Roman: Hubble's mirror, a hundred times more sky

On July 25, technicians at Kennedy Space Center pumped 290 gallons of hydrazine into a telescope.² On August 7, it was mated to the adapter that fastens it to the second stage of a Falcon Heavy.³ Encapsulation inside the rocket's fairing came next, with liftoff targeted for 7:26 a.m. Florida time on August 30 from Launch Complex 39A.⁴

Nine months early. NASA's formal commitment for the Nancy Grace Roman Space Telescope is May 2027, and April's budget request still carried that date as the program's baseline.⁵ In a field where flagship observatories routinely slip by years, arriving ahead of schedule is the surprising part of the story.

Even so, launch day tells you very little about this mission. What Roman will do with the next five years has already been decided, published and parceled out by committee, which makes it a stranger object than "NASA's new telescope" suggests.

Hubble's mirror, a different machine behind it

Roman's primary mirror is 2.4 meters across, exactly Hubble's, and weighs 186 kilograms: under a quarter of the original, thanks to three decades of progress in optical manufacturing. L3Harris built it in Rochester, New York.¹

What sits behind the mirror is where the two diverge. The Wide Field Instrument is a 300-megapixel infrared camera built around 18 mercury-cadmium-telluride detectors, each about the size of a saltine cracker, at 0.11 arcseconds per pixel, covering 0.281 square degrees between 0.48 and 2.3 microns, with eight filters and two slitless spectroscopy modes.¹ ⁶ A single exposure swallows a patch of sky larger than the full Moon.¹

One quirk of the optical design explains the instrument's shape. Roman's sharpest imaging falls in a ring around the center of its field rather than at the center itself, so the detectors were laid out in an arch that follows part of that ring. The same arrangement lets both instruments observe at once.¹

NASA quotes two different numbers for the size of that field. The press kit says "at least 100 times" Hubble's; the 2027 budget document says 200 times that of Hubble's infrared instrument.¹ ⁵ Both are accurate, and the gap is a question of which Hubble camera you compare against. The practical result is the same either way: Roman gathers data up to a thousand times faster than Hubble, and over five years it should image more than 50 times the sky Hubble covered in thirty.¹

Why this takes nothing away from Webb

NASA's own analogy is photographic: Roman is the wide-angle lens, Webb the zoom.¹ Webb's mirror is roughly three times Roman's diameter and sees deeper into a far smaller patch of sky.⁷ They occupy different links in the same chain. Roman sweeps enormous areas and turns up the rare objects; Webb goes back to those targets and measures them in detail; and Roman can image the surroundings of objects Webb has already studied, supplying context.¹

The same logic applies to this decade's other large surveys. ESA's Euclid will cover roughly 15,000 square degrees, about a third of the sky, in less detail, reaching back to when the universe was around 3 billion years old. Roman's largest survey covers 5,100 square degrees, some 12% of the sky, but deeper and more precisely, reaching back to 2 billion years. Because the footprints overlap, Roman's finer data can be used to correct Euclid's and then extended across Euclid's much larger area.¹ With the Vera C. Rubin Observatory in Chile the synergy runs differently: Roman's sharpness helps pull apart sources that blur together in ground-based images.¹

Three surveys settled by committee

Up to 75% of the observing time in the five-year primary mission is already committed to three surveys defined through an open process, with the astronomical community spending roughly two years arguing about what to do with the telescope.¹

The High-Latitude Wide-Area Survey points away from the Milky Way's dusty plane and combines imaging with spectroscopy to map more than a billion galaxies. Two dark-energy measurements come out of that one dataset: the systematic distortion of galaxy shapes by weak gravitational lensingThe slight distortion of distant galaxy shapes by mass along the line of sight. Measured across millions of galaxies, it maps where dark matter sits., which yields a 3D map of dark matter, and the pattern of baryon acoustic oscillations, the fossilized ripples of pressure waves that once ran through the primordial plasma.¹

The High-Latitude Time-Domain Survey returns to the same stretch of sky again and again and turns the sequence into a movie. NASA expects roughly 100,000 explosions and other transients, including the subset of type Ia supernovaA stellar explosion that always peaks at roughly the same intrinsic brightness, which makes it usable as a ruler for cosmological distances. events that serve as a ruler for the expansion of the universe.¹

