to predict optimal testing parameters based on historical data and focus on achieving the best temperature conditions and throughput rates. The inspection algorithm may use historical data, such as data on wafer characteristics, including material composition, layer thickness, and, EL testing parameters, including temperature settings, electrical input, etc., observed temperature stability deviations, resulting throughput rates, etc. The inspection algorithm may identify features, or variables, that impact throughput, such as initial temperature settings, electrical input, material properties, historical temperature deviations, etc. The inspection algorithm is trained on the historical data to learn the relationship between the selected features and the throughput rate. The historical data may be formatted, normalized, standardized, etc., and is split into training and testing sets. The inspection algorithm is trained using the input of the selected features from the historical data and the output of the algorithm is the throughput rate. The algorithm is then evaluated in which the algorithm’s performance is assessed using the testing set and metrics like R-squared, and R2, to determine how well the model predicts throughput rates based on the input features. Then before the EL test, the current wafer characteristics and initial testing parameters are inputted into the algorithm and the algorithm predicts the adjustments needed to optimize the throughput, such as suggesting the ideal temperature settings. In some embodiments, thermal imaging technology may be used to capture and analyze the infrared radiation emitted by the wafer to visualize its temperature distribution. The testing method may include preparation for thermal imaging through wafer cleaning and setting up the environment, using infrared cameras and imaging software for analysis, conducting the test by collecting a baseline measurement, performing powered state imaging, and capturing the thermal images. The data is analyzed, such as identifying the temperature variations, mapping the hot and cold spots, and correlating the mapping to the wafer features. Then the EL testing parameters may be adjusted, such as altering the electrical input, adjusting the testing environment's temperature, or implementing localized cooling or heating solutions.

[0029]

[0025] Further, embodiments may include an ET process database 122, which provides an example of the results of the control algorithm performed in the ET process module 112. The ET process database 122 may contain the substrate ID, initial alignment, final alignment after the optimization adjustments by the control algorithm, etc. For example, the ET process database 122 illustrates the alignment specifications for ten different substrates in a micro LED manufacturing scenario, where precise alignment is needed for accurate testing and overall device performance. The Initial Alignment column may represent the starting alignment values, with the specification of 0.1 pm as a reference point. The column illustrates an example of variances encountered in manufacturing and handling which is displayed by the substrate's initial alignment value being altered throughout the process. The Final Alignment After Al-based Adjustments column reflects the alignment values after making necessary corrections to counter initial deviations outputted by the control algorithm. These adjustments ensure that substrates are aligned with the requisite precision for micro LED testing, in which misalignments may impact test accuracy and potentially cause damage to the micro LEDs themselves. For example, the control algorithm may optimize alignment based on a dataset of previous alignments, outcomes, and various parameters affecting alignment quality to prevent misalignment which may lead to inaccurate test results or damage to the micro LEDs. For example, the alignment process may include sample preparation, such as cleaning and inspection, securing the micro LED, utilizing alignment equipment, such as high-resolution microscopes, optical alignment systems, vision systems, fiducial markers, detection systems, etc. to initially position the micro LED sample based on visual markers or reference points. The micro LED sample’s position may be fine-tuned through micromanipulators, piezoelectric positioners, nanopositioning stages, laser interferometer systems, alignment stages with fine adjustment screws, etc. Once the sample is aligned properly it is securely attached to the mounting platform or holder, which may involve vacuum sealing, adhesives, etc., and the attachment is verified. The alignment is re-checked using the alignment equipment and the electrical connections are verified. The control algorithm may analyze visual imagery by being trained on a dataset comprising images of aligned and misaligned micro LED substrates, along with corresponding alignment parameters and outcomes allowing the control algorithm to recognize patterns and features associated with optimal alignment. For example, the control algorithm may compile a dataset of high-resolution images of micro LED substrates at various stages of the alignment process, including pre-alignment, postmanual alignment, and post-AI alignment stages and each image may be tagged with metadata detailing the alignment parameters used and the success rate of the alignment. The control algorithm may be trained on the dataset which may involve feeding the images into the algorithm, which may learn to identify features and patterns correlating with successful alignments. The output of the control algorithm may be to predict the optimal alignment parameters. During the micro LED manufacturing process, live images of the substrate alignments are fed into the trained control algorithm, and it may assess the images in real time and predict the most effective adjustment parameters to achieve optimal alignment. The predicted parameters from the control algorithm may be used to automatically adjust the alignment tools, which may involve adjusting the position of the substrate, modifying the angle of incidence for alignment lasers, or altering the pressure applied by mechanical aligners. The outcomes of these adjustments may be continuously monitored and fed back into the control algorithm which allows it to refine its predictions over time. The control algorithm may process complex image data, recognize patterns, and learn from the visual cues associated with successful and unsuccessful alignments. The control algorithm may influence the control mechanism of the alignment tools by predicting the necessary adjustments and sending signals to the mechanical or optical components responsible for making precise movements, which may include motorized stages for lateral adjustments, tilt platforms for angular corrections, or laser steering systems for precise positioning.

[0030]

[0026] Further, embodiments may include an ET post-process database 124, which provides an example of the results of the post-process algorithm performed in the ET post-process module 114. The example data illustrates the detection of small-scale defects in various micro LED substrates, highlighting the defect type(s), the substrate's compliance with final specifications, and the applicability of Demura fix capabilities using the post-process algorithm learning on historical data. In some embodiments, the defect key for the second column may be 1. Dead Pixels, 2. Micro LED Shorts, 3. Non-Uniform Brightness (Mura), 4. Color Shifts, 5. Crosstalk, 6. Efficiency Droop, 7. Crystalline Defects, 8. Point Defects. The examples provide an overview of the presence and type of small-scale defects detected in a sample of micro LED substrates, assessing their compliance with final product specifications. The defects, which range from micro-crack formation to thin film peeling, illustrate the importance of inspection processes in the production of micro LEDs. The example also evaluates whether Demura's fix capabilities, leveraging the post-process algorithm learning from historical data, may address non-conformities. For example, the post-process algorithm may be used to detect small-scale defects, such as dead pixels, micro LED shorts, non- uniform brightness or mura, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. The post-process algorithm may use high-resolution images of the micro LEDs that capture the micro LEDs in their operational state, allowing for observation of luminance, color uniformity, and other characteristics that would indicate the presence or absence of defects. The post-process algorithm may use a historical dataset of high-resolution images of the micro LEDs that includes defect-free examples and defective examples. The dataset may be used to train the post-process algorithm, such as a convolutional neural network, to recognize patterns and features associated with defects and may involve feeding the images into the algorithm, which learns to differentiate between normal and defective micro LED characteristics. The post-process algorithm may then be validated and tested on a separate set of images not used during training to ensure it accurately identifies defects and the algorithm may be adjusted to improve accuracy and reduce false positives / negatives. Then resolution images of the micro LEDs from the electroluminescence testing post-process 150 are inputted into the post-process algorithm and each image is analyzed to identify features that indicate defects based on what the algorithm learned during training. The post-process algorithm may output the locations and types of defects detected and provide detailed information for further analysis or repair, which may include quantifying defect sizes, categorizing defect types, and determining defect distribution across the micro LED array. In some embodiments, the post-process algorithm may improve the accuracy, efficiency, scalability, etc. of the defect inspection post-process. In some embodiments, the post-process algorithm may utilize supervised learning models, such as decision trees, random forests, gradient boosting machines, etc. which may be trained on historical data where the input features include types and numbers of defects, and the target variable is whether the substrate meets specifications and may learn complex patterns and interactions among defects. In some embodiments, the post-process algorithm may utilize anomaly detection algorithms which may identify unusual patterns that do not conform to expected behavior, such as detecting unusual combinations or intensities of defects that are likely to result in a substrate failing to meet specifications. In some embodiments, the post-process algorithm may utilize regression analysis or time-series forecasting to predict the future occurrence of defects based on trends and patterns, which may preemptively adjust manufacturing processes to reduce the likelihood of producing substrates that do not meet specifications. In some embodiments, the ET post-process database 124 may store other parameter data collected or calculated from the ET post-process module 114, such as dead pixels, micro LED shorts, non- uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc.

[0027] Further, embodiments may include an ET network database 126, which may contain the historical data from the various processes performed by the electroluminescence testing pre- process 130, electroluminescence testing process 140, and electroluminescence testing postprocess 150. The ET network database 126 may contain the data parameters collected during the inspection process, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc., control parameters of the process, such as exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc., and data parameters collected during post-processing, such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc.

[0028] Further, embodiments may include a cloud 128, which may be a network of remote servers that provide on-demand computing resources and services over the internet. The Cloud 128 may consist of a collection of servers, storage devices, and networking equipment. Users may access the cloud 128 through a variety of devices, such as computers, smartphones, and tablets, using internet connectivity. In some embodiments, the architecture of the cloud 128 may be based on a distributed computing model, with multiple servers working together to provide services to users.

[0031]

[0029] Further, embodiments may include electroluminescence testing pre-process 130, which may be the processes, such as photoluminescence testing, visual inspection, mechanical and bonding checks, etc. utilized before the electroluminescence testing process 140, designed to prepare the micro LED processed wafers for optimal performance during the electroluminescence process. The pre-process may involve collecting a plurality of parameter data including wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. In some embodiments, the electroluminescence testing pre-process 130 may send and receive data from the 3rd party electroluminescence testing Al network 102, including parameter adjustments, control adjustments, material or workpiece data, decisions on material inspections, inputs to be used by the electroluminescence testing pre-process Al module 136, etc.

[0032]

[0030] Further, embodiments may include a communication interface 132, which may be a hardware or software component that enables communication between the electroluminescence testing pre-process 130 and the 3rd party electroluminescence testing Al network 102. In some embodiments, the electroluminescence testing pre-process 130 may communicate with the electroluminescence testing process 140, and electroluminescence testing post-process 150. The communication interface 132 may include a set of protocols, rules, and standards that define how information is transmitted and received between the devices. The communication interface 132 may be a physical connector, wireless network, or software application and may include components such as drivers, software libraries, and firmware that may be used to control and manage the communication process. In some embodiments, the communication interface 132 may be compatible with USB, Bluetooth, or Wi-Fi. The communication interface 132 may communicate with a network. Examples of networks may include but are not limited to, the Internet, a cloud network, a Wireless Fidelity (Wi-Fi) network, a Wireless Local Area Network (WLAN), a Local Area Network (LAN), a telephone line (POTS), Long Term Evolution (LTE), and / or a Metropolitan Area Network (MAN).

[0033]

[0031] Further, embodiments may include a memory 134 may include suitable logic, circuitry, and / or interfaces that may be configured to store a machine code and / or a computer program with at least one code section executable by a processor. Examples of implementation of the memory 134 may include, but are not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Hard Disk Drive (HDD), and / or a Secure Digital (SD) card.

[0034]

[0032] Further, embodiments may include an electroluminescence testing pre-process Al module 136 in which a predictive model may be performed to predict the final product of the electroluminescence testing process based upon the data collected from the electroluminescence testing pre-process 130, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. Predictive modeling involves the systematic analysis of historical data to identify patterns, correlations, and trends that can be used to build models capable of making accurate predictions about future outcomes. In some embodiments, the electroluminescence testing pre-process Al module 136 may receive a predictive model from the ET integration module 118. In some embodiments, the electroluminescence testing pre-process Al module 136 may receive and send data to the ET inspection module 110. In some embodiments, the data received from the ET inspection module 110 may be inputted into the predictive model to determine if any parameters of the pre-process should be adjusted, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc.

[0035]

[0033] Further, embodiments may include an electroluminescence testing pre-process database 138 which may include data parameters collected from the electroluminescence testing pre-process 130, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. In some embodiments, the data stored in the electroluminescence testing pre-process database 138 may be sent to the ET inspection module 110 and / or the ET historical module 116.

[0036]

[0034] Further, embodiments may include electroluminescence testing process 140, which may be systems or equipment for automated electroluminescence testing of micro LEDs, including sample preparation, configuration, electroluminescence induction, data capture, analysis, reporting, decision making, etc., in which a controlled electrical current is applied to the micro LEDs, causing them to emit light through the electroluminescence phenomenon to assess the operational characteristics of the micro LEDs, such as their efficiency and uniformity. The electroluminescence testing process 140 may involve collecting parameter data, such as exposure dose, alignment accuracy, development time, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. The electroluminescence testing process 140 may include characterizing the optical and electrical performances of the devices. The micro LED samples are prepared by being mounted onto a test fixture or substrate to ensure proper electrical contact with the device's electrodes. Following sample preparation, the testing setup is established, comprising a precision current or voltage source for applying electrical signals to the micro LED and measurement instruments such as multimeters or source measure units for control and monitoring of electrical parameters during testing. The EL testing process begins with the application of a forward bias voltage to the micro LED, surpassing the device's turn-on voltage to induce electroluminescence. This voltage is selected to ensure it falls within the safe operating limits of the device and minimizes the risk of damage. Concurrently, electroluminescence emitted by the micro LED is captured using specialized imaging equipment, such as CCD cameras, photomultiplier tubes, etc., enabling the visualization and analysis of the emitted light pattern, intensity distribution, and spatial uniformity across the device surface. The emitted light spectrum is measured using spectroradiometers or spectrometers to analyze the wavelength distribution and color characteristics of the emitted light accurately. The spectral analysis provides critical insights into the color accuracy and consistency of the micro LED's emission. Additionally, electrical parameters such as forward voltage, forward current, and reverse leakage current may be monitored and recorded during the testing process to assess the device's electrical performance, efficiency, and reliability under operation. The collected data, including electroluminescence images, spectral measurements, and electrical parameters, may be subjected to analysis to evaluate the micro LED's performance and quality. Any deviations from desired specifications or performance metrics may be identified and analyzed to determine root causes and potential areas for improvement. The results of the EL testing process are compiled into a comprehensive report detailing the micro LED's characteristics, providing information for quality control, process optimization, and product validation in micro LED manufacturing. In some embodiments, the electroluminescence testing process 140 may send and receive data from the 3rd party electroluminescence testing Al network 102, including parameter adjustments, control adjustments, inputs to be used by the electroluminescence testing process Al module 146, etc.

[0037]

[0035] Further, embodiments may include a communication interface 142, which may be a hardware or software component that enables communication between the electroluminescence testing process 140 and the 3rd party electroluminescence testing Al network 102. In some embodiments, the electroluminescence testing process 140 may communicate with the electroluminescence testing pre-process 130, and electroluminescence testing post-process 150. The communication interface 142 may include a set of protocols, rules, and standards that define how information is transmitted and received between the devices. The communication interface 142 may be a physical connector, wireless network, or software application and may include components such as drivers, software libraries, and firmware that may be used to control and manage the communication process. In some embodiments, the communication interface 142 may be compatible with USB, Bluetooth, or Wi-Fi. The communication interface 142 may communicate with a network. Examples of networks may include but are not limited to, the Internet, a cloud network, a Wireless Fidelity (Wi-Fi) network, a Wireless Local Area Network (WLAN), a Local Area Network (LAN), a telephone line (POTS), Long Term Evolution (LTE), and / or a Metropolitan Area Network (MAN).

[0038]

[0036] Further, embodiments may include a memory 144 may include suitable logic, circuitry, and / or interfaces that may be configured to store a machine code and / or a computer program with at least one code section executable by a processor. Examples of implementation of the memory 144 may include, but are not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Hard Disk Drive (HDD), and / or a Secure Digital (SD) card.

[0039]

[0037] Further, embodiments may include an electroluminescence testing process Al module 146 in which a predictive model may be performed to predict the final product of the electroluminescence testing process based upon the data collected from the electroluminescence testing process 140, such as exposure dose, alignment accuracy, development time, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. Predictive modeling involves the systematic analysis of historical data to identify patterns, correlations, and trends that can be used to build models capable of making accurate predictions about future outcomes In some embodiments, the electroluminescence testing process Al module 146 may receive a predictive model from the ET integration module 118. In some embodiments, the electroluminescence testing process Al module 146 may receive and send data to the ET process module 112. In some embodiments, the data received from the ET process module 112 may be inputted into the predictive model to determine if any parameters of the electroluminescence testing process 140 should be adjusted, such as exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc.

