imec: The Semiconductor Watering Hole
imec stands for Inter-University Micro-Electronics Center.
Located in Belgium, imec is one of the world's most advanced institutes for microelectronics research. Perhaps, THE most.
Historically, the semiconductor equipment manufacturer ASML has collaborated very closely with imec to research, develop, and commercialize their tools.
This long-standing relationship has been very successful for both parties.
In this video, I want to talk about this interesting organization from Belgium and the unique way it helped ASML become one of the world's leading semiconductor manufacturing equipment makers.
Beginnings
In 1982, the Flemish government of Belgium founded a program for advanced research in microelectronics. The goal was to strengthen the small semiconductor industry in Flanders - a small district of about 6.6 million people.
Local governments in Belgium have unusually large control over their industrial development. So in this case, they had the authority to invest in building local industries.
The program involved setting up three things: A laboratory for advanced research, a semiconductor foundry, and a training program for VLSI design.
Two years later, the advanced research laboratory part of the program became imec - a nonprofit organization located in the city of Leuven.
At the time, it had about 70 people led by a professor named Roger Van Overstraeten.
He was a Stanford PhD at the Catholic University at Leuven.
imec's stated mission was "to perform R&D, ahead of industrial needs by 3 to 10 years". Mostly in microelectronics and design, but later expanded to other fields like energy.
Funding
The whole project began with an initial investment of 62 million euro. That money helped invest in the latest microelectronics infrastructure like offices, 200 millimeter cleanroom fabs, and advanced research labs.
At its start in 1984, imec funded these expenses through public grants from the Flemish government and privately generated annual revenue of about 6.5 million euro.
The Flemish government is generous, but every pocket has its bottom. So obviously, the research lab needed some way to generate money for itself. And in the early 90s, imec's VP of business development came up with an innovative one.
Thus in 1991, imec introduced its R&D Partnership model - the Industrial Affiliation Program.
Industrial Affiliation Program
The Industrial Affiliation Program is a business partnership model that helps share the costs of development between a number of partners.
It brings together a group of industrial researchers and imec researchers to focus on a specific technology. imec offers those guys a chance to work together on a neutral ground with clearly defined project scope and IP ownership rules.
Different parts of the value chain - the foundries, fabless designers, tool makers, EDA providers, and so on - all get to communicate and work together to create a single platform.
The costs of this technology development - which can be sizable - are all shared. While at the same time, the ownership of valuable intellectual properties are either shared or exclusively kept based on how it is discovered.
It kind of reminds me of a Hollywood movie. You bring together a talented director, actors, set designers, cinematographers, costume designers, a big pile of money, and so on to closely collaborate on a single project.
The value of the IPs are shared amongst each other based on pre-defined rules. And at the end of the project, everyone splits up and goes their own separate ways.
imec makes revenues from these partnerships - both for participating and for the original IP it brings into the venture. Whatever IP imec finds and owns from those partnerships, it turns around and uses to kickstart new partnerships.
ASML
ASML makes tools for the semiconductor industry. But the company has long considered its core competency to be a system designer and integrator rather than a deep research institute.
So for R&D, they closely collaborate with very competent organizations like Zeiss, Philips, and of course - imec. As described back in 2004 by their future President, Martin van den Brink:
> ASML has been collaborating with IMEC on advanced lithography processes for many years. The partnership provides ASML with valuable, early feedback from IMEC members, some are also ASML customers, which enables us to speed products to market. In exchange, IMEC members benefit from early access to new processes and a faster learning curve once new products are installed in their facilities
ASML and imec's relationship is one of the oldest in the industry, dating back three decades to the 1st generation Krypton Fluoride stepper in 1989.
Lithography is a very important piece of the semiconductor manufacturing puzzle, but it is not the only piece. And they all have to fit together seamlessly. So not knowing how your lithography tools might potentially play along with the other processes in the workflow can lead to bad outcomes.
Nothing feels worse than you spending millions of dollars building a 200 millimeter sized pizza oven just to find out at the customer site that all of their pizzas are 300 millimeters wide.
So ASML uses imec sort of like a watering hole to gather together other suppliers in the industry to coordinate on how to get everything to work together properly.
For instance, in 1999, imec coordinated a project involving its 193 nanometer ArF stepper. The project goal was to develop and use different resist materials and masks for the 180 nanometer process node.
