Abstract :
As integrated circuit (IC) market requirements drive technology advance within the semiconductor manufacturing sector, reuse of available high yielding established processes is a cost effective method of achieving a quick response to market requirements. Defect density numbers are regularly calculated for different products as a monitor of process and product performance. For new product introductions to an established process flow, the defect density target is usually assumed to be close to other products running on the same process. To give a better indication of expected defect density, modelling was performed on volume production over several months. Inline defectivity monitor data was extracted along with final-yield results for several products running on one established flow. Mask area data was also extracted far each layer and for critical areas. Critical area data was determined by performing boolean operations on mask layers. This modelling highlighted a method of being able to predict a more accurate defect density number for a new product and in addition could be used to specify the allowed percentage of killer defects at automatic inspection steps within the process flow, per product