[0040]

[0038] Further, embodiments may include an electroluminescence testing process database 148 which may include data parameters collected from the electroluminescence testing process 140, such as exposure dose, alignment accuracy, development time, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. In some embodiments, the data stored in the electroluminescence testing process database 148 may be sent to the ET process module 112 and / or the ET historical module 116.

[0041]

[0039] Further, embodiments may include electroluminescence testing post-process 150, which may be processes or methods for inspection to identify defective LEDs and subsequent repair or rework processes, which may include camera-based microscopic imaging systems to assess and measure the luminance of micro LED chips accurately to enhance production yield by efficiently repairing or replacing bad dies, employing methods such as ultraviolet irradiation, laser welding, and selective pick or laser repairing technologies. The post-process may involve parameters such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. In some embodiments, the electroluminescence testing postprocess 150 may send and receive data from the 3rd party electroluminescence testing Al network 102, including parameter adjustments, control adjustments, inputs to be used by the electroluminescence testing post-process Al module 156, etc.

[0040] Further, embodiments may include a communication interface 152, which may be a hardware or software component that enables communication between the electroluminescence testing post-process 150 and the 3rd party electroluminescence testing Al network 102. In some embodiments, the electroluminescence testing post-process 150 may communicate with the electroluminescence testing pre-process 130, and electroluminescence testing process 140. The communication interface 152 may include a set of protocols, rules, and standards that define how information is transmitted and received between the devices. The communication interface 152 may be a physical connector, wireless network, or software application and may include components such as drivers, software libraries, and firmware that may be used to control and manage the communication process. In some embodiments, the communication interface 152 may be compatible with USB, Bluetooth, or Wi-Fi. The communication interface 152 may communicate with a network. Examples of networks may include but are not limited to, the Internet, a cloud network, a Wireless Fidelity (Wi-Fi) network, a Wireless Local Area Network (WLAN), a Local Area Network (LAN), a telephone line (POTS), Long Term Evolution (LTE), and / or a Metropolitan Area Network (MAN).

[0042]

[0041] Further, embodiments may include a memory 154 may include suitable logic, circuitry, and / or interfaces that may be configured to store a machine code and / or a computer program with at least one code section executable by a processor. Examples of implementation of the memory 154 may include, but are not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Hard Disk Drive (HDD), and / or a Secure Digital (SD) card.

[0043]

[0042] Further, embodiments may include an electroluminescence testing post-process Al module 156 in which a predictive model may be performed to optimize the future final products based upon the data collected from the electroluminescence testing post-process 150, such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. Predictive modeling involves the systematic analysis of historical data to identify patterns, correlations, and trends that can be used to build models capable of making accurate predictions about future outcomes In some embodiments, the electroluminescence testing post-process Al module 156 may receive a predictive model from the ET integration module 118. In some embodiments, the electroluminescence testing post-process Al module 156 may receive and send data to the ET post-process module 114. In some embodiments, the data received from the ET post-process module 114 may be inputted into the predictive model to determine if any parameters of the electroluminescence testing post-process 150 should be adjusted, such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc.

[0044]

[0043] Further, embodiments may include an electroluminescence testing post-process database 158 which may include data parameters collected from the electroluminescence testing postprocess 150, such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. In some embodiments, the data stored in the electroluminescence testing post-process database 158 may be sent to the ET post-process module 114 and / or the ET historical module 116.

[0045]

[0044] FIG. 2 illustrates the base module 108. The process begins with the base module 108 initiating, at step 200, the ET inspection module 110. For example, the ET inspection module 110 begins by being initiated by the base module 108. The ET inspection module 110 connects to the electroluminescence testing pre-process 130. The ET inspection module 110 connects with the electroluminescence testing pre-process 130, such as photoluminescence testing, visual inspection, mechanical and bonding checks, etc. utilized before the electroluminescence testing process 140, designed to prepare the micro LED processed wafers for optimal performance during the electroluminescence testing process. In some embodiments, the ET inspection module 110 may transmit and receive data from the electroluminescence testing pre-process 130. In some embodiments, the electroluminescence testing pre-process 130 may measure wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. In some embodiments, the ET inspection module 110 may control or send inputs to control the electroluminescence testing pre-process 130. The ET inspection module 110 collects the data from the electroluminescence testing pre-process 130. The ET inspection module 110 collects the preprocessing data, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. For example, photoluminescence testing may evaluate the optical properties of micro LEDs before applying any electrical current and may identify defects or variations in luminescence efficiency across the wafer or panel without causing damage to the micro LEDs. Visual inspections may be conducted to check for obvious physical defects or alignment issues post-mass transfer which may include checking for missing LEDs, misalignment, or other visible defects that may impact the performance of the micro LED display. Mechanical and bonding checks may be performed to ensure that the micro LED chips are securely attached to the substrate before moving on to EL testing, which may involve mechanical stress tests or specific inspections to assess the integrity of the bond between the micro LED chips and the display panel. The ET inspection module 110 performs the inspection algorithm. For example, the inspection algorithm may achieve a balance between high throughput and the accuracy of defect detection. High throughput is essential for cost-effective manufacturing, but it may compromise the detection accuracy which may allow defective units to proceed to later stages of production and may result in increased costs if defects are detected only at the EL testing phase or beyond. For example, the inspection algorithm may reduce temperature stability deviations and potentially improve throughput results for EL testing. The inspection algorithm may predict optimal EL testing parameters, such as temperature settings, that maximize throughput while ensuring high-quality testing outcomes. The inspection algorithm may use a linear regression model to predict optimal testing parameters based on historical data and focus on achieving the best temperature conditions and throughput rates. The inspection algorithm may use historical data, such as data on wafer characteristics, including material composition, layer thickness, and, EL testing parameters, including temperature settings, electrical input, etc., observed temperature stability deviations, resulting throughput rates, etc. The inspection algorithm may identify features, or variables, that impact throughput, such as initial temperature settings, electrical input, material properties, historical temperature deviations, etc. The inspection algorithm is trained on the historical data to learn the relationship between the selected features and the throughput rate. The historical data may be formatted, normalized, standardized, etc., and is split into training and testing sets. The inspection algorithm is trained using the input of the selected features from the historical data and the output of the algorithm is the throughput rate. The algorithm is then evaluated in which the algorithm’s performance is assessed using the testing set and metrics like R-squared, and R2, to determine how well the model predicts throughput rates based on the input features. Then before the EL test, the current wafer characteristics and initial testing parameters are inputted into the algorithm and the algorithm predicts the adjustments needed to optimize the throughput, such as suggesting the ideal temperature settings. In some embodiments, the inspection algorithm may utilize decision trees and random forests to categorize wafers into different quality levels or predict throughput rates based on a wide range of testing parameters and wafer characteristics. In some embodiments, the inspection algorithm may utilize support vector machines to predict optimal testing parameters for achieving desired throughput rates. In some embodiments, the inspection algorithm may utilize reinforcement learning which may dynamically adjust EL testing parameters in real-time and learn the optimal testing strategy through continuous interaction with the testing environment to maximize throughput and yield. The ET inspection module 110 sends the determined process data from the inspection algorithm to the ET process module 112. For example, the ET inspection module 110 may send the optimal temperature settings to the ET process module 112 to optimize the electroluminescence process. In some embodiments, the ET inspection module 110 may send wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. The ET inspection module 110 stores the data in the ET inspection database 120. The ET inspection module 110 stores the data outputted from the inspection algorithm in the ET inspection database 120, such as the wafer ID, the initial throughput percentage, the initial temperature stability, the adjusted throughput percentage, the temperature stability optimized by the inspection algorithm, the throughput percentage optimized by the inspection algorithm, etc. In some embodiments, the ET inspection database 120 may include other parameter data collected or calculated during the ET inspection module 110 process, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. The ET inspection module 110 returns to the base module 108. The base module 108 initiates, at step 202, the ET process module 112. For example, the ET process module 112 begins by being initiated by the base module 108. The ET process module 112 receives the pre-process data, such as the data on the micro LED processed wafer, from the ET inspection module 110. The ET process module 112 receives the outputted data from the inspection algorithm performed in the ET inspection module 110, such as the optimized temperature, allowing the electroluminescence testing process 140 to determine the optimal control parameters for the electroluminescence testing process. In some embodiments, the ET process module 112 may receive wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. The ET process module 112 connects to the electroluminescence testing process 140. The ET process module 112 connects with the electroluminescence process 140, such as systems or equipment for automated electroluminescence testing of micro LEDs, including sample preparation, configuration, electroluminescence induction, data capture, analysis, reporting, decision making, etc., in which a controlled electrical current is applied to the micro LEDs, causing them to emit light through the electroluminescence phenomenon to assess the operational characteristics of the micro LEDs, such as their efficiency and uniformity. In some embodiments, the ET process module 112 may transmit and receive data from the electroluminescence testing process 140. In some embodiments, the ET process module 112 may control or send inputs to control the electroluminescence testing process 140, such as controlling parameters of the electroluminescence testing process, for example, exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. The ET process module 112 collects the data from the electroluminescence testing process 140. The ET process module 112 collects the processing data, such as exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. The ET process module 112 performs the control algorithm. For example, the control algorithm may optimize alignment based on a dataset of previous alignments, outcomes, and various parameters affecting alignment quality to prevent misalignment which may lead to inaccurate test results or damage to the micro LEDs. For example, the alignment process may include sample preparation, such as cleaning and inspection, securing the micro LED, utilizing alignment equipment, such as high-resolution microscopes, optical alignment systems, vision systems, fiducial markers, detection systems, etc. to initially position the micro LED sample based on visual markers or reference points. The micro LED sample’s position may be fine-tuned through micromanipulators, piezoelectric positioners, nanopositioning stages, laser interferometer systems, alignment stages with fine adjustment screws, etc. Once the sample is aligned properly it is securely attached to the mounting platform or holder, which may involve vacuum sealing, adhesives, etc., and the attachment is verified. The alignment is re-checked using the alignment equipment and the electrical connections are verified. The control algorithm may analyze visual imagery by being trained on a dataset comprising images of aligned and misaligned micro LED substrates, along with corresponding alignment parameters and outcomes allowing the control algorithm to recognize patterns and features associated with optimal alignment. For example, the control algorithm may compile a dataset of high-resolution images of micro LED substrates at various stages of the alignment process, including pre-alignment, post-manual alignment, and post-AI alignment stages and each image may be tagged with metadata detailing the alignment parameters used and the success rate of the alignment. The control algorithm may be trained on the dataset which may involve feeding the images into the algorithm, which may learn to identify features and patterns correlating with successful alignments. The output of the control algorithm may be to predict the optimal alignment parameters. During the micro LED manufacturing process, live images of the substrate alignments are fed into the trained control algorithm, and it may assess the images in real time and predict the most effective adjustment parameters to achieve optimal alignment. The predicted parameters from the control algorithm may be used to automatically adjust the alignment tools, which may involve adjusting the position of the substrate, modifying the angle of incidence for alignment lasers, or altering the pressure applied by mechanical aligners. The outcomes of these adjustments may be continuously monitored and fed back into the control algorithm which allows it to refine its predictions over time. The control algorithm may process complex image data, recognize patterns, and learn from the visual cues associated with successful and unsuccessful alignments. The control algorithm may influence the control mechanism of the alignment tools by predicting the necessary adjustments and sending signals to the mechanical or optical components responsible for making precise movements, which may include motorized stages for lateral adjustments, tilt platforms for angular corrections, or laser steering systems for precise positioning. In some embodiments, the control algorithm may utilize support vector machines, which may classify alignment errors based on feature spaces derived from sensor data or alignment outcomes and control mechanisms may then adjust alignment parameters based on the control algorithm recommendations to correct for specific types of misalignment. In some embodiments, the control algorithm may utilize reinforcement learning which may learn the best alignment strategies by receiving feedback, or rewards, based on the success of each alignment attempt and the control algorithm may adjust the alignment mechanisms, such as stage position, tilt, rotation, etc., to improve alignment outcomes overtime and learn an optimal alignment policy. The ET process module 112 adjusts the control parameters of the electroluminescence testing process 140. In some embodiments, the ET process module 112 may adjust the exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. The ET process module 112 executes the electroluminescence testing process. The electroluminescence testing process 140 may include characterizing the optical and electrical performances of the devices. The micro LED samples are prepared by being mounted onto a test fixture or substrate to ensure proper electrical contact with the device's electrodes. Following sample preparation, the testing setup is established, comprising a precision current or voltage source for applying electrical signals to the micro LED and measurement instruments such as multimeters or source measure units for control and monitoring of electrical parameters during testing. The EL testing process begins with the application of a forward bias voltage to the micro LED, surpassing the device's turn-on voltage to induce electroluminescence. This voltage is selected to ensure it falls within the safe operating limits of the device and minimizes the risk of damage. Concurrently, electroluminescence emitted by the micro LED is captured using specialized imaging equipment, such as CCD cameras, photomultiplier tubes, etc., enabling the visualization and analysis of the emitted light pattern, intensity distribution, and spatial uniformity across the device surface. The emitted light spectrum is measured using spectroradiometers or spectrometers to analyze the wavelength distribution and color characteristics of the emitted light accurately. The spectral analysis provides critical insights into the color accuracy and consistency of the micro LED's emission. Additionally, electrical parameters such as forward voltage, forward current, and reverse leakage current may be monitored and recorded during the testing process to assess the device's electrical performance, efficiency, and reliability under operation. The collected data, including electroluminescence images, spectral measurements, and electrical parameters, may be subjected to analysis to evaluate the micro LED's performance and quality. Any deviations from desired specifications or performance metrics may be identified and analyzed to determine root causes and potential areas for improvement. The results of the EL testing process are compiled into a comprehensive report detailing the micro LED's characteristics, providing information for quality control, process optimization, and product validation in micro LED manufacturing. The ET process module 112 stores the data in the ET process database 122. The ET process module 112 stores the data in the ET process database 122, such as the data collected from the electroluminescence testing process 140, the output of the inspection algorithm, and the output of the control algorithm. The ET process database 122 may contain the substrate ID, initial alignment, final alignment after the optimization adjustments by the control algorithm, etc. The ET process database 122 may contain the data for each wafer lot that is being processed by the electroluminescence testing process 140. The ET process module 112 returns to the base module 108. The base module 108 initiates, at step 204, the ET post-process module 114. For example, the ET post-process module 114 begins by being initiated by the base module 108. The ET post-process module 114 connects to the electroluminescence testing postprocess 150. The ET post-process module 114 may connect to the electroluminescence testing postprocess 150, such as processes or methods for inspection to identify defective LEDs and subsequent repair or rework processes, which may include camera-based microscopic imaging systems to assess and measure the luminance of micro LED chips accurately to enhance production yield by efficiently repairing or replacing bad dies, employing methods such as ultraviolet irradiation, laser welding, and selective pick or laser repairing technologies. The ET post-process module 114 collects the post-process data. The ET post-process module 114 may collect the parameter data, such as defect data. The post-process may involve parameters such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. The ET post-process module 114 extracts the data from the ET inspection database 120 and ET process database 122. The ET post-process module 114 extracts the data, such as material data, parameter data, etc. of the pre-process and the electroluminescence testing process. The ET post-process module 114 performs the post-process algorithm. The post-process algorithm may be used to detect small-scale defects, such as dead pixels, micro LED shorts, non-uniform brightness or mura, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. The post-process algorithm may use high-resolution images of the micro LEDs that capture the micro LEDs in their operational state, allowing for observation of luminance, color uniformity, and other characteristics that would indicate the presence or absence of defects. The post-process algorithm may use a historical dataset of high-resolution images of the micro LEDs that includes defect-free examples and defective examples. The dataset may be used to train the post-process algorithm, such as a convolutional neural network, to recognize patterns and features associated with defects and may involve feeding the images into the algorithm, which learns to differentiate between normal and defective micro LED characteristics. The post-process algorithm may then be validated and tested on a separate set of images not used during training to ensure it accurately identifies defects and the algorithm may be adjusted to improve accuracy and reduce false positives / negatives. Then high resolution images of the micro LEDs from the electroluminescence testing post-process 150 are inputted into the post-process algorithm and each image is analyzed to identify features that indicate defects based on what the algorithm learned during training. The post-process algorithm may output the locations and types of defects detected and provide detailed information for further analysis or repair, which may include quantifying defect sizes, categorizing defect types, and determining defect distribution across the micro LED array. In some embodiments, the post-process algorithm may improve the accuracy, efficiency, scalability, etc. of the defect inspection post-process. In some embodiments, the post-process algorithm may utilize supervised learning models, such as decision trees, random forests, gradient boosting machines, etc. which may be trained on historical data where the input features include types and numbers of defects, and the target variable is whether the substrate meets specifications and may learn complex patterns and interactions among defects. In some embodiments, the post-process algorithm may utilize anomaly detection algorithms which may identify unusual patterns that do not conform to expected behavior, such as detecting unusual combinations or intensities of defects that are likely to result in a substrate failing to meet specifications. In some embodiments, the post-process algorithm may utilize regression analysis or time-series forecasting to predict the future occurrence of defects based on trends and patterns, which may preemptively adjust manufacturing processes to reduce the likelihood of producing substrates that do not meet specifications. The ET postprocess module 114 performs the post-process inspection. For example, the electroluminescence testing post-process may include defect inspection, repair technologies application, efficiency improvement, etc. The defect inspection may utilize advanced imaging systems to detect luminance variations and identify defective LEDs. The repair technologies may be methods such as ultraviolet irradiation, laser welding, and selective repairing either pick or laser-based, to address and fix identified defects. The efficiency improvement may aim to increase production yield by efficiently repairing or replacing bad dies, ensuring high-quality micro LED output. The ET post-process module 114 stores the data in the ET post-process database 124. The ET post-process module 114 stores the data in the ET post-process database 124, such as the data collected from the electroluminescence testing post-process 150, the output of the post-process algorithm, etc. The ET post-process database 124 may contain the substrate ID, the defects, if the substrate meets the final specification, if the demura algorithm has the capability to fix the defects, etc. The ET process database 122 may contain the data for each wafer lot that is being processed by the electroluminescence testing post-process 150. The ET post-process module 114 returns to the base module 108. The base module 108 initiates, at step 206, the ET historical module 116. For example, the ET historical module 116 begins by being initiated by the base module 108. The ET historical module 116 connects to the electroluminescence testing pre-process 130, the electroluminescence testing process 140, and the electroluminescence testing post-process 150. The historical module 116 connects to the electroluminescence testing pre-process 130, such as the processes, such as photoluminescence testing, visual inspection, mechanical and bonding checks, etc. utilized prior to the electroluminescence testing process 140, designed to prepare the micro LED processed wafers for optimal performance during the electroluminescence process, the electroluminescence testing process 140, such as systems or equipment for automated electroluminescence testing of micro LEDs, including sample preparation, configuration, electroluminescence induction, data capture, analysis, reporting, decision making, etc., in which a controlled electrical current is applied to the micro LEDs, causing them to emit light through the electroluminescence phenomenon to assess the operational characteristics of the micro LEDs, such as their efficiency and uniformity, and the electroluminescence testing post-process 150, such as processes or methods for inspection to identify defective LEDs and subsequent repair or rework processes, which may include camerabased microscopic imaging systems to assess and measure the luminance of micro LED chips accurately to enhance production yield by efficiently repairing or replacing bad dies, employing methods such as ultraviolet irradiation, laser welding, and selective pick or laser repairing technologies. The ET historical module 116 aggregates the data from the various types of processes and stores the data in the ET network database 126. The ET network database 126 may contain the historical data from the various processes performed by the electroluminescence testing pre- process 130, electroluminescence testing process 140, and electroluminescence testing postprocess 150. The ET network database 126 may contain the data parameters collected during the inspection process, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc., control parameters of the process, such as exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc., and data parameters collected during post-processing, such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. The ET historical module 116 performs the historical machine learning algorithm on the historical data stored in the ET network database 126. The historical machine learning algorithm may improve each stage of the micro LED manufacturing process, such as electroluminescence testing pre-process, electroluminescence testing process, and electroluminescence testing post-process. For example, the historical machine learning algorithm may be implemented in various stages of the electroluminescence testing process to optimize performance, accuracy, and efficiency. The historical machine learning algorithm may prepare the micro LED substrates for electroluminescence testing by ensuring optimal conditions and alignments by analyzing images of the micro LED substrates to detect any physical anomalies or misalignments before testing to ensure that only substrates in proper condition proceed to the electroluminescence testing phase, reducing the risk of inaccurate results due to pre-existing physical defects. The historical machine learning algorithm may process images of substrates and identify defects such as scratches, cracks, or contamination and determine whether the substrate requires cleaning, repair, or realignment before testing. The historical machine learning algorithm may conduct the electroluminescence testing, capturing and analyzing the electroluminescence of the micro LEDs to identify functional defects by dynamically adjusting electroluminescence testing parameters, such as current levels, exposure times, etc. in real-time based on the immediate feedback from the testing results to maximize the detection efficiency of functional defects while minimizing the risk of damaging the micro LEDs. The historical machine learning algorithm may learn the best testing strategies over numerous testing cycles and continuously improve its policy to achieve optimal testing outcomes, such as by adjusting the electrical current to the optimal level that balances between adequate excitation of the LEDs and preventing overheating. The historical machine learning algorithm may analyze the data collected from the electroluminescence testing to identify defects, evaluate the quality of the micro LED substrates, and recommend further actions by processing large datasets to classify substrates based on defect types and severity and identify patterns correlating with specific types of electroluminescence defects, such as dead pixels, non- uniform brightness, or color shifts. The historical machine learning algorithms may analyze testing data to classify substrates into categories based on their quality, identify substrates that pass the quality threshold and those that require rework or are deemed as waste, and for substrates with correctable defects, the historical machine learning algorithm may recommend specific postprocessing actions, like local repair or adjustment through techniques like Demura. In some embodiments, the historical machine learning algorithm may use the historical data stored in the ET network database 126 to determine trends to understand how varying parameters affect the final quality and identify optimal parameter settings to create the best outcomes. The historical machine learning algorithm adjusts the parameters in real-time for the ET inspection module 110, ET process module 112, and ET post-process module 114 to correct any defects. The ET historical module 116 sends the process adjustments to the ET inspection module 110, the ET process module 112, and the ET post-process module 114. The historical machine learning algorithm may prepare the micro LED substrates for electroluminescence testing by ensuring optimal conditions and alignments. The historical machine learning algorithm may conduct electroluminescence testing, capturing and analyzing the electroluminescence of the micro LEDs to identify functional defects. The adjustments are sent to the ET inspection module 110 and ET process module 112 to create a more consistent final product which is then verified by the ET post-process module 114. The ET historical module 116 returns to the base module 108. The base module 108 initiates, at step 208, the ET integration module 118. For example, the ET integration module 118 begins by being initiated by the base module 108. The ET integration module 118 performs the integration machine learning algorithm. The integration machine learning algorithm may facilitate interactions between the electroluminescence testing pre-process 130, electroluminescence testing process 140, and electroluminescence testing post-process 150 stages of micro LED manufacturing. For example, the integration machine learning algorithm may be integrated across the electroluminescence testing providing a synergistic workflow between pre-processing, processing, and post-processing, to enhance the testing process. The integration machine learning algorithm may use the initial substrate condition, such as cleanliness, detected defects, alignment, etc. from the pre-process to inform the post-process analysis by utilizing the data on substrate preparation quality to help the post-process classify defects more accurately by correlating specific pre-test conditions with observed electroluminescence anomalies. For example, the quality of substrate preparation directly impacts the types of defects detected during electroluminescence testing and by understanding the correlation between pre-test conditions and electroluminescence outcomes, the post-process analysis can more accurately identify issues that are fixable versus inherent substrate flaws. The integration machine learning algorithm may continuously refine testing strategies based on the success of current and past tests from the electroluminescence testing process, such as if certain parameter settings consistently lead to better defect detection, the system may adjust future tests, accordingly, enhancing efficiency and accuracy over time. For example, real-time adjustments to testing parameters based on immediate feedback and historical data may ensure that the testing process becomes more efficient and accurate, reducing the likelihood of retests and enhancing the quality of data for post-process analysis. The integration machine learning algorithm may use insights from the pre-process about substrate condition, such as areas with minor physical imperfections, and adjust the focus and sensitivity of the electroluminescence testing to ensure that testing parameters are finely tuned to the substrate's specific characteristics, leading to more accurate defect detection. For example, the condition and preparation of the substrate may significantly influence the settings and focus areas for electroluminescence testing and information on detected physical imperfections may allow for adaptive testing strategies, prioritizing areas of concern and adjusting parameters to mitigate potential interference from known defects. In some embodiments, the integration machine learning algorithm may use the data stored in the ET network database 126 to perform a predictive modeling algorithm to make predictions or forecasts of the final product based on historical data and patterns from the previously created products. Predictive modeling involves the systematic analysis of historical data to identify patterns, correlations, and trends that can be used to build models capable of making accurate predictions about future outcomes. For example, the integration machine learning algorithm may be used for identifying defects across the electroluminescence testing pre-process 130, electroluminescence testing process 140, and electroluminescence testing post-process 150. The historical data in the ET network database 126 may include information on defects and relevant operational parameters such as contamination, alignment, wafer size, uniformity, color temperature, dead pixels, color shifts, etc. The data is cleaned and preprocessed, addressing any missing values and ensuring that all variables are in a suitable format for modeling. This may involve normalization or scaling of numerical features. The dataset is split into training and testing sets. The training set is used to train the predictive model, and the testing set assesses its performance on unseen data. The model used may be linear regression, decision trees, ensemble methods, neural networks, etc. Then key features are identified that influence the defects and engineer new features if necessary. For example, exposure dose from the electroluminescence process 140 may have a significant impact on the defects identified. The selected predictive model is trained using the training dataset. The model learns patterns and relationships between operational parameters and defects. The model's performance is validated on a separate validation dataset and hyperparameters are fine-tuned to optimize its accuracy and generalization. The model is then evaluated and deployed by being sent to the electroluminescence testing pre-process 130, electroluminescence testing process 140, and electroluminescence testing post-process 150. The ET integration module 118 connects to the electroluminescence testing pre-process Al module 136, the electroluminescence testing process Al module 146, and the electroluminescence testing post-process Al module 156. The ET integration module 118 connects to the electroluminescence testing pre-process 130, such as the processes, such as photoluminescence testing, visual inspection, mechanical and bonding checks, etc. utilized prior to the electroluminescence testing process 140, designed to prepare the micro LED processed wafers for optimal performance during the electroluminescence process, the electroluminescence testing process 140, such as systems or equipment for automated electroluminescence testing of micro LEDs, including sample preparation, configuration, electroluminescence induction, data capture, analysis, reporting, decision making, etc., in which a controlled electrical current is applied to the micro LEDs, causing them to emit light through the electroluminescence phenomenon to assess the operational characteristics of the micro LEDs, such as their efficiency and uniformity, and the electroluminescence testing post-process 150, such as processes or methods for inspection to identify defective LEDs and subsequent repair or rework processes, which may include camera-based microscopic imaging systems to assess and measure the luminance of micro LED chips accurately to enhance production yield by efficiently repairing or replacing bad dies, employing methods such as ultraviolet irradiation, laser welding, and selective pick or laser repairing technologies. The ET integration module 118 sends the process adjustments to the electroluminescence testing pre-process Al module 136, the electroluminescence testing process Al module 146, and the electroluminescence testing postprocess Al module 156. For example, the ET integration module 118 may send the uniformity, exposure dose, color shifts, etc. adjustments to each of the processes to minimize the defects of the substrates during the electroluminescence testing process. In some embodiments, the ET integration module 118 may send the integration machine learning algorithm to the processes, allowing the systems to further enhance the optimization. The ET integration module 118 returns to the base module 108.

