The air in the laboratory was thick with a tension that felt almost solidified, as if each breath required extra effort. Xiuxiu stood before the observation window of the EUV light‑source prototype "Candle Dragon," her gaze locked onto the constantly flickering red curve on the real‑time monitoring screen—the live thermal‑deformation data of the mirror. The curve resembled a dying viper, writhing frantically above the danger threshold; each peak stabbed her retina like a needle. Seventy‑two consecutive hours had passed. The team had tried seven different passive cooling schemes, all with the same outcome: when the light‑source power reached one‑hundred‑twenty watts—close to half the two‑hundred‑fifty‑watt threshold needed for mass production—the core Bragg reflector would undergo microscopic deformation due to the immense thermal load that could not be dissipated in time. This deformation, seemingly negligible at mere nanometers, was enough to cause fatal deviation in the meticulously designed reflection path, resulting in blurred patterns projected onto the silicon wafer and zero yield. The pearl on the crown of industry was being scorched dim by the heat it itself generated.
"Failed again." Engineer Wang's voice was dry, raspy from sleepless nights. He removed his glasses and rubbed his blood‑shot eyes hard. "The traditional heat‑sink material supplied by Zeiss in Germany has reached the limit of thermal conductivity. The heat flux density is too high; heat accumulates at the mirror substrate, like... like covering the mirror with a thick cotton quilt."
Xiuxiu did not reply immediately. She simply stared at the red curve, her mind drifting back to that late‑night call with Mozi a few days earlier. When she was almost drowning in frustration, her voice trembling with an imperceptible shiver, Mozi on the other end fell silent for a moment, then spoke in an unusually calm tone: "Xiuxiu, remember my shock model? Gradient descent. We never know where the true valley bottom lies; we can only probe repeatedly, measure the current slope, then take a step toward the direction that 'feels' downhill. Sometimes you think you've hit the lowest point, yet the next step might be lower still. The key is not to stop where you are, nor to give up because of one wrong step. Right now, you are measuring the 'slope of thermal management' for the whole team." He offered no hollow encouragement; instead, using knowledge from his own field, he gave her a new perspective to understand the problem. In that moment, what she felt was not just financial support but a spiritual anchor. A sense of dependency quietly sprouted, transcending mere gratitude.
She took a deep breath, turned, and faced the team members—equally weary but with eyes still unextinguished. "The passive‑cooling path may truly be at its end," her voice rang exceptionally clear in the silent lab. "We must introduce a more radical approach—active cooling."
"Active cooling?" Engineer Li frowned. "In an ultra‑high‑vacuum environment? And still guarantee extreme stability, with no vibration introduced? That difficulty..."
"The difficulty is great, I know." Xiuxiu cut him off, walking to the central console and calling up the three‑dimensional structural model of "Candle Dragon," highlighting the mirror‑mount region that enveloped the Bragg reflector. "That's why we cannot proceed recklessly. Before starting modifications, we must first 'see' how heat flows and how it causes deformation. We need to perform multi‑physics coupled simulation."
Her words brought pensive looks to several senior engineers. Multi‑physics coupled simulation was a relatively novel—and, to these veterans accustomed to relying on experience and experimentation to "tough it out"—somewhat "abstract" concept.
"Xiuxiu, simulation is after all simulation; there's always discrepancy with actual conditions..." Engineer Wang cautiously raised the objection.
"Precisely because it can reveal details invisible to our naked eyes and simple sensors, it is crucial." Xiuxiu opened a complex software interface; a finely meshed model of the mirror mount and reflector appeared on screen. "What we want to simulate is the interaction among thermal, mechanical, and optical physical fields. First, 'thermal.'" Her fingers tapped the keyboard, setting boundary conditions like source power and thermal‑load distribution. "Extreme‑ultraviolet light is absorbed by the reflector; although efficiency is already extremely high, the remaining few percent of energy converts into heat, injected into the mirror surface. How does this heat conduct through the mirror material itself, through the contact interface between it and the mirror mount?" She launched the thermal‑conduction module; a colorful temperature cloud map began to form on screen, from red representing high temperature to blue representing low temperature, clearly showing areas of heat accumulation.
"Look here," she zoomed in on the edge where the mirror contacted the mount. "Conventional design considers this the main cooling path, but our simulation shows that due to incomplete matching of thermal‑expansion coefficients and microscopic contact irregularities, a huge 'contact thermal resistance' forms here. Heat piles up, like a highway encountering a toll booth—throughput efficiency plummets." This was a detail they hadn't clearly recognized before.
"Next, 'mechanical.'" Xiuxiu switched the coupling settings, feeding the thermal‑analysis results as input into the structural‑mechanics analysis module. "Non‑uniform temperature distribution generates thermal stress. The material 'wants' to expand but is constrained by the surrounding structure, creating internal stress." On screen, the temperature cloud gradually gave way to a stress cloud; red high‑stress zones spread like fierce blood vessels across the mirror back and key junctions of the mount. "These thermal stresses are the direct cause of microscopic warping and distortion in the mirror. The simulation can precisely calculate, at different power levels, how many nanometers of deformation occur at specific mirror locations, and the deformation mode—whether the whole surface bulges like a lid, or local areas twist like potato chips." She pulled up measured deformation data from a previous failed trial, overlaying it with the simulation results; the curves matched closely in trend, only the simulation had predicted the deformation pattern and magnitude ahead of time.
