A behind the scenes look at inverse lithography, M&A activity and next-generation lithography.
Semiconductor Engineering sat down to discuss trends in the lithography and photomask business with Linyong “Leo” Pang, the new chief product officer and executive vice president at D2S, which focuses on model-based mask data preparation as well as other mask writing technologies. What follows are excerpts of that conversation.
SE: Before you arrived at D2S you were at Luminescent, which was a pioneer in inverse lithography. What is that?
Pang: Inverse lithography is one of the core technologies in the computational lithography field. Let’s start with OPC. OPC would be something you try to perturb the mask patterns with. So, you can print what you want to print. Inverse lithography is taking that one step further. The basic idea is that you have the patterns you want to print on the wafer. Then, we can start from what we call target patterns. Then, we model the entire scanner optics and also resist development. In addition, we have a mathematical model. So, we have a method to inversely calculate what would be the mask pattern and the scanner source. So, that would give you the pattern fidelity and the best process window.
SE: Where is inverse lithography going in the industry?
Pang: Looking out to the 10nm node, EUV will not be ready. Then, you look at other NGLs. It looks like they won’t make it for 10nm either. So what you have to rely on is basically optical lithography based on 193nm immersion. You also have multiple patterning. Multiple patterning will help you print those line-space patterns. You also have to cut those line-space patterns, plus the contact layer. Those will be the layers that inverse lithography and source-mask optimization can really help, because those layers are not very regular. They are the most difficult layers to print.
SE: It appears that KLA-Tencor has just acquired Luminescent, right?
Pang: In 2012, Synopsys acquired the computational lithography business from Luminescent. After that, Luminescent started looking at how to help the mask house to inspect and measure patterns. We (at Luminescent) called this business computational metrology and inspection. That business was just acquired by KLA-Tencor.
SE: So KLA-Tencor has acquired the remaining portions of Luminescent. Has that deal been completed?
Pang: That is already done. What KLA-Tencor did was an asset purchase. They basically purchased the products, technology and IP.
SE: What is your view on the next-generation lithography (NGL) landscape?
Pang: I could be wrong, but it looks like EUV will miss the 10nm node. We all wish EUV will come in. Then, you look at other NGLs like multibeam and others. I hope they can come in and do the cutting layers for 10nm, but it looks like there are challenges for them as well.
SE: Can you talk about mask complexity?
Pang: In the old days, mask patterns were single patterns. You could basically use rectangle shapes to assemble the mask patterns. But now, the shapes are much more complicated.
SE: What are some of the challenges and solutions for mask complexity?
Pang: Going forward, the patterns on the mask are getting smaller. They are getting more complicated. So the capacity and data volumes you have to deal with in mask data preparation will increase dramatically. That actually creates a challenge, but also an opportunity, especially for D2S to tackle such a problem. For example, there is model-based data preparation. That will improve the writing pattern fidelity. At the same time, it can actually reduce the number of shots. Therefore, it can reduce mask write times.
SE: For some time, the industry has talked about lithography hotspots. Now, there is talk about mask hotspots. What are the issues?
Pang: There are two types of hotspots. One is litho hotspots. The other one is the hot spots on the mask. On the mask side, you need more complicated patterns, especially for the assist features. They are much smaller than the main features. Therefore, they create the hotspots for the mask. If you don’t take care it by either increasing the dose margin or doing the simulation to figure out what kind of shot you need to use, then those complicated ones, especially those small assist features, won’t be printed correctly on the mask. If they are not printed correctly on the mask, then you will get a hot spot on the wafer. So, in other words, you need to have the correct pattern on the mask to help you to get a good process window for that pattern on the wafer.
SE: So given your background, is D2S expanding into new markets?
Pang: We are looking at getting into new fields. But at this moment, we are focusing on our existing products.