[0045] FIG. 3 illustrates the ET inspection module 110. The process begins with the ET inspection module 110 being initiated, at step 300, by the base module 108. The ET inspection module 110 connects, at step 302, to the electroluminescence testing pre-process 130. The ET inspection module 110 connects with the electroluminescence testing pre-process 130, such as photoluminescence testing, visual inspection, mechanical and bonding checks, etc. utilized before the electroluminescence testing process 140, designed to prepare the micro LED processed wafers for optimal performance during the electroluminescence testing process. In some embodiments, the ET inspection module 110 may transmit and receive data from the electroluminescence testing pre-process 130. In some embodiments, the electroluminescence testing pre-process 130 may measure wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. In some embodiments, the ET inspection module 110 may control or send inputs to control the electroluminescence testing pre-process 130. The ET inspection module 110 collects, at step 304, the data from the electroluminescence testing pre-process 130. The ET inspection module 110 collects the pre-processing data, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. For example, photoluminescence testing may evaluate the optical properties of micro LEDs before applying any electrical current and may identify defects or variations in luminescence efficiency across the wafer or panel without causing damage to the micro LEDs. Visual inspections may be conducted to check for obvious physical defects or alignment issues post-mass transfer which may include checking for missing LEDs, misalignment, or other visible defects that may impact the performance of the micro LED display. Mechanical and bonding checks may be performed to ensure that the micro LED chips are securely attached to the substrate before moving on to EL testing, which may involve mechanical stress tests or specific inspections to assess the integrity of the bond between the micro LED chips and the display panel. The ET inspection module 110 performs, at step FIG, the inspection algorithm. For example, the inspection algorithm may achieve a balance between high throughput and the accuracy of defect detection. High throughput is essential for cost-effective manufacturing, but it may compromise the detection accuracy which may allow defective units to proceed to later stages of production and may result in increased costs if defects are detected only at the EL testing phase or beyond. For example, the inspection algorithm may reduce temperature stability deviations and potentially improve throughput results for EL testing. The inspection algorithm may predict optimal EL testing parameters, such as temperature settings, that maximize throughput while ensuring high-quality testing outcomes. The inspection algorithm may use a linear regression model to predict optimal testing parameters based on historical data and focus on achieving the best temperature conditions and throughput rates. The inspection algorithm may use historical data, such as data on wafer characteristics, including material composition, layer thickness, and, EL testing parameters, including temperature settings, electrical input, etc., observed temperature stability deviations, resulting throughput rates, etc. The inspection algorithm may identify features, or variables, that impact throughput, such as initial temperature settings, electrical input, material properties, historical temperature deviations, etc. The inspection algorithm is trained on the historical data to learn the relationship between the selected features and the throughput rate. The historical data may be formatted, normalized, standardized, etc., and is split into training and testing sets. The inspection algorithm is trained using the input of the selected features from the historical data and the output of the algorithm is the throughput rate. The algorithm is then evaluated in which the algorithm’s performance is assessed using the testing set and metrics like R-squared, and R2, to determine how well the model predicts throughput rates based on the input features. Then before the EL test, the current wafer characteristics and initial testing parameters are inputted into the algorithm and the algorithm predicts the adjustments needed to optimize the throughput, such as suggesting the ideal temperature settings. In some embodiments, the inspection algorithm may utilize decision trees and random forests to categorize wafers into different quality levels or predict throughput rates based on a wide range of testing parameters and wafer characteristics. In some embodiments, the inspection algorithm may utilize support vector machines to predict optimal testing parameters for achieving desired throughput rates. In some embodiments, the inspection algorithm may utilize reinforcement learning which may dynamically adjust EL testing parameters in real-time and learn the optimal testing strategy through continuous interaction with the testing environment to maximize throughput and yield.