The Galactic Bulge Time-Domain Survey does the opposite and looks inward, toward the galactic center. Six fields, 1.7 square degrees in total, each imaged every 12 minutes, across six seasons: three early in the mission and three near the end.⁸ Outside the core surveys, the first general astrophysics program already selected is the Galactic Plane Survey, which aims to map up to 20 billion stars using 29 days spread across the first two years.¹

What microlensing finds that nothing else does

That 12-minute cadence exists because of one specific effect. When a star passes almost exactly in front of a more distant one, the nearer star's mass bends spacetime and magnifies the background star's light. A planet orbiting the nearer star acts as a smaller lens inside the larger one and shows up as a passing distortion in the brightness curve. That is gravitational microlensingA nearby star acts as a lens and magnifies the light of a more distant one. Planets orbiting the lensing star show up as brief distortions in that brightening., and it is sensitive to precisely the range other methods miss: cold worlds far from their star, out beyond the snow line.

The numbers come from peer-reviewed simulations rather than institutional promises. Penny and colleagues estimated about 1,400 bound planets above roughly 0.1 Earth masses, some 200 of them at 3 Earth masses or less, with sensitivity reaching down to about 0.02 Earth masses, near the mass of Ganymede.⁹ NASA rounds this to "more than a thousand" microlensing planets.¹ ⁸

The same images serve the opposite effect. When a planet crosses in front of its star, the brightness dips slightly; that is a transit. Wilson and colleagues projected between 60,000 and 200,000 transiting planets in that field, almost all of them giants on tight orbits, with something like 7,000 to 12,000 below 4 Earth radii.¹⁰ The press kit settles on roughly 100,000.¹

Launch coverage tends to drop one caveat. Microlensing is a chance alignment: it happens once, it doesn't repeat, and nobody goes back later to reobserve that system. The product is what NASA calls a statistical census, meaning the distribution of planets across the galaxy rather than a catalog of targets to study one by one. As for the transiting giants, distant and faint, the agency's own estimate is that perhaps a few thousand will have measurable atmospheres.¹

The coronagraphAn instrument that blocks a star's light so that whatever is far too faint beside it, such as planets and dust disks, becomes visible. flies as an experiment

Roman's second instrument is a coronagraph, and the label is worth reading closely: it flies as a technology demonstration. The stated goal is to detect planets 100 million times fainter than their stars, 100 to 1,000 times better than existing space-based coronagraphs, using deformable mirrors that correct for minute flaws in the telescope's own optics in flight.¹¹ It will be the first active coronagraph to operate in space.¹¹

What it can actually photograph is reflected light from a planet comparable to Jupiter in size, temperature and distance from its star.¹¹ No Earths. And it shrank along the way: in 2019, cost and schedule pressure removed the integral field spectrograph from the design in favor of a prism. The observing plan is modest too, at three months of pre-planned observations spread across the first year and a half of operations.¹ Its value lies downstream. This is the in-orbit rehearsal of the techniques the Habitable Worlds Observatory would need to photograph planets like ours.¹ ¹¹

1.4 TB a day, and nobody holds the key

Roman should downlink about 1.4 terabytes of raw data every day through a 1.7-meter antenna transmitting at up to 500 megabits per second to ground stations in New Mexico, Australia and Japan.¹ Across the five-year primary mission the processed data adds up to 20 petabytes; the budget document puts the archive at over 10 terabytes per day of operations.¹ ⁵ It is the largest data volume of any NASA astrophysics mission so far.¹ Julie McEnery, the mission's senior project scientist, put it concretely at a July briefing: displaying a single Roman image in full would take more than half a million 4K televisions.⁷

The most consequential decision, though, is administrative. Everything the surveys produce becomes public as soon as it is processed, with no exclusive-access window for the team that proposed the observation.¹ That inverts the usual arrangement, where a team sits on its own data for months or a year. And since nobody downloads 20 petabytes, the analysis moves to a cloud platform called the Roman Nexus: instead of pulling the data onto your machine, you bring your code to it.¹

