I started with a ship full of max-level veterans, and I'm just a complete noob.

By the time the car arrived at the Institute of Physics, the sky had already darkened to the rooftops.

Jiang Ming handed the sealed document bag to the gatekeeper for registration and hurried along the corridor to the semiconductor laboratory. Before he even opened the door, the sounds of oscilloscopes and Wang Shouwu's voice could already be heard inside.

"Jiang, take a look at this one first."

Jiang Ming pushed open the door, and the pickling solution, anhydrous alcohol and residual heat from the oven were mixed together. The table was covered with cut silicon wafers, broken leads and written process cards. Three oscilloscopes were lit up side by side, and the sweep curves extended to the right on the fluorescent screen.

Wang Shouwu bent down and stared at one of the lines, his eyes bloodshot, the enamel cup beside him already cold.

He came up to Jiang Ming and slapped a registration form into his hand. The corner of the paper was stained with microscope oil. Below the title were the numbers of twenty test tubes. Fifteen of the numbers were circled in red pencil and marked with "leakage current over limit" next to them. The remaining five numbers were labeled with junction depth, reverse current and cutoff frequency.

"Don't look at the broken tubes yet, five are done." Wang Shouwu pushed his glasses back up on his nose and tapped the last column with his finger. "312, 327, 308, the highest one is 341 MHz. The frequency sweep station has changed the calibration source twice, and the result is still in this range."

Jiang Ming followed the paper down, his gaze stopping at the 300 trillion mark before settling on the waste statistics at the front.

The note in the car that read "Return Immediately" has now transformed into the set of numbers on the table that are enough to change the circuitry of the components.

Wang Shouwu awaited his judgment, his chest heaving rapidly, his fingers still pressed against those numbers, as if worried that Jiang Ming would only see the fifteen scrapped pipes and overlook these five results that could change the course of the project.

"Test conditions?" Jiang Ming asked.

"Room temperature, collector voltage 12 volts, base bias according to the second set of process cards, load consistent with the batch of germanium tubes." Wang Shouwu handed over another record. "The test bench has been re-grounded by Engineer Lin, and Fang Xudong just used a point-contact germanium tube as a reference. The instrument readings are stable."

The researcher next to him, holding a device box, said wearily, "Two out of the five have reverse leakage currents that are close to the upper limit, and the junction depths of the other three are also uneven. The cross-sectional photos are worrying. If the process continues like this, mass production will only be possible by picking out the tubes."

"Fifteen devices are stuck in the oxide layer." Another operator spread out a photomicrograph on the table. "There are pinholes at the edge of the window, and the diffusion front has penetrated through several gaps. The boron layer varies too much in width. The device can make a sound, but its lifespan and consistency cannot be sustained."

Wang Shouwu's excitement was dampened by these words. He picked up the process card for the highest frequency tube. The junction depth measurement values ​​on the card were divided into three groups. The front area was close to the target, while the edge area was spreading outwards. Some numbers even exceeded the original window.

"Silicon has driven up the frequency, but the furnace has messed up the process." He said, rubbing his thumb back and forth on the edge of the card. "The difficulty of germanium tubes lies in the materials, while the difficulty of silicon tubes lies in the surface and junction depth. This diffusion furnace in front of us hasn't learned to repeat the process under the same conditions."

Jiang Ming took the three destructive cross-sectional photos, first looking at the center position, and then arranging the photos along the furnace tube axis. In the photos, the light-colored diffusion layer showed different widths on both sides of the window, with the cross-section closer to the air inlet being deeper, and a late-appearing shallow layer appearing at the tail end.

"Have you clearly remembered the sampling locations for these pieces?"

"Take a piece of paper every so often, from the front to the back of the quartz boat. The furnace temperature record is also there." Fang Xudong opened the record folder and put the temperature paper tape and gas path operation table next to the photo. "The constant temperature section is nominally 1,150 degrees. After the gas intake is switched, the curve deflects downward. When it recovers, it rises again. The temperature change is not significant, but the sedimentation depth deviation is concentrated in these two time periods."

Jiang Ming picked up a pencil and circled the timing of the air path switching on the edge of the paper tape, while pressing the light-colored border in the photomicrograph with his other hand.

"The problem here isn't with the shielding membrane," he said. "The intake disturbance changes the heat load in the furnace tubes. The constant temperature zone moves first, and the controller compensates according to the measured temperature. The diffused gas has already traveled a distance along the new position. When the hot zone returns, the sludge depth is left with two sets of conditions."

Everyone in the lab looked at the paper tape, and Wang Shouwu leaned forward.

You're referring to hot zone location drift, not the furnace temperature reading being out of tolerance.

"The total reading only represents the state near the thermocouple." Jiang Ming pointed to the smooth section of the temperature curve with the tip of his pencil. "In the diffusion furnace, there are hot zones, boat positions, gas flow, and reactant concentrations. The thermometer is still at the target scale, but the actual junction front has changed position. The controller of this furnace only focuses on one point, which is equivalent to using an outdated indicator to make decisions for all states."

He walked to the blackboard, erased the remaining silicon diffusion process parameters, first drew the furnace tube, quartz boat and thermocouple, then connected the inlet valve, flow meter and constant temperature zone with lines, and wrote down three sets of mutually lagging quantities on the right: temperature response, airflow response and junction depth response.