[0046]

[0046] Edge Al and TinyML Example 1: In some embodiments, the system may utilize Edge Al and TinyML. For example, the ET inspection module 110, ET process module 112, and ET postprocess module 114 may collaboratively employ Edge Al to perform real-time adjustments during the electroluminescence testing of micro LEDs. The ET inspection module 110 may initially gather input data from the electroluminescence testing pre-process 130, including wafer thickness, surface defects, and cleanliness. This data is processed locally using TinyML models optimized for low- power edge devices, allowing for immediate analysis without the need for cloud connectivity. The decision logic within the ET inspection module 110 may involve a TinyML-based linear regression model trained to predict optimal temperature settings that maximize throughput while maintaining defect detection accuracy. Once the model processes the input data, it outputs a control signal that adjusts the temperature settings of the electroluminescence testing process 140. This adjustment is communicated to the ET process module 112, which then modifies the control parameters such as exposure dose and alignment accuracy to align with the optimized temperature settings. The ET process module 112 may also receive real-time feedback from the electroluminescence testing process 140, including luminous intensity and emission spectrum data, which are processed using a TinyML-based convolutional neural network (CNN) to detect any misalignments. If misalignments are detected, the ET process module 112 sends a control signal to adjust the alignment tools, such as motorized stages or tilt platforms, ensuring precise positioning of the micro LEDs. Following the electroluminescence testing, the ET post-process module 114 receives high- resolution images of the micro LEDs, capturing their operational state. These images are analyzed using a TinyML-based anomaly detection algorithm to identify small-scale defects such as dead pixels or color shifts. The algorithm outputs defect locations and types, which are stored in the ET post-process database 124. This data is then used to generate a feedback loop to the ET inspection module 110 and ET process module 112, allowing for continuous improvement of the testing parameters. A prompt that would trigger the Al system could be the detection of a temperature deviation beyond a predefined threshold during the electroluminescence testing process 140. Upon detecting such a deviation, the Edge Al system would immediately process the input data to recalibrate the temperature settings, ensuring optimal testing conditions are maintained. This integration of Edge Al and TinyML within the ET inspection module 110, ET process module 112, and ET post-process module 114 facilitates a seamless, real-time adaptive testing environment that enhances the accuracy and efficiency of micro LED manufacturing. The system's ability to operate independently of cloud resources ensures low latency and high reliability, making it suitable for deployment in environments where connectivity may be limited or where immediate decisionmaking is crucial.

[0047]

[0047] The ET inspection module 110 sends, at step 308, the determined process data from the inspection algorithm to the ET process module 112. For example, the ET inspection module 110 may send the optimal temperature settings to the ET process module 112 to optimize the electroluminescence process. In some embodiments, the ET inspection module 110 may send wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. The ET inspection module 110 stores, at step 310, the data in the ET inspection database 120. The ET inspection module 110 stores the data outputted from the inspection algorithm in the ET inspection database 120, such as the wafer ID, the initial throughput percentage, the initial temperature stability, the adjusted throughput percentage, the temperature stability optimized by the inspection algorithm, the throughput percentage optimized by the inspection algorithm, etc. In some embodiments, the ET inspection database 120 may include other parameter data collected or calculated during the ET inspection module 110 process, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. The ET inspection module 110 returns, at step 312, to the base module 108.

[0048]

[0048] The ET inspection module that is being presented delineates a multifaceted procedure that is essential to the electroluminescence (EL) testing of micro LED wafers. It primarily centers on the use of a basic artificial intelligence (Al) algorithm at step 306 to optimize testing parameters and guarantee superior results. The aforementioned methodology utilizes linear regression, a fundamental method in the domain of machine learning, to anticipate the optimal EL testing circumstances. Because of its effectiveness and the simple relationship it creates between input variables and a continuous output, linear regression is the preferred method for predicting ideal temperature settings and other testing parameters that impact throughput and quality in manufacturing settings. The linear regression model in this case works with a dataset that includes historical information on wafer properties (material composition, layer thickness), EL testing parameters (temperature settings, electrical input), and observed results (temperature stability deviations, throughput rates). The model learns from this data to estimate adjustments required for improving testing throughput. It detects and uses features such as initial temperature settings and material qualities that have a substantial impact on throughput. In order to refine and validate the model's predictions, the gathered data is split into training and testing sets. Performance is assessed using metrics such as R-squared. The datasets needed for the linear regression model's implementation of step 306 would include thorough records of previous testing and inspection procedures. These databases may contain, but are not restricted to, characteristics of the epitaxial layer, wafer size and shape, thickness, surface imperfections, cleanliness, and electrical and physical properties. Furthermore, information about the actual EL testing procedure, including electrical inputs, temperature settings, and the performance metrics that are produced, such as temperature stability and throughput rates, is essential for algorithm training. The program can identify trends and correlations in this large dataset, which helps optimize EL testing parameters. The ultimate goal is to balance throughput efficiency with defect detection precision and overall quality assurance in the manufacturing of tiny LED displays. The objective of this example is to optimize the electroluminescence (EL) testing parameters for micro LED wafers in order to balance accurate defect detection with high throughput. A Decision Tree Regression combined with a Random Forests approach is an example of an Al algorithm that would be suitable for this task. In addition to offering decision trees' ease of use and interpretability, this combination makes use of Random Forests' ability to increase prediction accuracy and manage dataset complexity. By using a number of input variables, a Decision Tree algorithm builds a model that forecasts the value of a target variable. Every terminal node in a regression tree represents a continuous quantity. The relationship between different wafer and testing parameters (such as wafer thickness, cleanliness, temperature settings, and electrical input) and the desired results (like throughput rates and temperature stability deviations) could be modeled for the ET inspection module using a Decision Tree. This approach is especially helpful since it is simple to understand; it makes it evident which factors have the biggest effects on the results and offers insightful information about how wafer quality and throughput are impacted by EL testing settings. Typically trained using the "bagging" approach, the Random Forests algorithm is an ensemble technique that builds a "forest" of several decision trees. The fundamental idea is to control over-fitting and increase overall prediction accuracy by combining the predictions of many decision trees. Every tree in the forest was constructed using a replacement sample that was taken from the training set. Moreover, the split that is selected when dividing a node throughout the tree-building process is no longer the optimum split across all features. Rather, the optimal split from a random subset of the traits is selected. This method increases the model's level of randomness, which makes it more resilient and improves its ability to generalize to new data. Using Random Forests in the context of the ET inspection module would enable the model to manage the intricacies and interactions between different testing parameters and wafer attributes without requiring a high level of domain expertise to choose and adjust these interactions by hand. Because it can automatically take into account a wide range of variables and their possible interactions, it is especially well-suited for optimizing EL testing parameters. It can forecast optimal settings that increase throughput while maintaining the quality of the micro LED wafers. The dataset would need to contain a wide range of variables in order to train this Al model effectively. These variables could include: (1) Wafers' physical and electrical characteristics (such as thickness, cleanliness, surface defects, size, shape, epitaxial layer characteristics, resistivity, and doping concentration); (2) Past EL testing parameters (such as electrical inputs and temperature settings); and (3) Observed results (such as throughput rates, temperature stability deviations, and defect detection rates). With the help of this extensive dataset, the Random Forests algorithm may investigate a wide range of potential decision trees and take use of their combined forecasts, improving the model's accuracy and dependability when it comes to forecasting the best EL testing settings for novel wafers.

[0049]

[0049] FIG. 4 illustrates the ET process module 112. The process begins with the ET process module 112 being initiated, at step 400, by the base module 108. The ET process module 112 receives, at step 402, the pre-process data, such as the data on the micro LED processed wafer, from the ET inspection module 110. The ET process module 112 receives the outputted data from the inspection algorithm performed in the ET inspection module 110, such as the optimized temperature, allowing the electroluminescence testing process 140 to determine the optimal control parameters for the electroluminescence testing process. In some embodiments, the ET process module 112 may receive wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc. The ET process module 112 connects, at step 404, to the electroluminescence testing process 140. The ET process module 112 connects with the electroluminescence process 140, such as systems or equipment for automated electroluminescence testing of micro LEDs, including sample preparation, configuration, electroluminescence induction, data capture, analysis, reporting, decision making, etc., in which a controlled electrical current is applied to the micro LEDs, causing them to emit light through the electroluminescence phenomenon to assess the operational characteristics of the micro LEDs, such as their efficiency and uniformity. In some embodiments, the ET process module 112 may transmit and receive data from the electroluminescence testing process 140. In some embodiments, the ET process module 112 may control or send inputs to control the electroluminescence testing process 140, such as controlling parameters of the electroluminescence testing process, for example, exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. The ET process module 112 collects, at step 406, the data from the electroluminescence testing process 140. The ET process module 112 collects the processing data, such as exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc.

[0050]

[0050] Edge Al and TinyML Example 2: In some embodiments, the system may utilize Edge Al and TinyML. For example, the ET inspection module 110, ET process module 112, and ET postprocess module 114 may be integrated with TinyML models to perform real-time analysis and decision-making directly at the edge, minimizing latency and bandwidth usage. The ET inspection module 110 may employ a TinyML model to analyze photoluminescence testing data, specifically focusing on wafer thickness and surface defects, which are captured via high-resolution imaging sensors. The model, trained on historical data stored in the ET inspection database 120, processes this input to predict optimal temperature settings for the subsequent electroluminescence testing process. This prediction is then transmitted as a control signal to the ET process module 112, which adjusts the temperature settings of the electroluminescence testing equipment to maintain optimal conditions, ensuring accurate luminosity and color accuracy testing. The ET process module 112, equipped with its own TinyML model, further refines the testing parameters by analyzing real-time electroluminescence data such as luminous intensity and chromaticity coordinates. This data is collected by CCD cameras and spectroradiometers during the testing phase. The TinyML model processes these inputs to dynamically adjust the exposure dose and alignment accuracy, sending control signals to the electroluminescence testing apparatus to fine-tune these parameters. This ensures that the micro LEDs are tested under ideal conditions, thereby enhancing the reliability of the test results. Meanwhile, the ET post-process module 114 utilizes a TinyML model to evaluate the post-testing data, including defect types such as dead pixels and color shifts, which are identified using high-resolution imaging systems. The model, trained on defect data from the ET post-process database 124, processes this input to determine the applicability of repair processes like ultraviolet irradiation or laser welding. If a defect is deemed repairable, the model sends a signal to initiate the appropriate repair process, thereby improving the yield of functional micro LEDs. This interaction between the modules is facilitated through a communication interface 104, which ensures seamless data flow and control signal transmission across the system. A specific condition that would trigger the Al system is the detection of a significant deviation in temperature stability during the ET inspection module 110's analysis. Upon detection, the TinyML model immediately processes the deviation data and predicts the necessary adjustments to the temperature settings, which are then implemented by the ET process module 112. This real-time adjustment capability, enabled by Edge Al and TinyML, allows the system to maintain high throughput and defect detection accuracy without relying on cloud-based processing, thus ensuring efficient and localized decision-making within the micro LED testing environment.

[0051]

[0051] The ET process module 112 performs, at step 408, the control algorithm. For example, the control algorithm may optimize alignment based on a dataset of previous alignments, outcomes, and various parameters affecting alignment quality to prevent misalignment which may lead to inaccurate test results or damage to the micro LEDs. For example, the alignment process may include sample preparation, such as cleaning and inspection, securing the micro LED, utilizing alignment equipment, such as high-resolution microscopes, optical alignment systems, vision systems, fiducial markers, detection systems, etc. to initially position the micro LED sample based on visual markers or reference points. The micro LED sample’ s position may be fine-tuned through micromanipulators, piezoelectric positioners, nanopositioning stages, laser interferometer systems, alignment stages with fine adjustment screws, etc. Once the sample is aligned properly it is securely attached to the mounting platform or holder, which may involve vacuum sealing, adhesives, etc., and the attachment is verified. The alignment is re-checked using the alignment equipment and the electrical connections are verified. The control algorithm may analyze visual imagery by being trained on a dataset comprising images of aligned and misaligned micro LED substrates, along with corresponding alignment parameters and outcomes allowing the control algorithm to recognize patterns and features associated with optimal alignment. For example, the control algorithm may compile a dataset of high-resolution images of micro LED substrates at various stages of the alignment process, including pre-alignment, post-manual alignment, and post-AI alignment stages and each image may be tagged with metadata detailing the alignment parameters used and the success rate of the alignment. The control algorithm may be trained on the dataset which may involve feeding the images into the algorithm, which may learn to identify features and patterns correlating with successful alignments. The output of the control algorithm may be to predict the optimal alignment parameters. During the micro LED manufacturing process, live images of the substrate alignments are fed into the trained control algorithm, and it may assess the images in real time and predict the most effective adjustment parameters to achieve optimal alignment. The predicted parameters from the control algorithm may be used to automatically adjust the alignment tools, which may involve adjusting the position of the substrate, modifying the angle of incidence for alignment lasers, or altering the pressure applied by mechanical aligners. The outcomes of these adjustments may be continuously monitored and fed back into the control algorithm which allows it to refine its predictions over time. The control algorithm may process complex image data, recognize patterns, and learn from the visual cues associated with successful and unsuccessful alignments. The control algorithm may influence the control mechanism of the alignment tools by predicting the necessary adjustments and sending signals to the mechanical or optical components responsible for making precise movements, which may include motorized stages for lateral adjustments, tilt platforms for angular corrections, or laser steering systems for precise positioning. In some embodiments, the control algorithm may utilize support vector machines, which may classify alignment errors based on feature spaces derived from sensor data or alignment outcomes and control mechanisms may then adjust alignment parameters based on the control algorithm recommendations to correct for specific types of misalignment. In some embodiments, the control algorithm may utilize reinforcement learning which may learn the best alignment strategies by receiving feedback, or rewards, based on the success of each alignment attempt and the control algorithm may adjust the alignment mechanisms, such as stage position, tilt, rotation, etc., to improve alignment outcomes overtime and learn an optimal alignment policy.

[0052]

[0052] Generative Video Al Example 1: In some embodiments, the system may utilize Generative Video Al. For example, the ET inspection module 110, ET process module 112, and ET postprocess module 114 may interact with a Generative Video Al system to enhance defect detection and process optimization. The Generative Video Al receives high-resolution video inputs from the electroluminescence testing process 140, capturing detailed visual data of micro LED substrates under test. This video data includes frames that show the electroluminescence emission patterns, spatial uniformity, and any visible defects such as dead pixels or color shifts. The Al processes these frames to generate a predictive model of the substrate's condition over time, utilizing temporal changes and spatial patterns to identify potential defects that may not be evident in static images. The decision logic of the Generative Video Al involves analyzing the video data to detect deviations from expected emission patterns, using a trained neural network model that has learned from historical video datasets stored in the ET network database 126. This model is capable of recognizing complex defect patterns by comparing real-time video data against the learned baseline of defect-free and defective states. Upon detecting anomalies, the Al generates control signals that are sent to the ET process module 112 to adjust testing parameters dynamically. For instance, if the Al identifies a potential alignment issue based on the observed emission pattern, it may send a control signal to adjust the alignment accuracy parameter, ensuring the electroluminescence testing process 140 maintains optimal conditions. Additionally, the Al may instruct the ET post-process module 114 to perform targeted inspections on specific areas of the substrate, focusing on regions where the video analysis indicated potential defects. This interaction is facilitated through a data flow that includes real-time video analysis results and historical data comparisons, enabling the ET post-process module 114 to prioritize inspection tasks effectively. A specific condition that would trigger the Al system is the detection of a rapid change in emission intensity across a substrate section, which may indicate a developing defect. The Al's control outputs are grounded in the hardware by interfacing with the electroluminescence testing process 140's control systems, allowing for immediate parameter adjustments that are crucial for maintaining the integrity of the testing process. By integrating video data analysis with real-time control, the Generative Video Al provides a robust mechanism for enhancing the precision and reliability of micro LED testing, ensuring that any defects are promptly identified and addressed.

[0053]

[0053] Multimodal Al Example 1: In some embodiments, the system may utilize Multimodal Al. For example, the automated electroluminescence testing equipment may integrate multimodal Al to enhance the precision and efficiency of micro LED testing by simultaneously analyzing visual, thermal, and electrical data streams. The ET inspection module 110 may initially receive high- resolution optical images of micro LED wafers from a camera-based imaging system, capturing detailed visual information about the wafer's surface, including potential misalignments or surface defects. Simultaneously, a thermal imaging device may provide infrared data indicating temperature distribution across the wafer, which is critical for identifying thermal hotspots that could affect LED performance. The ET inspection module 110 processes these multimodal inputs using a convolutional neural network (CNN) trained to detect anomalies by correlating visual defects with thermal inconsistencies. This CNN may output a control signal to the ET process module 112, indicating specific areas on the wafer that require adjusted alignment or temperature stabilization. The ET process module 112, upon receiving this control signal, may adjust the wafer's position using precision alignment stages and modify the thermal environment using localized cooling systems to ensure optimal testing conditions. Concurrently, the ET process module 112 may monitor electroluminescence data, such as emission spectrum and luminous intensity, captured by photodetectors, and compare these metrics against expected values stored in the ET process database 122. If discrepancies are detected, indicating potential defects, the ET process module 112 may trigger a feedback loop, sending data back to the ET inspection module 110 for further analysis and corrective action. Post-processing, the ET post-process module 114 may aggregate data from the inspection and process phases, including defect types and locations, into the ET post-process database 124. This module may employ a decision tree algorithm to classify defects and determine the most effective repair strategy, such as laser welding or selective pick repair, based on historical success rates stored in the ET network database 126. The ET postprocess module 114 may then send a control signal to initiate the appropriate repair action. Throughout this process, the ET historical module 116 continuously updates the ET network database 126 with new data, refining predictive models used by the inspection and process modules. This module may also perform cross-modal analysis to identify patterns that singlemodality analysis might miss, such as correlating specific visual defects with thermal anomalies that predict certain electrical failures. If a pattern is identified that consistently leads to defects, the ET historical module 116 may send a process adjustment recommendation to the ET integration module 118. The ET integration module 118 may then update the ET process module 112's operational parameters, such as modifying alignment precision or thermal management protocols, to preemptively mitigate these defects in future testing cycles. A prompt that could trigger this Al system might be the detection of a thermal hotspot exceeding a predefined threshold, indicating potential overheating that could lead to LED failure. This integration of multimodal data streams allows the system to perform comprehensive analysis and real-time adjustments, ensuring high- quality micro LED production with minimal defects.

