ASML is always on high alert.

Last year, the company raised its production targets twice: it hopes to ship about 600 DUV and 90 EUV lithography machines by 2025, up from fewer than 200 and 35 respectively last year.

 

In a typical week for ASML, hundreds of people start their new jobs in Veldhoven. Delivery issues are a daily occurrence due to the ongoing chip shortage. Not to mention setbacks like the fire at the Berlin plant at the beginning of this year, which temporarily paralyzed the production of EUV wafer fixtures.

Nevertheless, Martin van den Brink (editor's note: ASML CTO) was content and relaxed. The productivity of EUV lithography machines has not yet met customer expectations, but for several years, EUV lithography machines have become an integral part of the production process of the world's most advanced chips.

“ After ten more years than planned, it was a relief. ” Martin van den Brink.

The latest generation of EUV equipment. — The development of the High NA system is also progressing well. “ I visited the factory floor this afternoon. That machine was a highlight of my career. van den Brink says in his office, where a series of lithographic products are on display.

According to ASML's roadmap, the delivery date for the world's first High-NA lithography machine is sometime next year, and Van den Brink believes that goal will be met, although supply chain issues could still upset the timing.

 

As it turns out, van den Brink even more or less charted the technological course for the next decade. This could be his last strategic contribution to ASML, as the executive, who has been with the company for many years, is expected to retire in 2024.

 

In any case, anything at van den Brink seems manageable right now. — Pretty much business as usual, which is possible for a company that assembles some of the world's most complex production machines. The only thing that seemed to bother van den Brink was his office chair. “ Look at this piece of shit. It has too many adjustment options, ” "He complained as he sat down, demonstratively shaking the armrest.

<

van den Brink was in an equally good mood when he spoke with his visitor five years ago. At a time when it still takes all hands on deck to get EUV on track, he has moved forward with his High-NA plan. ASML's management and supervisory boards supported the required $1 billion investment, with key optics partner Zeiss also joining in. “ They didn't want to at first. Van den Brink admitted.

“ At that time, I thought that the High NA would be the last NA of EUV. But there is considerable risk because it comes too late and there is not enough time period to recoup the investment. I don't think we can delay the development of high numerical aperture until we know EUV. ”

“ The transition period for lithography was terrible. If you mess up, things get messy. Especially now that the organization has become so big, I feel a great responsibility. I'm very paranoid. Do we have a successful proposal? Can we fix this? ” Van den Brink eventually became convinced that high numerical aperture was the right way forward and won over skeptics.

But he missed one thing. “ In my arrogance, I thought customers would line up, but it really wasn't easy. ”

Around April 2018, everything is ready. EUV was poised for mass production, and soon after, the first orders for High NA systems followed. Since then, preparations for another transition period have generally gone quite smoothly. The chief technology officer revealed that this is more than any previous lithographic conversion. Much easier. .

“ Why? First, we now have a good understanding of the role of EUV photons. Today, we still have problems with the stability of shortwave systems, and productivity is not up to par. But there are major physical issues that we need to understand and clean up. — We've basically left it behind. ”

In addition, ASML needs to rely on external innovation in any lithographic transformation. “ The photoresist changes, the mask changes, and you get new types of defects. All of these things affect infrastructure. For the High NA, the infrastructure changes are relatively minor. So the risks involved are much lower. ”

< Despair

Van den Brink said that by far the biggest challenge in developing High NA technology has been building metrology tools for EUV optics. The High NA mirror is twice the size of its predecessor and needs to be kept within 20 picommeters of flatness. This needs to be done in a “ Can accommodate half the company ” Verify in a vacuum container.

 

“ The problem with building this tool is that you can't be sure it's accurate enough. There are all sorts of tests you can do to provide some assurance, but you can never be absolutely sure. That's where we are now. We think it works, but the truth won't come out until we get the first footage next year. ”

If the lens does not meet the specifications, we will take emergency procedures. “ We have a backup plan. If it doesn't work, we have plenty of latitude to fix it. We can repolish the surface to a certain extent and replace individual mirrors if necessary. ” An EUV lens consists of several mirrors. — The exact amount is a trade secret.

