Support portal Careers Contact Investors
EN
EN · English JA · 日本語 KO · 한국어 ZH · 简体中文
Request a Demo
Request a Demo
Request a Demo
Support portal Careers Contact Investors
Language EN JA KO ZH

Process Proximity Compensation: correct-by-construction OPC and mask pattern synthesis

Get a mask that prints your layout as drawn, computed from how your process actually moves each edge and checked in simulation first.

Talk to our Technical Experts Downloads and documentation

The mask you draw isn't the pattern you print

Silvaco's Process Proximity Compensation (PPC) is a third-generation computational lithography suite, and its whole approach is correct-by-construction OPC. Optics, etch and the mask each pull a printed edge off target. PPC models all three, then computes the mask that cancels their combined shift. That gets harder at every node, because you're expected to hold tighter dimensions without a longer mask cycle or a bigger OPC team. Silvaco acquired PPC, and the engineers behind it, from Cadence in March 2025.

Five stages from process data to a checked mask

  1. Model calibration Fit the models to measured data from your own process. Tool: PPC
  2. Predictive modeling and inversion A physics-based predictive process model (PPM) that can be inverted, target to mask. Tool: PPC
  3. Correct-by-construction OPC Compute the mask from the model instead of tuning it over repeated runs. Tool: PPC; curvilinear mask synthesis in DimensionPPC
  4. Source and SRAF optimization Tune illumination and assist features with the mask to widen the process window. Tool: PPC
  5. Verification Simulation checks the synthesized mask; mask rule checks confirm it can be made. Tool: PPC; PPV and curvilinear MRC in DimensionPPC

Inside the PPC suite

  • Compute and cost of ownership

    Synthesis and verification stay on general-purpose hardware, distributed over the cluster you already maintain. That keeps mask turnaround short and cost of ownership low.

  • Models

    Production-proven lithography, etch and mask models. With all three in the correction, the target is the post-etch critical dimension your metrology reports, not the resist image.

  • Predictive process model

    The PPM describes patterning physically, well enough to run in reverse. Give it the shape you need on the wafer and it works out the mask that will print it. That inversion is what correct-by-construction means in practice, and the optimization methods behind it are patented.

  • Source and SRAF optimization

    Source optimization across multiple areas, and source-mask optimization, shape the illumination and the mask together. Hybrid SRAF placement adds the assist features. The aim is the widest process window your depth of focus allows.

  • Recipe creation

    Build and optimize recipes in the GUI or in scripts. Someone new to a layer can start in the interface. The engineers who own the flow keep recipes as scripts, under version control, ready to rerun on the next tapeout. So the correction flow belongs to the team and doesn't leave when one specialist does.

Layers where compensation earns its keep

PPC is in production at both ends of the node range. The hard part moves as you go.

  • Critical layers at advanced nodes A few nanometers of edge-placement error can fail a layer, so no critical layer goes out uncorrected.
  • DRAM arrays Dense layouts outgrow what Manhattan mask shapes can print, and curvilinear masks take over.
  • Etch-dominated layers Etch bias moves the final edge further than the optics do.
  • Layers short on process window The depth of focus the layer needs has to be won back through the source and assist features.
  • Mature and large-geometry nodes The same models and flow keep established processes on target.

Published work on PPC

Bring the layer that won't close

Pick the one that keeps drifting off its CD target, or one you're qualifying at a new node, and talk it through with the team that builds PPC.

Talk to our Technical Experts Downloads and documentation

Questions about process proximity compensation

Process proximity compensation adjusts a photomask so the pattern that reaches the wafer matches the layout. It works out how each feature will shift between mask and wafer, then pre-distorts the mask to cancel the shift. Silvaco's PPC suite is named for the job.

PPC is an OPC suite; optical proximity correction, or OPC, is the wider technique it belongs to. The difference is what gets modeled: PPC brings etch and mask effects into the correction alongside the optics. That points the result at the dimension you measure after etch, rather than just the printed image.

It means calculating a mask that prints correctly, rather than arriving at one by trial. Conventional model-based OPC moves edges, re-simulates and repeats until the print converges. PPC inverts a physics-based model of the patterning process, so it solves directly for the mask that hits the target.

SRAFs, or sub-resolution assist features, are small mask shapes too fine to print on their own. They change how the features next to them image, so those features keep their size across a wider spread of focus and dose. PPC places them with a hybrid method, in the same optimization as the illumination and the mask.

PPC is production-proven from large geometries through the most advanced processes. Micron has presented curvilinear OPC on several DRAM layers using DimensionPPC. For a specific node or layer type, ask our Technical Experts what PPC has already corrected on processes like yours.

It runs on general-purpose hardware. Each mask synthesis and verification step is broken into pipelined jobs that share out across your existing compute cluster. To size a job against your own layers, talk to our Technical Experts with the number of layers and the cluster you run.

Related

Explore AI-Driven Manufacturing