[0054]

[0054] Synthetic Data and AI-Generated Digital Twins Example 1: In some embodiments, the system may utilize Synthetic Data and AI-Generated Digital Twins. For example, the ET process module 112, ET post-process module 114, and ET historical module 116 may collaboratively function to enhance the electroluminescence testing of micro LED substrates. The system begins by generating synthetic data that mirrors real-world scenarios of micro LED substrate conditions, including various defect types, alignment issues, and temperature deviations. This synthetic data is fed into the ET process module 112, which is responsible for conducting the electroluminescence testing process. The ET process module 112 uses this data to simulate the application of a forward bias voltage to the micro LEDs, capturing electroluminescence emissions using CCD cameras. The synthetic data allows the module to adjust parameters such as exposure dose and alignment accuracy in real-time, utilizing Al-driven decision logic to predict optimal settings based on historical performance patterns stored in the ET network database 126. This simulation enables the ET process module 112 to preemptively identify potential deviations and adjust control outputs, such as modifying the alignment laser's angle or the electrical input, ensuring that the testing conditions remain within optimal ranges. The ET post-process module 114 then receives the simulated electroluminescence data, including high-resolution images of the emitted light patterns and intensity distributions. The module employs ALgenerated digital twins to create virtual replicas of the micro LED substrates, allowing for a detailed inspection of potential defects such as dead pixels or non-uniform brightness. The digital twin technology enables the module to perform a virtual inspection algorithm that cross-references the synthetic data with real-world defect patterns stored in the ET post-process database 124. This comparison allows the module to output precise defect locations and types, which are then used to generate control signals for repair processes, such as adjusting laser welding parameters or initiating selective pick repairs. The ET post-process module 114 also sends feedback to the ET historical module 116, detailing the defect types and repair outcomes. The ET historical module 116 aggregates this feedback with historical data from previous testing cycles. It employs machine learning algorithms to analyze trends and patterns, identifying correlations between initial substrate conditions and defect occurrences. This analysis informs the creation of updated synthetic data sets, which are used to refine the Al models in the ET process module 112 and ET post-process module 114. For instance, if the historical analysis reveals a consistent correlation between specific alignment deviations and color shifts, the system generates synthetic scenarios that emphasize these conditions, allowing the Al models to learn and adapt their decision-making processes. This iterative cycle of data generation, testing, and analysis enables the system to continuously improve its predictive accuracy and control precision. A specific trigger for this Al system could be the detection of a sudden increase in defect rates during real-world testing, prompting the generation of new synthetic data scenarios to explore potential causes and solutions. By leveraging synthetic data and digital twins, the system provides a robust framework for optimizing the electroluminescence testing of micro LEDs, ensuring high- quality output and efficient defect management.

[0055]

[0055] The ET process module 112 adjusts, at step 410, the control parameters of the electroluminescence testing process 140. In some embodiments, the ET process module 112 may adjust the exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc. The ET process module 112 executes, at step 412, the electroluminescence testing process. The electroluminescence testing process 140 may include characterizing the optical and electrical performances of the devices. The micro LED samples are prepared by being mounted onto a test fixture or substrate to ensure proper electrical contact with the device's electrodes. Following sample preparation, the testing setup is established, comprising a precision current or voltage source for applying electrical signals to the micro LED and measurement instruments such as multimeters or source measure units for control and monitoring of electrical parameters during testing. The EL testing process begins with the application of a forward bias voltage to the micro LED, surpassing the device's turn-on voltage to induce electroluminescence. This voltage is selected to ensure it falls within the safe operating limits of the device and minimizes the risk of damage. Concurrently, electroluminescence emitted by the micro LED is captured using specialized imaging equipment, such as CCD cameras, photomultiplier tubes, etc., enabling the visualization and analysis of the emitted light pattern, intensity distribution, and spatial uniformity across the device surface. The emitted light spectrum is measured using spectroradiometers or spectrometers to analyze the wavelength distribution and color characteristics of the emitted light accurately. The spectral analysis provides critical insights into the color accuracy and consistency of the micro LED's emission. Additionally, electrical parameters such as forward voltage, forward current, and reverse leakage current may be monitored and recorded during the testing process to assess the device's electrical performance, efficiency, and reliability under operation. The collected data, including electroluminescence images, spectral measurements, and electrical parameters, may be subjected to analysis to evaluate the micro LED's performance and quality. Any deviations from desired specifications or performance metrics may be identified and analyzed to determine root causes and potential areas for improvement. The results of the EL testing process are compiled into a comprehensive report detailing the micro LED's characteristics, providing information for quality control, process optimization, and product validation in micro LED manufacturing. The ET process module 112 stores, at step 414, the data in the ET process database 122. The ET process module 112 stores the data in the ET process database 122, such as the data collected from the electroluminescence testing process 140, the output of the inspection algorithm, and the output of the control algorithm. The ET process database 122 may contain the substrate ID, initial alignment, final alignment after the optimization adjustments by the control algorithm, etc. The ET process database 122 may contain the data for each wafer lot that is being processed by the electroluminescence testing process 140. The ET process module 112 returns, at step 416, to the base module 108.

[0056]

[0056] For step 408, which focuses on performing a control algorithm within the Electroluminescence (ET) process module, a suitable Al algorithm could be a Convolutional Neural Network (CNN) tailored for image analysis. This selection is due to the specific need for optimizing alignment in the ET process module, where visual imagery plays a crucial role in determining the success of micro LED substrate alignment. The CNN can analyze complex image data, recognize patterns, and learn from visual cues associated with successful and unsuccessful alignments, making it an ideal choice for this application. A CNN is adept at handling image data due to its architecture, which is designed to automatically and adaptively learn spatial hierarchies of features from images. This capability is particularly useful for the alignment process described, where the algorithm needs to discern between aligned and misaligned micro LED substrates. The process begins with the CNN being trained on a dataset comprising high-resolution images of micro LED substrates at various stages of the alignment process. This dataset includes images taken before alignment, after manual alignment, and after adjustments guided by the Al, with each image tagged with metadata detailing the alignment parameters used and the outcome's success rate. The training process allows the CNN to identify features and patterns correlating with successful alignments. These features could include the positioning of micro LEDs relative to fiducial markers, the uniformity of spacing between LEDs, or specific visual indicators of misalignment. Once trained, the CNN can analyze live images of substrate alignments during the manufacturing process, assess them in real-time, and predict the most effective adjustment parameters to achieve optimal alignment. The predicted parameters by the CNN are then used to automatically adjust the alignment tools. Adjustments could involve modifying the position of the substrate, altering the angle of incidence for alignment lasers, or changing the pressure applied by mechanical aligners. The effectiveness of these adjustments is continuously monitored, and feedback is provided to the CNN, allowing it to refine its predictions and improve alignment accuracy over time. For this approach to be implemented, a comprehensive dataset is necessary. This dataset would need to include high-resolution images of micro LED substrates at different alignment stages, each associated with metadata about the alignment parameters and the resulting success rate. This data enables the CNN to learn the nuances of alignment, including identifying subtle cues that may indicate an alignment is off by a minuscule but critical amount. Additionally, the dataset should be continuously updated with new data to ensure the CNN adapts to any changes in the manufacturing process or new types of misalignments that may emerge. This dynamic learning process ensures that the control algorithm remains effective over time, continually improving the precision and reliability of the micro LED manufacturing process. An appropriate Al algorithm for step 408 — which deals with executing a control algorithm within the Electroluminescence (ET) process module — might be an image-analyzing Convolutional Neural Network (CNN). This choice was made in response to the particular requirement for alignment optimization in the ET process module, where the effectiveness of micro LED substrate alignment is greatly influenced by visual imaging. The CNN is a perfect fit for this application because it can evaluate complicated image data, identify patterns, and learn from visual cues associated with successful and failure alignments. Because of its architecture, which is built to dynamically and adaptively learn spatial hierarchies of information from images, a CNN is skilled at handling image data. This feature is very helpful for the alignment procedure that is explained, as the algorithm must be able to distinguish between micro LED substrates that are aligned and those that are not. First, a dataset containing high- resolution photos of micro LED substrates at different alignment phases is used to train the CNN. This dataset contains photos that were taken prior to alignment, following manual alignment, and following ALguided modifications. Each image has metadata labeled with information about the alignment parameters that were employed and the success rate of the results. The CNN can recognize characteristics and patterns associated with successful alignments thanks to the training process. These characteristics could be the way in which tiny LEDs are positioned in relation to fiducial markers, how uniformly spaced out the LEDs are, or whether there are particular visual cues that indicate misalignment. After being trained, the CNN can evaluate real-time photos of substrate alignments taken during manufacturing, identify the best adjustment parameters to achieve optimal alignment, and analyze live photographs of these alignments. The alignment tools are then automatically adjusted based on the CNN's projected parameters. Adjustments may include shifting the substrate's location, adjusting the alignment lasers' angle of incidence, or adjusting the pressure used by the mechanical aligners. The efficacy of these modifications is continually assessed, and the CNN receives input to help it hone its forecasts and increase alignment accuracy over time. A large dataset is required in order to apply this strategy. High- resolution photos of micro LED substrates at various alignment phases, each accompanied by metadata describing the alignment settings and the success rate that resulted, would need to be included in this dataset. With the use of this data, CNN is able to learn the finer points of alignment, such as recognizing faint indications that point to alignments that may be slightly but significantly incorrect. In order to guarantee that the CNN adjusts to any modifications in the manufacturing process or emerging new kinds of misalignments, the dataset should also be updated on a regular basis with fresh data. The micro LED manufacturing process' accuracy and dependability are continuously increased by this dynamic learning process, which guarantees that the control algorithm will continue to function effectively over time.

[0057]

[0057] FIG. 5 illustrates the ET post-process module 114. The process begins with the ET postprocess module 114 being initiated, at step 500, by the base module 108.

[0058]

[0058] Al Agents Example 1: In some embodiments, the system may utilize Al Agents. For example, the automated electroluminescence testing equipment system for micro LED substrates may employ Al Agents to manage the integration of the ET inspection module 110, ET process module 112, and ET post-process module 114. The Al Agent begins by receiving input data from the ET inspection module 110, which includes parameters such as wafer thickness, surface defects, and cleanliness. This data is collected through precision sensors and imaging devices that capture high-resolution images of the micro LED substrates. The Al Agent processes this data using a decision tree algorithm to classify the substrates into different quality categories. The decision logic involves analyzing the input features to determine if the substrates meet predefined quality thresholds or if they require adjustments before proceeding to the ET process module 112. Once the classification is complete, the Al Agent generates control signals that adjust the handling equipment to either redirect substrates for rework or allow them to proceed to the ET process module 112. In the ET process module 112, the Al Agent receives real-time data on parameters such as alignment accuracy, luminous intensity, and emission spectrum from sensors integrated into the testing equipment. The Al Agent uses a convolutional neural network (CNN) to analyze this data, identifying patterns that indicate optimal alignment and emission characteristics. The decision logic here involves comparing current measurements against historical data stored in the ET network database 126 to detect deviations from expected performance. If misalignments or suboptimal emission patterns are detected, the Al Agent sends control outputs to adjust the alignment mechanisms, such as motorized stages or laser steering systems, ensuring precise positioning of the micro LEDs. The Al Agent also communicates with the ET post-process module 114, where it receives data on detected defects such as dead pixels and color shifts. Using this data, the Al Agent employs a regression analysis model to predict the likelihood of successful defect repair using techniques like laser welding or ultraviolet irradiation. The decision logic involves evaluating the severity and type of defects against historical repair success rates stored in the ET post-process database 124. If the Al Agent determines that repair is feasible, it sends control signals to activate the appropriate repair mechanisms. Additionally, the Al Agent continuously updates the ET historical module 116 with new data, enabling ongoing learning and refinement of decision models. A specific condition that triggers the Al system is the detection of a substrate with alignment deviations exceeding 0.1 pm, prompting the Al Agent to initiate real-time corrective actions. This integrated approach ensures that the Al Agent effectively coordinates the interactions between the ET inspection module 110, ET process module 112, and ET post-process module 114, optimizing the overall electroluminescence testing process for micro LED substrates.

[0059]

[0059] The ET post-process module 114 connects, at step 502, to the electroluminescence testing post-process 150. The ET post-process module 114 may connect to the electroluminescence testing post-process 150, such as processes or methods for inspection to identify defective LEDs and subsequent repair or rework processes, which may include camera-based microscopic imaging systems to assess and measure the luminance of micro LED chips accurately to enhance production yield by efficiently repairing or replacing bad dies, employing methods such as ultraviolet irradiation, laser welding, and selective pick or laser repairing technologies. The ET post-process module 114 collects, at step 504, the post-process data. The ET post-process module 114 may collect the parameter data, such as defect data. The post-process may involve parameters such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. The ET post-process module 114 extracts, at step 506, the data from the ET inspection database 120 and ET process database 122. The ET post-process module 114 extracts the data, such as material data, parameter data, etc. of the pre-process and the electroluminescence testing process. The ET post-process module 114 performs, at step 508, the post-process algorithm. The post-process algorithm may be used to detect small-scale defects, such as dead pixels, micro LED shorts, non-uniform brightness or mura, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. The post-process algorithm may use high- resolution images of the micro LEDs that capture the micro LEDs in their operational state, allowing for observation of luminance, color uniformity, and other characteristics that would indicate the presence or absence of defects. The post-process algorithm may use a historical dataset of high-resolution images of the micro LEDs that includes defect-free examples and defective examples. The dataset may be used to train the post-process algorithm, such as a convolutional neural network, to recognize patterns and features associated with defects and may involve feeding the images into the algorithm, which learns to differentiate between normal and defective micro LED characteristics. The post-process algorithm may then be validated and tested on a separate set of images not used during training to ensure it accurately identifies defects and the algorithm may be adjusted to improve accuracy and reduce false positives / negatives. Then high resolution images of the micro LEDs from the electroluminescence testing post-process 150 are inputted into the post-process algorithm and each image is analyzed to identify features that indicate defects based on what the algorithm learned during training. The post-process algorithm may output the locations and types of defects detected and provide detailed information for further analysis or repair, which may include quantifying defect sizes, categorizing defect types, and determining defect distribution across the micro LED array. In some embodiments, the post-process algorithm may improve the accuracy, efficiency, scalability, etc. of the defect inspection post-process. In some embodiments, the post-process algorithm may utilize supervised learning models, such as decision trees, random forests, gradient boosting machines, etc. which may be trained on historical data where the input features include types and numbers of defects, and the target variable is whether the substrate meets specifications and may learn complex patterns and interactions among defects. In some embodiments, the post-process algorithm may utilize anomaly detection algorithms which may identify unusual patterns that do not conform to expected behavior, such as detecting unusual combinations or intensities of defects that are likely to result in a substrate failing to meet specifications. In some embodiments, the post-process algorithm may utilize regression analysis or time-series forecasting to predict the future occurrence of defects based on trends and patterns, which may preemptively adjust manufacturing processes to reduce the likelihood of producing substrates that do not meet specifications. The ET post-process module 114 performs, at step 510, the post-process inspection. For example, the electroluminescence testing post-process may include defect inspection, repair technologies application, efficiency improvement, etc. The defect inspection may utilize advanced imaging systems to detect luminance variations and identify defective LEDs. The repair technologies may be methods such as ultraviolet irradiation, laser welding, and selective repairing either pick or laser-based, to address and fix identified defects. The efficiency improvement may aim to increase production yield by efficiently repairing or replacing bad dies, ensuring high-quality micro LED output. The ET post-process module 114 stores, at step 512, the data in the ET post-process database 124. The ET post-process module 114 stores the data in the ET post-process database 124, such as the data collected from the electroluminescence testing post-process 150, the output of the post-process algorithm, etc. The ET post-process database 124 may contain the substrate ID, the defects, if the substrate meets the final specification, if the demura algorithm has the capability to fix the defects, etc. The ET process database 122 may contain the data for each wafer lot that is being processed by the electroluminescence testing post-process 150. The ET post-process module 114 returns, at step 514, to the base module 108.

