What does it take to measure a three-meter drift in a satellite orbiting 12,265 kilometers up?
FreeAstroScience read the full Nature paper behind this result, not the press release alone. This piece works through the frame-dragging measurement in plain numbers.
It is evening on Italy’s Adriatic coast as this goes to print, and LARES-2 is somewhere overhead right now, mid-orbit, doing exactly what it was built to do: bouncing laser pulses back to ground stations that fix its position to a fraction of a millimeter.
Two satellites, one cancellation trick
Two satellites can cancel out Earth’s biggest source of orbital noise simply by tilting their orbits in opposite directions. LARES-2 and LAGEOS fly in nearly identical orbits, but their inclinations are supplementary. Tilt one satellite’s orbital plane by an angle, and the other sits at 180 degrees minus that angle. Our planet is not a perfect sphere, and that imperfection, mostly its equatorial bulge, pushes both orbital planes around the pole at a rate that depends on the cosine of the inclination. For two orbits with supplementary tilts, that push comes out equal and opposite.
Frame-dragging refuses to cancel the same way. Both orbital planes get nudged in the same direction no matter the inclination, so adding the two drifts together doubles the relativistic signal instead of erasing it. Sum two numbers, and the giant Newtonian term disappears while the tiny Einsteinian one survives.
- The bulge, which cancels.
- Frame-dragging, which adds instead of canceling, so it survives the sum.
- Higher-order gravitational terms, corrected separately using each satellite’s tracked orbit.
- Everything non-gravitational, from sunlight pressure to thermal drag, calibrated satellite by satellite in its own separate correction.
LARES-2 itself is almost absurdly plain. Picture a sphere just 42 centimeters across, machined from solid Inconel 718 and studded with small corner-cube mirrors, with no electronics, no fuel, and no moving parts. Italy’s own space agency built it that way on purpose, since a passive metal ball is not pushed around by sunlight pressure or thermal drag nearly as much as a satellite bristling with solar panels. Ordinary orbital mechanics, the same kind covered in our explainer on why satellites stay in orbit at all, still governs both satellites; they just carry a far subtler signal buried in how their orbital planes drift over time. LARES-2 rode into medium Earth orbit on the very first Vega C launch, in 2022, alongside a much older partner, NASA’s LAGEOS, which has been circling since 1976.
How precise is 0.2 percent, really?
Nature published this result online on 8 July 2026. Point two percent means the measured frame-dragging and Einstein’s predicted frame-dragging now agree to two parts in a thousand, roughly ten times tighter than any earlier test managed. Researchers express the result as a single number, μ, which general relativity fixes at exactly 1. Their fit returned μ equal to 1.001, with a statistical uncertainty of 0.001 and a systematic uncertainty of 0.002 layered on top. A t-statistic near 1,200 and a P-value near 10−135 leave almost no room, if any, for the signal being a coincidence of noisy data.
Nothing about this is armchair physics. GPS satellites already rely on the same kind of relativistic bookkeeping to stay synchronized, a correction whose size we’ve worked through separately.
Two independent stress tests backed the headline number up. Three hundred Monte Carlo simulations, each built from randomly perturbed orbital parameters, returned μ equal to 1.0001 plus or minus 0.0019. A separate covariance analysis, propagating the same list of uncertainties by hand rather than by simulation, returned μ equal to 1.0007 plus or minus 0.0019. Three different methods. Each landed within a hair of 1.
Turn that agreement into something you can picture. LARES-2’s share of the frame-dragging signal is a nodal precession of 30.678 milliarcseconds a year, an angle so small it means nothing on its own. Multiply it by the satellite’s orbital radius of about 12,265 kilometers, and the node itself creeps sideways by roughly 1.8 meters every year, a shift about the length of a small car. Twelve thousand kilometers up, tracked to within a fraction of a millimeter and averaged over nearly three years of laser pulses, that car-length drift is the entire measurement.
Table 1 — Relative uncertainty in frame-dragging tests, by mission (Ciufolini et al., Nature, 9 July 2026)
| Test | Satellites | Relative uncertainty |
|---|---|---|
| 2004 | LAGEOS, LAGEOS 2 | About 10 percent |
| Since then | LAGEOS, LAGEOS 2, LARES, Gravity Probe B | Several percent |
| 2026, this result | LARES-2, LAGEOS | 0.2 percent |
What the paper admits it doesn’t know
The paper is refreshingly specific about what it left uncertain, starting with how LARES-2’s own orbit was modeled. Its Methods section states plainly that along-track empirical accelerations were re-estimated every five days to soak up unmodeled forces, while cross-track and radial components were never estimated at all and were simply fixed at zero. Modeling choices like that carry real consequences, and they rarely survive into a press release.
A second, smaller gap sits open on the authors’ own terms. LARES-2’s radiation-pressure coefficient, called Cr, came out to 1.071, cleaner and more stable than LAGEOS’s long-established 1.130. Even so, the team flags a long-period wobble in that number that might be a tidal alias tied to the K1 tide, though as they put it themselves, “a longer time series will be needed to confirm that.” We cannot settle that question either. Neither can they, yet.
One more correction matters for how you read the timeline. Secondary coverage of this result, including at least one science outlet we read while preparing this piece, describes the campaign loosely as spanning about three years. Exactly 1,050 days passed, from 17 July 2022 to 1 June 2025, and that figure is not a rounding of convenience. Averaging over one full period of the two satellites’ nodal precession cancels the K1 tide’s disturbance instead of leaving a residue in the fit, which is precisely why the window is that long and no longer.
