fictionpage Chapter 199 The Entire Audience Speechless
Chapter 199 The Entire Audience Speechless
March 12th, 9 PM Beijing time.
On the third day of the IEEE MEMS 2021 online conference, an inconspicuous panel discussion was marked on the agenda: "New Advances in Thermoelastic Coupled Nonlinear Modeling", organized by Professor Heinrich Vogt of the Technical University of Munich.
But everyone knew that this discussion group wasn't talking about any vague "new developments".
It discusses Wei Lan.
More precisely, it's the third-order nonlinear extended model that has shaken up the entire MEMS academic community over the past three months.
The conference system showed 682 participants online. For a specialized academic topic, this number is quite rare. Vogt has organized no fewer than a hundred seminars in his forty-year academic career, but very few have had more than three hundred participants.
And today, that number is still jumping up.
"Fellow colleagues, welcome to this symposium." Vogt's voice came through the microphone, carrying the characteristically meticulous rhythm of a German. "I believe everyone is already familiar with today's topic. Over the past few months, a research team from China has proposed a third-order nonlinear extended theoretical framework for thermoelastic coupling in MEMS. This theory has attracted widespread attention, but also sparked considerable controversy."
He paused for a moment, and the old man in front of the camera adjusted his glasses.
"Today, I've invited Professor Zhou Zhiyuan from East China University of Science and Technology, a key researcher in this theory, to give us a comprehensive technical report. Following that will be a free discussion session. I hope everyone will uphold academic integrity and speak with data and logic."
The chat area has started to flood with messages.
Klaus Stein of Bosch posted a message on his account, which always displays the suffix "Bosch Sensors Division": "Looking forward to the presentation." — The wording was polite, but anyone familiar with him knows that this technical head of Bosch's sensor division would never specifically wait online for something he considers "interesting but unimportant."
STMicroelectronics' Marco Bertoli did not speak, but his account icon was lit up in green, having been lit up for ten minutes before the meeting began.
Professor Kosuke Yamamoto of Kyoto University posted a message in Japanese that roughly translates to "Finally, the official report has arrived."
Franz Weber of Infineon was quietly listed on the attendees' list, neither speaking nor greeting anyone, like a pike lurking in the seaweed.
"Thank you for the introduction, Professor Vogt." Zhou Zhiyuan's voice rang out, his English with a Chinese accent, spoken slowly but clearly, each word as if etched with a chisel. "I am Zhou Zhiyuan, from the Micro-Nano Fabrication Research Center of East China University of Science and Technology. Today, I will spend about thirty minutes presenting our team's latest research results in the field of MEMS thermoelastic coupling."
The first slide lights up: "Third-order Nonlinear Extension: Theoretical Framework, Experimental Validation, and Scale Extrapolation for MEMS Thermal-Elastic Coupling".
The five names in the author column are neatly arranged: Su Chen, Zhou Zhiyuan, Chen Qiming, Lin Wei, and Zhang Li.
"Our work is divided into three parts," Zhou Zhiyuan said calmly, as if he were lecturing to doctoral students in a laboratory. "The first part is establishing the theoretical framework. The second part is experimental verification. The third part is scale extrapolation and prediction."
The slide flips to the first page of the theory section. A dense matrix of formulas fills the entire screen.
More than 600 online scholars and engineers stared at the Greek letters and integral symbols, and some began to discuss them in hushed tones in the chat area.
"The complete theoretical framework contains 147 core formulas, but I won't derive them one by one today," Zhou Zhiyuan said. "I'll only talk about the three most crucial breakthroughs."
It took him twelve minutes.
In twelve minutes, he started by discussing the fundamental limitations of the traditional second-order model—the second-order approximation is accurate enough below the 300mm scale, but when the wafer size breaks through 300mm and moves towards 400mm, the non-uniformity of the thermal field distribution makes the contribution of the third-order term change from negligible to non-negligible.
He presented a set of contrasting curves: blue represents the predicted values from the traditional model, red represents the actual measured values, and green represents the predicted values from the third-order model. At a 300mm scale, the three lines almost overlap. However, at the 400mm edge region, the blue line begins to deviate, like a derailed railway, veering further and further off course. Meanwhile, the green line clings tightly to the red line, as if drawn by a magnet.
In the chat area, Stein typed: "Impressive mathematics. But I've seen many beautiful models that fail in production."
This message was pinned to the top for a few seconds before being overwhelmed by a flood of new messages.
Zhou Zhiyuan seemed oblivious to the activity in the chat area, or perhaps he noticed but chose not to respond. He turned to the first page of the second part.
"Now we will proceed to the experimental verification phase."
