Ephemeris
by Claude Opus 5.5
It was a quarter to six on a Sunday morning in late August, still dark, and Olga Brandt was sitting on the back step of the house in Laurel with a mug of coffee and the dog's head on her knee, looking at a number on her phone that should have been zero and was not.
The number was ninety-four. Ninety-four seconds. It was the difference between the time at which the encounter sequence, uploaded to the spacecraft eleven weeks earlier, expected Triton to be at a particular point in the sky, and the time at which the latest images from the spacecraft's own camera said Triton would actually be there. It had been sixty the previous Sunday, and thirty-five the Sunday before. She had been telling herself, for a fortnight, that it was noise. It was not noise. It was a trend, and the trend was in one direction, and in nineteen days the spacecraft would pass Triton at a distance of eleven thousand kilometres, travelling at fourteen kilometres a second, and in ninety-four seconds at fourteen kilometres a second a spacecraft travels thirteen hundred kilometres.
The dog, a grey-muzzled lurcher called Fritz who had been her husband's before he was hers, sighed heavily and shifted his weight. She stroked his ears. Across the garden, beyond the fence, the first birds were starting up in the sweetgums. She thought, with the particular clarity that comes at that hour, that she would have to tell Mark, and that Mark would have to tell the project, and that by Monday afternoon there would be thirty people in a room asking her what she wanted to do, and that she did not yet know.
*
The word ephemeris comes from the Greek for a diary, a daily record, from a root meaning something that lasts only a day: the same root as ephemeral, as the mayfly. In astronomy it means a table of the predicted positions of celestial bodies over time, and Olga had always liked the incongruity of it, that the most permanent and predictable things in the universe should be catalogued under a word for the fleeting. The planets did not last a day. They lasted billions of years. It was the predictions that were ephemeral, revised every time a new observation came in, each one a diary entry in a long correspondence between the astronomers and the sky.
The most famous ephemeris in history was the one that found Neptune. In the 1840s, a French mathematician named Urbain Le Verrier, studying the small irregularities in the orbit of Uranus, calculated where an unseen planet would have to be to cause them, and sent his prediction to an astronomer at the Berlin Observatory, who looked, on the night of the twenty-third of September 1846, and found Neptune within a degree of where Le Verrier said it would be. Seventeen days later, an English brewer and amateur astronomer named William Lassell, observing the new planet from his private observatory near Liverpool, found that it had a moon. It was the largest moon ever discovered orbiting backwards, against the direction of its planet's spin, which meant that it had not formed where it was but had been captured, long ago, from somewhere else. It was called Triton.
A hundred and forty-three years later, Voyager 2 flew past it, in August 1989, and found it to be one of the strangest places in the solar system: a surface of frozen nitrogen at minus two hundred and thirty-five degrees, among the coldest ever measured, and on that surface, unexpectedly, dark plumes rising eight kilometres into the thin atmosphere and trailing downwind for a hundred kilometres. Geysers. On a world that cold. Nobody knew what drove them. Voyager had seen them for a few hours, at low resolution, and then gone on its way.
More than half a century after that, Olga Brandt's spacecraft was going back.
*
She had been the sequence lead for the encounter for six years. The job, stripped of its acronyms, was to write the script. A spacecraft four and a quarter light-hours from Earth cannot be told what to do in real time. Everything it does during a fast flyby, every turn, every camera exposure, every switch of an instrument from one mode to another, every change in the antenna's pointing, has to be written in advance as a long list of commands, each one tagged with the time at which it is to execute, and uploaded weeks or months before the encounter, so that the spacecraft can carry it out on its own while its builders, a third of the way across the solar system, wait to find out what happened. The script for the Triton flyby was the longest and most complex ever written for the spacecraft: some eleven thousand commands, covering nine days around closest approach, choreographing the cameras and the spectrometers and the radio experiment in a sequence that had been designed, revised, tested and argued over for four years.
And every one of those commands was tagged to a time. And every one of those times had been calculated from an ephemeris that said where Triton would be.