What can still go wrong

First the rocket, then three months of waiting. The planned commissioningThe stretch after launch when an observatory is deployed, aligned, calibrated and tested before it starts producing science. sequence deploys the solar panels and sunshade five hours after launch, releases the antenna and aperture cover within two days, powers on the coronagraph in the first week and the Wide Field Instrument in the third, finishes alignment and focus by the second month, and declares the observatory ready around the third, when the first images are released and science operations begin. Insertion into the orbit around L2The second Sun-Earth Lagrange point, about 1.5 million km out, where a spacecraft keeps pace with Earth while spending very little fuel. follows in the third to fourth month.¹ In other words, even a clean launch on August 30 means no science images before late November.

The schedule history sits in the budget itself, unvarnished. The commitment signed in 2020 called for an October 2026 launch; the current baseline date is May 2027, seven months later. The development cost estimate rose from $2.898 billion in the 2021 base year to $3.038 billion in 2026, up 5%. Total lifecycle cost, operations included, runs around $4.3 billion.¹ ⁵

The risk that remains sits in the operating budget rather than on the pad. The FY2027 request sets aside $166.8 million for Roman and then drops to $91.4 million, $78.5 million, $73.0 million and $70.9 million in the years that follow,⁵ a descending curve arriving exactly as the surveys would be running. NASA calls the adjustment small and says it does not expect an impact on the launch readiness date,⁵ which is a claim about liftoff rather than about the five years after it. In the previous request, Roman's development line had fallen to $156.6 million against the $376.5 million that had been planned,¹² inside a proposal that cut NASA science by roughly 47% and ended dozens of missions.¹³

None of this is new for this telescope. Back when it was still called WFIRST, it showed up unfunded in two consecutive budget requests, and both times Congress put the money back: $312 million one year, $510.7 million the next.¹⁴

What the astronomical community asked for

Roman was the top-ranked large space mission in the 2010 decadal survey, and NASA's budget document still describes the project in exactly those terms.⁵ Sixteen years on, the telescope is launching with an observing plan written by the community that asked for it.

At the same briefing, McEnery summed up what the field expects to find, and the summary is less triumphant than launch marketing: over the past five to ten years, hints have accumulated that all is not well with the standard model of cosmology.⁷ Roman's stated job is to collect enough data to establish how large that problem is. She was blunt about the rest, too: the telescope's reach is there to find the weird, the rare and the unusual.⁷

Verdict

Roman is a catalog factory, designed so that discovery happens years later, on the computers of people who never set foot in Kennedy Space Center. That is the bet, and it carries an uncomfortable corollary: the mission's value depends far less on August 30 than on five years of operations, archiving and funding that have not happened yet. In a survey the heavy lifting is never in the first image; it's in the millionth, taken with the same care.

Sources

1. NASA. "Nancy Grace Roman Space Telescope Press Kit," June 2026, https://assets.science.nasa.gov/content/dam/science/missions/rst/education/Final%20Roman%20Press%20Kit-508compliant.pdf — 2.4 m, 186 kg mirror from L3Harris; 300-megapixel WFI with 18 detectors; field "at least 100 times" Hubble's; arch-shaped detector layout; 1.4 TB/day and 20 PB over five years; 1.7 m antenna at 500 Mbit/s; ground stations in New Mexico, Australia and Japan; public data with no exclusive access and the Roman Nexus; core surveys capped at 75% of observing time; comparisons with Hubble, Webb, Euclid and Rubin; post-launch timeline; $4.3 billion lifecycle cost.
2. NASA. "NASA Fuels Roman Space Telescope for Late August Launch," Jul 27, 2026, https://science.nasa.gov/blogs/roman/2026/07/27/nasa-fuels-roman-space-telescope-for-late-august-launch/ (290 gallons of hydrazine loaded on July 25).