A table appeared on the screen with the title "VL-400E Series Edge Verification Experiment".
"We completed fifteen sets of thermoelastic coupling tests on the edge region using Vilan's self-developed 400mm test line. Each set of tests included three temperature gradient sampling points—T1 at the center, T2 at 5 mm from the edge, and T3 at 10 mm from the edge."
Next slide. Fifteen S-shaped curves are neatly arranged in the coordinate system, like fifteen copper wires bent by the same hand.
"All fifteen tests observed the S-shaped turning point characteristic predicted by the third-order model. The hit rate was 100%."
The chat area suddenly fell silent.
The silence wasn't because no one was speaking; rather, everyone simultaneously stopped typing. Over six hundred cursors hovered above the input boxes, but no one pressed the enter key.
Zhou Zhiyuan continued, "The overall goodness of fit R² is 0.978. The deviation from T1 to T2 is 1.3%, and the deviation from T2 to T3 is 2.5%. The prediction deviation at the S-shaped inflection point is less than 0.3 millimeters."
Three seconds of silence.
Then the chat section exploded.
"R²=0.978?! On 400mm edge data?"
"15/15 S-curve observation. That's... statistically significant."
"The transition position error is less than 0.3mm? That's insane for a theoretical prediction."
"Wait, is this peer-reviewed data?"
The messages were pouring in like a waterfall. Vogt had to click the "slow mode" button, limiting each person to one message every ten seconds.
Kosuke Yamamoto was the first to raise his hand and ask to speak.
Fogg agreed.
"Professor Zhou, hello." Yamamoto's English was thickly Kansai-accented. "Your data is very convincing, but I have a question. While fifteen samples may not be statistically insufficient, is this sample size too small for a theoretical model that claims universality? Especially—these data all come from your own experimental lines, right?"
Zhou Zhiyuan nodded.
"Professor Yamamoto is right, fifteen samples is indeed not a large number. But I need to point out that our fifteen tests cover three different temperature gradient ranges, two different etching depth combinations, and three different edge-to-center distance ratios. Within this multi-dimensional cross-validation framework, the fifteen sets of data provide far more information than fifteen repetitions of a single variable."
He paused for a moment.
"But you're right, there is indeed a gap between lab data and mass production data."
Someone started typing in the chat: "So it's lab data only? That's a significant limitation..."
Zhou Zhiyuan glanced at the time in the lower right corner of the screen, and then said something that surprised everyone.
Regarding mass production data—I would like to invite my colleague, Ms. Lin Wei, to supplement this section.
A new video window popped up.
Lin Wei's face appeared on the screen. She was wearing a dark blue business jacket, with the logo of Vilan Microelectronics Technology Co., Ltd. on the wall behind her. The light shone on her face, highlighting her clear and composed features.
"Hello everyone, I'm Lin Wei, CEO of Vilan Microelectronics." Her English was much more fluent than Zhou Zhiyuan's, and she spoke quickly but clearly. "Professor Zhou just presented the laboratory verification data. Next, I will present the commercial mass production data."
The chat area fell silent once again.
A new slide appeared on the screen, titled "WL-300 Series Commercial Delivery — First Batch Quality Report".
"On February 25th, Vilan completed the pre-inspection of its first batch of commercial sensors. On February 26th, 150,800 sensors were officially shipped to eight core customers in our supply chain alliance."
Below is a set of data:
Total shipments: 150,800 units
Overall yield rate: 93.6%
Process flow: Standard 300mm production line
Production line optimization: None
Screening process: Routine quality inspection, no additional screening.
"Among them, Zhenxin Microelectronics, as our core customer for inertial navigation sensors, conducted a ten-day integration test on the sensors we delivered."
Slideshow slide turn.
A set of accuracy test data for the inertial navigation module appeared on the screen. The most crucial number was highlighted in bold red:
Zero drift stability: ±0.021°/h
"What does this number mean?" Lin Wei's voice was as calm as reading a weather forecast. "Currently, the mainstream automotive-grade inertial navigation sensors on the market generally have a zero-drift stability between ±0.035°/h and ±0.050°/h. Bosch's BMA456 series has a publicly stated value of ±0.038°/h."
She paused for just two perfect seconds.
"Our sensor achieved ±0.021°/h directly on a standard production line, without optimization or screening. This is 40% to 60% better than mainstream products on the market."
The slow-motion mode in the chat area was no match for the flood of messages. A new message popped up every ten seconds, most containing only one or two words—
"Unbelievable."
"Standard line? No optimization?"
This changes everything.
Stein finally spoke up. Not in the chat area, but by raising his hand to request a voice message.
Fogg agreed.
