We’re right on the cusp of the SEMulator3D 2014 release. This has been a big release in the making, and I know I’m not alone in my excitement as we approach release day. You can read the press release and get an updated data sheet, but I wanted to take the opportunity to give you my personal engineer-to-engineer perspective on why this is so exciting to anyone doing advanced process development.
The big item to trumpet about is our new Pattern-Dependent Etch models. We’ve enabled multiple different forms of pattern dependence, including short-range effects like Aspect Ratio Dependent Etching (ARDE) and also longer-range effects like Pattern Density Loading (PDL). We’ve applied this capability to the Basic Etch model and also the very popular MultiEtch model. MultiEtch, which made huge improvements in modeling multiple different types of etch physics in multi-material stacks, was released in SEMulator3D 2013. Pattern-Dependence might seem a bit daunting to model, so we’ve taken a few specific actions to help different types of users into this feature. First, Pattern-Dependence can just be switched off, leaving Basic Etch and MultiEtch exactly as users remember them in the last release. Second, for the more process-oriented users, we have a “Calibration Wizard” view, which lets users enter a few measurement points, and has SEMulator3D calculate the numerical dependence coefficients. Finally, for the hard-core modeling community, we let you directly specify modeling coefficients… and even material-dependent coefficients separately. This should keep all of our users happy!
The results of the Pattern-Dependence are really exciting. The model below is a simple via calibration layout which includes an isolated small via, a dense array of small vias and a single large via shape.
With simple calibration, a few critical effects become clear:
• The isolated small via etched shallowest, but with the most vertical sidewalls
• The dense array etches somewhat deeper, but shows a clear dependence between the vias in the center of this array and those on the edge.
• The large shape etches dramatically deeper, with large lateral bias and higher hardmask consumption
These effects match the calibration data well, and the model can then be used to predict the behavior of these processes over a large set of design constructs. This capability is going to let our customers be even more predictive in their virtual fabrication efforts, specifically due to design sensitivities which have become a really hot topic in the industry.
Structure Search was actually released in its first form in October of 2013, but SEMulator3D 2014 represents the first official production release of this powerful checking feature. In contrast with Virtual Metrology, released in 2013, which performs a specific measurement at a specific location of the 3D model, Structure Search examines the entire 3D model, and then reports the locations of specific criteria. This can be used to find, for example, the location of minimum dielectric thickness, a critical reliability criteria for advanced technologies. Output includes a text output listing violations, and also a graphical representation where error markers are superimposed over the fully built 3D model. The examples below show a bunch of other checks that can be performed on this 3D example model of a 48nm pitch integrated BEOL. Now, users can find problems that they didn’t even know to look for. These are the most insidious problems in complex integrated technology.
The SEMulator3D 2014 release also includes 400 other enhancements, improvements and bug fixes (according to the database), but I think the development team may be padding their numbers! In any event, there’s a lot to cheer about in this release. I know the customers have been waiting eagerly for these features, so I can’t wait to let it loose!
Figure 1: A virtual model of a GAA FET showing residual SiGe after the channel release step. Process engineers have to make a trade-off between silicon loss and residual SiGe.(b) Variation in residual SiGe as a function of the channel width and etch lateral ratio. The higher the channel width, the higher the lateral ratio needed to etch away all the SiGe. Channel widths are shown as delta values from the nominal value of 30 nm.