Optical Detection Method Of Microparticles On Wafer Surface

The rapid development of the semiconductor industry has caused the industrial field to put forward higher requirements for the fineness of the semiconductor processing technology.In the front-end inspection process,the defect detection on the wafer surface is an important research field.The most common defect is processing.Some fine particles generated on the surface of the wafer during the process,and the existence of these particles will lead to circuit breakage or short circuit during the lithography process,thereby affecting the performance of the chip.Therefore,for the detection of microparticles on the wafer surface,the following work has been done in this paper:First,we studied Maxwell's equation,the basic theory of light scattering,and calculated Mie scattering and Rayleigh scattering due to the different particle sizes of the physical scattering formulas.The relationship between the particle shape and defect type on the detection of scattered light intensity is analyzed,and a set of wafer surface microparticle detection system is built according to the above research theory.The bright field and dark field are mainly considered in the system design process.The influence of the light source incident method on the system imaging,the dark field illumination method is selected through the comparison of specific examples.The hardware construction part analyzes the influence of each part of the device selection on the system detection capability,and then selects a more suitable device.The system software part is based on the Visio Studio 2015 platform and uses the C plus plus language to write the graphical operation control interface under the MFC framework.The software can be connected to the industrial camera to detect the collected images in real time,and can perform image denoising,enhancement,threshold segmentation,etc.on the original image.operation,it is easy to identify defects in the inspection area and count the number of particles.Then a set of laser scattering microscopy detection device was built,and based on the detection system with laser as the incident light source,a detection scheme for microparticles on the surface of a large-area wafer was designed.Using the LED array as the incident light source,it could be To achieve large-area illumination,in order to better absorb the stray light generated by the LED light source,the entire detection system is placed in a light hood covered with high-strength light-absorbing materials around the inner surface.Using standard-sized polystyrene microspheres as the detection samples,the detection capabilities of the two schemes were compared.When using a 405 nm highpower laser as the incident light source,the system's detection capability can reach 500-300 nanometers.Under certain circumstances,its detection capability can reach about 1micron.Finally,different from the traditional image processing detection methods,this paper uses the deep learning training method for the collected data based on the U-Net model under the fully convolutional neural network(FCN).Use a laser inspection system to collect data on wafers coated with particle samples,manually calibrate the collected raw data images to create a data set,and perform enhancement operations on the data set so that the amount of data meets the training requirements of the network model.After success,it is tested,and the training results are evaluated according to the data evaluation indicators and direct visual effects.