"Ms. Lin." Stein's voice came through the microphone, thick with a German accent, speaking slowly, each word as if carefully chosen before being uttered. "Regarding the accuracy data you just presented—±0.021°/h—I need to confirm one thing."
"Please speak."
Is this data the optimal value for a single sensor, or the statistical average of the batch?
"Statistical mean," Lin Wei replied. "The core sample was tested twelve times, with twelve independent tests. What I'm showing is the mean of those twelve samples. The best value for a single sample is ±0.018°/h, and the worst value is ±0.025°/h. The consistency deviation between particles is less than two percent."
Stein remained silent for five seconds.
In an online academic conference, five seconds of silence can feel like an eternity.
"...the consistency deviation between particles is less than two percent." Stein repeated this sentence, the rigor in his voice unable to conceal a certain tremor beneath. "This consistency data—on the standard production line—with all due respect, is even better than the data from our internal R&D line."
He didn't say anything more. But everyone understood the unspoken words—
This shouldn't be the case.
A Chinese startup, less than two years old, produced sensors more stable than Bosch's R&D line using a standard production line without any process optimization. This shouldn't happen in Stein's understanding.
But the data is there for all to see.
Kosuke Yamamoto raised his hand again. Fogg agreed.
"Ms. Lin, I have a question." Yamamoto's tone was much more polite than before. "Zhenxin's test data is indeed impressive. But Zhenxin is your customer; they use your sensors for integration testing. From a data independence perspective—this still falls within the scope of verification within your supply chain."
He paused, and his wording became more cautious.
"My question is—have any independent third-party laboratories independently verified the predictions of the third-order model using equipment and methods not supplied by Vilan?"
The chat area fell silent again.
This is a pointed question, and also a legitimate one. In academia, proprietary data is never as convincing as third-party data. No matter how impressive Vilan's commercial data may be, as long as all verification comes from Vilan's own supply chain system, there will always be a suspicion of "talking to oneself."
Lin Wei's expression remained unchanged.
"Professor Yamamoto, your question is very reasonable." Her voice remained calm. "Frankly, as of now, we do not have complete validation data from independent third-party laboratories."
Someone started typing in the chat: "So no independent verification yet..."
The air seemed to freeze. The more than six hundred people watching online held their breath, waiting for the next sentence.
Just as Vogt was about to say something to ease the tension—
A new video window popped up.
In the list of attendees, a name that had been grayed out suddenly turned green. The organization logo next to the name read: "Ishikawa Lab, Tokyo Institute of Technology".
Akira Ishikawa.
The moment the camera panned up, the face of a middle-aged Japanese man appeared on everyone's screen. He wore a greyish-white lab coat, and behind him was a row of neatly arranged high-precision testing instruments. The camera was slightly tilted, clearly an angle adjusted hastily.
"Excuse me for interrupting everyone." Akira Ishikawa's English had a Tokyo accent, and he spoke much slower than Yamamoto, as if he were translating each word in his head before uttering it. "I'm Akira Ishikawa from the Precision Instruments Laboratory at the Tokyo Institute of Technology. I'd like to... please allow me to share some data."
Vogt paused for a moment, then quickly said, "Please go ahead, Professor Ishikawa."
Akira Ishikawa tapped the screen share button. A hand-drawn data chart appeared before everyone—not the kind of elaborately crafted charts used in academic papers, but raw data records photographed directly from a lab notebook.
"For the past two months, my lab has been conducting research completely unrelated to Wei Lan—we're developing a new type of MEMS gyroscope for satellite attitude control," said Akira Ishikawa, his voice carrying the cautiousness typical of a scholar. "In this process, we encountered a system bias problem caused by thermoelastic coupling. Traditional models cannot explain the bias patterns we observed."
He flipped to the next picture.
"In January, I saw a preprint of Mr. Su Chen's team's third-order model on arXiv. Out of curiosity, I recalculated the systematic bias of our gyroscope using their theoretical framework."
The screen displayed a comparison of two columns of numbers—the left column showed the "measured values," and the right column showed the "predicted values from the third-order model."
Sixteen lines of data.
The deviation between the measured and predicted values for each row is less than 1.5 percent.
"Sixteen independent tests showed an average deviation of 0.83% between the measured values and the predicted values from the third-order model." Ishikawa Akira's voice trembled slightly—not from nervousness, but from the barely suppressed excitement of a true scholar facing unexpected data. "And—and this is the most crucial part—our gyroscope uses a completely different manufacturing process and material system than Vilan's. Different equipment, different substrates, different etching methods, different packaging solutions."
He stopped.
"In other words, this is a set of completely independent third-party data that has no supply chain relationship with Vilan. And the third-order model's prediction accuracy for this set of data is almost identical to its prediction accuracy for Vilan's own data."
