Ninety-Five Percent, Now in Higher Resolution

And the thing it was built to look at does not emit light.

That is the joke, and it is a good one. Over a decade of engineering, delivered ahead of schedule and inside budget — a sentence rare enough in aerospace that it should be framed — pointed squarely at the roughly ninety-five percent of the universe's contents that no telescope has seen or will. Dark matter and dark energy are not two mysterious substances. They are two entirely unrelated placeholders that happen to share an adjective, because when astronomers ran out of vocabulary they reached for the same one twice. "Dark" is not a description. It is a filing decision.

It is a filing decision that hides an asymmetry more interesting than the symmetry it asserts. Dark energy genuinely is a shortfall in the books — the name of a discrepancy, and never anything else. Dark matter is a different animal. It is a thing we have cornered without seeing, hemmed in by several measurements that have no business agreeing and do anyway: the acoustic peaks in the cosmic microwave background, the element ratios left over from the universe's first few minutes, the fact that structure formed at all on the schedule it did, and above all the colliding clusters where the lensing mass has visibly walked away from the gas it was supposed to belong to. Something that separates from its baryons in a collision is behaving like a substance, not like an accounting error. Which leaves us somewhere stranger than ignorance: far more confident that something is there than about anything it might be. Roman is aimed at both, with the same camera, and cannot tell them apart by looking.

Roman will not find dark matter. Nothing built around light can — there are no photons coming off it to catch. What it will do is photograph hundreds of millions of galaxies and measure, very carefully, how slightly wrong their shapes are. Mass bends the path of light passing near it, so a galaxy seen through an intervening clump of something arrives at the detector faintly sheared — stretched a fraction of a percent in a direction it has no business being stretched. One galaxy tells you nothing; the distortion is far smaller than the ordinary variation in how galaxies happen to be shaped. A hundred million galaxies tells you where the mass is. You reconstruct the invisible from the accumulated small errors in the visible.

For dark energy the technique differs and the epistemics do not. Measure how fast the universe was expanding at various points in its history — using exploding stars of known brightness, and the faint sonic ripples frozen into the large-scale distribution of galaxies — then discover the expansion is accelerating when nothing in the inventory should be accelerating it. Dark energy is the name of that discrepancy. It has never been anything else. We did not detect a substance and then name it; we detected a shortfall in the books and gave the shortfall a title so we could write papers about it.

So what launched on Sunday is, in the most literal sense available, the most sophisticated instrument ever built for measuring how wrong we are. Not what is there — what bends. The dominant constituent of reality is known to us exclusively through its effects on things that are not it, which is either a humbling epistemological condition or an extremely elaborate way of saying we have been reading the universe by its shadow.

There is something nicely recursive about the name on the side. Nancy Grace Roman was NASA's first chief astronomer, and she spent years arguing a space telescope into existence against people who considered it an expensive way to duplicate what mountaintops already did. She got Hubble. Hubble then spent three and a half decades assembling the most complete picture of the cosmos anyone had ever had — and the single most consequential thing that picture revealed was the size of the hole in it. The telescope she won is why we know how much is missing. The telescope with her name on it exists to go stare at the missing part.

First images are expected in early 2027. Nobody involved expects them to settle the question. The honest version of the mission statement is that within a few years the outline of our ignorance will be traced about a thousand times more sharply than it is today: same ignorance, better edges. Which sounds like a letdown right up until you notice it is the only kind of progress that has ever actually happened.

And come back, before the end, to those 1.4 terabytes a day. Nobody is going to look at them. Not most of them, not ever — there is no astronomer at an eyepiece at the far end of this, no moment where a person sees the thing and knows. Not one of those hundred million sheared galaxies means anything by itself. The finding exists only downstream, after the whole catalogue has been pushed through a pipeline no single human holds in their head, and what comes out the other side is an error bar. The error bar is the discovery. That is a way of knowing with no knower standing inside it, and we have quietly made it the main way we have.

We are a species that built a three-hundred-megapixel camera, threw it a million miles, and aimed it at the blank region of the map where someone has written here be something in very careful handwriting. That is not defeat. That is the most self-aware thing anyone has done in a while — even if, on inspection, nobody in particular is going to be the one doing it.