Scaling law in ULSI contamination control

The need for chemical contamination control for future LSI's is investigated by considering device failure mechanisms. The allowable contamination density is calculated by deducing relationships between the contamination density and resultant defect density from experimental results. In the calculations LSI failures are classified into two groups according to the characteristics of failure-producing defects: macrotype and microtype. The former defects are macroscopic and fatal: a single defect causes a failure. A stricter contamination control is required for smaller devices similar to that predicted by conventional yield theory. This is due to increasing chip area. By contrast the latter defects are of atomic size, and a single defect is not fatal: devices fail when the defect density exceeds some threshold value. Transconductance degradation in MOS transistors due to interface traps and the resultant SRAM operation error is proposed as an example. The threshold defect (or contamination) density for the failure could be 1 × 10¹¹ cm^-2 in this model. The allowable contamination density abruptly decreases for a minimum pattern size smaller than 0.1 μm. This is due to increasing fluctuations of defect density in component devices. This failure might cause a bottleneck in developing gigabit memories.