René Raaijmakers
14 April

System engineers and architects at ASML make technical choices to create maximum value. This second article on systems engineering at the litho giant focuses on product generation – the actual development, assembly and integration of the system parts. The central themes are splitting up, keeping an overview, making choices and aligning.

You’d think that ASML’s product generation process (PGP) begins with nanometer budgets and supersonic accelerations. But no. Something as basic as logistics is all-consuming when systems engineers in Veldhoven start with a white sheet of paper for a new machine.

Right now, ASML transports each EUV system to customers on three Boeing 747 cargo planes. How are they going to move their new high-NA machines in 2025? The Twinscan EXE:5000 and EXE:5200 scheduled for that year are larger, and that means the product team in Veldhoven is again putting all transport options on the table. The decisions have already been made and customers are making preparations, but details haven’t yet been shared.

So we can only guess how big the successor to the double-decker-sized NXEs will be. Will it take five or six 747s to carry an EXE system? Will Zeiss be disassembling the optics for transport? Or is ASML switching to its dream scenario of drop shipments with last-minute assembly of modular systems in the wafer fab? In that case, the modules leave the manufacturing sites and fly directly to customers in Asia and the US – the EUV source laser from Ditzingen, the lenses from Oberkochen, the reticle handlers from Wilton and the frames with wafer stages and electronics from Veldhoven.

The drop shipment option has to be intriguing for ASML’s systems thinkers. Potentially, it gives a shorter time to market, and a reduction in kerosene consumption is a nice bonus. After all, sustainability is of increasing importance in the story to shareholders.

Lead times

The volume of the Twinscans also directly determines the size of very expensive chip factories. So the machine footprint must be as small as possible. Only when dimensions and connections are known can the industry proceed with its planning.

Technologically, the optics are the conductor of the development orchestra. Zeiss projection lenses are at the heart of ASML’s litho systems and therefore all-important. Their size and shape dictate the rest of the architecture. “For EUV, realizing this lens design was the first and most important activity,” says Frank de Lange, senior manager of systems engineering at ASML. “Only after the concepts for the EUV lens were ready could we start creating scenarios for the system architecture.”

The lens design triggers a cascade of design decisions. De Lange: “Some parts like lens elements or machine frames have very long lead times. Sometimes special manufacturing or test equipment is needed.” As an example, De Lange points to the extensive test systems that Zeiss uses to measure its EUV mirrors with atomic precision. “You want to start those activities as soon as possible during the system design phase. A wafer handler is less on the critical path than a lens element. We have to pick our battles.”

De Lange gives another EUV example. “The chosen lens concepts led to pretty steep requirements for the acceleration of the reticle stage. That resulted in a new concept with very challenging technology. We decided to build a prototype for that.” For another subsystem, which required fewer changes, it was decided to keep the existing NXE design in the EXE system. “Not much design work was done for that because it was less critical.”

ASML PGP
Overview of the product generation proces at ASML. Credit: ASML

Whole world

Zooming out, the product generation process revolves around product development and everything needed to order, build, ship, install, maintain and dispose of a DUV or EUV scanner or a metrology system. “We don’t have much experience with the last step because just about all the systems we’ve built are still working,” notes De Lange. The Lithocruiser process optimization software is also an ASML product, but in this article, we’ll stick to hardware.

Whereas ASML involves pretty much the whole world in its product roadmap, the product generation process is primarily a task for ASML internally and design partners like Zeiss, Trumpf and VDL ETG. Veldhoven also keeps an eye on design progress at subsidiaries and smaller external suppliers.

Everyone and everything is involved. Not only transport but all sorts of disciplines that aren’t directly related to development either: representatives from the factories, purchasing, customer support, financial experts – they all play a role in the PGP orchestra. So does the project management office, a department that watches over things like best practices, gathering statistical information and creating documentation, manuals and training materials.

Three constraints

In addition to these disciplines, three main roles can be distinguished in the product team that leads the product generation process for a particular machine type. These are the product development manager (the actual leader of the system development), the product manager (responsible for marketing) and the product system engineer (responsible for the system design, together with his team of system engineers and system architects).

Broadly speaking, the product team deals with three constraints: specifications (dictated by Moore’s Law, among others), time to market and the available budget to realize the machine. It should be noted, however, that ASML has acquired a rather special position. In any other technology-driven market, there’s competition, but because it has absolute leadership in chip lithography, these constraints have proven to be quite elastic for ASML – but not infinitely so. Just look at the years of delay in EUV and the 2 billion euros the company annually pays to shareholders and therefore doesn’t invest in innovation.

ASML key components
The function of a lithographic scanner has remained quite stable over the past decades. This uniformity makes it possible to reuse the functional architecture of the system. Credit: ASML

Two views

Viewed from a distance, the function of a lithographic scanner has remained pretty stable over the past few decades. The machines still consist of a light source, optics and a wafer stage, among other things. These modules have become more complicated, though. The lamp in an i-line machine has turned into a very complex EUV source in the latest systems.

This uniformity makes it possible to reuse the functional architecture of the system. In doing so, ASML’s system engineers look at their machines in two ways. The first is the functional view with the product features, such as wafer measurement and exposure, wafer transport, light transport, light distribution, light projection and alignment of successive slices. To enable these product features, you have to develop components that are needed to create functional assemblies that can be used to build the production modules. These so-called engineering blocks are assemblies that you can test in advance, allowing you to bolt the system together with more certainty that everything works.

Putting the system together system from the engineering blocks leads to the production system view. Production modules can differ from functional subsystems. De Lange: “The wafer stage production module, for example, hosts sensors that are part of different functional subsystems.” This makes the wafer stage production module (including the sensors) different from the wafer stage functional module (the hardware needed to position the wafers). The wafer stage project team has no functional responsibility for the sensors, but it is responsible for integrating and testing the sensors.

