The Weighting Scheme

09 FEB 06:12, St. Catherine's ICU, Vane, H., stroke 03:40, stable.

“Stable” was the word. Helen Vane died at 08:12, and I’ve been repeating the minute ever since, trying to make it mean something. Seventy-four. Stroke in February, dead by eight in the morning, in a hospital room with the television on. That’s a better ending than her field usually hands out, and I tell myself that too, as if saying it twice makes either part a measurement.

I drove to the institute. Not the hospital, where they were doing whatever you do with a person who is already gone. Not home. Work. Fifteen years and I have not once not gone to work, and a dead teacher was not going to be the day I started. That’s the sentence I said to myself.

Every distance in the observable universe hangs from a number nobody has ever seen directly: the absolute brightness of a Cepheid variable, the zero point of the period luminosity relation. Set it one way and the cosmos is 13.6 billion years old and expanding at 73 kilometers per second per megaparsec. Set it the other way and it’s older, calmer, 68. The two versions disagree by five sigma, which is a crisis, and the crisis is the most fundable problem in astronomy, and my salary is the denomination it pays in. The whole field has chosen, is the thing. Helen Vane was the anchor of the side I work for.

I was the housekeeper. It’s not a title, it’s a description. The calibration pipeline, the quality reports, the log: Helen wrote the astronomy and I kept the books, and for nine years the books listed a weighting scheme that I certified, quarterly, by hand, and I never once asked what it was for, because housekeepers don’t. Her machine was still on when I got there. Thirty-four days of uptime, and the last login was February 8th, the night before the stroke, and her password was 2013, the year she’d set the number. I sat in her chair, because mine was in my office and it felt wrong in a way I filed under grief, and I opened her last session.

The file was called run 5.1.3 and it had been modified at 03:47 on the third of February. I knew it. I’d helped her queue it in November, when the new sample came down, forty-one Cepheids with parallaxes good enough to bounce a career off. She called it housekeeping. Helen never, in her life, called anything housekeeping, and I didn’t notice. That’s the whole of my confession.

The title line said: Z FREE. CONVERGED. 69.8 ± 1.0.

I read it twice, standing up. The third time I sat down. This was not a number. It was an instruction, and it had been sitting in the machine for six days before the stroke, with the annotation beneath it. The only reason it was mine to find is that she knew I was the only person alive who could read it and know what it meant.

I didn’t read the annotation. Not that day. I closed the session, logged out, and drove home, and for five weeks I did not open the file again.

My sister called at six. I let it ring. She’ll call again. She always calls again.

13 FEB, her house.

The department asked me to collect her things. There was no family to ask, or there was and they didn’t want the job, and it fell to me.

The house was a rental in the hills, kept for the view and nothing else. The calendar by the phone had the March run circled in her hand. The coffee cup in the sink had been there a while. Above the desk, pinned to the wall, was the residual plot from 2016, the long period points circled in red. I took it down, rolled it, and put it in my bag. I don’t know what I meant to do with it. I still don’t.

28 FEB, no run.

I have not run anything since the ninth. The archive doesn’t care; the frames will keep. I certified the February quality report on the fifteenth because it was due, and I did not open run 5.1.3 again. The annotation is still below the fold. I tell myself I know what it says, because I know Helen, and because the title line already told me the only thing that matters. I have not opened it.

17 MAR, no run. Reading.

The annotation was below the fold. I read it three times, and the third time is the one that matters, so that’s the one I’ll tell you.

Z free. Converged. 69.8 ± 1.0. The long period parallaxes resolve the color term. They were resolving it in 2016 and I weighted them down, because the answer was 70 and there is no program, no grant, no building for 70. The zero point was bright by a tenth of a magnitude. That’s the whole of my life’s work, M.: a tenth of a magnitude, and the universe is half a billion years older and the crisis is a bookkeeping error. I knew it by 2016 and I kept it, because the field runs on the tension and I ran on the field, and by the time that sounded like a reason instead of an excuse I had a reputation that ran on it too.

Don’t publish this under my name. If it goes out, it goes out ugly, or not at all, and you’ll live with whichever you choose. I’ve lived with the alternative for twenty years. You can survive it. That’s the worst thing I know how to tell you: you can survive it.

H.

The third read was where I understood who M. was. She had written it knowing I would be the one. Not the department. Not Dana. Me. The housekeeper. Helen left the numbers and took the truth, and left me holding the place where the truth used to be. I sat in her chair. A confession is also a contract, and I had signed mine fifteen years ago, without reading it, like a credit card form.

She didn’t foresee the stroke. She foresaw the decision.

18 MAR, observatory night. NGC 4258. Seeing 0.8".

The mountain doesn’t read the obituaries. The night assistant meant it as a joke. One block of the program was left, Helen’s name on the proposal, three nights every spring on the galaxy whose distance is pure geometry, water masers in a disk around its black hole, Kepler plus trigonometry, no physics arguing with you. I’d reduced the same field every year for fifteen years. The frames came down like they always had.

Spiral arm, ten thousand stars, forty of them pulsing on schedules set by their own size, the ones Leavitt counted by hand a century ago on glass plates, before anyone knew why they pulsed. She just wrote down what she saw and let it be a law. Longer period, brighter star. Anyone can measure the brightness of a star. The only lie in the whole machine is that one star out of ten thousand is where it seems to be, and somebody has to decide who that star is.

The night assistant was a kid named Ortiz who had never met Helen and didn’t care, which was the best company I’d had all month. He made coffee at midnight, the kind that stands a spoon in, and I drank it because it was something to do with my hands. The dome creaks all night, every night, and I know the creaks now, which ones mean the wind and which ones mean the bearing, and I sat in the warm room and listened to them and reduced the frames.

