Wet-etch sequence optimisation incorporating time dependent chemical maintenance

Wafer fabrication is the major cost contributor in semiconductor manufacturing. One of the steps in the fabrication is the removal of exposed layers in an automatic wet-etch station with chemicals. In time, these chemicals get polluted and their effectiveness decreases. Therefore, the chemicals in these baths need to be renewed or maintained on a regular basis. This can be required after a fixed number of processed wafers, or after a specified period of time. It has been observed in the factory that the wet-etch station can get completely blocked when one of the baths goes into maintenance. This happens when the first batch in queue needs the bath maintained. All subsequent batches then have to wait and the tool is blocked. Placing a different batch in the load lock might have prevented this blockage. In this paper we investigate the influence of the sequence of the batches inserted on the total processing time, when time dependent maintenance is incorporated. In order to do this, a detailed model is constructed based on an actual wet-etch tool. The model parameters were tuned to the values of the actual tool and it was verified that the resulting behaviour was correct. Based on this model a (sub)optimal sequence was constructed. Three sequences were considered: i) an arbitrary sequence, ii) an optimised sequence without considering the time-dependent behaviour of the maintenance, and iii) an optimised sequence with the time-dependent behaviour. It was found that the mean processing time could decrease with about 10% when the time dependent maintenance was incorporated, in comparison with the original, arbitrary, sequence. When a (sub)optimal sequence was constructed without incorporating the time-dependent maintenance, a mean decrease of 4% was realised. This shows that the time-dependent maintenance plays an important role in the total processing time and should be incorporated in the optimisation.