The Scheduler Clock
The scheduler clock on the wall said 03:42, and Lale Demir had the 0.72-micron edge pulled up on her screen for the eleventh time that night.
“Run it again,” Nuri Yilmaz said from the next chair without looking up.
“It does not change when I run it again, Nuri. The edge is there.”
“It is there in your reduction. That is not the same as there.”
The Paranal control room was half dark, the way it always was at this hour, the only light the blue-white glow of the four consoles and the red numbers of the scheduler clock. The Lazuli coronagraph’s first-light list sat on the big monitor in the center, ranked by the optimizer, a column of numbers that moved when you breathed on them. Lale’s target, the rocky planet twelve light-years out, was number fourteen. The cutoff was thirteen.
“Nuri. The 0.72 edge is real. I found it in my own reduction, and I have been looking at it for two years.”
“And in two years you have not found a mineral that makes it.”
“No mineral in the BRDF library makes it. That is the point.”
Nuri turned his chair. He was a small man with a careful face, the kind of face that had spent twenty years learning not to show what it was thinking. He was the PI. The coronagraph was his project, his name on the first-light paper, his reputation in the committee’s hands.
“Lale. The lab’s reflectance measurements are incomplete. We are missing the feldspar series above sixty degrees phase angle, and the entire olivine solid-solution set for wet surfaces. An unexplained edge in a spectrum where the BRDF library has holes is not a discovery. It is a calibration artifact wearing a discovery’s coat.”
“So you cut it.”
“I cut it. The optimizer gives the slot to a known hot Jupiter with a clean spectrum and a published ephemeris. Signal-to-noise four point two versus your one point nine. The committee will ask why I gave first light to a target I cannot explain, and I will not have an answer.”
“The committee will ask why you did not image the only rocky planet in the list.”
“The committee will ask why I imaged a calibration artifact and put my name on it.”
The scheduler clock ticked. 03:43.
Lale had found the edge in 2024, in a reduction she had run herself on a Sunday because the pipeline had flagged an anomaly and nobody else had wanted to look at it. A shallow dip at 0.72 microns in the reflected-light spectrum of a rocky planet at twelve light-years, a depth of about three percent, narrow, repeatable across three epochs. She had gone through the BRDF library mineral by mineral. Basalt, no. Olivine, no. The pyroxenes, no. The feldspars, no, and the feldspars were the problem because the lab had never measured them above sixty degrees phase angle, and the planet’s orbital inclination meant the observations sat at seventy-one.
“Let me add the calibration,” she said.
“You have been adding calibrations for two years.”
“One more. The wet-surface olivine set. If I can close that gap, the edge either matches a mineral or it does not.”
“And while you close the gap, the optimizer re-ranks the list and your target drops to fifteen, because the hot Jupiter has a brighter reflected signal and a known ephemeris and the optimizer does not care about your edge.”
“Then tell the optimizer to hold the slot.”
“The optimizer does not take instructions from sentiment, Lale. It takes signal-to-noise, ephemeris confidence, and target visibility. Your planet has one point nine. The hot Jupiter has four point two. The difference is not sentiment. It is a factor of two in integration time.”
The clock said 03:44. Somewhere outside the window, the desert was dark and cold and very large, and the telescopes on the ridge were doing their work without opinion.
Round one was the cut, and Lale went home at dawn and slept badly and came back the next night and ran the olivine calibration herself, because the lab technician who was supposed to do it had quit and the position was frozen.
The lab’s spectrometer was a benchtop instrument in a basement room that smelled of solvent and old coffee. Lale had never liked it. The phase-angle goniometer was a rotating stage with a manual crank, and the sample holder had a wobble she had reported twice and nobody had fixed. She set up the wet-surface olivine samples, three grains sizes, two hydration states, and ran them from zero to seventy-five degrees in five-degree steps.
It took her four nights. She did not sleep well during any of them. On the third night she called Nuri at 02:00 and said, “The wet olivine set does not match. The edge is not olivine.”
“I know,” Nuri said. “I read your preliminary. The edge is not olivine. The edge is not anything in the library. That is not evidence that it is real. It is evidence that the library is incomplete.”
“The library is incomplete in a specific direction. Feldspar above sixty degrees. If I close that—”
“Then the optimizer re-ranks and your target drops. I have told you this.”
“Then tell the committee.”
“Tell the committee what? That you have an unexplained edge and the lab cannot reproduce it?”
“That the lab cannot reproduce it because the lab has not measured the right angles.”
Nuri was quiet for a moment. “Lale. I am not your enemy.”
“I know you are not my enemy. I am asking you to be my colleague.”
The line went quiet. Then, “Come to the control room tomorrow. We will look at the optimizer together.”
Round two was the 04:00 argument over a 0.02-magnitude difference.
