192: Chapter 192 "The Epic of Chips": 14-Nano Mass Production

Xingchen Academy of Sciences Lithography Lab · March 5th

3:47 PM, Dapeng, Shenzhen, third basement level of the Xingchen Academy of Sciences.

The cleanliness here reached an incredible ISO Class 1—no more than 10 particles with a diameter greater than 0.1 microns per cubic meter of air.

By comparison, an operating room has a cleanliness of ISO Class 5, a difference of four orders of magnitude.

Luo Qinghe was wearing a full set of anti-static cleanroom suit, staring at the production line through the observation window.

He was 44 years old, and under such high-intensity working conditions, his eye bags were heavy, but his gaze remained sharp as a knife.

At the end of the production line, a robotic arm grabbed the 1000th wafer and fed it into the testing machine.

The entire laboratory was terrifyingly quiet, with only the low hum of the air conditioning system and the faint electrical sound of the instruments running.

More than thirty engineers stood at their respective stations like prisoners waiting for judgment.

The 1000th one—this was a number with a sense of ritual.

Three months ago, when they started the 14nm process R&D, no one in the industry believed they could achieve an intergenerational leap within six months.

TSMC took 22 months to go from 20nm to 14nm, Intel took 24 months, and Samsung took 20 months.

Six months? That was science fiction.

But Xinghuo did just that—using three-shift rotations, 100-hour work weeks, and near-manic trial-and-error iterations.

Most importantly, the system provided the technical data, otherwise they would have been grasping at straws.

Test data began to jump on the big screen.

First item: Line width uniformity.

Numbers scrolling: 14.2 nm, 14.1 nm, 13...

Standard deviation: 0.15 nm.

"Good!" someone whispered.

Second item: Transistor performance.

Carrier mobility, switching speed, leakage current...

All within specification range.

Third item: Defect density.

This was the most critical data—defect density directly determines the yield.

The numbers stopped jumping: 0.08 defects/cm².

Luo Qinghe's heart almost stopped.

The international standard, the mass production threshold for the 14nm process, is a defect density of less than 0.15 defects/cm².

0.08—this was already the top level in the industry.

Finally, the comprehensive yield calculation result popped up: 92.3%.

The laboratory was dead silent for three seconds.

Then, a female engineer suddenly squatted in the corner, covering her face with the sleeve of her lab coat, her shoulders shaking violently.

She was Wang Yuxin, 28 years old, a Ph.D. in Microelectronics from Tsinghua University, who hadn't been home for three consecutive months, living in the dormitory next to the laboratory.

Then came the second, the third...

Some cried, some laughed, some hugged, others slumped into chairs.

Luo Qinghe took off his glasses and wiped the lenses with trembling hands.

He was nearsighted by 600 degrees; without his glasses, the world was a blur, but at this moment, he didn't need to see clearly—the numbers said everything.

"From 16nm to 14nm, international giants take an average of 18 months." His voice was exceptionally clear in the quiet laboratory.

"We took... 182 days."

He turned around, looking at the group of young people behind him, whose average age was under 30.

"Children," he said, his voice choking with emotion, "you have made history."

The surveillance camera recorded this moment: engineers in white cleanroom suits, surrounded by equipment worth billions, crying and laughing like children.

The mass production of the 14nm chip officially completed all technical verification; now it just waited for the lithography machine to complete its final prototype testing.

In the blink of an eye, another 5 days passed. This day, March 10th, "Dragon Chant" Laboratory.

One level deeper underground, the "Dragon Chant" 16nm lithography machine R&D center.

The atmosphere here was even more oppressive—if chip manufacturing was climbing Mount Everest, then the lithography machine was the equipment for the climb.

And it had to be built while climbing, during the ascent.

André Dupont, Academician of the French Academy of Sciences and Chief Optical Scientist.

He joined Xinghuo Chip a year ago through the system.

This old Frenchman had a quirky temper, but his technical attainments were unfathomable.

He had a habit: he had to listen to classical music while working; today, he was playing Debussy's "Clair de Lune."

"Mr. Dupont, the final check is complete." The assistant reported in Chinese.

Dupont nodded and walked to the main console.

His hands were steady, but the beads of sweat on his forehead revealed his nervousness.

The lithography machine is the most core and precise equipment in chip manufacturing.

Worldwide, only the Netherlands' ASML can build high-end lithography machines—and it took thirty years, integrating Zeiss lenses from Germany, light sources from America's Cymer, and materials from Japan, to reach today's level.

Xinghuo had to start from scratch and build its own lithography machine within a year.

This was called an "impossible mission" by the industry.

But today, the "Dragon Chant" prototype was to undergo its first full-process test.

"Start." Dupont pressed the red button.

The huge machine emitted a low hum, like a giant dragon waking up.

The laser lit up, emitting a ghostly blue beam, which passed through a complex set of mirrors and finally through the projection objective lens, carving nanometer-scale patterns onto the photoresist-coated silicon wafer.

The whole process was automated, but in the monitoring room, everyone held their breath.

24 hours.

No one left.

Lunch boxes were brought in, and no one ate them even when they got cold.

The coffee pot was emptied and filled, filled and emptied.

Dupont sat in his chair, eyes fixed on the screen.

He was old and couldn't pull all-nighters, but today he insisted on witnessing the entire process.

At 4 AM, the test ended.

The robotic arm took out the wafer and sent it into the electron microscope.

The first image appeared on the big screen.

16nm line width, lines clear and sharp, edges neat.

"Measure the line width." Dupont's voice was hoarse.

The software measured automatically: 15.8 nm, 1...

Error: less than 0.5 nanometers.

"Measure the overlay accuracy again."

Second item, third item, fourth item...

All passed.

When the last piece of data popped up as "Qualified," Dupont took off his reading glasses and muttered in broken Chinese: "We did it... using Xinghuo lenses, Xinghuo light sources, Xinghuo control systems."

He turned around and looked at the Chinese engineers behind him.

"Children," he said, "you have created a dragon. Now, it's time to let it fly."

The laboratory erupted in cheers—more enthusiastic than upstairs, because the pressure here was greater and the sense of despair was heavier.

When the lithography machine project started, many people said it was "nationalistic fanaticism" or a "money-wasting fantasy."

Even within Xinghuo, there were voices of doubt.

But today, they proved everything.

Dupont walked to the window, watching the eastern sky turn white with the dawn.

He remembered thirty years ago, when he first joined ASML.

At that time, ASML was still an obscure little company, suffocated by Japan's Nikon and Canon.

But the Dutch persisted, and in thirty years, they became the world's number one.

"Now it's China's turn." He said softly, "And... they are faster."