The air in the cleanroom carried a peculiar, chilled scent—multilayer‑filtered, as if even the flow of time here were deliberately slowed and purified. Xiuxiu stood beside the massive prototype lithography machine, gazing through the observation window at that crucial subsystem inside: the **dual‑stage**. It was not a single platform but two completely independent, ultra‑high‑speed, ultra‑precision motion platforms, like two ballet dancers about to perform an exquisite duet in the microscopic world, encapsulated in a vacuum and constant‑temperature absolute environment, awaiting fate's command.
A taut mixture of anticipation and anxiety pervaded the lab. Over the past few months, even years, most of the team's energy had been poured into this system, regarded as the soul of modern lithography machines. Breakthroughs in light sources, polishing of objective lenses—these were arduous enough, but without the dual‑stage delivering the wafer to the exposure position with atomic‑scale precision and lightning speed, achieving near‑perfect synchronization, all resolution improvements would be meaningless. It was like possessing a magical brush capable of drawing the finest lines, yet lacking a hand steady to the extreme, tirelessly delivering the canvas.
She had heard snippets about the market storm Mozi recently endured. That world seemed infinitely distant from her environment filled with robotic arms, laser interferometers, and the hum of vacuum pumps, yet she could imagine the pressure when order built on probability and models abruptly collapsed. She sent him that message about "iterative optimization" and "coffee"—the most direct empathy she could think of. In her world, however difficult, problems were concrete technical points that could be measured, attempted, and conquered. His battlefield had infinite variables, unpredictable human minds. She was grateful she chose to deal with tangible physical laws, even if these laws often proved unruly at extreme precision.
"Engineer Xiuxiu, final system self‑check completed. All parameters within theoretical range." A young engineer's voice came through the internal comms, carrying a barely perceptible tremor.
Xiuxiu took a deep breath, banished personal distractions, her gaze sharpening with focus. "All units confirm readiness. Prepare to initiate 'Dance' sequence." Her voice, transmitted via microphone throughout the control room, was calm and firm, like a steadying anchor.
**The core idea of the dual‑stage system lies in separating the two most time‑consuming steps in the lithography process—exposure and measurement/alignment—assigning them to two stages executing in parallel.** Traditional single‑stage lithography machines, after completing exposure of a chip region (die) on a wafer, need to move the wafer to the measurement position for alignment measurement of the next region, then move back for the next exposure. This "move‑measure‑move‑expose" serial process severely limits the machine's throughput.
The dual‑stage, like an efficient production line: One stage (say, Stage A) carries a wafer undergoing exposure beneath the objective lens. Simultaneously, the other stage (Stage B) carries another wafer, or a different region of the same wafer, performing high‑speed alignment and leveling & focusing measurements under an independent measurement system. When Stage A finishes exposure, Stage B completes measurement precisely. Then, under the control system's command, the two stages perform a high‑speed, precise exchange—Stage A moves away for measurement, Stage B moves in for exposure. Cycling thus, exposure and measurement operate in parallel, theoretically doubling throughput.
The principle sounds simple, but implementation is a challenge to the limits of precision engineering. The success of this "dance" depends on the absolute quality of the two "dancers" and their perfect synergy.
First, **the stage's own precision.** This is not merely static positioning accuracy but dynamic accuracy and stability during high‑speed motion. Stages must achieve near‑mythical performance metrics in several key dimensions:
**Nanometer‑level positioning accuracy:** To project circuit patterns accurately onto the wafer, the stage must stabilize the wafer at the target position, with errors controlled within a few nanometers—equivalent to a few ten‑thousandths of a hair's diameter. This demands the drive system—typically linear motors—to possess extremely smooth thrust and resolution, along with ultra‑precise guides or magnetic‑levitation technology to eliminate mechanical‑friction‑induced creep and vibration.
**Extremely high motion speed and acceleration:** To boost throughput, stages must complete long‑distance moves (e.g., from exposure to exchange positions) in minimal time, requiring very high acceleration and deceleration, often reaching several Gs. This imposes extreme demands on structural rigidity, material lightweighting, and drive power.
**Ultra‑low vibration and deformation:** Any tiny vibration, whether from external environment (ground micro‑tremors) or internal motors, pumps, acts like an earthquake disrupting exposure stability. Thus, stages themselves need active damping or passive vibration‑isolation techniques; structural materials must have extremely high rigidity and low thermal‑expansion coefficients, ensuring minimal deformation under high‑speed motion and varying thermal loads.
**Six‑degree‑of‑freedom precise control:** Stages must move precisely in X‑Y horizontal directions, control Z‑direction lifting (focusing), and minute rotations about X, Y, Z axes (leveling). These six‑degree motions require independent yet coordinated control, ensuring the wafer surface always perfectly coincides with the projection objective's focal plane.