The optimizer’s re-ranking logic was simple and stated in its documentation, which Lale had read three times. For each target it computed a figure of merit: signal-to-noise ratio divided by the square root of integration time, multiplied by a priority weight that the PI set. The priority weight for known hot Jupiters was 1.0. For rocky planets with unconfirmed spectral features, it was 0.8. Nuri had set that weight himself, the week before, and had not told her.
Lale found it at 03:50 on a Tuesday.
“You weighted it,” she said.
“The priority weight is a standard parameter.”
“You weighted rocky planets with unconfirmed features at 0.8. That is a twenty percent penalty. That is the difference between rank thirteen and rank fourteen.”
Nuri looked at the screen. “The penalty reflects the uncertainty in the target. The uncertainty is real. The BRDF gap is real.”
“The BRDF gap is your reason for cutting it, and now it is also the reason for weighting it down. You are using the missing data twice.”
“I am using it once. The cut is a consequence of the weight. The weight is a consequence of the uncertainty.”
“The uncertainty is a consequence of the lab not measuring feldspar above sixty degrees.”
“And the lab has not measured feldspar above sixty degrees because the goniometer wobbles and the technician quit and the position is frozen. That is not my fault, Lale. That is the system.”
The scheduler clock said 03:58. The optimizer had re-ranked while they argued. The rocky planet was now rank fifteen. The hot Jupiter had moved to rank eleven.
“Look at the difference,” Lale said. “Zero point zero two magnitudes. That is the difference between your target and mine. Zero point zero two. The optimizer is splitting hairs on a system that cannot measure feldspar above sixty degrees.”
“The optimizer is doing what the optimizer does.”
“Then the optimizer is wrong.”
“The optimizer is consistent. Consistency is what first light requires.”
Lale stood up. Her chair rolled back and hit the wall. “Consistency is not the same as correctness, Nuri. You know that. You have published on this. In 2019 you wrote that target-selection optimization systematically underweights targets with incomplete calibration data, because the optimizer treats missing data as zero signal rather than as uncertainty. You wrote that.”
Nuri’s face did something complicated. “I wrote that as a caveat. Not as a policy.”
“You wrote it as a finding. And now you are the PI and the finding is a caveat and the caveat is a weight of 0.8 and the weight puts my planet at rank fifteen.”
“Lale.”
“Zero point zero two magnitudes, Nuri. At 04:00 in the morning. Over a number that would not survive peer review if the reviewer knew the BRDF gap.”
The clock said 04:00 exactly. Neither of them said anything for a moment. The optimizer ticked on the screen, re-ranking, re-ranking, its column of numbers indifferent to the hour.
“I will not change the weight,” Nuri said. “Not without a calibration that closes the gap. That is the condition. Close the gap, and I will remove the penalty.”
“The gap is feldspar above sixty degrees. The goniometer wobbles.”
“Then fix the goniometer.”
“With what money? The position is frozen.”
“With the money you get from your next contract, if you get one. Which you will not get if you publish a false positive under my name.”
Lale sat back down. She did not say anything for a while. The clock moved to 04:01, 04:02.
“Fine,” she said. “I will fix the goniometer.”
She fixed the goniometer with a shim cut from a beer can, which she did not tell Nuri about because he would have objected on grounds of scientific rigor and she would have agreed with him and they would have argued about it for another week. The shim held. She ran the feldspar series from sixty to seventy-five degrees over three nights, and the edge at 0.72 did not match anything in the expanded library.
She took the result to Nuri on a Thursday afternoon, not at 04:00, because she had learned that the hour mattered. Nuri looked at the data for a long time. Then he looked at the optimizer.
“The weight stays,” he said. “The edge is still unexplained. An unexplained edge after a more complete calibration is more suspicious, not less.”
“More suspicious? The calibration is more complete.”
“The calibration is more complete and the edge persists. That means either the edge is real or the library is still incomplete in another direction. I cannot distinguish those two. The optimizer cannot distinguish them. The committee will not be able to distinguish them.”
“Then let the committee decide.”
“The committee will decide what I put in front of them. I am putting the hot Jupiter in front of them.”
Lale went to the time-allocation committee on her own. She wrote a three-page request, which she rewrote four times because the first three versions were angry and the fourth was not. She asked for a single orbit of the coronagraph on the rocky planet, to be scheduled in the first-light window, on the grounds that the 0.72-micron edge was a candidate biomarker that could not be evaluated without imaging and that the BRDF gap had been substantially closed.
The committee met on a Monday. Nuri was there, because he was the PI and the committee required the PI’s presence for any target request. Lale presented her three pages. Nuri presented the optimizer’s ranking and the 0.8 priority weight and the remaining BRDF gap, which was now the olivine solid-solution set for wet surfaces, which Lale had measured, and the feldspar series above seventy-five degrees, which nobody had measured because the goniometer’s maximum angle was seventy-five.
The committee asked Lale whether the edge could be a calibration artifact.
“It could be,” she said. “I do not think it is. But I cannot prove it is not, because the BRDF library is incomplete in directions we have not measured.”
