Technology invades the modern world

Chapter 407 Problem Setting and Solving

"Mr. Lin, I want to emphasize to you again that Canon's NIL technology has major flaws, and it is definitely not as user-friendly as Canon claims!"
It has a crucial, almost, in my opinion, insurmountable flaw!

Dr. Chen Lei, from the Advanced Technology Center of Shanghai Jiao Tong University, who led the team to Utsunomiya City in Tochigi Prefecture, Japan, and was responsible for receiving the complete NIL technology from the Canon Optical Technology Research Institute, said.

Dr. Chen Lei is a top expert in condensed matter physics in China. He previously worked at the Texas Electronics Institute in America and was a talent that China invested heavily in to bring back to the country.

He was also the only Chinese engineer among those present who had actually worked with Canon's NIL lithography machine.

In October 2023, Canon officially announced the launch of its first NIL device for commercial production, the FPA-1200NZ2C, and successfully delivered it to major storage manufacturers such as Kioxia and SK Hynix, as well as America's R&D institutions.

This allows the equipment to be validated, tested, and integrated into the processes at the customer's factory.

The Texas Electronics Research Institute (TEI) is responsible for receiving NIL equipment from America, and it is also one of the core members of the America Semiconductor Research and Development Alliance.

The Texas Electronics Research Institute acquired this equipment not for production, but for the research and development and prototyping of advanced semiconductors.

Lin Ran nodded, signaling Chen Lei to continue.

This was the first time the only Chinese engineer in the NIL field had met the other party since returning to China.

After all, Lin Ran was too busy and had just returned to Shanghai, and Chen Lei had just returned to China.

Chen Lei took a deep breath and nodded.

He took off his glasses and carefully wiped them with a soft cloth. This habitual action helped clear and organize his chaotic thoughts.

"Okay, Mr. Lin." He put his glasses back on and glanced at everyone present.

"First of all, I must admit that Canon's advertising is not a lie."

Under ideal conditions, N-ILs can indeed do everything they say.

I have even personally witnessed the amazing precision and yield of single-layer wafer samples imprinted using NIL technology in their laboratory.

But it's like a concept sports car; it can break speeds on a professional racetrack, but once you drive it on a congested city street, it might not even be as fast as a regular family car.

Its flaws lie not in its upper limits, but in its lower limits, in every seemingly insignificant step from the laboratory to mass production.

I've summarized it into the following points, the first of which I consider the most critical, what I call the "Flaw Waterfall."

The moment the word was uttered, it caused a small commotion.

SMIC's representative, Liang Mengsong, was clearly sensitive to the word "defect," and he immediately frowned.

Chen Lei held up a finger, his tone becoming extremely serious: "Traditional lithography, whether it's DUV or EUV, is non-contact."

I believe everyone here is very clear about this.

Light passes through the photomask and is projected onto the wafer.

Even if there is a speck of dust on the photomask, its impact on a single chip is relatively controllable after being reduced in size by the optical system.

But NIL is a contact type, a physical imprint!
This means, what happens if even a single nanoparticle falls onto the template, or is generated during the imprinting process?

He paused, giving everyone time to think about the problem.

"The result is that this defect will be stamped onto every chip unit it comes into contact with, like a template stamping it."

Even more frightening is that this particle may contaminate the template, turning the template itself into a source of contamination, and repeatedly replicating this defect on every subsequent wafer.

This is not a problem at one point, but a disaster that extends along a line and across a whole.

We call this Defect Propagation.

At the Texas Electronics Research Institute, we once had a project where a minor contamination resulted in the scrapping of 13 wafers, causing a direct loss of over $80.

What is Canon's solution? Enhanced cleaning and real-time monitoring.

But this means that the production line needs to be stopped frequently for inspection and cleaning of the templates.

The lost production capacity and time costs will frantically devour the money it saved on equipment procurement!
This is the waterfall of defects; once the floodgates are opened, it instantly overwhelms your yield and cost defenses.

Chen Lei continued, "Then there's the black box of materials science."

The core of NIL technology, besides the equipment, also includes photoresist, or what they call resin.

This is fundamentally different from our past understanding of photoresist.

The formulation of this resin is the soul of the NIL process.

Its viscosity, fluidity, curing speed, and peeling characteristics from the template—each parameter directly determines the success or failure of the imprinting process.

This formula is a top secret of Canon, developed in collaboration with a chemical company. What does this mean?

Chen Lei's gaze sharpened: "It means that once we choose the NIL route, we're not just buying a machine, but binding ourselves to Canon's ecosystem! We have to keep buying their specified, expensive, patented resin."

We want to develop our own alternatives? That would be extremely difficult, because we don't know what kind of resin properties their machine parameters are designed for.

We want to change suppliers? Sorry, other suppliers have not been verified by Canon, and they will not be responsible for any problems that arise.