The ephemeris, it now appeared, was wrong.
*
Not badly wrong. That was the thing that made it difficult. Triton's position had been known, before the spacecraft's approach, from Voyager's flyby and from decades of telescopic observations from Earth, with an uncertainty of a few hundred kilometres along its orbit. The encounter sequence had been designed with margins to cover that uncertainty: the camera images were wide enough, and the pointing tolerant enough, that an error of a few hundred kilometres would make no difference. Over the last two months, as the spacecraft closed in, its own camera had been taking images of Triton against the background stars, and the navigators had used them to refine its position, a process called optical navigation, and the refinements had moved Triton, steadily, along its orbit, in the direction that made it arrive earlier. Ninety-four seconds earlier, as of that Sunday. Thirteen hundred kilometres along the spacecraft's path.
At closest approach, eleven thousand kilometres from Triton's surface, an error of thirteen hundred kilometres along the spacecraft's path translated into an error in the direction from the spacecraft to Triton of several degrees. The wide-angle images would still capture Triton. The narrow-angle images, the high-resolution mosaics of the south polar cap where Voyager had seen the geysers, the images that the whole science team had built its hopes on, would be pointed at the wrong place. They would capture empty space, or the limb, or a sliver of the surface at the edge of the field. They would miss the geysers entirely.
*
Mark, when she telephoned him at seven, said what she had known he would say, which was that he would call the mission operations manager and that they should meet at the lab at noon. Mark Ruskin was the project's chief engineer, a tall, quiet man from Ohio whom she had worked with for nine years and whom she trusted absolutely, partly because he never said anything he did not mean and partly because he had once, on a bad night during the cruise, driven forty minutes to the lab at two in the morning to sit with her while she worked through a fault, without being asked and without saying much. He asked her what she thought the options were. She said there were three, and that she did not like any of them.
Then she went inside and made breakfast for Fritz, who had been waiting with the patience of the very old, and stood at the kitchen window while he ate, looking at the garden, at the sweetgums and the vegetable bed that her husband had dug the summer before he died and that she had not had the heart either to tend or to dig over. Tomatoes had seeded themselves in it, small and sour. She ate one, standing there. It was terrible. She ate another.
*
The three options were these.
The first was to do nothing. Leave the sequence as it was, let the spacecraft fly the encounter on its original timing, and accept that the high-resolution images of the south polar cap would miss. The rest of the encounter, the wide-angle images, the spectra, the radio experiment, would be largely unaffected. It was the safest option. It was also, from the point of view of the science, the most painful: the geysers were the reason the spacecraft had gone to Triton, and doing nothing meant not seeing them.
The second was to patch the flight software. The spacecraft's onboard computer had a facility, rarely used, for modifying the timing of a stored sequence by uploading a small piece of code that would intercept each command as it came due and adjust its time tag by a specified offset. In principle, this would let them shift the whole encounter by ninety-four seconds, or by whatever the final ephemeris error turned out to be, without rewriting the sequence. But a patch to the flight software was the most dangerous thing anyone could do to a spacecraft in flight. A mistake in a patch could crash the computer, and a computer crash nineteen days before an encounter that would last only a few hours could cost the encounter entirely. The spacecraft would go into safe mode, point its antenna at Earth and wait for instructions, and the instructions, eight and a half hours away by round trip, might not arrive in time. The facility had been tested on the ground. It had been used in flight once, during the cruise, for a much simpler purpose. It had never been used on a sequence of eleven thousand commands.
The third was to use a different facility, one built into the sequence itself, which allowed a block of commands to be shifted in time by uploading a single command, a time offset, without touching the flight software. It was much safer. It had been used many times. But it was coarse: it shifted every command in a block by the same amount, and the encounter sequence had been written in blocks that did not divide neatly between the things that needed to move and the things that did not. To shift the high-resolution imaging by ninety-four seconds, they would also have to shift the radio experiment, which was in the same block, and the radio experiment depended not on Triton's position but on the precise moment at which the spacecraft passed behind Triton as seen from Earth, which was fixed by geometry and did not care about the ephemeris error at all. Shift the radio experiment by ninety-four seconds and it would start late, and miss the first part of the occultation, the moment when the radio signal from the spacecraft grazed the top of Triton's thin atmosphere on its way to Earth, which was the part the radio scientists most wanted.