The pipeline ran aperture photometry overnight. Circular apertures, sky annuli, the same nine steps, and at three in the morning I watched the light curves square up and I thought about 1952, when Baade revised the zero point and the universe doubled, twice as big, twice as old, and the field blinked, absorbed it, moved on. A discovery has no author. A confession does, and the author was still the problem, lying dead in the ground and still the problem.

I reduced the night with the 2013 anchor, because the program called for it and the schedule doesn’t care and the telescope doesn’t care and I am the housekeeper, which is to say the thing that does not care on behalf of the building. Output: 73.2 ± 1.2, like every other spring.

My sister called at noon. I said I’m fine. I’m fine. She said you don’t sound fine. I said the mountain air. She said it’s March. I said I know. We let it go; that’s what we do.

19 MAR 23:00, her notes.

Thursday night I went through the logs. Fifteen years of pipeline runs, and Helen’s hand in the margins of the first ones, learning the rooms of the house I’d learn after her.

2013, the anchor year. Gone to press. 73.2. The maser holds; the nine parallaxes hold. Let the CMB camp call it dust. They’re welcome to try. Bold, and true, and I remember that Helen, because she was worth believing, and you can’t believe a person you haven’t met, so I never met the later one.

2016, scrawled beside a residual plot in a hand that had already started to get old: Z correlated with period again. Short periods: 73. The long ones want 70. Sample too small to trust. Weight them down.

And below, in the same ink, smaller, so I almost missed it: the long ones are not going away.

Then nothing. I checked the dates. Nine years of clean runs, not a single mark in the margin. Nine years of quarterly certifications by me, nine years of the weights, nine years of a correlation she had seen and folded away like a napkin. And the worst part is that the machinery was perfect. Helen kept the books clean. I kept them cleaner.

The choice wasn’t a door, it was a ledger. One line, then the page, then the year. She built the economy that paid her, and then she couldn’t stop it. Two and a half sigma is a footnote. Five is a crisis, and a crisis is a budget line.

She taught me the calibration. She forgot to teach me that the calibration was the payroll.

20 MAR, draft. Not sent.

Friday the acting director emailed. The white paper, he said, the one arguing the CMB side has systematics, the institute’s marquee result, needed the anchor line finalized, and with Helen gone, would I certify. We’re calling it the Vane calibration, he wrote. In memoriam.

I opened a new document at two in the afternoon, and I wrote:

We present a free parameter refit of the Cepheid zero point on the new parallax sample. The zero point converges to 69.8 ± 1.0, 0.10 magnitudes fainter than the 2013 anchor. The residuals of the 2013 fit are period correlated. We conclude the anchor was in error and the local determination of H₀ requires revision.

That’s all I could write, because there was no we. There was a dead woman and a housekeeper, and either name on the byline was an execution order. Hers, posthumous. Mine, the kind where I live and explain myself until the grant runs out, and the grant names the anchor in the first line of the abstract, in the first sentence.

Nobody funds a solved problem. The crisis is worth more than the truth. I know, because I’ve been paid in it for nine years. I saved the draft as unsent. Then I opened the director’s email, attached the draft, and sat with the cursor over send until the light outside went gray. The pothos on the windowsill was drooping, and I thought, water it tomorrow, and I closed the email. The draft stayed where I put it.

21 MAR, memorial, 11:00.

The chapel was full, which surprised me, and shouldn’t have. Twenty years of students. The program had a photograph from 2013, the anchor year, before the hand got old. I folded it into my coat. The director read the obituary: the Vane zero point, the anchor of the local ladder, 73.2, the crisis that defined a decade, the hundred careers built on it, mine among them. Spoken aloud in the register of a eulogy, which is how the treaty gets read, every time, to a full chapel.

Dana Reyes found me afterward. Dana went over to the CMB side fifteen years ago, Helen’s student, defected, and the two of them kept a fifteen-year civil war because the argument was real once, and then it was just their shared calendar. She hugged me and I let her.

What was she working on? she said.

The zero point, I said. A recheck, on the new parallaxes.

And?

Nothing conclusive, I said. I said it clean. It was the first time all week I’d lied without being asked to.

Dana said: she was having a fight with her own numbers those last months. I’d never seen her do that. I hope she won it.

She won it, I said.

Dana nodded, moved on to the next person, and the sentence went into the room and stayed there.

My sister called at six. It was a good service, I said. She was loved. I hung up and went back to the office and looked at the drooping pothos and thought, water it tomorrow.

22 MAR.

Sunday I called my sister before she could call me. She said: what are you going to do. I said nothing. She said you always say nothing. Remember Mom. I said I remember. She said you made that funeral fine. You make everything fine. That’s what you do. I said I know.

Then I certified the anchor. Confirmed: 73.2 ± 1.2. The 2013 Vane calibration is correct as stated and certified for the white paper. I typed it, hit send, and the machine chimed, and nothing in the building noticed.

Late, I reduced Wednesday’s frames with the 2013 anchor, because the archive needs the correct calibrations or the person after me has to redo it, and I know who the person after me is now. Output: 73.2 ± 1.2.

I typed the log entry, and the operator line came out VANE, and I deleted it, and I typed DELGADO, and I hit enter, and that is the entire record of the decision. The draft went into the drawer. The drawer is not a drawer. It’s a directory on a machine with one password, and the password is 2013, and the file in it is true, and I am the only person alive who can read it, which is the same sentence it’s always been, with only the operator changed.

I watered the plant. It had grown a new leaf sometime in the week.

The certification I sent this morning has my name on it, and it’s a lie, and the file in the directory is the truth, and both of them are mine now.