Mr. Lin, do you understand? We've jumped from ASML's hardware monopoly into Canon's double lock-in of hardware and consumables! Strategically, this is extremely passive.

Of course, this time they will transfer the technology, and the only restriction we will be subject to is on consumables, but that is equally serious.

We can hardly avoid consumables; this is something they've accumulated over twenty years, and there's no other way but to work it out over time.

One of the accompanying technical officials couldn't help but ask, "Dr. Chen, what about the engraving precision? Canon showed that their interlayer alignment precision is extremely high."

Chen Lei looked at him and shook his head: "This is the third trap I'm going to talk about next: cumulative error."

Yes, with only a few layers stacked, NIL's overlay accuracy looks beautiful.

However, the manufacturing of logic chips often requires the stacking of dozens or even hundreds of layers.

NIL is a physical contact process. Each imprint and demolding generates micron-level stress deformation and heat changes on the wafer.

This deformation is extremely small; you can hardly detect it in the first or second layer.

But what about when you add layers 30 or 40?

Such tiny errors can be amplified like a snowball, eventually leading to catastrophic misalignments in higher-level circuitry.

This is like using a mold with tiny errors to build a 100-story building.

When the building reaches the tenth floor, it looks perfect; when it reaches the fiftieth floor, problems begin to appear; by the time it is topped out, the entire building may be twisted, tilted, and on the verge of collapse.

Canon always shows us the perfect view of the first ten floors, but they will not, and dare not, show us what it looks like after the top is completed.

Because in Texas, no top logic chip company dares to use NIL technology in critical processes with more than 20 layers.

The reason Canon is willing to sell this solution is probably...

The three major pitfalls proposed by Chen Lei—the defect waterfall, the materials science black box, and cumulative error—are interconnected. They construct a flawed framework behind a seemingly perfect technical solution, encompassing factors from process and cost to strategic security and the reliability of the final product.

This also impressed everyone present, who remarked that the other party truly deserved to be the only expert who understood NIL technology, as their points were spot on.

Lin Ran waved his hand: "Dr. Chen, I understand. These are all minor issues."

Don't worry, all you need to do is go there, learn their techniques, and bring them back.

Chen Lei said softly, "The reason I was able to come back is because in America's circle, I, and other Chinese scientists like me, were always outsiders."

He paused, seemingly organizing his thoughts, before calmly expressing the words that had been weighing on his heart for so long. "Why did I return to China? Not because of the offers I couldn't refuse, but because in America, scientists like me have no place to stand."

All my projects were reviewed, my team was dismantled, and my security clearance was downgraded. They marginalized me in a very civilized way.

That's why I had to come back.

But my return does not mean that I can ignore the problems with the NIL technology route, or that I can comfortably lie on the NIL technology route and enjoy countless benefits while we invest a lot of resources in this route.

EUV has closed the door to us. We can see it, we can touch it, and everyone knows that we need to build our own lithography machine.

But NIL is different!
NIL is a soft trap!

It won't kill you immediately, but it will slowly drain your blood!

It will instantly deprive our domestic lithography machine project, which has just begun to show promise and into which we have invested heavily, of internal support and confidence.

Everyone asks: Why should we spend hundreds of billions and spend ten years developing our own technology? Isn't buying Canon cheaper and better?
Once we believe in it, once we place our bets on NIL, we are essentially giving up on our own efforts.
I don't want our country, or the hard work of countless colleagues, to fall into the same trap that I experienced firsthand.

We can buy one or two to study, disassemble, and analyze, treating them as a formidable adversary to learn from.

However, we must never entrust the fate of our nation's semiconductor industry to the goodwill of others or to a seemingly attractive shortcut.

I've walked this road before, and I know it ends at a cliff.

Lin Ran clapped his hands and said, "Dr. Chen, don't worry. You know, we all know. I don't know if this road is a trap, but I believe in this road. We will definitely bring them a surprise, an unprecedented surprise."

I'm the kind of person who likes sweet talk; I'll eat the sugar coating and then shoot the bullet back.

Chen Lei entered Canon's heavily guarded research institute with a feeling of unease.

He and his team of twelve, comprised of top experts in materials science, precision mechanics, optics, and software engineering from across China, were all filled with excitement; they were there to collect their spoils.

Although Chen Lei had warned of the risks before his departure, Lin Ran's guarantee made it all worthwhile.

Apart from Chen Lei, no one else had any experience with this technology. According to the promotional materials, Canon was exaggerating its capabilities.

In their eyes, Lin Ran was omnipotent.

Everyone was really looking forward to it.

They were received by Mr. Suzuki, the chief engineer of the Canon NIL project.

