Parametric Anchoring--Matching TVT Process Models to Your Process

This technical note shows some representative examples of matching TVT process models to process data provided by various TVT customers. Generally, the TVT models are lumped parameter two-dimensional (2D) models suitable for representing 2D proximity corrections. The examples shown here demonstrate that these simplified 2D process models are very effective in matching the process data to within the measurement noise.

In the examples here, no attempt was made in TVT's process models to model thin film reflection effects. Due to process variations, thin-film effects are often not reproducible. Correcting for thin-film effects in the reticle might be ill-advised. In addition, that degree of modelling would have been overkill and beyond the correction ability of reticle manufacturing. The TVT models are used in the SimRule tools to generate correction rule lookup tables for the OPRX corrector.
Note: Specific values of CD and other technological information have been deliberately omitted to honor proprietary data agreements with our customers.

In the diagrams which follow, a "measured value" appears as a solid diamond, the companion TVT modelled value appears as an open square, and the two are connected by a vertical line. The left-to-right sequencing of the measured/modelled data pairs is not significant to the process matching. Generally there are several "data series" of increasing CD or period as identified in the figure captions.

The first example (Figure 1) shows that superior matching can be achieved if the matching software is allowed to choose its own value of sigma (partial coherhece factor) rather than the one stated by the customer. This may be due partly to some vagueness in the meaning or interpretation of sigma or a difference between the actual illumination configuration and the model. The final fitted result is an RMS deviation of only 6.4nm.


Figure 1: Process matching using customer-provided sigma (top) and allowing sigma to self select (bottom). There are 4 data series in each diagram from left to right: 1. Line=Space vs Period; 2. Fixed CD1 vs Period; 3. Fixed CD2 vs Period; 4. Iso Line vs CD.

The second example (Figure 2) shows process matching for a conventional I-Line process and for a new DUV top-surface imaging process. For the former the proximity effects are primarily optical and resist development. For the latter, the optical proximity effect is under 15nm, while the total effect is over 60nm. The major proximity effect is an etch effect. Again matching results are excellent with RMS deviations of 4.5nm and 5.6nm.






Figure 2: Process matching to a conventional I-Line process (top: 1. Iso Line vs CD [6]; 2. Line=Space vs CD [6]; 3. Fixed CD vs Period [9]) and to a top-surface imaging DUV process (bottom: 1. Fixed CD1 vs Period; 2. Fixed CD2 vs Period).

The last example (Figure 3) shows matching of TVT process models to sophisticated process simulators rather than to direct measurements. In these cases the customers have obtained excellent matching between their simulators and their processes, including agreement with cross-sectional profiles. The model in their simulators represents the process they want corrected. That being the case, their simulators are the sources of the process matching data. The excelent matching (RMS of 3.2nm and 0.9nm) owes to the absense of measurement noise. The residual error in the first case is due to customer-modelled thin-film effects. In the second case the residual error is at the linewith quantization limit of 1nm.




Figure 3: Process matching to a proprietary process simulator modelling a non-ARC process (top: 1. Line=Space vs CD; 2. Iso Line vs CD) and to a commercial process simulator modelling an ARC process (bottom: 1. Line=Space vs CD [4]; 2. Fixed CD1 vs Period [3]; 3. Fixed CD2 vs Period [10]; 4. Fixed CD3 vs Period [8]).