Inverse lithography technology (ILT) creates ideal lithography results through a mathematically rigorous inverse approach that determines the mask shapes that will produce the desired on-wafer results. The semiconductor industry has long recognized the value of curvilinear ILT for improving process windows by about 2X. However, when ILT output is not an Entirely Manufacturable™ mask, the wafer simulation used by ILT is making an incorrect assumption that the mask shapes produced by ILT will actually be what is manufactured on the physical masks. This causes a cascade of problems. In addition to the physical mask not being able to match the assumed mask shape, mask uniformity is worse, and therefore wafer uniformity is systematically worse on every exposure on every wafer. Since 90-degree corners are definitely not manufacturable, Entirely Manufacturable masks are always entirely curvilinear.
D2S TrueMask® ILT combines a GPU-accelerated, patented frequency-domain, stitchless approach to ILT to make Entirely Manufacturable ILT in a day a practical reality.
It has long been established that unconstrained curvilinear ILT results in the best process windows. However, until 2018, two major obstacles kept curvilinear ILT from being widely applied. One of these barriers – the ability to write curvilinear mask patterns – was removed by the introduction of multi-beam mask writers, which can write any shape without time penalty. The other major barrier – conventional ILT runtimes – was seemingly insurmountable. This barrier was overcome with TrueMask ILT by leveraging pixel-based computing on GPUs. Unlike ILT solutions that claim to be full-chip, but only address hotspots throughout a mask, TrueMask ILT provides a solution for Entirely Manufacturable, therefore entirely curvilinear, mask features throughout the mask.
Micron and D2S presented a joint paper at SPIE Photomask Technology 2019 (Best Paper, Oral Presentation — 2nd Place) that confirmed that TrueMask ILT produces a much larger process window than OPC and demonstrated the high quality of TrueMask ILT mask results in terms of continuity and symmetry.
Linyong Pang, Ezequiel Vidal Russell, Bill Baggenstoss, Michael Lee, Jennefir Digaum, Ming-Chuan Yang, P. Jeffrey Ungar, Ali Bouaricha; Kechang Wang, Bo Su, Ryan Pearman, Aki Fujimura, “Study of mask and wafer co-design that utilizes a new extreme SIMD approach to computing in memory manufacturing: full-chip curvilinear ILT in a day,” Proc. SPIE 11148, Photomask Technology 2019 (October 3, 2019). DOI: https://doi.org/10.1117/12.2534629
In August, 2021, SPIE published Leo Pang’s review of 30 years of ILT development, “Inverse lithography technology: 30 years from concept to practical, full-chip reality,” SPIE 2021, DOI: 10.1117/1.JMM.20.3.030901
By embracing a “from the ground up” approach, D2S was able to conceive a holistic, purpose-built system of GPU-accelerated hardware and software that is designed and implemented specifically for entirely curvilinear ILT computation. Every aspect of the physics and chemistry of wafer lithography and processing, including litho simulations, mask and wafer models, has been optimized synergistically throughout the system to reap the largest potential run-time benefits without compromising computational accuracy.
Most significantly, TrueMask ILT embodies a unique, stitchless approach. Although the D2S computational design platform (CDP) employed by TrueMask ILT includes many GPU/CPU pairs, it emulates a single, giant GPU/CPU pair that can optimize the entire chip as a whole, making Entirely Manufacturable ILT in a day a practical reality.
Partitioning and stitching – the conventional approach to runtime improvement – hasn’t worked for entirely curvilinear ILT, as they result in lengthy and recursive computations to correct the inevitable errors on partition borders. The stitchless approach of TrueMask® ILT iterates on the entire chip at once, eliminating stitching errors.