Second, and the most challenging part of dual‑stage systems—**scan synchronization control.**
When a stage performs scanning exposure (modern lithography widely uses scanning: the objective projects a slit of light, the stage carries the wafer uniformly scanning through this slit to complete exposure), demands on motion smoothness reach pathological levels. Not only must speed be extremely stable (speed error less than a few ten‑thousandths), but the trajectory must be nearly ideal straight line (straightness error at nanometer level).
At the moment of dual‑stage exchange, the system is most vulnerable. Two precision platforms each weighing hundreds of kilograms, moving at high speed, must within extremely short time (milliseconds), in limited exchange space, complete trajectory crossover and role swap. This process is called the "swap sequence."
**Exchange synchronization** is critical. Both stages must strictly follow preset time‑space trajectories; any tiny asynchrony—whether a microsecond delay or nanometer‑scale misalignment—could lead to catastrophic consequences. Light: wafer mispositioning, scrapping the entire wafer. Severe: physical collision of two high‑speed massive platforms, utterly destroying core components worth tens or even hundreds of millions of RMB.
This demands the control system have:
* **High‑bandwidth servo controllers:** Able to process real‑time massive data from position sensors like laser interferometers (sampling frequency often MHz‑level), quickly compute drive signals, correcting any minute trajectory deviation.
* **Precise trajectory‑planning algorithms:** Pre‑compute two absolutely safe, efficient, smooth motion trajectories, ensuring the two stages "brush past" yet never collide during exchange.
* **Powerful real‑time computing capability:** Continuously monitor actual position, speed, acceleration of both stages throughout motion, compare with target values, performing feedforward and feedback control, resisting various internal/external disturbances.
Xiuxiu's team faced precisely this final hurdle of sync control. They had solved individual‑stage precision, but during dual‑stage joint debugging, always encountered tiny, random asynchrony at exchange moments, causing yield fluctuations, unable to meet stable mass‑production requirements. The problem was elusive, as if between the two dancers existed an ineffable yet tangible subtle interference.
"Initiate!" Xiuxiu gave the final command.
In the control room, on the large display, data streams began refreshing cascadingly. Light‑points representing the two stages' positions started moving along preset complex trajectories on a simulated 2D plane. Real‑time footage from high‑speed cameras showed those massive, metallic‑cold‑glowing platforms beginning a breath‑holding dance.
Initially, all smooth. Stage A accelerated steadily, carrying that silicon wafer bearing countless hopes, sliding toward the exposure area under the objective. Laser‑interferometer data showed its motion trajectory smooth as if guided by an invisible hand. Simultaneously, Stage B in the measurement area, using alignment marks and leveling sensors, rapidly and precisely acquired position and orientation data for the next wafer.
"Exposure commencing… scanning speed stable… trajectory deviation <0.5 nm…"
"Stage B measurement completed, data valid."
…
Green lights flashed frequently; the control‑room atmosphere eased slightly. But everyone's hearts were in their throats, awaiting that critical node—exchange.
"Preparing exchange sequence… 3… 2… 1… Exchange initiate!"
On screen, light‑points of the two stages began moving at higher speed along preset, interweaving complex curves. Real‑time data‑stream velocity surged abruptly; servo‑controller load indicator bars instantly spiked.
Xiuxiu stared intently at the sync‑error monitor data—a curve magnified ten‑thousand‑fold, originally should be a stable zero‑line. Yet, as the exchange action reached mid‑point, the curve jerked: a tiny but clear spike appeared abruptly!
"Sync‑error alarm! X‑axis, peak 3.2 nm!" The system emitted a cold prompt.
Dead silence in the control room. This problem again! That ghost‑like issue haunting them for weeks. Though no collision this time, 3.2‑nm sync error was enough to cause unacceptable misalignment on chip circuits; yield would inevitably fail standards. Dejection and fatigue surged like tides threatening to overwhelm everyone.
Xiuxiu didn't speak immediately. Closing her eyes, she rapidly replayed every aspect of the system in her mind. Electromagnetic interference from drive motors? Control‑algorithm delay? Structural‑vibration coupling? Or… She snapped her eyes open.
"Check phase noise of linear‑motor power supplies for both stages!" She almost shouted. A previously overlooked detail flashed: due to layout constraints, power‑supply cables for the two stages ran parallel very close for a segment. At extremely high switching frequencies and currents, they might generate exceedingly weak electromagnetic coupling. This coupling's effect was negligible when a single stage operated, but during the two stages' high‑speed, complex, strictly synchronized exchange action, that minute disturbance was amplified, fed back through the servo system, causing those randomly appearing sync‑error spikes!