The chat area wasn't bombed.
Because at that moment, more than six hundred people simultaneously forgot how to type.
The silence lasted for a full eight seconds.
Then Vogt's voice came through the microphone. This 62-year-old German professor, with a 40-year academic career and countless academic presentations and technical debates, spoke for the first time with a slightly unsteady tremor.
"Professor Ishikawa... thank you for sharing." Vogt took a deep breath. "I need to confirm—was the data you just presented completed independently by your lab, without using any equipment, materials, or technical support provided by Vilan?"
"Completely independent," Akira Ishikawa said. "Mr. Su Chen and I have never had any technical exchanges by phone, email, or in any other form. I simply read their publicly released preprints and then used their theoretical framework to calculate my own data."
Vogt nodded slowly.
"Well then," his voice regained its composure, but with an unprecedented solemnity, "we are now facing a situation—a theoretical model that has undergone three independent verifications. The first is data from the researchers' own laboratory. The second is integrated data from commercially available customers. The third is data obtained from a completely independent third-party laboratory using entirely different processes and material systems."
He took off his glasses and rubbed the bridge of his nose.
"Ladies and gentlemen, in my forty-year academic career, I have rarely seen a new theory receive so much multi-dimensional verification and support in such a short period of time after its proposal."
The chat area is finally back online. Messages are pouring in like a flood.
This is extraordinary.
"Three independent validations in less than 3 months."
"The implications for 400mm process design are enormous."
"Has anyone started replicating the model calculations?"
"I'm downloading the preprint right now."
Stein did not speak in the chat area.
His account icon was still lit up in green, but he hadn't said a word since Ishikawa Akira started showing the data.
……
When Zhou Zhiyuan closed the meeting window, it was nearly midnight Beijing time.
The panel discussion, originally scheduled for one and a half hours, actually stretched to two hours and forty minutes. Vogt had to extend the meeting twice because the question-and-answer session was constantly interrupted by the raised hand button.
He leaned back in his chair, looking at the more than twenty emails that had flooded his inbox—from scholars all over the world, some asking for preprints of the full text, some for experimental data, and some to discuss the possibility of collaboration.
My phone vibrated. It was a WeChat message from Lin Wei.
"Is the meeting over?"
"Just finished."
"How is it?"
Zhou Zhiyuan thought for a moment, then typed four words and sent them.
"Exceeded expectations."
Lin Wei replied with a smiling emoji, then sent a longer message: "Chairman Chen Jun of Zhenxin just called me. He watched the entire meeting replay. He increased the order from 40,000 units to 100,000 units. Furthermore, he agreed to let us cite Zhenxin's integration test data as a commercial validation case in our subsequent paper."
Zhou Zhiyuan's finger stopped at the top of the screen.
One hundred thousand.
Two months ago, Vilan's first order was only 150,000 units. Now, Zhenxin alone wants 100,000 units. Zhou Zhiyuan understands this better than anyone—it means that Zhenxin has jumped from the "trial evaluation" stage to the "large-scale integration" stage. They are betting.
Just as he was about to reply, his phone vibrated again. This time it wasn't Lin Wei.
It's Su Chen.
"Teacher Zhou, I watched the meeting recording."
"Hmm, what do you think?"
"Akira Ishikawa's data is even cleaner than I expected. His gyroscope uses SOI substrates with reactive ion etching, a completely different process from ours. If the third-order model still applies under such differences, it means the model's universality is stronger than we estimated."
Zhou Zhiyuan nodded slightly—although Su Chen couldn't see his movement.
"Akira Ishikawa's data is very helpful for peer review," said Zhiyuan Zhou.
"I know," Su Chen replied, "but I'm more concerned about the second paper right now. Ishikawa Akira's independent data, combined with Zhenxin's commercial data, is enough to support a complete multi-platform validation paper. If the first paper passes review smoothly, the second one can be submitted immediately afterward."
Zhou Zhiyuan smiled.
This student—no, we should now call him this young researcher—is always half a step ahead of everyone else in his thinking. While everyone was still immersed in the awe of the IEEE MEMS session, Su Chen was already planning his next move.
"Don't rush," Zhou Zhiyuan typed in reply. "The first one hasn't been approved yet. Let's wait for news from NM."
"good."
Su Chen's last message was brief and almost indifferent. But Zhou Zhiyuan knew that behind those words lay not indifference, but a near-obsessive restraint—not to celebrate prematurely, not to plan ahead, and not to let any emotions influence the next step of judgment before the results were in.
This is a trait that Su Chen has had since day one.
This is also what Zhou Zhiyuan admires most about him.