"Finally, 'optical.'" Her operations grew even more focused, mapping the computed thermally‑induced deformation data into the optical model, simulating its effect on reflection of the 13.5‑nm extreme‑ultraviolet light. "Look, even a local bulge of just a few nanometers acts like a pebble dropped onto a perfect plane, disturbing the reflected wavefront. The eventual image point formed on the silicon wafer shifts, defocuses, even splits." On screen, a light spot that should have been a sharp point became blurred, trailing, matching exactly the defect patterns they'd seen in actual exposures.
The entire simulation process was like a meticulously choreographed surgical dissection, peeling back layer after layer of the black‑box process "heat causes deformation, deformation affects light path," laying it out clearly before everyone's eyes. The lab fell silent, save for the hum of server fans and the faint sound of data flowing on screens. The vague sense of defeat based on experience and intuition was replaced by this digitally precise insight down to nanometers and milliseconds. The problem had never been clearer.
"Thus, our goal is clear." Xiuxiu concluded, her voice carrying a strength born from insight. "First, reduce the total thermal resistance along the path from heat source to heat sink, especially this deadly 'contact thermal resistance.' Second, actively and precisely compensate for unavoidable thermal deformation."
Just then, the lab door opened; the logistics supervisor entered with several sealed temperature‑controlled containers. "Engineer Xiu, the diamond‑copper composite high‑thermal‑conductivity material you urgently ordered from German company MCT has arrived! Also, the samples of micro‑piezoelectric‑ceramic actuator arrays coordinated by CEO Mo through special channels have been delivered simultaneously."
A slight stir ran through the team. Diamond‑copper composite material, with theoretical thermal conductivity several times that of traditional copper‑tungsten alloy, was practically the top‑tier thermal‑conductive material obtainable in engineering. The piezoelectric‑ceramic actuators, capable of producing precise displacements at micron or even nanometer scales under voltage drive, were ideal actuators for active deformation compensation.
A glimmer of light flashed in Xiuxiu's eyes—the spark of hope. She stepped forward herself, donning sterile gloves, carefully opening the temperature‑controlled container and extracting that composite sample shimmering with a mix of metallic and carbon‑crystal luster. Cold to the touch, heavy, as if containing immense energy‑conduction potential. "Immediately sample for basic thermal‑property testing, and calibrate against material parameters in the simulation model."
Next, she opened another box, inside which dozens of match‑head‑sized piezoelectric‑ceramic chips were neatly arranged, silver electrode leads finer than hair. "These actuators will be the core of our 'active‑cooling system.' We'll embed them into the mirror mount, pressing closely against the back of the reflector."
A new battle began. Based on the calibrated high‑precision multi‑physics coupled simulation model, the team started intensive design iterations. Simulation replaced much trial‑and‑error experimentation, greatly shortening the development cycle. In the virtual world, they tried dozens of different mount structures, cooling‑channel layouts, actuator arrangements, and control algorithms.
"Plan A: add a heat‑spreader, but weight exceeds limit, affects wafer‑stage dynamic performance... reject."
"Plan B: adopt micro‑channel liquid cooling, but flow resistance too high, requires high‑pressure pump, vibration risk high... reject."
"Plan C: place actuators directly behind mirror, but thermal load directly affects actuator performance... high risk, keep as backup."
...
Each virtual failure meant saving weeks or even months of actual machining and testing time. Xiuxiu practically lived in the lab, dozing off in the adjacent rest room when exhausted, waking to continue scrutinizing simulation results, discussing optimization with team members. Her eyes grew blood‑shot from long hours staring at screens, yet her gaze grew brighter. Mozi occasionally sent messages—no excessive questioning, just simple "Progress?" or "Take care." She replied equally briefly: "Hacking away" or "Progress." But this silent companionship and understanding, like an undercurrent, supported her through each hurdle.
Finally, a scheme named "Nirvana‑1" stood out in simulation. Its core was a "graded‑heat‑conduction and active‑deformation‑compensation synergistic system." First, fabricate a completely new mirror mount using diamond‑copper composite material, and employ nano‑silver‑paste sintering technology at the contact interface with the reflector to form an ultra‑thin yet excellent‑thermal‑conductivity bonding layer, drastically reducing contact thermal resistance—this was "draining," efficiently guiding heat out. Second, inside the mount, design a "latent‑heat storage layer" based on phase‑change‑material microcapsules, absorbing or releasing heat during instantaneous power fluctuations, acting as a "buffer"—this was "storing." Lastly, the most innovative part: at specific locations within the mount, embed thirty‑six micro‑piezoelectric‑ceramic actuators, linked to a closed‑loop control system based on real‑time thermal‑deformation feedback.
Simulation results showed that when the source power reached one‑hundred‑eighty watts, the system could keep the temperature rise in critical mirror regions within fifteen percent, while the actuator system could apply opposing forces based on predicted thermal‑deformation models and real‑time sensor feedback, compensating mirror deformation down to under half a nanometer, fully meeting lithographic‑imaging requirements.
Simulation success was just the first step. The real challenge lay in turning "Nirvana‑1" from digital model into reality. Precision machining, sterile assembly, vacuum packaging... each step demanded extreme precision and cleanliness. During integration of the actuator array, a micron‑level alignment error caused the first prototype to fail during testing, sending team morale plummeting.
"Gradient descent." Xiuxiu looked at the dejected team members and uttered only these two words. Everyone paused, then understood her meaning. No complaints, no discouragement; disassemble, analyze failure, adjust assembly process, start over.
When the second "Nirvana‑1" mirror‑mount assembly was carefully installed into "Candle Dragon's" vacuum chamber, all wiring and cooling lines connected, the control system completed final integration—two more weeks had passed. The night of final testing, the lab was packed; even some technicians who had clocked out heard the news and returned, standing quietly at the periphery. An air of near‑sacred solemnity filled the space.