INTEGRATED MODELING OF NANOTOPOGRAPHY IMPACT IN PATTERNED STI CMP

摘要

As advancing technologies increase the demand for planarity in integrated circuits, nanotopography has emerged as an important concern in shallow trench isolation (STI) on wafers polished by means of chemical-mechanical planarization (CMP). Nanotopography ― starting silicon surface height variations 100 nm in amplitude extending across millimeter-scale lateral distances ― can result in CMP-induced localized thinning of surface films such as the oxides or nitride used in STI. In previous work, a two-step analysis of the effect of nanotopography on CMP has been reported. First, a contact wear model is used on a blanket wafer to understand how the underlying nanotopography impacts the time to clear oxide over nitride during CMP. Second, the additional dishing and erosion in patterned STI wafers due to the additional nanotopography-induced clearing time are simulated using a pattern-dependent STI CMP model. In this paper, we present two alternative approaches for simultaneous simulation of both pattern and nanotopography effects to better understand the impact of nanotopography on STI CMP. In the first approach, we perform a "pure" contact wear simulation on a finely discretized grid. Because of high computational demands, this approach is only feasible for one dimensional cut lines. In the second approach proposed here, an integrated model incorporating contact wear and density/step-height effects in STI CMP is used. A contact wear component accounts for pressure differentials across the chip due to long-range nanotopography and other surface height variations. Patterned feature effects are captured by a pattern density and step-height dependent component. This makes feasible 2D simulations of patterned structures over underlying uneven surfaces to study the effect of realistic nanotopography maps on the CMP of STI patterns. The simulation shows that nanotopography results in longer oxide clearing time, in agreement with previous work. Depending on the specific random configuration of any given nanotopography map, either more or less nitride erosion may result despite the increased polishing time. Thus nanotopography effects should be considered an additional component of STI nitride loss budgets, in addition to that required by layout pattern dependencies.