[0060]

[0060] A suitable Al algorithm for the post-processing control method described in step 508 would be Convolutional Neural Networks (CNNs), which are specifically made for tasks related to picture categorization and anomaly identification. The necessity to identify minute flaws in micro LEDs, like dead pixels, micro LED shorts, uneven brightness or mura, color shifts, crosstalk, and efficiency droop, among others, is what motivated this decision. Because CNNs can automatically extract and learn spatial hierarchies of characteristics from images, they are especially well-suited for evaluating high-resolution images to find features indicative of faults. To find operational flaws, the post-process algorithm would have to examine pictures taken during the electroluminescence testing post-process, under this procedure, a historical dataset containing high-resolution photos of micro LEDs under a range of conditions — including both flawless and flawed examples — would be used to train the CNN. By identifying patterns and properties linked to flaws, the training process enables the CNN to learn how to distinguish between normal and defective micro LED characteristics. In order for the CNN to correctly detect flaws when examining fresh images from the post-process, this learning process is essential. In order for this algorithm to work well, it has to have access to a large dataset that contains high-resolution photos of the tiny LEDs that were taken after the electroluminescence testing was completed. In order to guarantee that the CNN can be trained to recognize a diverse range of possible problems, the dataset ought to comprise a multitude of samples representing different kinds of flaws. Every image in the collection would include a label indicating whether or not faults were present, what kinds of defects there were, and perhaps even how severe they were or other pertinent details. In order for the algorithm to learn from the labeled instances and make predictions on unlabeled data, supervised learning requires this labeling. Furthermore, it would be ideal for the dataset to be updated on a regular basis with new samples to account for modifications to production procedures, the arrival of novel micro LED technologies, or the appearance of new kinds of errors. The post-process algorithm's accuracy and applicability over time depend on this continuous learning process. In real -world applications, fresh images from the electroluminescence testing post-process can be analyzed by the CNN after it has been trained and proven. The program would analyze every picture to find characteristics that might be signs of flaws, and it would then provide comprehensive information on the flaws that were found, such as their types, locations, and maybe even their sizes or other pertinent characteristics. After problems have been found, this output can be used to direct additional analysis, repair, or rework procedures, ultimately increasing manufacturing yield. This method offers a clever solution to the problem of identifying and controlling flaws in micro LEDs and serves as an example of how artificial intelligence and machine learning technologies may be used to enhance the quality control procedures in semiconductor production.

[0061]

[0061] FIG. 6 illustrates the ET historical module 116. The process begins with the ET historical module 116 being initiated, at step 600, by the base module 108. The ET historical module 116 connects, at step 602, to the electroluminescence testing pre-process 130, the electroluminescence testing process 140, and the electroluminescence testing post-process 150. The historical module 116 connects to the electroluminescence testing pre-process 130, such as the processes, such as photoluminescence testing, visual inspection, mechanical and bonding checks, etc. utilized prior to the electroluminescence testing process 140, designed to prepare the micro LED processed wafers for optimal performance during the electroluminescence process, the electroluminescence testing process 140, such as systems or equipment for automated electroluminescence testing of micro LEDs, including sample preparation, configuration, electroluminescence induction, data capture, analysis, reporting, decision making, etc., in which a controlled electrical current is applied to the micro LEDs, causing them to emit light through the electroluminescence phenomenon to assess the operational characteristics of the micro LEDs, such as their efficiency and uniformity, and the electroluminescence testing post-process 150, such as processes or methods for inspection to identify defective LEDs and subsequent repair or rework processes, which may include camerabased microscopic imaging systems to assess and measure the luminance of micro LED chips accurately to enhance production yield by efficiently repairing or replacing bad dies, employing methods such as ultraviolet irradiation, laser welding, and selective pick or laser repairing technologies.

[0062]

[0062] Synthetic Data and AI-Generated Digital Twins Example 2: In some embodiments, the system may utilize Synthetic Data and AI-Generated Digital Twins. For example, the system can simulate the electroluminescence testing process by generating synthetic datasets that replicate real-world scenarios using Al-generated digital twins of micro LED substrates. The ET inspection module 110 can receive synthetic data inputs that mimic variations in wafer thickness, surface defects, and epitaxial layer thickness. These inputs are generated by the Al to simulate a range of manufacturing conditions and potential defects. The ET inspection module 110 uses this synthetic data to predict the initial throughput percentage and temperature stability, calculating the optimal temperature settings for the electroluminescence testing process. This prediction is based on a linear regression model trained on historical data from the ET network database 126. Once the inspection algorithm determines the optimal settings, it sends this data to the ET process module 112. The ET process module 112, upon receiving these inputs, adjusts the electroluminescence testing parameters such as exposure dose and alignment accuracy. The module interfaces with the hardware through a control signal that modifies the voltage applied to the micro LEDs, ensuring that the settings align with the predicted optimal conditions. This control signal is crucial for maintaining the desired luminous intensity and chromaticity coordinates during testing. Simultaneously, the ET post-process module 114 utilizes the Al-generated digital twin to simulate the post-process scenario. It inputs synthetic defect data, such as dead pixels and color shifts, to assess the effectiveness of repair strategies like laser welding. The post-process algorithm employs a convolutional neural network trained on both real and synthetic high-resolution images to identify defect patterns and predict the success of repair actions. This module extracts data from the ET inspection database 120 and ET process database 122 to validate its predictions against historical outcomes. If the synthetic data indicates potential deviations in defect repair outcomes, the ET post-process module 114 sends feedback to the ET historical module 116. The ET historical module 116 aggregates this feedback with historical data, performing a machine learning analysis to refine the predictive models used by the ET inspection module 110 and ET process module 112. This analysis identifies trends and correlations that inform future process adjustments, ensuring continuous improvement in defect detection and repair accuracy. The integration of synthetic data allows the system to anticipate and mitigate issues before they occur in real-world production, enhancing the reliability and efficiency of the electroluminescence testing process. This comprehensive interaction between the ET inspection module 110, ET process module 112, and ET post-process module 114, facilitated by Al-generated digital twins, exemplifies how synthetic data can be leveraged to optimize micro LED manufacturing processes.

[0063]

[0063] The ET historical module 116 aggregates, at step 604, the data from the various types of processes and stores the data in the ET network database 126. The ET network database 126 may contain the historical data from the various processes performed by the electroluminescence testing pre-process 130, electroluminescence testing process 140, and electroluminescence testing postprocess 150. The ET network database 126 may contain the data parameters collected during the inspection process, such as wafer thickness, surface defects, cleanliness, wafer size, wafer pitch, wafer shape, epitaxial layer thickness, composition, uniformity, resistivity, doping concentration, breakdown voltage, etc., control parameters of the process, such as exposure dose, alignment accuracy, development time, alignment, luminous intensity, emission spectrum, chromaticity coordinates, color temperature, color uniformity, spatial distribution, peak wavelength, forward voltage, reverse leakage current, electroluminescence images, luminous efficacy or lumens per watt, external quantum efficiency, etc., and data parameters collected during post-processing, such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc.

[0064]

[0064] The ET historical module 116 performs, at step 608, the historical machine learning algorithm on the historical data stored in the ET network database 126. The historical machine learning algorithm may improve each stage of the micro LED manufacturing process, such as electroluminescence testing pre-process, electroluminescence testing process, and electroluminescence testing post-process. For example, the historical machine learning algorithm may be implemented in various stages of the electroluminescence testing process to optimize performance, accuracy, and efficiency. The historical machine learning algorithm may prepare the micro LED substrates for electroluminescence testing by ensuring optimal conditions and alignments by analyzing images of the micro LED substrates to detect any physical anomalies or misalignments before testing to ensure that only substrates in proper condition proceed to the electroluminescence testing phase, reducing the risk of inaccurate results due to pre-existing physical defects. The historical machine learning algorithm may process images of substrates and identify defects such as scratches, cracks, or contamination and determine whether the substrate requires cleaning, repair, or realignment before testing. The historical machine learning algorithm may conduct the electroluminescence testing, capturing and analyzing the electroluminescence of the micro LEDs to identify functional defects by dynamically adjusting electroluminescence testing parameters, such as current levels, exposure times, etc. in real-time based on the immediate feedback from the testing results to maximize the detection efficiency of functional defects while minimizing the risk of damaging the micro LEDs. The historical machine learning algorithm may learn the best testing strategies over numerous testing cycles and continuously improve its policy to achieve optimal testing outcomes, such as by adjusting the electrical current to the optimal level that balances between adequate excitation of the LEDs and preventing overheating. The historical machine learning algorithm may analyze the data collected from the electroluminescence testing to identify defects, evaluate the quality of the micro LED substrates, and recommend further actions by processing large datasets to classify substrates based on defect types and severity and identify patterns correlating with specific types of electroluminescence defects, such as dead pixels, non- uniform brightness, or color shifts. The historical machine learning algorithms may analyze testing data to classify substrates into categories based on their quality, identify substrates that pass the quality threshold and those that require rework or are deemed as waste, and for substrates with correctable defects, the historical machine learning algorithm may recommend specific postprocessing actions, like local repair or adjustment through techniques like Demura. In some embodiments, the historical machine learning algorithm may use the historical data stored in the ET network database 126 to determine trends to understand how varying parameters affect the final quality and identify optimal parameter settings to create the best outcomes. The historical machine learning algorithm adjusts the parameters in real-time for the ET inspection module 110, ET process module 112, and ET post-process module 114 to correct any defects.

[0065]

[0065] Large Language Models (LLMs) Example 1: In some embodiments, the system may utilize Large Language Models (LLMs). For example, the ET historical module 116 may employ an LLM to analyze and synthesize historical data stored within the ET network database 126 to generate insights that guide the electroluminescence testing process. The input data to the LLM includes a comprehensive dataset encompassing parameters such as wafer thickness, surface defects, cleanliness, wafer size, epitaxial layer thickness, and various electroluminescence testing parameters like exposure dose, alignment accuracy, luminous intensity, and emission spectrum. The LLM processes this data to identify patterns and correlations that may not be immediately apparent through traditional analytical methods. The decision logic of the LLM involves parsing the historical data to generate predictive models that estimate the likelihood of defects occurring under specific testing conditions. This predictive model is then used to inform the ET integration module 118, which adjusts the electroluminescence testing process 140 parameters in real-time. For instance, if the LLM predicts a high probability of non-uniform brightness defects due to a specific combination of wafer thickness and exposure dose, the ET integration module 118 can modify the exposure dose to mitigate this risk. The control outputs from the LLM include specific parameter adjustments communicated to the ET process module 112, such as altering the alignment accuracy or modifying the development time to ensure optimal electroluminescence testing conditions. These adjustments are transmitted via a control signal to the hardware interface of the electroluminescence testing process 140, ensuring precise implementation of the recommended changes. The ET post-process module 114 also interacts with the LLM by providing feedback on the defect types and frequencies detected during the post-process inspection. This feedback loop allows the LLM to refine its predictive models continuously, enhancing the accuracy of future predictions. A specific condition that would trigger the LLM system could be the detection of a significant deviation in temperature stability during the ET inspection module 110 phase. Upon detecting such a deviation, the LLM would analyze historical instances of similar deviations to predict potential impacts on throughput and defect occurrence. The LLM would then provide recommendations for adjusting the electroluminescence testing parameters to compensate for the deviation, ensuring that the testing process remains within optimal operational thresholds. This interaction between the LLM, ET historical module 116, ET integration module 118, and ET postprocess module 114 demonstrates a cohesive system where data-driven insights lead to actionable process improvements, ultimately enhancing the reliability and efficiency of micro LED electroluminescence testing.

[0066]

[0066] The ET historical module 116 sends, at step 610, the process adjustments to the ET inspection module 110, the ET process module 112, and the ET post-process module 114. The historical machine learning algorithm may prepare the micro LED substrates for electroluminescence testing by ensuring optimal conditions and alignments. The historical machine learning algorithm may conduct electroluminescence testing, capturing and analyzing the electroluminescence of the micro LEDs to identify functional defects. The adjustments are sent to the ET inspection module 110 and ET process module 112 to create a more consistent final product which is then verified by the ET post-process module 114. The ET historical module 116 returns, at step 612, to the base module 108.

[0067]

[0067] Linear regression is a good and reasonably simple algorithm for the historical machine learning algorithm that is detailed in step 606 of the ET historical module procedure. This decision is influenced by the requirement to spot patterns and forecast outcomes using the enormous amount of previous data gathered during the pre-, during, and post-process phases of electroluminescence (ET) testing procedures. By fitting a linear equation to observed data, linear regression is a fundamental technique in statistics and machine learning that models the relationship between a dependent variable and one or more independent variables. It can reveal which factors have the greatest impact on those outcomes and is especially helpful for making predictions about continuous data. The historical data collected in the ET network database 126 would be subjected to the Linear Regression method in this scenario. Numerous variables, including wafer features (e.g., thickness, surface defects), process parameters (e.g., alignment precision, light intensity, exposure dose), and post-process data (e.g., defect kinds, repair outcomes) would likely be included in this extensive dataset. The Linear Regression model was able to forecast the results of next manufacturing batches by examining these factors and finding connections between particular process parameters and the performance or quality of the micro LEDs. For example, the model may show how differences in alignment accuracy or exposure dose during the ET process step relate to the overall yield of micro LEDs without faults or the prevalence of specific flaws. This realization would enable real-time modifications to process parameters with the goal of maximizing manufacturing efficiency and product quality. This situation calls for the use of a well- organized and comprehensive dataset, which can only be obtained by accessing the historical records kept in the ET network database 126. Throughout the whole ET testing cycle, a wide range of data points, including both numerical and categorical data, covering every facet of the manufacturing process — from initial wafer preparation to final post-process inspection and repair — would need to be covered in these records. The ET historical module may forecast the effects of different process parameters on the quality of the end product by using historical production data and linear regression. The production of micro LEDs would benefit from increased efficiency, decreased waste, and better results as a result of this predictive capabilities, which would facilitate better informed decision-making and process optimization initiatives. A more sophisticated Al algorithm that might be especially helpful for the thorough analysis and optimization tasks outlined in the ET historical module that involve data from the pre-process, process, and post-process stages of electroluminescence (ET) testing is a Deep Learning model, more specifically a Multi-Layer Perceptron (MLP). Multiple layers of nodes, each fully connected to the next, make up the MLP class of feedforward artificial neural networks. They are appropriate for the multidimensional analysis needed in this situation because they can learn feature hierarchies and capture complicated relationships in vast datasets. The capacity of MLPs to learn non-linear correlations sets them apart and is essential for managing the diversity and complexity of data used in ET testing procedures. Wafer properties, process parameters, and the results of several testing and repair activities are only a few of the many variables that are involved in these processes. Because MLPs are non-linear, they may represent the complex relationships that exist between these factors and the micro LEDs' ultimate quality, providing information that more straightforward linear models are unable to provide. Using previous data gathered over the course of the ET testing cycle, the MLP would be trained as part of this deep learning technique. Wafer thickness, surface imperfections, cleanliness, wafer size, exposure dose, alignment precision, and defect types are a few examples of the input features that would be included in the dataset. A quality parameter, like the production of micro LEDs without defects or the electroluminescence process's efficiency, could be the target variable. The MLP can be trained on this data to determine the best process parameters by learning to predict how different factors will affect the quality of the end product. A large and rich dataset is necessary for an MLP to be successful in this application, as deep learning models often need a significant quantity of data to learn well. The ET network database 126 offers an appropriate basis for creating and refining an MLP model since it aggregates extensive historical data from the electroluminescence testing cycle. After being trained, the MLP model might be used to evaluate fresh or continuing production data, forecast results based on the state of the process, and suggest changes meant to boost output and efficiency. This feature would facilitate ongoing manufacturing process development, assisting in guaranteeing the reliable manufacture of high- quality tiny LEDs.