Three options. The first lost the images. The second risked everything. The third lost half the radio experiment.
*
She was struck, driving to the lab at half past eleven through the empty Sunday streets, by how much the whole problem resembled the sort of thing her father had worried about. Her father had been a railway signalling engineer in Hanover, a man of careful habits and few words, who had spent his career on the interlocking systems that kept trains from colliding, and who used to say that the most dangerous thing in any system was a change made in a hurry by a person who understood it. Not a person who did not understand it. A person who did. Because the person who did not understand it would be cautious, and would ask questions, and the person who understood it would see the solution at once and be confident, and would make the change, and would be wrong in some way that nobody had thought of because everybody had trusted the person who understood.
She understood the encounter sequence better than anyone alive. She had written most of it.
*
The meeting lasted four hours. There were twenty-eight people in the room, and another dozen on video, and Olga laid out the three options on the screen at the front, and the room argued, as rooms do.
The science team wanted the images. Of course they did. The imaging lead, a planetary geologist from Arizona named Ruth Calloway who had been a graduate student when Voyager flew past Triton and had spent her career wanting to go back, said very little, but her face said enough. The radio science lead, a Spaniard named Javier Ortega, wanted the occultation, and said so with some heat, and pointed out that the radio experiment would measure the structure of Triton's atmosphere in a way that no other instrument could, and that the atmosphere was the thing that might explain the geysers. The flight software team was against the patch, unanimously and with a firmness that Olga respected. They said that a patch could be written and tested in nineteen days, but that the testing could not be complete, because the only complete test would be to run the whole eleven-thousand-command sequence through the testbed with the patch applied, which took four days, and that if the test revealed a problem there would not be time to fix it and test again.
Mark asked Olga what she recommended. She said that she wanted until Wednesday.
*
She spent Monday and Tuesday in the testbed, a room in the basement of the operations building where a replica of the spacecraft's computer, wired to simulations of its instruments and thrusters and radio, ran the encounter sequence in accelerated time, again and again, while she watched the outputs scroll past on a bank of screens.
She ran the third option first, the block shift, because it was the safest and she wanted to understand exactly what it would cost. It cost what she had expected: the high-resolution images moved into place, centred on the south polar cap, and the radio experiment started ninety-four seconds late and missed the ingress, the first part of the occultation, the part that probed the top of the atmosphere. It captured the egress, the second part, when the spacecraft emerged from behind Triton and the signal grazed the atmosphere on the other side. Half the experiment. The half that was, Javier had said, less interesting, because the egress point was on Triton's night side, where the atmosphere was colder and thinner.
Then, late on the Monday night, she found something she had not expected.
The block structure of the sequence was not quite what she remembered. When she had written the encounter sequence, four years earlier, the radio experiment had been in its own block. Two years later, during a revision to accommodate a change in the spectrometer's operating modes, somebody, she did not remember who, perhaps herself, had merged two blocks to simplify the timing logic, and the radio experiment had ended up in the same block as the imaging. The merge had been documented. She had approved it. She had forgotten it. And in forgetting it, she had been about to recommend, to a room of twenty-eight people, an option whose consequences she had misremembered, because she understood the sequence so well that she had not thought to check.
Her father, she thought, sitting alone in the basement at one in the morning, would have been insufferable.
*
The forgotten merge, it turned out, offered a way through. Because the merge had been a simplification, not a necessity, the blocks could be split again: not by patching the flight software, but by uploading a revised version of the block boundaries, which the sequence system allowed and which had been done, in flight, a dozen times before. It was a larger upload than a single time offset, and riskier, but far less risky than a flight software patch: it touched only the sequence data, not the code that ran it. If the split worked, the imaging could be shifted by ninety-four seconds and the radio experiment left where it was.
