Yield Enhancement for Micro-LED Production

Automatically optimizing fabrication processes like etching or deposition using machine learning and statistical inference to guide process development. The technique involves using models to simulate the fabrication process, characterizing the uncertainty in the model predictions, and then using an acquisition function to select optimal process parameters for the next experiment. The function balances exploration to further constrain model uncertainty with exploitation to drive closer to target specifications. By iteratively selecting parameters, optimal sets can be found even with multiple competing specifications and many adjustable parameters.

Source

HEMT driven MicroLED integrated backplane and manufacturing method for high resolution and large size MicroLED displays that avoids the challenges of transfer bonding and mass transfer methods. The method involves growing a HEMT epitaxial structure along with the RGB MicroLED chips on the same substrate. This allows direct integration of the HEMT and MicroLEDs at the epitaxial growth stage, eliminating the need for separate wafer processing and transfer steps. The integrated HEMT-MicroLED structure is then transferred to the display backplane. This provides a compact, efficient and reliable MicroLED display solution with improved yield and performance compared to separate wafer processing and transfer techniques.

Source

A method for monitoring annealing uniformity of semiconductor chips to improve yield by detecting voltage in the annealed chips instead of depositing thick metal layers. The method involves forming P-type contacts on epitaxial layers, annealing, then forming N-type contacts and annealing again. Voltage is measured in the N-type contact areas to monitor annealing uniformity across the chips. This reduces steps compared to depositing thick metal layers for voltage measurement.

Source

Method to sort and process sapphire substrate wafers after cutting to reduce waste and improve yield. The method involves classifying the cut wafers based on warpage and bending. Wafers with higher defects undergo annealing and grinding to improve quality. Then, the wafers are sorted again and polished using parameters specific to their defect level. This targeted processing reduces loss compared to uniform polishing.

Source

A manufacturing system that dynamically adjusts processing parameters of each step in a multi-step manufacturing process to achieve a desired final quality metric for the product. The system uses a monitoring platform to track product progress at each step and a control module to adjust parameters based on projected final quality. It leverages reinforcement learning to project final quality based on product state at each step. This allows real-time optimization of processing conditions as the product moves through the steps.

Source

Automatically identifying and correcting deviations in chip manufacturing processes like glue coating to improve yield and reliability. The method involves decomposing the target glue thickness into multiple processes with single thicknesses, monitoring those processes in real-time, identifying deviations using a neural network, and adjusting parameters for the next process based on the deviation impact. This iterative correction improves glue uniformity and reduces defects. The network analyzes factors like platform speed, glue dispense rate, and amount to determine deviation impact on photoresist uniformity.

Source

Using AI to detect defects in LED wafers during production to improve yield by catching defects earlier in the process. The method involves capturing images of the wafers at various stages using machine vision, identifying defects using an AI module that learns defect characteristics, processing the images to mark areas with and without defects, calculating yield based on the marked areas, and displaying the results. This allows identifying areas prone to defects and making process adjustments to prevent them.

Source

Monitoring chip annealing uniformity in semiconductor manufacturing to improve yield and performance. The method involves annealing contacts and the light-emitting layer separately on multiple chips in multiple wafers, then voltage testing the electrode areas on all the chips to monitor annealing uniformity across the wafers. This provides more comprehensive annealing monitoring compared to just testing a few points.

Source

Method to measure flaws in multi-environment Micro LED epitaxial wafers for improving overall yield and quality by detecting and addressing defects in different environmental conditions. The method involves processing the wafers, fixing them in a jig, and moving them between temperature, humidity, and gas chambers for comparison measurements. By subjecting the same wafer to variations in environment, it allows observing how flaws change and integrally detecting defects under continuous environment changes.

Source

Detecting defects in epitaxial wafers during production to reduce waste and improve yield. The method involves monitoring heating power data during epitaxial growth and looking for fluctuations. If fluctuations are detected, it indicates potential pin mark defects in the wafer. This allows stopping the production process to avoid producing more defective wafers.

Source

Optimizing semiconductor manufacturing process parameters to improve critical dimension quality and yield. The method involves establishing a comprehensive model based on process parameters and quality parameters, and using it to find the optimal process parameter values for critical dimensions. The model can predict quality based on process parameters, allowing selection of the best values for optimized critical dimensions. By optimizing process parameters based on critical quality attributes, it reduces wafer loss and improves yield.