[0068]

[0068] Al Agents Example 2: In some embodiments, the system may utilize Al Agents. For example, an Al Agent can be deployed to coordinate the interaction between the ET inspection module 110, the ET process module 112, and the ET post-process module 114. Upon initiation, the Al Agent receives input data from the ET inspection module 110, which includes parameters such as wafer thickness, surface defects, and cleanliness. This data is collected during the electroluminescence testing pre-process 130 and is stored in the ET inspection database 120. The Al Agent analyzes this data using a decision tree model to classify wafers into categories based on quality metrics. If the Al Agent detects a wafer with a cleanliness level below a predefined threshold, it generates a control signal to the ET process module 112, instructing it to adjust the exposure dose and alignment accuracy during the electroluminescence testing process 140. This adjustment is crucial to compensate for the initial deficiencies detected by the ET inspection module 110. The Al Agent sends this control signal via the communication interface 104, ensuring real-time adjustments are made to the testing parameters. Concurrently, the Al Agent monitors the data flow from the ET process module 112 to the ET process database 122, where it records the adjusted alignment parameters and exposure settings. The Al Agent then interfaces with the ET post-process module 114, which collects post-process data such as defect types and locations from the electroluminescence testing post-process 150. This data is stored in the ET post-process database 124. The Al Agent employs a convolutional neural network (CNN) to analyze high- resolution images of the micro LEDs, identifying patterns that indicate defects such as dead pixels or color shifts. Upon detecting a pattern that suggests a defect, the Al Agent triggers a repair protocol within the ET post-process module 114, which may involve laser welding to correct the identified issues. The Al Agent ensures that these corrective actions are logged in the ET postprocess database 124 for future reference. Additionally, the Al Agent continuously updates the ET historical module 116 with aggregated data from the ET inspection database 120, ET process database 122, and ET post-process database 124. This historical data is used to refine the decision tree model, enhancing its predictive accuracy for future wafer assessments. The Al Agent's integration of data across these modules allows for a dynamic and responsive testing environment, where real-time adjustments are made based on the initial conditions and ongoing process feedback. This system is particularly effective when prompted by a condition such as a sudden increase in detected surface defects, which would trigger the Al Agent to initiate a comprehensive review and adjustment of the testing parameters across the modules. By maintaining a continuous feedback loop, the Al Agent ensures that the electroluminescence testing equipment operates at optimal efficiency, reducing the likelihood of defective micro LEDs proceeding through the manufacturing process. This detailed interaction between the Al Agent and the system modules provides a robust framework for maintaining high-quality standards in micro LED production.

[0069]

[0069] FIG. 7 illustrates the ET integration module 118. The process begins with the ET integration module 118 being initiated, at step 700, by the base module 108. The ET integration module 118 performs, at step 702, the integration machine learning algorithm. The integration machine learning algorithm may facilitate interactions between the electroluminescence testing pre-process 130, electroluminescence testing process 140, and electroluminescence testing post-process 150 stages of micro LED manufacturing. For example, the integration machine learning algorithm may be integrated across the electroluminescence testing providing a synergistic workflow between preprocessing, processing, and post-processing, to enhance the testing process. The integration machine learning algorithm may use the initial substrate condition, such as cleanliness, detected defects, alignment, etc. from the pre-process to inform the post-process analysis by utilizing the data on substrate preparation quality to help the post-process classify defects more accurately by correlating specific pre-test conditions with observed electroluminescence anomalies. For example, the quality of substrate preparation directly impacts the types of defects detected during electroluminescence testing and by understanding the correlation between pre-test conditions and electroluminescence outcomes, the post-process analysis can more accurately identify issues that are fixable versus inherent substrate flaws. The integration machine learning algorithm may continuously refine testing strategies based on the success of current and past tests from the electroluminescence testing process, such as if certain parameter settings consistently lead to better defect detection, the system may adjust future tests, accordingly, enhancing efficiency and accuracy over time. For example, real-time adjustments to testing parameters based on immediate feedback and historical data may ensure that the testing process becomes more efficient and accurate, reducing the likelihood of retests and enhancing the quality of data for post-process analysis. The integration machine learning algorithm may use insights from the pre-process about substrate condition, such as areas with minor physical imperfections, and adjust the focus and sensitivity of the electroluminescence testing to ensure that testing parameters are finely tuned to the substrate's specific characteristics, leading to more accurate defect detection. For example, the condition and preparation of the substrate may significantly influence the settings and focus areas for electroluminescence testing and information on detected physical imperfections may allow for adaptive testing strategies, prioritizing areas of concern and adjusting parameters to mitigate potential interference from known defects. In some embodiments, the integration machine learning algorithm may use the data stored in the ET network database 126 to perform a predictive modeling algorithm to make predictions or forecasts of the final product based on historical data and patterns from the previously created products. Predictive modeling involves the systematic analysis of historical data to identify patterns, correlations, and trends that can be used to build models capable of making accurate predictions about future outcomes. For example, the integration machine learning algorithm may be used for identifying defects across the electroluminescence testing pre- process 130, electroluminescence testing process 140, and electroluminescence testing postprocess 150. The historical data in the ET network database 126 may include information on defects and relevant operational parameters such as contamination, alignment, wafer size, uniformity, color temperature, dead pixels, color shifts, etc. The data is cleaned and preprocessed, addressing any missing values and ensuring that all variables are in a suitable format for modeling. This may involve normalization or scaling of numerical features. The dataset is split into training and testing sets. The training set is used to train the predictive model, and the testing set assesses its performance on unseen data. The model used may be linear regression, decision trees, ensemble methods, neural networks, etc. Then key features are identified that influence the defects and engineer new features if necessary. For example, exposure dose from the electroluminescence process 140 may have a significant impact on the defects identified. The selected predictive model is trained using the training dataset. The model learns patterns and relationships between operational parameters and defects. The model's performance is validated on a separate validation dataset and hyperparameters are fine-tuned to optimize its accuracy and generalization. The model is then evaluated and deployed by being sent to the electroluminescence testing pre-process 130, electroluminescence testing process 140, and electroluminescence testing post-process 150.

[0070]

[0070] Multimodal Al Example 2: In some embodiments, the system may utilize Multimodal Al. For example, the system integrates data from multiple sensory inputs to optimize the electroluminescence testing process for micro LEDs. The ET inspection module 110 collects initial data on wafer characteristics such as thickness, surface defects, and cleanliness using high- resolution cameras and spectrometers. This data is then processed by a convolutional neural network (CNN) to identify any potential defects or anomalies that might affect subsequent testing stages. The CNN analyzes visual patterns and correlates them with historical defect data stored in the ET inspection database 120, enabling it to predict the likelihood of defects affecting throughput. If the CNN identifies a potential issue, it triggers a control signal to the ET process module 112, instructing it to adjust the alignment parameters of the electroluminescence testing process 140. This adjustment is achieved by modifying the position of the substrate using nanopositioning stages, ensuring precise alignment before the testing begins. The ET process module 112 then uses the adjusted alignment parameters to conduct the electroluminescence testing, applying a controlled electrical current to the micro LEDs and capturing the emitted light using spectroradiometers. The data collected, including luminous intensity and emission spectrum, is fed into a recurrent neural network (RNN) that continuously learns from the historical data stored in the ET network database 126. The RNN evaluates the real-time testing data against historical performance metrics, such as luminous efficacy and external quantum efficiency, to determine if the current testing setup yields optimal results. If deviations are detected, the RNN sends a feedback loop to the ET integration module 118, which performs a decision tree analysis to determine whether further adjustments are necessary. The ET integration module 118 then communicates with the ET post-process module 114, providing it with refined parameters for defect inspection, such as specific focus areas for high-resolution imaging based on the initial substrate conditions and testing outcomes. The ET post-process module 114 utilizes these parameters to perform a detailed inspection, identifying defects such as dead pixels or color shifts, and stores the findings in the ET post-process database 124. This data is then analyzed by a gradient boosting machine within the ET historical module 116, which compares it against historical defect patterns to predict future occurrences and suggest process optimizations. The integration of these Al models allows the system to dynamically adjust testing parameters in real-time, ensuring high-quality micro LED production. A specific condition that would trigger this Multimodal Al system is the detection of a temperature stability deviation exceeding 2 degrees Celsius during the initial inspection phase, prompting the system to recalibrate the testing parameters to maintain optimal throughput and defect detection accuracy.

[0071]

[0071] Large Language Models (LLMs) Example 2: In some embodiments, the system may utilize Large Language Models (LLMs). For example, the ET integration module 118 may employ an LLM to facilitate seamless interaction between the ET inspection module 110, the ET process module 112, and the ET post-process module 114. The LLM is trained on a comprehensive dataset comprising historical electroluminescence testing data, including parameters such as wafer thickness, surface defects, alignment accuracy, luminous intensity, and defect types like dead pixels and color shifts. This training enables the LLM to understand complex relationships and dependencies between various process parameters and their impact on micro LED quality. Upon receiving real-time data from the ET inspection module 110, such as wafer thickness and surface defect measurements, the LLM analyzes this input to predict potential issues that may arise during the electroluminescence testing process. The LLM uses its learned model to generate a set of recommended adjustments to the testing parameters, such as modifying the exposure dose or alignment accuracy, which are then communicated to the ET process module 112. This module utilizes the LLM's output to fine-tune the electroluminescence testing parameters, ensuring optimal conditions for defect detection and process efficiency. Concurrently, the ET post-process module 114 receives data from both the ET process module 112 and the LLM, including the adjusted testing parameters and the predicted defect types. The ET post-process module 114 uses this information to perform a detailed inspection of the resultant micro LED substrates, focusing on areas identified by the LLM as having a higher likelihood of defects. The post-process module employs high- resolution imaging systems to capture electroluminescence images, which are then compared against the LLM's predictions to validate the accuracy of the defect detection process. Control signals generated by the ET process module 112, based on the LLM's recommendations, adjust the alignment mechanisms and exposure settings in real-time, ensuring that the testing environment is dynamically optimized for each batch of substrates. A specific condition that would trigger the LLM system is the detection of a significant deviation in wafer thickness beyond a predefined threshold during the ET inspection module 110's initial assessment. This deviation prompts the LLM to analyze historical data for similar instances and predict the optimal parameter adjustments required to mitigate potential defects. By integrating the LLM's predictive capabilities across multiple modules, the system achieves a cohesive and adaptive testing workflow, where data- driven insights guide real-time process adjustments, thereby enhancing the overall quality and yield of the micro LED manufacturing process.

[0072]

[0072] Generative Video Al Example 2: In some embodiments, the system may utilize Generative Video Al. For example, the ET integration module 118 may employ a Generative Video Al system to enhance the electroluminescence testing process by creating synthetic video data that simulates the behavior of micro LEDs under various testing conditions. The input data for this Al system includes high-resolution images and videos from the electroluminescence testing process 140, capturing the emitted light patterns, intensity distributions, and spatial uniformity of micro LEDs. The decision logic of the Generative Video Al involves training on historical datasets stored in the ET network database 126, which contain detailed records of previous testing scenarios, including variations in exposure dose, alignment accuracy, and luminous intensity. The Al system uses this data to generate video simulations that predict how micro LEDs will perform under different electroluminescence testing parameters. These simulations are then analyzed to identify potential defects or anomalies that may not be evident in static images. The control outputs of the Generative Video Al are integrated into the ET post-process module 114, where the Al-generated video data is used to refine the defect detection algorithms. Specifically, the ET post-process module 114 utilizes the video simulations to perform a more comprehensive post-process inspection, allowing for the identification of transient defects such as flickering or intermittent brightness variations. This interaction between the ET integration module 118, ET network database 126, and ET postprocess module 114 facilitates a dynamic feedback loop, where the Generative Video Al continuously updates its models based on real-time data from ongoing testing processes. A specific condition that would trigger the Generative Video Al system is the detection of a deviation in the luminous intensity beyond a predefined threshold during the electroluminescence testing process 140. Upon detecting such a deviation, the ET process module 112 sends a control signal to the ET integration module 118, prompting the Generative Video Al to simulate the potential impact of this deviation on the overall performance of the micro LEDs. The Al system then generates a video simulation that is analyzed to determine whether the deviation could lead to defects such as color shifts or non-uniform brightness. The results of this analysis are communicated back to the ET process module 112, which may adjust the testing parameters accordingly to mitigate the risk of defects. This interaction ensures that the electroluminescence testing process is adaptive and responsive to real-time data, thereby enhancing the accuracy and reliability of defect detection. The Generative Video Al system thus plays a crucial role in optimizing the electroluminescence testing process by providing predictive insights and enabling proactive adjustments to testing parameters. The detailed integration of the Generative Video Al with the ET integration module 118, ET network database 126, and ET post-process module 114 exemplifies a sophisticated approach to leveraging Al in micro LED manufacturing, offering a clear pathway for engineers to implement and refine this technology within the system.

[0073]

[0073] The ET integration module 118 connects, at step 704, to the electroluminescence testing pre-process Al module 136, the electroluminescence testing process Al module 146, and the electroluminescence testing post-process Al module 156. The ET integration module 118 connects to the electroluminescence testing pre-process 130, such as the processes, such as photoluminescence testing, visual inspection, mechanical and bonding checks, etc. utilized prior to the electroluminescence testing process 140, designed to prepare the micro LED processed wafers for optimal performance during the electroluminescence process, the electroluminescence testing process 140, such as systems or equipment for automated electroluminescence testing of micro LEDs, including sample preparation, configuration, electroluminescence induction, data capture, analysis, reporting, decision making, etc., in which a controlled electrical current is applied to the micro LEDs, causing them to emit light through the electroluminescence phenomenon to assess the operational characteristics of the micro LEDs, such as their efficiency and uniformity, and the electroluminescence testing post-process 150, such as processes or methods for inspection to identify defective LEDs and subsequent repair or rework processes, which may include camerabased microscopic imaging systems to assess and measure the luminance of micro LED chips accurately to enhance production yield by efficiently repairing or replacing bad dies, employing methods such as ultraviolet irradiation, laser welding, and selective pick or laser repairing technologies. The ET integration module 118 sends, at step 706, the process adjustments to the electroluminescence testing pre-process Al module 136, the electroluminescence testing process Al module 146, and the electroluminescence testing post-process Al module 156. For example, the ET integration module 118 may send the uniformity, exposure dose, color shifts, etc. adjustments to each of the processes to minimize the defects of the substrates during the electroluminescence testing process. In some embodiments, the ET integration module 118 may send the integration machine learning algorithm to the processes, allowing the systems to further enhance the optimization. The ET integration module 118 returns, at step 708, to the base module 108.

[0074]

[0074] For the integration machine learning algorithm required in step 702 within the ET integration module, a suitable and relatively simple Al algorithm is the Decision Tree algorithm. This algorithm is chosen for its simplicity and interpretability, making it an excellent tool for integrating and facilitating interactions between the various stages of micro LED manufacturing, namely the electroluminescence testing pre-process, process, and post-process stages. The Decision Tree algorithm works by breaking down a dataset into smaller subsets while simultaneously developing an associated decision tree incrementally. The final result is a tree with decision nodes and leaf nodes, where each decision node represents a test on an attribute, each branch represents the outcome of the test, and each leaf node represents a class label or decision taken after computing all attributes. This structure makes Decision Trees particularly useful for the ET integration module’s objective of creating a synergistic workflow that enhances the entire testing process through the integration of data across different stages. In this context, the integration machine learning algorithm could leverage a Decision Tree to analyze data from the pre-processing stage, such as substrate cleanliness and detected defects, and use this information to inform decisions in the post-processing analysis. By understanding the correlation between pre-test conditions and electroluminescence outcomes, the algorithm can more accurately pinpoint issues that are either fixable or indicative of inherent substrate flaws. Moreover, by continuously refining testing strategies based on the successes of current and past tests, the algorithm can facilitate realtime adjustments to testing parameters, thus enhancing efficiency and accuracy over time. The dataset for this algorithm would comprise a comprehensive collection of variables from the ET network database 126, encompassing initial substrate condition, processing parameters, and postprocessing outcomes. This dataset would include detailed records on contamination, alignment, wafer size, uniformity, color temperature, dead pixels, and color shifts, among others. Through systematic preprocessing to clean and normalize the data, followed by division into training and testing sets, the Decision Tree model can be trained to identify key features that influence defects and operational parameters that lead to successful outcomes. By employing a Decision Tree algorithm, the ET integration module can not only predict the likelihood of defects based on pretest conditions but also recommend specific adjustments to the electroluminescence testing process and post-process defect classification. This leads to a more streamlined and efficient manufacturing process, where insights from one stage directly inform and enhance the operations of another, all facilitated through an understandable and straightforward Al-driven approach.