ASML functional and production views
Breaking down the system into subsystems and modules and looking at them from different angles is helpful. Credit: ASML

Clearly delineated

Breaking down into subsystems and modules and looking at them from different angles is helpful. Manufacturing and assembly view the system in a different way than the service organization. This results in different design requirements.

Testability is also a conundrum. A small local component is easy to test. But it can also be extremely complex. Take the reticle alignment, which is only really testable after the whole machine has been assembled. Furthermore, clearly delineated components provide a good handle when it comes to mass and size.

Modular thinking is also useful when agreeing on ownership. The machine decomposition is such that all functionality for a subsystem is fixed and there’s no overlap with other subsystems. “It’s evident that collaboration between subsystems is necessary,” says De Lange. “There are functions or circumstances where multiple subsystems need to work together, like moving a wafer from the wafer handler to the wafer stage. There’s always one party owning that overarching function, but it must coordinate with the others on its design.”

Line and project organizations

Modularity ensures that projects can run as independently as possible, but it has many other advantages. It makes collaboration with partners easier, it’s good for manufacturability and maintenance, and, last but not least, it makes system upgrades easier – given the growing installed base an ever-expanding part of ASML’s business.

The functional and production system decompositions also provide boundaries for ASML’s line and project organizations. Within the line organizations lies the responsibility for technology development and achieving future performance. This is also input for the technology roadmap. The line organizations are responsible for securing and developing the knowledge of the subsystems.

The project organizations need to deliver. De Lange: “Just delivering a piece of hardware isn’t enough; it has to work and interact well with other parts of the system. When it doesn’t work, it’s immediately clear who we need to get hold of because the ownership is clear.”

In the first part of this diptych, we noted that ASML is less constrained by regulations and administrative burdens than, say, the aircraft or automotive industries. That makes it possible to work in parallel in the product generation process, compressing the six phases feasibility, system design, detailed design, realization, integration and verification. “We accelerate by working concurrently,” states De Lange. “In doing so, we improve our time to market. Some see this approach as risky. But we actually see risk reduction because it saves us time and there’s intensive contact between architects and engineers at an early stage.”

ASML concurrent working
ASML is able to work concurrently on different phases in the product generation process. Credit: ASML

Four o’clock

The PGP starts with an initial system architecture that the systems engineering team defines in close collaboration with the architects of the subsystems. From the system properties follow the properties for the functions. The teams start with designs, make budgets and look at the most important aspects. “Where the risks are high and the models don’t provide enough certainty, we have to make very detailed designs or even prototypes.”

While engineers hone the designs, the teams share knowledge “upwards and downwards.” De Lange: “This naturally generates quite a bit of discussion. A lot of people are involved and the big challenge is to keep everyone aligned and the pace up.” Where necessary, adjustments will be made to the design. Even the specs can change along the way. “This iterative process goes on and on and on as the system design grows.”

The process from a raw architecture to a production-ready system is an adventure that’s becoming ever more exciting. Speed must be maintained, but as soon as sketches turn into prototypes or orders, the risks also increase. “As long as you’re playing with models, there’s not much to worry about, but when you change the system architecture or the lens concept in such a way that you’d have to change material orders for detailed frames or glass, then it becomes painful. But you have to move forward because if you start too late, you lose time. Doing nothing is not an option.”

The problem can also be a technical issue or a supplier who can’t deliver on time. “That happened to us recently. A supplier informed us that they couldn’t deliver an important part of the wafer stage within the agreed time. In the short term, that gives you an integration problem and in the long run, you have to think about an alternative.”

To speed up decision-making, there are the renowned four o’clock sessions. These meetings start at four in the afternoon. “At the end of the day, people want to go home,” explains De Lange. “This prevents the meetings from running late.” The four o’clock sessions are aimed at making important decisions. All kinds of issues can be put on the table – lingering differences of opinion in teams, problems with suppliers or a system engineer with a persistent dilemma.

There are many reasons to call for a four o’clock session. All important stakeholders from the product team are present and all aspects are discussed. So it requires a very good preparation from the persons requesting the session. “If you do this well, you have a discussion with all the stakeholders and you can come to a balanced decision.”

20210923 ASML Frank de Lange Incose event RRA_9923
“We talk a lot about MBSE, but I’d like to see it evolve to model-supported systems engineering,” says Frank de Lange, senior manager of systems engineering at ASML.

People

De Lange needs few words to outline the future of systems engineering at ASML. “We’re dealing with a growing organization, growing complexity and rapid job rotation. That means we have to share a lot of knowledge with many people. So access to system design must be easy. The baseline should be non-negotiable and available to all involved. We also need to do more system modeling and tie together data, design tools and people to have a single source of truth for the design baseline.”

“Our DNA remains the same, it’s all about the art of keeping complex system development manageable.” According to De Lange, people remain the most important success factor in this and they’ll need to have a thorough knowledge of systems but also of systems that consist of multiple systems – systems of systems.

Advanced tools such as those for model-based systems engineering (MBSE), requirements engineering and product lifecycle management play a role. But these tools are still mainly supportive, for sharing information, clarifying information and getting people to communicate better. “We talk a lot about MBSE, but I’d like to see it evolve to model-supported systems engineering,” concludes De Lange. “That does more justice to the crucial position of humans in our beautiful systems engineering work!”

This is the second of two articles on systems engineering at ASML, based on talks Tom Castenmiller and Frank de Lange gave last year at Incose, the international council for systems engineering, and the Bits&Chips system architecture webinars. Main picture credit: ASML