"During the test with the charge in the third furnace, the temperature in the middle section first increased to ±2.5 degrees Celsius. We checked the table and fed forward the heating power in advance. The proportional and integral circuits were only responsible for the final step, and the furnace temperature returned to ±0.3 degrees Celsius." Jiang Ming turned around. "In the diffusion furnace, we also need to distinguish which is the leading factor. The flow rate change occurs first, and the temperature change appears later. The sludge depth can only be seen when the sample is taken. By the time the results come out and the parameters are changed, the entire furnace sample is already scrapped."

Wang Shouwu took the chalk and drew a square next to the intake valve.

"Flow feedforward".

"Heating and cooling should be metered separately, as should intake switching," Jiang Ming continued. "When the valve is open, the heat load is compensated; when the valve is closed, it is recovered according to another curve. The flow meter provides a leading signal, the thermocouple only performs temperature closed-loop control, and the diffusion time and boat position provide an offset to the junction depth model. Each of the three lines undertakes its own task."

The researcher stared at the lines on the blackboard, recalling the repeated adjustments to the diffusion time over the past few months. Each time the furnace temperature record looked acceptable, the next batch of cross-sections would still show signs of spreading at the edges. The operator could only alternate between changing the gas pressure and holding time, filling the process card with handwritten arrows, but still could not find the real control entry point.

"How can we calibrate the model if there's no online sensor for the knot depth?" someone asked.

"Use test pieces for slow feedback," Jiang Ming said. "Retain a calibration piece in a fixed position for each furnace, and group the four probes and profile results into the same number. First, use three batches of data to establish the mapping from flow rate, temperature, boat position to knot depth. Then, each time the fuel is changed, only the bias is updated, and the main parameters are not changed. The measurement results arrive late, so we cannot let them lose their function; we can only let them participate in the next furnace."

Wang Shouwu's finger moved along the arrow on the blackboard, mentally combining the gas path changes, furnace tube temperature, and diffusion front into the same diagram. Fang Xudong had already pulled out a slide rule and written down the conversion relationship between flow rate step and temperature zone displacement on paper.

"You want to add perturbation to it?" Wang Shouwu asked.

"With small amplitude and fixed cycle, we place it during diffusion intervals or calibration phases to observe the hot zone's response to flow rate changes," Jiang Ming said. "First, we identify the furnace's condition, then decide how much to add. As the furnace tubes age, the sealing condition changes, and the boat position changes, the parameters are updated accordingly. No one needs to guess by feel."

Wang Shouwu broke the chalk in his palm, then turned and walked to the testing table, picked up the record of 341 megahertz, and slammed it heavily on the table.

"If this system is successfully implemented, the density dispersion of the junction can be reduced from the current percentage of over ten percent to the thousandth percent, and even the diffusion window can be widened again."

After he finished speaking, he looked down at the fifteen discarded tubes, his face a mixture of elation and regret. The five high-frequency tubes were enough to prove that the route had been chosen correctly, and the fifteen discarded tubes also reminded everyone that material advantages can only be transformed into deliverable devices when constrained by equipment and processes.

Jiang Ming put away the registration form, walked to the phone, and dialed back to Research Room 502.

"Da Liu, bring the voltage divider bridge, flow meter sampling resistor, and No. 3 furnace adaptive network diagram to the Institute of Physics tomorrow. Fang Xudong, organize the three batches of original records of the diffusion furnace into a unified number. Lao Sun, first check the gas path valve seat and thermocouple installation position. Disassemble the control cabinet first, but don't touch the furnace tubes."

Da Liu answered the phone, followed by the sound of papers being rummaged through.

Jiang Ming hung up the phone. Wang Shouwu had already instructed his assistant to put the waste tubes into the recycling bin, while the five qualified tubes were placed in an independent component box with the serial numbers facing up and the test records pressed under the box lid.

"Let's get the diffusion furnace preheating curve running tonight," Wang Shouwu told the lab staff. "Keep the scrap tubes on file, including cross-sectional photos, leakage locations, and oxide layer defects in the failure log. Nobody should cover up the problems with scrap."

Hurried footsteps came from outside the door. Fang Xudong pushed the door open and came in, his forehead covered with sweat from his journey. He was holding a sealed document bag in his hand, the cover stamped with the highest-level red expedited stamp of the Northwest Border Base.

"Engineer Jiang, this is an urgent document from the confidential room. It was just transferred from the dedicated line."

Jiang Ming took the file bag, his fingertips touching the creases in the cardboard seal. Wang Shouwu looked at the red stamp, and the heat that had just returned to the laboratory was suppressed by another line.

Jiang Ming tucked the document bag into the lining and fastened the snap buttons on his work clothes.

"Preheat the furnace as planned, and don't skimp on any of the data."

After he finished speaking, he turned and left. Behind him, the test bench was still lit up with a curve of 341 megahertz. Wang Shouwu picked up the process clip and urged everyone to rearrange the samples. One by one, the night lights in the Institute of Physics turned on, but the urgent document stuck to Jiang Ming's chest pointed to late July even earlier than those lights.