[0075]

[0075] FIG. 8 illustrates the ET inspection database 120. The ET inspection database 120 provides an example of the results of the inspection algorithm performed in the ET inspection module 110. The ET inspection database 120 may contain the wafer ID, the initial throughput percentage, the initial temperature stability, the adjusted throughput percentage, the temperature stability optimized by the inspection algorithm, the throughput percentage optimized by the inspection algorithm, etc. For example, the ET inspection database 120 provides an example of the inspection algorithm optimizing the temperature to optimize the throughput percentage. The inspection algorithm may achieve a balance between high throughput and the accuracy of defect detection. High throughput is essential for cost-effective manufacturing, but it may compromise the detection accuracy which may allow defective units to proceed to later stages of production and may result in increased costs if defects are detected only at the EL testing phase or beyond. The first example illustrates a "Temperature Stability (degrees C)" column which displays deviations from the optimal temperature condition (0 degrees C), for normal process variations such as temperature that may occur in EL Testing, the positive values indicate a temperature above the optimal and the negative values indicating below optimal. The "Adjusted Thruput for EL Testing (%)" column reflects the throughput adjustment based on the temperature stability, where each degree of deviation adjusts the throughput by approximately -2 percent. This example illustrates how temperature variations can impact the efficiency of the EL testing process, to maintain as close to the optimal temperature as possible to ensure high throughput. The second example reflects how the inspection algorithm may reduce temperature stability deviations and improve throughput results for EL testing. The example data in the inspection database 120 illustrates approximately a 50 percent reduction, or improvement, in temperature stability deviations due to the inspection algorithm optimizations with the improved optimized throughput percentage. For example, the inspection algorithm may reduce temperature stability deviations and potentially improve throughput results for EL testing. The inspection algorithm may predict optimal EL testing parameters, such as temperature settings, that maximize throughput while ensuring high-quality testing outcomes. The inspection algorithm may use a linear regression model to predict optimal testing parameters based on historical data and focus on achieving the best temperature conditions and throughput rates. The inspection algorithm may use historical data, such as data on wafer characteristics, including material composition, layer thickness, and, EL testing parameters, including temperature settings, electrical input, etc., observed temperature stability deviations, resulting throughput rates, etc. The inspection algorithm may identify features, or variables, that impact throughput, such as initial temperature settings, electrical input, material properties, historical temperature deviations, etc. The inspection algorithm is trained on the historical data to learn the relationship between the selected features and the throughput rate. The historical data may be formatted, normalized, standardized, etc., and is split into training and testing sets. The inspection algorithm is trained using the input of the selected features from the historical data and the output of the algorithm is the throughput rate. The algorithm is then evaluated in which the algorithm’s performance is assessed using the testing set and metrics like R-squared, and R2, to determine how well the model predicts throughput rates based on the input features. Then before the EL test, the current wafer characteristics and initial testing parameters are inputted into the algorithm and the algorithm predicts the adjustments needed to optimize the throughput, such as suggesting the ideal temperature settings. In some embodiments, thermal imaging technology may be used to capture and analyze the infrared radiation emitted by the wafer to visualize its temperature distribution. The testing method may include preparation for thermal imaging through wafer cleaning and setting up the environment, using infrared cameras and imaging software for analysis, conducting the test by collecting a baseline measurement, performing powered state imaging, and capturing the thermal images. The data is analyzed, such as identifying the temperature variations, mapping the hot and cold spots, and correlating the mapping to the wafer features. Then the EL testing parameters may be adjusted, such as altering the electrical input, adjusting the testing environment's temperature, or implementing localized cooling or heating solutions.

[0076]

[0076] FIG. 9 illustrates the ET process database 122. The ET process database 122 provides an example of the results of the control algorithm performed in the ET process module 112. The ET process database 122 may contain the substrate ID, initial alignment, final alignment after the optimization adjustments by the control algorithm, etc. For example, the ET process database 122 illustrates the alignment specifications for ten different substrates in a micro LED manufacturing scenario, where precise alignment is needed for accurate testing and overall device performance. The Initial Alignment column may represent the starting alignment values, with the specification of 0.1 pm as a reference point. The column illustrates an example of variances encountered in manufacturing and handling which is displayed by the substrate's initial alignment value being altered throughout the process. The Final Alignment After Al-based Adjustments column reflects the alignment values after making necessary corrections to counter initial deviations outputted by the control algorithm. These adjustments ensure that substrates are aligned with the requisite precision for micro LED testing, in which misalignments may impact test accuracy and potentially cause damage to the micro LEDs themselves. For example, the control algorithm may optimize alignment based on a dataset of previous alignments, outcomes, and various parameters affecting alignment quality to prevent misalignment which may lead to inaccurate test results or damage to the micro LEDs. For example, the alignment process may include sample preparation, such as cleaning and inspection, securing the micro LED, utilizing alignment equipment, such as high- resolution microscopes, optical alignment systems, vision systems, fiducial markers, detection systems, etc. to initially position the micro LED sample based on visual markers or reference points. The micro LED sample’s position may be fine-tuned through micromanipulators, piezoelectric positioners, nanopositioning stages, laser interferometer systems, alignment stages with fine adjustment screws, etc. Once the sample is aligned properly it is securely attached to the mounting platform or holder, which may involve vacuum sealing, adhesives, etc., and the attachment is verified. The alignment is re-checked using the alignment equipment and the electrical connections are verified. The control algorithm may analyze visual imagery by being trained on a dataset comprising images of aligned and misaligned micro LED substrates, along with corresponding alignment parameters and outcomes allowing the control algorithm to recognize patterns and features associated with optimal alignment. For example, the control algorithm may compile a dataset of high-resolution images of micro LED substrates at various stages of the alignment process, including pre-alignment, post-manual alignment, and post-AI alignment stages and each image may be tagged with metadata detailing the alignment parameters used and the success rate of the alignment. The control algorithm may be trained on the dataset which may involve feeding the images into the algorithm, which may learn to identify features and patterns correlating with successful alignments. The output of the control algorithm may be to predict the optimal alignment parameters. During the micro LED manufacturing process, live images of the substrate alignments are fed into the trained control algorithm, and it may assess the images in real time and predict the most effective adjustment parameters to achieve optimal alignment. The predicted parameters from the control algorithm may be used to automatically adjust the alignment tools, which may involve adjusting the position of the substrate, modifying the angle of incidence for alignment lasers, or altering the pressure applied by mechanical aligners. The outcomes of these adjustments may be continuously monitored and fed back into the control algorithm which allows it to refine its predictions over time. The control algorithm may process complex image data, recognize patterns, and learn from the visual cues associated with successful and unsuccessful alignments. The control algorithm may influence the control mechanism of the alignment tools by predicting the necessary adjustments and sending signals to the mechanical or optical components responsible for making precise movements, which may include motorized stages for lateral adjustments, tilt platforms for angular corrections, or laser steering systems for precise positioning.

[0077]

[0077] FIG. 10 illustrates the ET post-process database 124. The ET post-process database 124 provides an example of the results of the post-process algorithm performed in the ET post-process module 114. The example data illustrates the detection of small-scale defects in various micro LED substrates, highlighting the defect type(s), the substrate's compliance with final specifications, and the applicability of Demura fix capabilities using the post-process algorithm learning on historical data. In some embodiments, the defect key for the second column may be 1. Dead Pixels, 2. Micro LED Shorts, 3. Non-Uniform Brightness (Mura), 4. Color Shifts, 5. Crosstalk, 6. Efficiency Droop, 7. Crystalline Defects, 8. Point Defects. The examples provide an overview of the presence and type of small-scale defects detected in a sample of micro LED substrates, assessing their compliance with final product specifications. The defects, which range from micro-crack formation to thin film peeling, illustrate the importance of inspection processes in the production of micro LEDs. The example also evaluates whether Demura's fix capabilities, leveraging the post-process algorithm learning from historical data, may address non-conformities. For example, the postprocess algorithm may be used to detect small-scale defects, such as dead pixels, micro LED shorts, non-uniform brightness or mura, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc. The post-process algorithm may use high-resolution images of the micro LEDs that capture the micro LEDs in their operational state, allowing for observation of luminance, color uniformity, and other characteristics that would indicate the presence or absence of defects. The post-process algorithm may use a historical dataset of high-resolution images of the micro LEDs that includes defect-free examples and defective examples. The dataset may be used to train the post-process algorithm, such as a convolutional neural network, to recognize patterns and features associated with defects and may involve feeding the images into the algorithm, which learns to differentiate between normal and defective micro LED characteristics. The post-process algorithm may then be validated and tested on a separate set of images not used during training to ensure it accurately identifies defects and the algorithm may be adjusted to improve accuracy and reduce false positives / negatives. Then high resolution images of the micro LEDs from the electroluminescence testing post-process 150 are inputted into the post-process algorithm and each image is analyzed to identify features that indicate defects based on what the algorithm learned during training. The post-process algorithm may output the locations and types of defects detected and provide detailed information for further analysis or repair, which may include quantifying defect sizes, categorizing defect types, and determining defect distribution across the micro LED array. In some embodiments, the post-process algorithm may improve the accuracy, efficiency, scalability, etc. of the defect inspection post-process. In some embodiments, the post-process algorithm may utilize supervised learning models, such as decision trees, random forests, gradient boosting machines, etc. which may be trained on historical data where the input features include types and numbers of defects, and the target variable is whether the substrate meets specifications and may learn complex patterns and interactions among defects. In some embodiments, the postprocess algorithm may utilize anomaly detection algorithms which may identify unusual patterns that do not conform to expected behavior, such as detecting unusual combinations or intensities of defects that are likely to result in a substrate failing to meet specifications. In some embodiments, the post-process algorithm may utilize regression analysis or time-series forecasting to predict the future occurrence of defects based on trends and patterns, which may preemptively adjust manufacturing processes to reduce the likelihood of producing substrates that do not meet specifications. In some embodiments, the ET post-process database 124 may store other parameter data collected or calculated from the ET post-process module 114, such as dead pixels, micro LED shorts, non-uniform brightness, color shifts, crosstalk, efficiency droop, crystalline defects, point defects, etc.

[0078]

[0078] The functions performed in the processes and methods may be implemented in differing order. Furthermore, the outlined steps and operations are only provided as examples, and some of the steps and operations may be optional, combined into fewer steps and operations, or expanded into additional steps and operations without detracting from the essence of the disclosed embodiments.

Claims

1. CLAIMSWhat is claimed is,1. An automated electroluminescence testing (AET) equipment method for micro LED chips system, comprising: an automated electroluminescence testing equipment inspection module that measures at least one parameter from a first group of parameters and calculates an output of at least one parameter from a second group of parameters to help control the automated electroluminescence testing equipment processing module; an automated electroluminescence testing equipment processing module which inputs the at least one parameter from said second group to control the electroluminescence testing equipment using at least one parameter from a third group of parameters; an automated electroluminescence testing equipment post-processing module which measures at least one parameter from a fourth group of parameters that determines how well the parameters of said first and second group were running; an automated electroluminescence testing equipment historical module which calculates a plurality of group parameters based upon parameters of the first group, second group, third group, and fourth group; and an automated electroluminescence testing equipment integration module to update the automated electroluminescence testing equipment processing module based upon the results of the automated electroluminescence testing equipment historical module.

2. The method of claim 1, wherein the AET process module acts as the core controller, utilizing inputs from the AET inspection module to govern the electroluminescence testing process to ensure optimal exposure dose, alignment accuracy, development time and luminous intensity.

3. The method of claim 1, wherein the AET post-process module conducts inspections to assess the execution of parameters and identify potential deviations, and the AET historical module analyzes data collected across various stages, allowing for continuous learning and patternrecognition and subsequently a AET integration module then employs historical insights to update and optimize the process, enabling adaptive improvements over time.

4. The method of claim 1, wherein a communication interface, which is a hardware or a software component that enables communication between an AET Al network and an AET pre-process, an AET process, and an AET post-process.

5. The method of claim 4, wherein the communication interface comprises a set of protocols, rules, and standards that define how information is transmitted and received between devices is a physical connector, wireless network, or software application and include components such as drivers, software libraries, and firmware that are used to control and manage the communication process.

6. The method of claim 4, wherein the communication interface is compatible with a USB, aBluetooth, or a Wi-Fi and the communication interface communicates with a network.

7. The method of claim 1, wherein there is a memory that comprises suitable logic, circuitry, and / or interfaces that are configured to store a machine code and / or a computer program with at least one code section executable by a processor.

8. The method of claim 1, wherein there is a base module, which initiates the AET inspection module, the AET process module, the AET post-process module, the AET historical module, and the AET integration module.

9. The method of claim 8, wherein the AET inspection module begins by being initiated by the base module wherein the AET inspection module connects to the AET pre-process wherein the AET inspection module collects data from the AET pre-process.

10. The method of claim 9, wherein the AET inspection module performs an inspection algorithm, and the AET inspection module sends the determined process data from the inspection algorithmto the AET process module wherein the AET inspection module stores the data in the AET inspection database and AET inspection module returns to the base module.

11. The method of claim 8, wherein the AET process module begins by being initiated by the base module wherein the AET process module receives the process data, such as the data on the microLED processed wafer, from the AET inspection module and the AET process module connects to the AET process.

12. The method of claim 11, wherein the AET process module collects the data from the AET process and the AET process module performs a control algorithm and the AET process module adjusts the control parameters of the AET process and executes the AET process wherein AET process module stores the data in the AET process database and the AET process module returns to the base module.

13. The method of claim 8, wherein the AET post-process module begins by being initiated by the base module and the AET post-process module connects to the AET post-process wherein the AET post-process module collects the post-process data.

14. The method of claim 13, wherein the AET post-process module extracts the data from the AET inspection database and AET process database and performs the post-process algorithm subsequently the AET post-process module performs the post-process inspection wherein the AET post-process module stores the data in the AET post-process database and the AET post-process module returns to the base module.

15. The method of claim 8, wherein the AET historical module begins by being initiated by the base module and the AET historical module connects to the AET pre-process, the AET process, and the WB post-process wherein the AET historical module aggregates data from the various types of processes and stores the data in the AET network database.

16. The method of claim 15, wherein the AET historical module performs a historical machine learning algorithm on the historical data stored in the AET network database and sends the processadjustments to the AET inspection module, the AET process module, and the AET post-process module and subsequently the AET historical module returns to the base module.

17. The method of claim 8, wherein the AET integration module begins by being initiated by the base module and the AET integration module performs the integration machine learning algorithm and connects to the AET pre-process Al module, the AET process Al module, and the AET postprocess Al module.

18. The method of claim 17, wherein the AET integration module sends the process adjustments to the AET pre-process Al module , the AET process Al module , and the AET post-process Al module and the AET integration module returns to the base module.

19. An automated electroluminescence testing (AET) equipment of micro LED substrates system, comprising: an automated electroluminescence testing equipment inspection module; an automated electroluminescence testing equipment process module; an automated electroluminescence testing equipment post-process module; an automated electroluminescence testing equipment historical module; and an automated electroluminescence testing equipment integration module; wherein, an automated electroluminescence testing equipment inspection module is used to determine the quality of incoming processing material and an automated electroluminescence testing equipment process module is used to process the processing material, an automated electroluminescence testing equipment post-process module is used to inspect the resultant process material for acceptability, and an automated electroluminescence testing equipment historical module is used to analyze the post process results of the most recent process, against all historical data to find enhanced process changes in the automated electroluminescence testing equipment process module and an automated electroluminescence testing equipment integration module to update the automated electroluminescence testing equipment process module based upon the results of the automated electroluminescence testing equipment historical module.

Images