Which alternative theories just got squeezed?
Chern-Simons gravity is the theory that took the hardest hit from this result. Rival to general relativity though it is, it agrees with every post-Newtonian test the Solar System has ever run except one: it predicts its own, different amount of frame-dragging. That single disagreement is the only place a frame-dragging measurement can actually corner it.
Back in 2008, working from the earlier LAGEOS-only test and its roughly 10 percent uncertainty, Smith and colleagues could only push the theory’s mass parameter above 0.001 per kilometer. Now that floor sits at 0.02 per kilometer, more than an order of magnitude tighter, alongside a fresh constraint on a related post-Newtonian parameter that keeps it below one part in ten thousand of a second.
Why keep hunting for gaps in a theory that keeps winning? Because general relativity cannot be the last word, regardless of how well it performs here. Quantum mechanics still refuses to talk to it, and Roger Penrose, a co-author on this very paper, proved back in 1965 that the theory predicts its own breakdown inside a singularity. Every bound this experiment tightens is one more place a successor theory cannot hide.
None of this proves Chern-Simons gravity is wrong. Only its room to maneuver just got a great deal smaller.
Better tide models, an unplanned bonus
The same dataset that pinned down frame-dragging also sharpened something far more down to earth, the tides. Buried inside the signal is a tide called K1, a once-a-day bulge in Earth’s shape driven by the Moon and Sun, cataloged under Doodson number 165.555. Groote Eylandt in northern Australia feels that tide directly, with the water there rising about 33 centimeters once a day. Up on the satellites’ orbits, that same daily tide masquerades as a slow, 1,050-day wave, which happens to be exactly the nodal period the whole experiment is built around.
A weaker experiment would have been swamped by it. This one used the tide’s precisely known period and phase to fit it out, leaving a residual of about 14 milliarcseconds on each satellite’s node, in opposite directions, against an estimated pre-fit amplitude near 1,760 milliarcseconds. Less than one percent of the tide survived the fit, with a standard error of just 0.2 milliarcseconds. Noise became calibration: what began as the single biggest disturbance in the whole experiment ended up sharpening how we model the Earth’s own lunisolar tides.
From 1918 to a Vega C launch pad
Frame-dragging is not a new idea. Its number simply took until now to become precise enough to matter, a full century after Josef Lense and Hans Thirring worked out the effect in 1918, just three years after Einstein published general relativity. A rotating mass, they showed, curves spacetime and drags it around as the body spins, the way a spinning ball dips into honey and drags the honey along with it. Near Earth the drag is tiny. Close to a rotating black hole, that same effect can dominate everything happening nearby.
Chasing tiny numbers against a noisy planet is not a new habit either. The very first experimental test of general relativity, a story we’ve told in detail elsewhere, was exactly this kind of hunt, and a century later the same hunt continues with a metal sphere bouncing laser light back from medium Earth orbit.
We are skipping the full post-Newtonian formalism here, the ten parameters and the Kerr-metric bookkeeping the paper leans on to connect this measurement to rotating neutron stars and black holes. Pages would be needed to reproduce it properly, and none of it changes the headline number. Anyone who wants that derivation can follow it through the paper’s own references, starting with the Hartle-Thorne transformation that ties a slowly rotating star’s exterior metric back to Kerr’s original solution.
- Frame-dragging confirmed to 0.2 percent, the tightest Solar System test yet.
- Chern-Simons gravity’s allowed parameter space just shrank by more than a factor of ten.
- A byproduct nobody asked for, better lunisolar tide models.
- LARES-2 keeps flying, and its data keeps accumulating for decades to come.
A century separates Lense and Thirring’s pencil-and-paper prediction from a solid metal ball bouncing laser light home from 12,265 kilometers up. What connects them is a habit worth keeping. Check the number again, more precisely than last time, and see if it still holds. It did. FreeAstroScience will keep watching for the next check.
Gerd Dani, President of FreeAstroScience, Science and Cultural Group
Sources
- Ciufolini, I., Paolozzi, A., Pavlis, E. C., Ries, J. C., Paris, C., Ortore, E., Matzner, R., Kuzmicz-Cieslak, M., Deka, D., Pavlis, D. E., Schreiner, P., Ni, W.-T., Penrose, R., and Gurzadyan, V. (2026). LARES-2 satellite measures frame-dragging effect around the Earth. Nature, 655, 332-335. https://doi.org/10.1038/s41586-026-10715-0
- Lense, J., and Thirring, H. (1918). Uber den Einfluss der Eigenrotation der Zentralkorper auf die Bewegung der Planeten und Monde nach der Einsteinschen Gravitationstheorie. Physikalische Zeitschrift, 19, 156-163.
- Ciufolini, I., and Pavlis, E. C. (2004). A confirmation of the general relativistic prediction of the Lense-Thirring effect. Nature, 431, 958-960.
- Smith, T. L., Erickcek, A., Caldwell, R., and Kamionkowski, M. (2008). Effects of Chern-Simons gravity on bodies orbiting the Earth. Physical Review D, 77, 024015.
- Meloni, D. (2026). Effetto Lense-Thirring: confermare Einstein con la precisione spaziale. Reccom Network, 17 July 2026.