In the end, van den Brink doesn't want to underestimate the complexity of a system larger than a typical bus. “ It's a monster. In the past, a lithography machine required several hundred kilowatts.

 

For EUV, it is 1.5 megawatts, mainly because of the light source. We use the same light source for High NA, but require an additional 0.5 MW. We use water-cooled copper wire to power it, and that has led to a lot of engineering progress. ”

Still, Van den Brink doesn't see any roadblocks, though supply chain issues have obscured the calendar. “ The timing is a bit of a problem. High NA machines contain quite a few components that we use in our production systems, with today's meal taking precedence over tomorrow's meal. It is still a crisis project in terms of time, but I believe we will be very far along by the end of next year. ”

The first system will remain in Veldhoven, where ASML and Imec have established a joint High NA research laboratory. We plan to have our customers own machines for research and development in 2024, with the first high-volume manufacturing tools to be delivered the following year.

 

This will largely spare ASML's customers the difficult times caused by delays in EUV development.

“ They're getting desperate, simple as that, &rdquo. van den Brink said in 2017, referring to his customer's announcement to put EUV into production. The number of chip layers that had to be patterned using two or more 193-nanometer (immersion) exposure steps grew so large that semiconductor manufacturers continued to use EUV lithography machines, even though they were not as productive at the time.

Although multiple patterns have been used in EUV. — Not for lithography, by the way. — But before this uncomfortable situation appears, producible High NA lithography machines are expected to appear. “ Clients are not desperate. But, to be fair, if the High NA were ready now, they would use it. ”

< It was all worth it

In addition to making High NA EUV lithography available as soon as possible, ASML's current priority is to continue to reduce the cost of EUV and High NA patterning. According to Van den Brink, it will take another decade.

“ As long as the performance has not yet reached the level of 193 nm lithography, there is a lot of room for improvement. We can still make a lot of gains in the transition, for example, maybe double. And we haven't squeezed every nanometer of resolution out of optics. With the 193 nm lithography machine, we were at the limit due to the operation of the lighting system. EUV lithography machines do not yet have the same level of complexity. ”

ASML will also continue to focus on overall lithography. This set of metering and computing technologies enables chip manufacturers to maintain tight control over their manufacturing processes. This reduces the defect rate and, like increased productivity, reduces costs.

What the semiconductor community is eager to know, however, is whether the high numerical aperture will find a successor. Jos Benschop, ASML's vice president of technology, had already revealed at last year's SPIE Advanced Lithography conference that a possible alternative, lowering the wavelength, was not an option. This is a matter of Angle: the efficiency with which EUV mirrors reflect light depends largely on the Angle of incidence. The reduction in wavelength changes the angular range such that the lens must become too large to compensate.

This phenomenon also appears with the increase of the numerical aperture (NA) of the lens. So is it possible for NA to increase again? van den Brink confirmed that ASML is looking into the matter.

But, personally, he doesn't believe hyper-NA will ever prove viable. “ We are working on it, but that doesn't mean it will go into production. For many years I suspected that high-NA would be the last, and that belief has not changed. ”

for “ Standard ” EUV, NA is 0.33, for High NA, it is 0.55, for hyper-NA, it will “ Above 0.7, maybe 0.75. In theory, it can be done. Technically, it can be done. But how much room is there for the larger lens market? Can we even sell these systems?

I'm paranoid about High NA, and I'm even more paranoid about hyper-NA. If the cost of hyper-NA grows as fast as we've seen with High NA, then it's almost not economically viable. Although this is also a technical issue in itself. And that's exactly what we're working on. ”

Van den Brink doesn't want to build bigger monsters, he says, pointing to miniature versions of his collection of metering vessels, but also mentioning lithographic components such as lenses. “ We're trying to make fundamental changes in manufacturing and design to make sure that if we're going to do this, it's economically viable. ”

“ So this is very different from our High NA method. We will ensure that high numerical apertures are achieved. For hyperNA, we acknowledge that there may be insurmountable cost constraints , at least because transistor shrinkage is slowing.

Because of system integration, it is still worthwhile to continue to develop a new generation of chips. — That's good news. But at this point, the question becomes very real: Which chip structures are too small to be manufactured economically? ”