The technical team acted immediately, pulling up power‑monitor data for analysis. Indeed, during exchange moments, extremely subtle, correlated noise spikes were detected on both power supplies' current waveforms!
"Re‑route power cables, add shielding! Adjust control algorithm, add feedforward compensation targeting this specific‑frequency noise!" Xiuxiu issued rapid instructions, voice hoarse with excitement. This was a "devil detail"—seemingly simple, requiring extremely keen insight and profound system‑wide understanding to uncover.
Adjustments and testing continued several more hours. When all was ready, restarting the "Dance" sequence, everyone's breath nearly stopped.
Exposure, measurement, smooth running… Again reaching exchange node.
"Exchange initiate!"
Data‑streams surged, trajectory light‑points interwove. Xiuxiu fixed her eyes on that sync‑error monitor curve. Time seemed stretched; each millisecond felt like an eternity.
The curve… remained mirror‑smooth! Throughout the exchange process, that ten‑thousand‑fold‑magnified error curve clung tightly to the zero‑line, almost no fluctuation!
"Exchange completed! Sync‑error peak… 0.8 nm!" The report carried unbelievable delight.
Subsequent exposure and measurement processes proceeded seamlessly. When the first wafer processed by the new dual‑stage system was conveyed out, preliminary inspection showed all key overlay‑accuracy data meeting or exceeding design targets, the control‑room silence broke into deafening cheers!
Young engineers hugged, high‑fived, some even jumped excitedly. Years of effort, countless sleepless nights, finally bore fruit. They had successfully tamed this pair of microscopic‑world precision dancers, making them perform perfectly synchronized waltzes.
Xiuxiu stood in place, a relieved smile spreading, but her eyes uncontrollably moistened, warming. She didn't cheer like others; just stood quietly, feeling that surge welling from deep within—a mix of immense accomplishment, post‑long‑pressure‑release exhaustion, and inexpressible excitement. Tears slid silently down her cheeks, pale from long nights, dripping onto the cleanroom suit's chest.
She walked alone to the observation window, gazing at the massive machine standing silently inside which just completed a heart‑stopping dance. Ten years of returning‑home journey, countless technical‑barrier breakthroughs, immersion‑technology conquest, EUV‑light‑source nirvana… scenes flashed before her. The dual‑stage conquest was another crucial milestone in this long march. It meant they not only mastered core lithography imaging technology but also the key "rhythm controller" enabling efficient, stable mass production.
Taking out her private communicator, her finger hovered momentarily, ultimately didn't dial any number. Beneath the huge success joy lay a deep weariness and a complex emotion wanting to share yet unsure how. She simply sent Mozi a message:
"The dance succeeded."
She didn't know if he was still handling market aftermath, or if he could understand the weight behind these five brief words. She just felt she should tell him.
To her surprise, mere hours later, dragging an utterly exhausted body, walking out of the lab building toward nearby apartment, a familiar figure leaned against a black car, quietly waiting.
It was Mozi.
Shanghai's autumn night had chill; he wore simple dark coat, face bearing traces of travel‑weariness, but eyes bright and warm, smiling at her.
Xiuxiu froze, steps halting. She hadn't expected him to come, especially so quickly.
Mozi stepped forward, saying nothing, just opened arms, gently yet firmly embracing her.
This hug held no ambiguity; more like comrade‑style, deep understanding and solace. He understood the immense effort behind her present success, understood all hardship and persistence contained in those silently flowing tears. He might not grasp specific technical details of "dual‑stage sync error," but fully understood the significance of conquering such a core challenge and the dedication it demanded.
Xiuxiu's stiff body slowly relaxed within his warm embrace. She didn't cry aloud, just rested forehead gently against his shoulder, feeling that silent support and affirmation. Night breeze brushed past, carrying distant city faint clamor, yet accentuating this moment's tranquility and reality.
"I saw the news," Mozi murmured, voice bearing a subtle hoarseness, "and received your message. The storm over there temporarily subsided; I thought… should come see you." He didn't say "congratulations," nor asked details; just with action showed he was here, he understood.
Xiuxiu nodded lightly within his embrace. All words seemed to melt in this embrace bridging physical distance, connecting different battlefields. She felt warmth for his arrival, felt more grounded for her own success. Capital fluctuations, lithography precision—at this moment seemed to find some common footnote: that indomitable spirit facing extreme challenges in respective domains, and the unspoken support among fellow travelers.
Thus they quietly held each other in the night, like two ships sailing respective storms, temporarily entering a calm harbor, sharing each other's weariness and glory.