Y. Tsang, C. Howard, Freescale Semiconductor, Austin, TX; X. D. Wang, Freescale Semiconductor, Tempe, AZ

Summary: With ever shrinking device dimensions of CMOS integrated circuit technology, related failure analysis becomes more and more challenging. One of such problems has been to find dopant related failures in a single device after all the high resolution microcopies failed to capture any physical evidence of the failure mechanism. While microscopic techniques are capable of imaging physical defects at atomic level, imaging dopant remains an obstacle on production chips. Although TEM holography and scanning probe microscopy (SPM) based techniques, such as SCM, scanning spread resistance microscopy (SSRM) and scanning microwave microscopy (SMM) have demonstrated impressive spatial resolution and dynamic range in test devices, the success was rarely duplicated in products due to complicated circumstances and small isolated devices for sample preparation and optimization of imaging conditions. Previously reported techniques to achieve precise x-sectioning are mainly based on the combination of FIB and mechanical polishing [1,2]. To avoid the high ion beam energy (Ga+ ion of 5KeV-15KeV) and charge effects, considerable amount of material (typically in the order of microns) has to be removed with mechanical polishing. This significantly limited the precision required to achieve x-section of a single transistor that is typically smaller than 100nm in modern CMOS chips. Here we report our approach to x-section through a specific device for SCM analysis, exemplified by addressing dopant related root causes in devices fabricated with 90nm CMOS technology. We were able to determine root causes related to anomalous doping not only in silicon, but also in poly gates. The key step in our approach is the development of a sample preparation technique that allows us to precisely x-section through a transistor without being affected by FIB ion beam. Our approach may be applied to other SPM techniques in the same category, i.e., SSRM or SMM.