Chapter 3: What Okafor Knows

The drive room sat aft of the reactor bulkhead, separated from the rest of the Ardent by two meters of shielded composite and a pressure door that never quite sealed flush. Okafor had meant to fix that seal for two years. She never had. The small leak of warm air it admitted — not warm by accident, but warm because the drive always ran warm — served as her first indication, every time she entered, of how the engine was doing.

Tonight the air was right. She noted it without thinking.

The Ardent's torch drive was a Series-7 chemical thruster with a deuterium cell array, rated to 0.8g sustained acceleration and now twelve years old. It was Okafor's project in the way that surgeons develop a precise familiarity with a particular patient's physiology over years of corrective procedures: intimate knowledge built without ownership. She knew which coupling ran three degrees high under load, which flow sensor had a 0.4-second lag that the system compensated for automatically, which particular harmonic the drive chamber produced when it hit sustained 0.3g and held there for more than six hours.

She had heard something at 21:40 during the standard evening status review. Not a sound — the Ardent ran near-silent at operational acceleration, drive noise absorbed by the composite hull — but a deviation in the sensor feed that lasted 0.8 seconds before the system smoothed it over. She had written down the timestamp. She had continued her review. She had gone to her quarters at 23:00, waited until 01:30 to confirm she would not sleep, and come here at 03:00 to run diagnostics that were not due for twelve hours.

She set her kit on the inspection deck and plugged her handheld unit into the secondary access port. The standard diagnostic interface glowed green across twelve subsystems. Flow rate nominal. Combustion chamber pressure nominal. Cell array temperature 312 Kelvin, within spec. The monitoring suite had nothing to tell her.

She trusted none of it.

The handheld unit ran a different protocol — not the system's own internal checks, but an external measurement layer that read the sensors directly and compared raw output to the processed values the monitoring suite reported. Okafor had written the comparison algorithm herself, seven years ago, after a tertiary coolant failure on the Brecht that the monitoring suite had flagged informational for eleven days before the coolant stopped flowing entirely. She kept the handheld calibrated to within 0.1 percent accuracy.

She started at the deuterium feed coupling, which was where the 21:40 anomaly had originated, by her read of the sensor geometry.

The coupling was a 20-centimeter connector assembly that regulated deuterium flow from the cell array into the primary combustion chamber. It was not a high-drama component. It ran at 440 Kelvin under normal operational load, with a tolerance ceiling of 520 Kelvin before thermal stress began accumulating in the wall material. The drive's thermal management kept it comfortably inside that range. The coupling was rated to 150,000 hours. The Ardent's drive had 89,000 hours on it. There was no reason to look there first.

She looked there first.

She ran the temperature differential across the coupling's twelve sensor nodes. Eleven read within 0.5 Kelvin of each other — normal variation. Node 7 read 4.2 Kelvin higher than its neighbors. The monitoring suite would have logged that as noise and averaged it out. Her handheld logged it as an observation and held it.

She took a physical reading with the contact probe, pressing it against the coupling housing directly above node 7's position. The probe read 448 Kelvin. The coupling's own sensor read 444 Kelvin.

Four Kelvin. Consistent. Not random variation.

She found the fracture eight minutes later with the acoustic probe — a hairline stress fracture in the coupling wall, 3 centimeters long, running longitudinal to the flow axis. At rest it was invisible to the eye. The acoustic probe found it by detecting the micro-vibration signature of a material boundary where the material should have been continuous. In the standard diagnostic, this registered as background structural vibration within acceptable range. In her handheld, it read as a distinct frequency signature at the exact coordinates of the anomaly she had noted at 21:40.

She photographed it, logged the acoustic data, and ran the thermal model.

The fracture was not currently leaking. The coupling wall remained structurally intact. But the fracture changed the thermal distribution — created a hotspot, reduced the coupling's effective tolerance ceiling from 520 Kelvin to approximately 490 Kelvin, depending on progression rate. She estimated progression rate based on the fracture geometry and the drive's operating history: slow, likely 0.1 to 0.3 millimeters per hundred operating hours at normal load.

At normal operational acceleration — 0.3g, the Ardent's standard cruise — the coupling temperature held at 448 Kelvin. Twelve Kelvin of margin below her revised tolerance ceiling.

She built the timeline on a piece of physical paper. Old habit. She preferred paper for calculations she did not want sitting in the system log.

Mission profile: 60-hour hold in the Corridor, then return burn. Total drive hours at operational load: approximately 80, accounting for transit already completed and the return. The fracture, at estimated progression rate, would extend to 3.2 to 3.5 centimeters by end of mission. Thermal hotspot would increase by 2 to 4 Kelvin. She wrote the number: 452 Kelvin maximum, maybe 452.5. Still inside the revised tolerance. Barely.

She drew a second line on the paper. The mission profile assumed 0.3g cruise. If command ordered a hard burn — anything above 0.5g sustained — the coupling temperature would climb toward 480, 490 Kelvin. She ran the math on fracture propagation under elevated thermal load. The fracture would accelerate. The numbers became imprecise. She wrote: uncertain above 0.5g sustained.

To repair the coupling properly, she needed four hours of cold drive. No acceleration. Mid-mission, that meant the Ardent sitting ballistic for four hours before resuming thrust. She knew what the mission was. She knew what a four-hour delay meant to a 60-hour hold window.

She also knew what she thought of the mission. She had known since the briefing, when she had listened to the objective and the stated intelligence and the confidence with which command described a situation that was, by her read, not fully understood by anyone who had briefed command. She had not said so. That was not her job. Her job was the drive.

She sat with the paper for two minutes.

The numbers worked. Sixty hours at 0.3g, return burn, the coupling would hold. That was the mission profile. If they stayed inside it, the coupling would hold.

She folded the paper and put it in her breast pocket.

She disconnected the handheld unit, closed the access port, and picked up her kit. At her terminal she opened the diagnostic log and filed a report: drive subsystem check completed 03:47 ship time, all systems nominal. Deuterium feed coupling node 7 thermal signature noted for monitoring. She appended the handheld data to the report and categorized it under maintenance flag priority: informational.

The system accepted the log. The maintenance flag appeared in the secondary queue, where informational flags waited until someone decided to schedule them. Standard scheduling intervals ran weekly. No one would look at it before the mission was complete.

She closed the terminal and walked back through the pressure door into the drive room.

She stood in the warm air and listened to the drive running at 0.3g, steady, the way it had been running for the past eighteen hours. The coupling was behind the primary housing, out of sight. She could not hear it directly — the coupling itself made no audible noise — but she could hear its position in the harmonic structure of the overall drive sound. The resonance pattern of the combustion chamber shifted by a small interval when the coupling's thermal balance changed.

It sounded the same as it had at 21:40.

She stood there for a while, listening.