Experimental design for complex systems

A method of developing a semiconductor device comprising:

• a development phase including;
• evaluating a plurality of processing alternatives;
• tuning a nominal process target for each of said plurality of processing alternatives;
• a pre-production phase including;
• acquiring process experience based upon said tuned nominal process targets comprising;
• designing a plurality of experiments including;
• determining a plurality of variables for a design of said semiconductor device and the manufacturing of said semiconductor device;
• creating a causal network diagram for said determined plurality of critical factors;
• transforming said causal network diagram into a causal map;
• identifying responses from said causal map;
• calculating map-based coefficients for a matrix A as a function of said causal map;
• initializing an experimental design matrix Uk wherein k=0;
• selecting a base design matrix X;
• running an optimal design algorithm to determine said design matrix Uk, wherein said running comprises:
• defining Y(P)=I−B(BTB)−1BT)[XP//U]A according to Wynn's criterion, wherein P is a permutation, I is an identity matrix, B is a blocking matrix, BT is a transposed matrix of B, and Y(P) is a possible solution for said design matrix Uk;
• running a first algorithm to choose the best of random column permutations matrices P;
• and
• running a second algorithm to choose the best column permutation matrix P that is near a previous solution for said design matrix Uk;
• and
• modeling responses of said plurality of experiments, wherein said plurality of experiments are designed using said design matrix Uk;
• determining a tolerance window for each tuned nominal process target;
• and
• a production phase including manufacturing a semiconductor device according to said process targets and said tolerance windows.