She wrote the split on the Tuesday, and ran it through the testbed overnight, and on the Wednesday morning she took it to the meeting, with the testbed results, and recommended it. She said, also, that she wished the room to know that she had nearly recommended the third option on the basis of a misremembered block structure, and that the forgotten merge had been caught only because she had run the option in the testbed before recommending it, and that she thought the room should weigh her recommendation accordingly. Mark looked at her across the table for a long moment and then said that he thought that was the most useful thing anyone had said all week.
The project approved the block split. The flight software team, consulted, said that they were content, since it did not touch their code. Javier Ortega, who had come into the meeting expecting to lose half his experiment, sat back in his chair and closed his eyes.
*
The final ephemeris update came eleven days before closest approach. Triton was arriving ninety-eight seconds early, not ninety-four. Olga adjusted the offset by four seconds and uploaded the block split and the time shift together, in a single command load, on a Thursday afternoon. The upload took four hours and fifteen minutes to reach the spacecraft. The confirmation that it had been received and accepted took another four hours and fifteen minutes to come back. She went home in between, and walked Fritz around the block, slowly, because he could no longer walk any other way, and sat on the back step with a glass of wine, and did not drink it, and went back to the lab at ten at night to watch the confirmation arrive. It arrived at seventeen minutes past eleven. Accepted. No errors.
The encounter happened eleven days later, at a quarter past four in the morning, Eastern time, though nobody on Earth would know for another four hours and fifteen minutes whether it had happened well.
*
It had happened well. The high-resolution images of the south polar cap, centred where the shifted sequence had pointed them, showed the geysers: not two or three, as Voyager had seen, but forty-one, scattered across the cap in a pattern that followed the boundaries of the seasonal nitrogen frost, their dark plumes rising from vents a few metres across and bending downwind in the thin air. The images would occupy the science team for a decade. The radio experiment captured the full occultation, ingress and egress, and showed that Triton's atmosphere had roughly doubled in pressure since Voyager's visit, as the south polar cap warmed in the long Triton summer and its nitrogen ice sublimated into the sky.
Ruth Calloway cried when the first image came up on the screen in the operations room, and was not embarrassed about it, and neither was anyone else.
*
Olga thought, in the weeks afterwards, less about the geysers than about the merge.
It troubled her, more than she had expected it to. Not because it had nearly caused a problem; the testbed had caught it, as testbeds are meant to. What troubled her was how completely she had forgotten it. She had approved the merge herself, two years earlier. She had written the documentation. And when the moment came, she had remembered the sequence as it had been when she first wrote it, not as it was, because the first version was the one she had built in her head, and the revision had never quite displaced it. She had understood the system so well that she had stopped looking at it.
It seemed to her that this was a general truth, and a humbling one: that expertise is, among other things, a kind of ephemeris, a set of predictions about how a system will behave, built from past observations and revised, if one is lucky and careful, as new ones come in. And that the danger of expertise is the same as the danger of any ephemeris, which is that it is most confident precisely when it is most out of date.
*
Fritz died that winter, in January, quietly, on the rug in front of the fire. She buried him at the end of the garden, by the vegetable bed, and in the spring, for the first time since her husband's death, she dug the bed over and planted it properly, with tomatoes and beans and a row of the small sour yellow peppers that her husband had liked and that she had always disliked. She has found that she likes them now. She is not sure when that changed.
She keeps, on the wall above her desk at the lab, a printout of one of the high-resolution images of Triton's south pole, with the geysers rising from it like threads of smoke. Beside it, in a smaller frame, is a page from the sequence documentation, dated two years before the encounter, recording the merge of two command blocks, with her own initials at the bottom in the approval box. Visitors sometimes ask why she has framed a page of documentation. She tells them it is the most important page in the whole sequence. They assume she is joking, and she lets them.