About

The Statistical Metrology Group focuses on the understanding and reduction of variation in advanced micro- and nano-fabrication processes, devices, and circuits. We develop new methods and approaches to measure, model, and mitigate the wide range of deviations observed in manufactured devices.

Figure 1: Comparison of an experimental imprinted topography with a feature-scale simulation, for a test-pattern imprinted into 200-nm-thick 495 kg/mol PMMA under a sample-average pressure of 40 MPa at 165 °C for 1 minute. The simulation tracks the differences between the residual layer thicknesses (RLT) of different features, as well as RLT variation within individual features. A simple Newtonian model of the resist and a linear-elastic model for stamp deflections capture much of the RLT variation observed.

One branch of our work is closely tied to important semiconductor fabrication processes, and emerging processes in MEMS and nanofabrication technologies. Processes of particular interest include chemical-mechanical polishing (CMP), electroplating, deep reactive ion etch (DRIE), hot embossing, and nanoimprint lithography. In each of these, we have developed test structures and masks, and approaches to measure systematic variation at the wafer scale as well as die scale (particularly layout pattern dependent variations). These measurements are coupled to empirical and physical models and simulation tools, for designers to predict manufacturing results for their particular layout. Finally, methods to reduce or mitigate these variations are being explored, such as through
dummy fill strategies.

Figure 2: Normalized contact plug resistance map for a representative test chip.

The second major branch of our work is tied to the design implications of manufacturing variation. We develop novel test circuits to measure variation, particularly to gather the large numbers of measurements to enable modeling of not only mean but also variance dependencies. We develop new approaches to model these variations, ranging from spatial correlation models, to systematic layout models, to random and other variation models in compact model form. We study the impact of variations in parameters such as Vt, Id(sat), and leakage on circuit and system (e.g. multicore) design. Finally, we consider yield optimization, circuit compensation and self-healing approaches to mitigate
these variations.