Source

Method to analyze and improve LED epitaxial wafer quality by reverse engineering the effect of patterned substrates. The method involves step-by-step etching and inspection of the epitaxial layers to connect substrate defects with layer anomalies. Macroscopic and microscopic analysis is done on the epitaxial surface and etched surfaces to identify abnormalities. By sequentially etching and inspecting layers, the correlation between substrate features and epitaxial defects is established. This enables understanding of substrate impact on epitaxy and optimizing processes to reduce abnormal products.

Source

A micro-LED display device that can meet high-resolution requirements and has a high micro-LED product yield for manufacturing. The device has a circuit substrate with micro-LED units on one side and a metal conductive layer on the other. The conductive layer contacts the epitaxial layer of the micro-LEDs and has light conversion regions. A light conversion layer in some regions converts emitted light. Light-shielding structures on the epitaxial layer cover some areas. The metal layer is thicker than the epitaxial layer.

Source

Display device using micro-LEDs with a method to selectively separate defective micro-LED chips before assembly on the display substrate. It involves arranging micro-LEDs in different types of assembly grooves on the substrate. Micro-LEDs from the chamber are moved magnetically into the grooves. The grooves and electrodes are designed to exert different dielectrophoresis forces on normal versus defective micro-LEDs. This allows only the defective chips to be recovered while the good ones remain in place.

Source

Reducing non-radiative recombination of micro LED to prevent the non-radiative recombination occurring on the etched sidewalls of the micro LEDs. The reduction includes supplying at least one etched LED epitaxial wafer, laying a first ALD to the etched LED epitaxial wafer in a first temperature range; performing a second ALD to the etched LED epitaxial wafer etched by the first ALD in a second temperature range, and forming a passivation layer on the etched sidewalls of the mesas later being the micro LEDs.

Source

Method for monitoring semiconductor wafer fabrication processes using optical inspection to improve yield management and defect analysis. The method involves using machine learning to extract features from optical inspection images, calculate monitoring metrics like defect counts, and determine separability metrics that quantify how well classified defects meet inspection criteria. By analyzing separability trends, it can distinguish between process variations within specification versus true defects. This allows isolating and monitoring the impact of variable defect capture rates and counts on wafer inspection results due to process variations, which improves process control and yield analysis.

Source

Manufacturing process for an ultra-high-resolution Micro-LED micro-display screen with a simple process and few pixel defects. The process includes: providing an LED epitaxial wafer; forming isolation columns on the wafer to divide it into chip regions; placing conductive solders in the chip regions; performing a first CMP process to remove the conductive solders outside the chip regions; forming electrodes and LED light-emitting structures on the wafer; performing a second CMP process to remove upper parts of the electrodes and LED structures outside the chip regions; and removing the isolation columns to separate the chips. The process enables producing micro-displays with ultra-high pixel density and resolution suitable for AR/VR headsets.

Source

MicroLED array with improved efficiency and yield for displays and lighting applications. The microLED array is formed by growing a monolithic microLED stack on a relaxed lattice constant substrate. The stack has layers with different compositions to introduce strain. This strain is relieved by deforming the top layer during processing. This reduces defects and improves efficiency compared to conventional microLEDs where the active layer is strained. The relaxed microLED array can be formed in a single process by selectively removing and heat treating layers.

Source

A production control method for LED devices that improves yield by optimizing process parameters and environmental conditions. The method involves collecting production step parameters and environmental indicators during LED manufacturing, stopping production after a set number of devices, inspecting them, and adjusting step parameters and environment if quality requirements are not met. This iterative optimization based on real-time data helps control factors that affect LED quality and yield.

Source

A method for manufacturing micro-LED display panels that simplifies the mass transfer of millions of micro-LEDs by bonding them directly to a backplane substrate through a conductive layer on the LED surface. The method starts with growing the LED epitaxial layers on a base substrate that can then be peeled off by laser lift-off. The peeled LEDs are then bonded to a backplane using a conductive layer. This eliminates the need for separate transfer steps and improves yield compared to conventional methods.

Source