In its Applications business line, ASML works on the essential ingredients for its holistic lithography approach. A completely different animal from wafer scanner development in terms of its methods, technologies, tools and culture, says software manager Maurice du Mée. ‘When I started at Applications two and half years back, I had the same feeling as when I joined ASML twenty years ago.’
‘If we want to keep following Moore’s Law, we’re going to have to do more and more correction outside the machine,’ Maurice du Mée says, quoting his technology boss, Martin van den Brink. In ASML’s Applications business line, Du Mée is responsible for the software running on the systems that perform that outside-the-machine correction. ‘That software can help us shave off a whole lot of nanometres.’
The litho machine is no longer an independent entity, but rather part of a larger whole. To facilitate this holistic lithography, ASML offers a growing package of options on its wafer scanners. The main ingredients are the simulation software designed by the company’s Silicon Valley office, formerly Brion, and three products made in Veldhoven: the Yieldstar metrology system, the Baseliner solution for scanner stability and the Litho Insight application for advanced correction calculations. This trio is the work of Applications.
The ‘holistic lithographic control loop’ runs from the scanner, which writes the wafer, through the Yieldstar, which measures the written wafers, and the Litho Computational Platform (LCP), which analyzes the measurement data and calculates the required corrections, back to the scanner, which adopts the calculated corrections. ‘The LCP is a computational powerhouse that runs several applications to keep the machine on track,’ Du Mée explains. ‘Using the simulation software and other tools, we can predict its behaviour, after which we can use our metrology equipment to measure the degree to which that prediction was correct. Using Baseliner we can adjust away the machine’s drift. Using Litho Insight we can remove dynamic process errors that are caused by other systems in the plant.’
Big brother
‘Holistic application development is a completely different animal from software development for the litho machine,’ Du Mée notes. One of the biggest differences is the clean slate with which Applications began. ‘We don’t have to take legacy software into account, the way you do with the scanner.’
PhD pitches at the Benelux RF Conference
Learn about the latest trends and developments in high-end RF techniques. On 24 May, the Benelux RF Conference will take place in Nijmegen. New this year are the PhD pitches, in which young professionals present their research results. Make sure to reserve your seat in time and register now.
Applications was also free to choose its technologies and tools. ‘And we made different choices than our big brother made. For the Yieldstar we use C#.Net, for Litho Insight Java Enterprise Edition. We can also switch much more easily to new technologies and tools, and we use a lot more third-party and open-source software.’
And so Applications has thrown itself fully into data analytics using off-the-shelf standard solutions. ‘Not only to set alarm bells ringing when we see something going wrong, but also to help us find our next products,’ says Du Mée. ‘By thoroughly analyzing the big data coming out of the machine with powerful tooling, we discover connections that weren’t visible before. If they turn out to be systematic, we can use them to shave off another few nanometres.’
Another difference is the importance of usability. ‘We’ve parametrized a great deal. But clients should lose the least possible time setting things up; the time-to-recipe should be a short as possible. We’ve created setup software as an aid. Usability is key for the success of this tooling. Two years ago we didn’t even know we’d be making setup software; now we’ve got fifty people working on it.’
Sometimes Applications moves so fast that it overshoots itself a little. ‘We put together the first version of the Yieldstar in no time, for example. But now we see that we do need a little platform management after all,’ Du Mée laughs. ‘And then it’s pretty useful that we have a big brother who’s paved the way before us.’
Schizophrenia
ASML may be the market leader by a mile in litho systems, but it’s just dipping its toes into ‘outside the machine’. ‘We’re a new kid on the block there,’ Du Mée says.
That’s why Applications is listening carefully to its clients. ‘Using teams of five to six people, we help them optimize their machines, and we see first-hand what they need. The people we have at our clients are simultaneously part of our development team, which means the feedback gets back to us rapidly and we can incorporate it quickly. We have a dedicated connection between the client and development, as it were.’
Applications is able to operate so quickly thanks in part to its use of Agile/Scrum. ‘For the Yieldstar we embed that in a V, because there’s also a lot of hardware development involved. For Litho Insight, a full software solution, the whole organization is Agile. We work with eighteen scrum teams totalling roughly two hundred people in two-week sprints. Right now we still have release cycles of four to eight weeks, but we want to shorten those. In addition, we’re looking at scaling models like Safe and Less, and we’re exploring exactly how we can apply those.’
Applications’ schizophrenia, as Du Mée calls it, is also easy to tame using Agile. ‘On one hand, we’ve got the Yieldstar for overlay in the adoption phase, just before acceptance. We’ve got more than 250 units in the field. A system like that has to keep doing its work properly even at high volumes, because our client’s production depends on it. The same is true for our Baseliner product. To keep from endangering acceptance, we have to keep from disappointing the client. So these systems have to be just as reliable as the scanner, and sometimes even more reliable. That means we’re doing endless regression tests to prove we haven’t broken anything.’
Other systems are in a much earlier phase. ‘Litho Insight, and the Yieldstar in terms of measuring the critical dimension, are still at the proof-of-concept stage. Quality is still of secondary importance; the main thing now is that we demonstrate the added value, that we can show that you can squeeze out nanometres with them. In this exploratory phase, we sometimes still send a simple Matlab script into the field.’
Like a weed
Applications’ unusual approach gives the division its own unique culture and atmosphere. That uniqueness is further heightened by the group’s location: not at De Run, but in its own accommodations at the Flight Forum. ‘We’re permeated by a real startup vibe,’ Du Mée says. ‘When I started here two and half years back, I had the same feeling as when I joined ASML twenty years ago. We’re in the same phase and are the same size as the company was then.’
Roughly five hundred people now work at Applications, of whom 270 work on software. ‘But we’re growing like a weed: last year we added fifty people, and in two years’ time we’ve doubled our size – it’s like we’re following Moore’s Law.’