Show 12 more sourcesHide sources

3. NASA. "NASA's Roman Space Telescope Begins Integrated Operations for Launch," Aug 10, 2026, https://science.nasa.gov/blogs/roman/2026/08/10/nasas-roman-telescope-team-begins-integrated-operations-for-launch/ (mated to the payload adapter on August 7; encapsulation next).
4. NASA. "Roman Launch Countdown," https://science.nasa.gov/mission/roman-space-telescope/roman-launch-countdown/ (August 30, 2026, 7:26 a.m. EDT, Falcon Heavy, Launch Complex 39A).
5. NASA. "FY 2027 Budget Estimates," April 2026, sections ASTRO-15 to ASTRO-21, https://www.nasa.gov/wp-content/uploads/2026/04/fiscal-year-2027-full-budget-request.pdf (field of view 200 times greater than Hubble's infrared instrument; top-priority large mission of the 2010 decadal survey; launch baseline October 2026 to May 2027; development cost from $2,898.1M to $3,037.6M; FY2027 through FY2031 budget lines; "a small adjustment to the FY 2027 funding level that is not expected to impact Roman's launch readiness date").
6. Roman @ IPAC/Caltech. "Wide Field Instrument," https://roman.ipac.caltech.edu/page/wfi (0.281 deg², 0.11 arcsec/pixel, 0.48–2.3 μm, eight filters, grism and prism).
7. Yahoo News. "NASA's new Nancy Grace Roman Space Telescope is about to launch. Here's why it's so cool," Aug 2026, https://www.yahoo.com/news/science/article/nasas-new-nancy-grace-roman-space-telescope-is-about-to-launch-heres-why-its-so-cool-154334619.html (July 29 news conference; remarks from Julie McEnery and Jackie Townsend; mirror comparison across Roman, Hubble and Webb).
8. NASA. "Galactic Bulge Time-Domain Survey," https://science.nasa.gov/mission/roman-space-telescope/galactic-bulge-time-domain-survey/ (six fields, 1.7 square degrees, 12-minute cadence, six seasons, more than a thousand microlensing planets and 100,000 transiting ones).
9. Penny, Gaudi, Kerins, Rattenbury, Mao, Robin, Calchi Novati. "Predictions of the WFIRST Microlensing Survey I: Bound Planet Detection Rates," The Astrophysical Journal Supplement Series, 2019, https://arxiv.org/abs/1808.02490 (~1,400 planets above ~0.1 M⊕; ~200 at ≤3 M⊕; sensitivity down to ~0.02 M⊕).
10. Wilson et al. "Transiting Exoplanet Yields for the Roman Galactic Bulge Time Domain Survey Predicted from Pixel-level Simulations," The Astrophysical Journal Supplement Series 269(1), 2023, https://iopscience.iop.org/article/10.3847/1538-4365/acf3df (60,000 to 200,000 transiting planets; 7,000 to 12,000 below 4 R⊕).
11. NASA/JPL. "The Roman Coronagraph Instrument," https://www.jpl.nasa.gov/missions/the-roman-coronagraph-instrument/ ("detect planets 100 million times fainter than their stars, or 100 to 1,000 times better than existing space-based coronagraphs"; first active coronagraph in space; Jupiter analog in reflected light; relationship to the Habitable Worlds Observatory). Technical background and the 2019 descope: Kasdin et al. "The Nancy Grace Roman Space Telescope Coronagraph Instrument (CGI) Technology Demonstration," 2021, https://arxiv.org/abs/2103.01980
12. American Astronomical Society. "The FY26 President's Budget Request: NASA and NSF Details," June 2025, https://aas.org/posts/news/2025/06/fy26-presidents-budget-request ($156.6 million against the $376.5 million previously projected).
13. The Planetary Society. "Analyzing the FY 2027 NASA budget request," https://www.planetary.org/articles/analyzing-the-fy-2027-nasa-budget-request — Universe Today, "NASA's FY 2026 Budget Request has been Released," https://www.universetoday.com/articles/nasas-fy-2026-budget-request-has-been-released (roughly 47% cut to NASA science).
14. SpaceNews. "NASA budget proposal seeks to cancel WFIRST," https://spacenews.com/nasa-budget-proposal-seeks-to-cancel-wfirst/ — Space.com, "Trump's 2020 NASA Budget Would Cancel Space Telescope, Earth Science Missions (Again)," https://www.space.com/trump-nasa-2020-budget-cancels-wfirst-earth-missions.html ($312 million and $510.7 million restored by Congress).

— Newsroom