Steady As She Blows

Every second, somewhere in the universe, a star blows up. You can look that up. You won’t, but you could.

The ones we use are Type Ia. A white dwarf that eats too much and goes off like a bomb. Same bang every time, or nearly. That’s the whole trick of the trade. You see the candle, you know how bright it started, you read off how far away it is. Do it a few thousand times and you’ve mapped the universe.

Only there’s a catch. The catch has a name in the journals. Evolution.

A candle that burns at one brightness now might have burned at a different brightness a billion years ago. Everything changes. Everything ages. Why shouldn’t the candles?

If the candles change, the map changes. The dark energy changes. The history of everything changes. And a lot of reputations go with it. Mine, for one.

I’m the one who ran the test. The big one. Forty-seven thousand candles, ten years of data, looking back ten billion years. The question: does the brightness change with time? The answer: no. The evolution parameter, alpha, came out zero. Zero at every decimal we had. Error bars small enough to choke on.

That’s the answer everybody wanted. It keeps the cosmology standing. It keeps the grants coming. Me in my job.

It made me sick.

Because the residuals were too quiet. The scatter left over after the fit was smaller than the noise of the instruments that made it. And you cannot get that. That’s like shouting down a well and getting back a whisper. Real data lies and argues and has dirt on it: cosmic rays, dust, bad seeing, a janitor walking past the dome on a bad night. But this data was clean. Clean as a whistle. Somebody had washed it.

So I went looking for the wash job. That’s what you do. I re-reduced. I recalibrated. I ran a second pipeline over the raw files. I ran the worst night from the worst telescope all by itself. Clean. I ran the best night. Clean. I brought in the instrument people. They tested the chips, the clock, the whole rig. Fine. All of it fine. All of it ordinary.

That’s when I got the feeling. You know the feeling. You look at a residual plot and every point sits on the line. Not near the line. On it. Like the data was drawn instead of measured.

The statistics were the worst of it. There’s a thing called chi-square. Skip it if you like. The point is, real data comes out messier than the theory says, because the theory doesn’t know about the janitor. Mine came out five times better than the theory allowed. Five times. As if the noise knew its part and played it quiet, so as not to disturb anybody.

Nothing in the universe is that smooth. Nothing. Dust is lumpy. Light comes in chunks. The universe is a dice game, and the dice are loaded toward noise, not away from it. You cannot get a cleaner answer than the universe allows.

Unless the universe is holding the dice steady. On purpose.

I spent a month proving the data was real. That was the bad month, because it was real. Every photon and every frame. The numbers came out of the sky that way. Nobody washed them.

So I put the question to the data. I stopped asking how the candles looked. I asked who was making them look that way. And the data answered. It had been answering all along, and I’d been reading the answer wrong.

A null result doesn’t just say nothing is happening. It says nothing is happening, and I’m sure. The sureness is the tell. The quieter the residuals, the harder somebody was holding still. And somebody has been holding still for a long time.

I pulled the old plates. Harvard, 1890 to 1955. Glass and dust. The candles sat in them at the same brightness, to the fourth decimal. I pulled the old photographic magnitudes, error bars you could drive a truck through, and the candles sat inside them like a man holding very still in a parked car with the engine off.

All the way back. As far as the light can carry. The sky had been steady since the first eye looked up at it.

Now comes the part I don’t write down. I think it at three in the morning, which is the correct time for thoughts like this.

The measurement that proves nothing changes is itself a measurement of somebody trying not to be noticed.

I sat with that for a week. Then I wrote the paper. The scariest null in the history of the field. Forty pages. Introduction, method, results. And then, the way you do, I read the method one more time. Line by line. Looking for the wash job.

Found it.

Not in the data. In the corrections. The standardization. Stretch, color, the nuisance parameters. Ten years of them, mine and the people before me. We tuned them, iteration after iteration, until the residuals went quiet. Every tuning pass stole a little scatter from the sky and gave the null a little more muscle. Nobody cheated. Nobody fudged. It was the most honest work of my life.

That’s the joke. The wash job was diligence.

We fit the corrections together with the cosmology, all at once, the way everybody does. The corrections ate the evolution. They were built to. The test that proves nothing changes can’t see the thing that changes, because we taught it, carefully, over a hundred person-years, not to.

It goes back to 1929, when somebody had a candle that didn’t fit and invented a correction that made it fit. A little nudge. An honest little nudge, the kind you make when the alternative is a wobble in your own result. And every survey since has been calibrated through that nudge. Through all the nudges. Passing them down like a lit candle, each generation adding one.

The plates. I told myself they proved it, back to 1890. They proved nothing. Their errors were fat enough to hide a war. I read steadiness into them because I’d already decided the sky was steady.

The sky never held still. We held still. All of us, for a hundred years, so the answer would come out the way we needed it to. The universe didn’t try not to be noticed. We did.

I could fix it. One night of work. Hold the corrections at their local values and let the distant candles speak. If the fit goes sour, the sky has been lying all along. If it holds, we’re clean. Either way, I’d know.

I closed the file instead.

The residuals are still a lullaby. The fit is still perfect. It always will be, now. The test can’t fail. It was built, brick by honest brick, to pass.