The committee asked Nuri whether the hot Jupiter’s spectrum was reliable.
Nuri said it was, that it had been taken in 2024 with the standard detector, that it was flat and featureless and therefore a clean first-light target, and that the ephemeris was published and the signal-to-noise was four point two.
The committee voted. The hot Jupiter kept the slot. Lale’s request was noted and filed.
She was off the project by the end of the month. Not fired, not formally, but her contract was not renewed, and the reason given was budget, and the budget reason was real enough that she could not argue it. She found a position at a university two hundred kilometers north, teaching, and she kept the edge in a folder on her laptop and opened it sometimes at night.
Three weeks after she left, she was running a literature search for a lecture on exoplanet spectroscopy when she found the paper. It was a 2024 observation of the hot Jupiter, the same target Nuri had given the slot to, taken with the same detector she had used in her own work. The paper reported a flat spectrum, featureless, which was why Nuri had chosen it. But the detector, a HgCdTe array on the VLT’s X-shooter, had been found in 2025 to have a flat-field error in the 0.7 to 0.9 micron range. The error was a multiplicative factor of about 1.04, smooth, slowly varying, and it flattened any shallow spectral feature in that range into the noise.
The hot Jupiter’s spectrum was not flat. It was flattened. The 2024 paper had reported a flat spectrum because the detector had erased the features, and Nuri had chosen it because it was flat, and it was flat because of an error that was known and published and that nobody on the Lazuli team had checked for, because the hot Jupiter was a known target and known targets were not checked.
Lale sat with this for two days. She thought about calling Nuri. She thought about writing to the committee. She thought about the fact that she had been right about the edge and wrong about nothing and that it did not matter, because the slot was gone and her contract was gone and the edge was still unexplained and the hot Jupiter’s spectrum was still flat in the literature.
She did not call Nuri. She wrote a short note to the committee, factual, three sentences, attaching the 2025 paper on the flat-field error. She did not ask for anything. She sent it and closed her laptop and went to bed.
Three months later, another team’s paper appeared on the arXiv. The title was “Specular Glint at 0.8 Microns from the Rocky Exoplanet at 12 Light-Years: Evidence for Surface Liquid.” The abstract reported a specular reflection feature at 0.8 microns, geometry-dependent, observed at high phase angle, consistent with a liquid surface on a rocky planet. The planet was the one Nuri had cut. The feature was a glint, not an absorption edge, which meant it was a reflection and not a mineral signature, which meant Lale’s 0.72-micron edge was a separate feature, a spectral edge that predicted a planet worth imaging, and the glint confirmed that the planet had a surface liquid.
The glint paper did not mention Lale. It did not mention Nuri. It cited the 2024 hot Jupiter spectrum as a calibration reference, the flat one, the one with the flat-field error, and used it as a baseline for the rocky planet’s contrast measurement, which meant the contrast was slightly wrong, but the glint was bright enough that it did not matter.
Lale read the paper at her desk, in her office, at the university, on a Tuesday afternoon. She read it twice. Then she closed the tab and looked at the scheduler clock on her wall, which was not the same scheduler clock but which told the same time, and she thought about the 0.02 magnitudes and the 0.8 priority weight and the beer-can shim and the fact that Nuri had said he was not her enemy and had meant it.
She did not call Nuri. She did not write to the authors. She opened her folder, the one with the edge in it, and she added the glint paper to it, and she closed the folder, and she went back to preparing her lecture.
The lecture was on exoplanet spectroscopy. She had added a slide about the 0.72-micron edge the week before, with a note in the margin that said “unexplained; possibly real.” She left the note. She did not change it to “confirmed.” The edge was still unexplained. The glint was confirmed. They were different features on the same planet, and one of them was hers and the other was not, and the distinction mattered in a way she could not explain to her students without sounding bitter, which she was not, exactly, but was close enough to it that she chose not to try.
She taught the lecture on Wednesday. A student asked whether the edge could be chlorophyll.
“It could be,” Lale said. “We do not know. The BRDF library is incomplete.”
“What would it take to know?”
“Measuring the minerals at the right angles. Which costs money, and time, and a position that is currently frozen.”
The student wrote that down. Lale watched her write it and thought that this was the part that would outlast the paper, the part where a student wrote down that the library was incomplete and that incompleteness had a cost, and that cost was measured in frozen positions and 0.02 magnitudes and a planet that turned out to have a surface liquid and whose finder had been off the project for three months when the glint was announced.
She finished the lecture. She packed her bag. She walked home. The sky was clear and very large, and she did not look up at it, because she had stopped looking up at it in the way she used to, and she was not sure whether that was loss or relief or just the ordinary consequence of being right about something that did not get you back what being right was supposed to get you.
She unlocked her door. She put her bag down. She opened the folder on her laptop and looked at the edge, the shallow dip at 0.72 microns, three percent deep, narrow, repeatable across three epochs, and she closed the folder, and she made dinner.