Examining the impact of pattern dependence on etch, deposition and CMP processes.
First order process modeling can help tremendously with process setup and integration challenges that occur in a semiconductor fabrication flow, by visualizing process variation problems “virtually” prior to actual fabrication.
In some instances, a deeper level of complexity needs to be added to the process model to capture the effects of variation in the process. Specific examples include pattern dependent processes and effects that can impact the flow and potentially the final device, including etch, deposition, CMP, and others. Understanding these pattern dependent processes can be important in understanding variations, macro to scribe effects, within array effects, and even feature scale effects. Through process simulation, unexpected results can be captured before expensive hardware, devices or wafers are manufactured or characterized. Even during yield ramps, these factors can be important in helping to understand and minimize yield detractors. A virtual fabrication platform can incorporate these complex effects and provide users with a deep level of understanding of their process flows, using rich process-accurate models.
To exemplify the level of complexity that can be captured, we will explore three different processes and the effect of pattern dependence. Etch modeling is quite intricate and can incorporate information both from within the modeling domain and external to the modeling domain (using design), which provides robust simulation results. For instance, with fin etching, density of fins between single and double pitch can strongly impact etch depth, as shown in Figures 1 and 2. This etch depth variation can have downstream impact, for instance modifying the effectiveness of solid state doping into the lower fin regions.
Figure 1. Post fin etch with pattern dependence invoked. Fins and substrate are depicted in red material and the other materials represent the hardmask. Between fins in single pitch, the etch is impacted by pattern dependence and does not etch to the depth of the double pitch areas.
Figure 2. Cross section of post fin etch from Figure 1 through the center of the build. Further exemplifying the single pitch to double pitch etch impact.
The second process of interest, deposition, can impact other aspects of the integration flow where device pattern density and topography are varied. This can be demonstrated by something as simple as a review of deposition process differences between an isolated trench and a set of dense trenches. For instance, we can model a spin on material deposition (as shown in Figure 3), with a nominal thickness of 50nm at the top most surface in the blocked area. If the deposition occurs in an area where trenches are placed in a dense array, the thickness of the material from the top most surface is reduced. For example, deposition modeling highlights that the film thickness is only 22.5nm just adjacent to the trench array. An isolated trench, however, behaves much differently than a trench array during the deposition process, producing a virtual metrology measurement of film thickness (just adjacent to the trench) of 46.5nm. These deposition differences can be critically important, for instance where a material deposited does not actually fill a dense trench array as was originally intended, due to the pattern density of the structures.
Figure 3. Post spin on material, shown in the purple color, deposition of 50nm deposited on a structure containing a dense trench array and isolated trench area. The top film thickness variation is highlighted adjacent to the dense trench array and isolated trench areas.
Chemical mechanical polishing, CMP, is the third example of where pattern dependence can be modeled. In this particular example, shown in Figure 4, the CMP step polishes down to the array top where normally the material removal would stop. In this case, however, our process model displays some interesting and unexpected behavior at the edge of the array. The simulation incorporates design data for the array beyond the build area as well where the design array extends well beyond the edge to the right of the build. This additional input to the model is demonstrated by the CMP “skating,” or not removing any additional material, on the right side of the build. In contrast, the left edge of the array begins to erode on the left side of our model, and CMP dishing is apparent into the target CMP material on the left. These effects can become even more important when trying to understand array to scribe polishing deltas and the effect of pattern dependence on CMP processes.
Figure 4. CMP of the blue material in animation of dishing and array erosion at the edge of a patterned area, gray material. The array extends beyond the build in this structure and is apparent in the right most side of the structure where the CMP skates on the surface.
Pattern dependent effects are important throughout the entire integration flow as shown in our examples in this article. Capturing these effects is important to understanding variations, macro to scribe effects, within array effects, and even feature scale effects. Even seemingly harmless design changes can lead to unexpected results due to these pattern dependent effects. Using a virtual fabrication platform, such as SEMulator3D, allows engineering teams to model and investigate these complex behaviors for expected—and more importantly unexpected— results, thereby saving time and money throughout the product lifecycle, from concept to yield ramp to volume production.