This project brought together researchers from ASML, 5 resist-makers, a resist coater-developer, 4 chipmakers including from Intel and Texas Instruments, one mask tester, and a partridge in a pear tree.
Another common project goal is to test a particular tool that is still in development. ASML makes a new photolithography tool for next generation lithography like 157 nanometer or EUV.
They want to see how it is used inside a real fab, but resources for such things are limited. So they bring in imec and a few members from the chipmakers to test the tool in a semi-real environment. They will attach actual industrial tools from Hitachi or Tokyo Electron to get the most honest review.
This has the side effect of being very effective marketing for ASML and helped ASML pull ahead of the Japanese lithography makers - Canon and Nikon - during the industry transition to immersion lithography.
When Intel dropped 157 nanometer lithography, ASML quickly recalibrated and decided to go all in on immersion lithography. They quickly rushed their immersion prototypes out to imec ahead of the Japanese.
Nikon and Canon unfortunately were not able to get one to imec as fast and so all the chipmakers - even the Japanese ones - got their first taste of immersion lithography with an ASML device. This eventually led to ASML overtaking Nikon and Canon and assuming lithography leadership.
This has led to some grumbling from the Japanese lithography toolmakers, who see imec as a glorified ASML product showcase - far from being a neutral space for research.
But losers whine about doing their best. Winners go home and well … take a nice bubble bath.
EUV at imec
imec played a large role in ASML's commercialization of EUV. Long before any EUV prototype existed, imec was collaborating with the industry's 6 biggest photoresist makers to find a suitable EUV photoresist.
Without a suitable tool, imec did their tests using synchrotron-generated EUV radiation at the Paul Scherrer Institute in Switzerland.
Finally in August 2006, ASML developed the first large-scale, full-field EUV lithography machine - the Alpha Demo Tool.
imec was one of the first two customers to receive a tool from ASML. The other customer was the College of Nanoscale Science and Engineering of the State University of New York at Albany, New York. What a mouthful of a name.
imec's goal with their demo tool was to evaluate its commercial usefulness using gradually more advanced process nodes and in the meanwhile identify critical issues in the tool, photoresist, and the photomask.
For some additional context on lithography and its various components, I recommend checking out my previous video on 157 nanometer lithography.
Then in 2011, imec received the NXE:3100 from ASML, and they worked closely with laser supplier Ushio to figure out how to increase the tool's wafer throughput and reduce defects.
And as EUV approached commercialization, imec used what they've learned to help integrate the tool into chipmakers' processes.
With EUV now in foundries, imec researchers are working on mapping out future process nodes as well as dealing with troublesome issues like stochastic printing failures in EUV. I think I will probably cover some of this stuff down the line in future videos, but not this one.
Expansion
Starting in the late 2000s, imec has been expanding beyond its European roots. It has been interesting to see the organization's global expansion in certain high-value areas. In 2008, they established their first overseas development center in Hsinchu, Taiwan's Silicon Valley.
That facility has since expanded into a full-fledged R&D center focusing on system development and product prototyping. Per a 2018 article, imec Taiwan focuses on helping Taiwanese companies produce highly advanced products using their semiconductor expertise.
After that, imec expanded to the United States - opening a design center specializing in photonics and high performance computing in partnership with Osceola County at NeoCity, Florida.
This peculiarly-named city also hosts a number of semiconductor firms including SkyWater Technology - an American-owned pure-play semiconductor foundry. SkyWater has a 200 millimeter facility there and is planning an advanced packaging facility so it makes a lot of sense for imec to be there.
Three years later in 2019, imec USA set up another R&D office near the University of California, Berkeley out in the SF Bay Area. This office - which apparently sits in a WeWork - focuses on studying hardware-software co-design and new structures for machine learning and the like.
imec has other offices around the world including in India, Tokyo, San Francisco, and Shanghai. However, it seems like those offices largely perform interfacing functions with local markets. Most of the R&D appears to happen in Europe, with the US and Taiwan to a far lesser extent.
Conclusion
A frequent theme that I have seen throughout the history of the semiconductor industry is the critical importance of collaboration. The semiconductor manufacturing process is so interdependent that no one company can go it alone.
imec is in the position it is as the world's most advanced semiconductor research center because of that collaboration. Their involvement across the value chain gives them high-level insights that few individual companies are in a position to glean.
I am fascinated at how imec has managed to turn itself into a watering hole for semiconductor collaboration. A platform for building teams of rivals working towards a more technologically advanced future. That's kind of special.