“Dr. Chen, experts, welcome to Utsunomiya.” Suzuki said in fluent English, “According to our intergovernmental agreement, over the next three months we will be giving you a comprehensive demonstration of the operation procedures, maintenance, and process parameter settings of the Canon FPA-1200NZ2C nanoimprint lithography machine.”

Before entering the enormous cleanroom, an uninvited guest joined them.

“Hello everyone, I’m David Davidson,” a white American woman in a KLA uniform smiled and extended her hand. “I’m a technical advisor from KLA to assist Canon in ensuring the perfect integration of our SpectraShape 11 optical inspection system with NIL equipment. It’s a pleasure to work with you all.”

Chen Lei shook hands with him, but alarm bells were ringing in his mind.

KLA is an American company and the lifeline of semiconductor yield control.

Given Chen Lei's experience working at America, the presence of a technical consultant here is definitely not a coincidence.

Upon entering the cleanroom, the FPA-1200NZ2C was quietly placed there.

Unlike ASML's EUV lithography machine, which resembles a huge and cumbersome monster, it looks more like a futuristic medical device.

Suzuki's explanation was remarkably generous.

He demonstrated the workflow of the NIL device in detail:
He demonstrated how the device, like a printer, precisely sprays nanoscale resin droplets onto a 300mm silicon wafer, with a flatness error controlled within 0.1 nanometers.

They saw how the robotic arm could cover a quartz template onto a wafer with micron-level precision and then physically contact it in a vacuum environment to stamp the circuit pattern onto it.

The entire imprinting process is completed in a few seconds. The resin is rapidly cured by ultraviolet light, and the circuit pattern is perfectly replicated.

Everything seemed so simple, elegant, and efficient.

Back at the hotel that evening, a relatively young engineer from the Chinese team excitedly told Chen Lei, "Professor Chen, this is simply a game-changer! The principle is so simple, and the cost is so low! If we could solve this..."

Chen Lei remained silent, his brows furrowing deeper and deeper. Was President Lin truly omnipotent, capable of solving problems that the entire America Semiconductor R&D Alliance combined couldn't resolve?

Can humans do this?

Canon's NIL devices were shipped in 2023, and were originally scheduled to enter the production stage in 2025. However, the devices to be produced were not complex logic chips, but only basic NAND flash memory chips.

Even so, neither Kioxia nor SK Hynix were able to solve this problem.

Needless to say, America's semiconductor research alliance was also unable to solve the problem.

So what solution does President Lin have? Could it be the brilliant idea Lin Benjian had back then to use water as a diaphragm? Chen Lei was filled with countless questions.

When the team began to delve into the details of the core technologies, just as Chen Lei said, they hit an invisible wall for the first time.

“Mr. Suzuki,” Chen Lei asked at a technical seminar, “the core of NIL technology is the perfection of the template.”

May we visit your template manufacturing and repair workshop?

Suzuki bowed politely: "I'm very sorry, Dr. Chen."

The master template was manufactured using America Applied Materials' electron beam writing equipment, and our agreement regarding its operation and defect detection is limited to the imprinting machine itself.

America's consultant, Davidson, who had been sitting in the back row, looked up and gave Chen Lei an impeccable smile.

In the afternoon, the team's materials science experts encountered another setback while analyzing the composition of the imprinting resin.

"Mr. Suzuki, the chemical composition and ratio of this photosensitive resin are key to achieving a low defect rate."

Can we obtain the relevant technical information?

Suzuki's reply was almost a carbon copy of what he had said that morning: "I'm very sorry, Doctor."

This resin is a patented compound of America Dow Chemical, and we directly import pre-packaged raw material boxes.

The formula is a core trade secret of America and is not included in this technology transfer.

When the weekend came and the team wanted to understand how to detect and classify nanoscale imprint defects, Davidson finally stepped in himself.

While operating the expensive KLA inspection equipment, he explained in a friendly manner: "So, as you can see, our SpectraShape system can clearly identify these defects smaller than 10 nanometers."

Of course, the algorithmic models for interpreting this data, and how to feed them back to upstream processes for correction, are all part of KLA's patented software services.

According to the agreement, we will provide installation and technical support services for your factory in mainland China.

Late at night, in a hotel room

Chen Lei and his core team members reviewed their progress over the past two weeks in the room. The initial excitement had completely vanished.

“I see,” murmured an engineer in charge of process integration. “They’re not selling us a machine, they’re selling us a niche.”

The unspoken implication was that this niche was so deeply embedded in their ecosystem that they couldn't escape it.

Chen Lei walked to the window and looked at the quiet night view of Utsunomiya City outside.

“I was right before.” Chen Lei’s voice was unusually calm. “Now we just have to wait for President Lin to solve the problem.”

I'm curious whether the problems America painstakingly set were more difficult, or whether the world's most outstanding mathematicians had stronger problem-solving abilities.

This is also the only remaining question in the minds of the team members after all this time.

(End of this chapter)