Why Roman’s Microlensing Survey Could Miss Systems Like TRAPPIST-1

Seven planets, almost one lens

The Nancy Grace Roman Space Telescope is expected to discover planets through gravitational microlensing: the temporary magnification of a background star by the gravity of an intervening object. But a system rich in planets can still leave almost no recognizable planetary signature.

In a numerical study, Stephen R. Kane of the University of California, Riverside, asks whether the seven terrestrial planets of a TRAPPIST-1-like system could produce a detectable collective signal. Under the representative conditions he examines, the answer is no—even when the planets are arranged to reinforce one another’s effects.

The work is available as an arXiv preprint submitted on November 3, 2026. Its landing page reports acceptance for publication in Publications of the Astronomical Society of the Pacific. These are calculations of hypothetical lensing events, not observations of Roman detecting or missing such a system.

A mismatch in scale

Microlensing is especially sensitive to planets near a characteristic gravitational scale called the Einstein radius. The conventional “lensing zone,” where sensitivity is greatest, extends from roughly 0.6 to 1.6 times that radius.

Kane places a hypothetical TRAPPIST-1-like host, with 9% of the Sun’s mass, halfway to a background star 8,000 parsecs away. The resulting Einstein radius is about 1.21 astronomical units, with one astronomical unit representing the Earth–Sun distance. Its lensing zone therefore spans approximately 0.73–1.94 astronomical units.

The actual TRAPPIST-1 planets orbit at about 0.012–0.062 astronomical units. In this modeled geometry, all seven sit 12–63 times closer to their star than the lensing zone’s inner boundary. Their tightly packed architecture, rather than simply their small masses, is the central obstacle.

Why the collective signal stays small

The relevant planetary signal is not every difference between a planetary system and a lens fixed at the host star’s position. Observers fit a single-lens model to an event. Much of the apparent difference disappears because a compact system acts approximately like one lens located at its center of mass.

Kane calculates what remains after that fit. He finds that the leading observable distortion decreases roughly with the cube of the planets’ separation in Einstein-radius units. In his analysis, a lower-order distortion cancels to first order between the two principal images of the background star. Moving planets inward therefore suppresses their recognizable signature particularly rapidly.

For planet g, the strongest individual case examined, the maximum fractional residual is only about 6.6 parts per million under an extreme, high-magnification point-source scenario. Modeling the background star as an extended disk reduces that already small signal further.

The calculation including the host and all seven planets fares little better. With the planets aligned to favor coherent addition, the combined detection statistic, delta chi-squared, is approximately 0.00001 at the adopted Roman-like precision. Kane uses 160 as the detection threshold. Even improving the precision tenfold, to one millimagnitude, leaves the statistic near 0.001—more than five orders of magnitude below that threshold.

Bigger planets are not an easy escape

Kane also increases every planet’s mass while holding the orbital architecture fixed. At one-millimagnitude precision, detection requires a mass multiplier of about 315. At the baseline ten-millimagnitude precision, the multiplier rises to roughly 1,900.

Those changes create a dynamical problem. The closest neighboring pair crosses an analytic two-planet Hill-stability boundary after a mass increase of only about sevenfold. Crossing that boundary removes a guarantee of stability; it does not prove immediate destruction. Kane notes that determining when the resonant chain actually breaks would require dynamical integrations beyond this study.

Expanding every orbit can restore detectability without changing the mutual Hill spacings, but it also removes the compact architecture the study set out to investigate.

What the result does—and does not—establish

The main calculation adopts one lens–source geometry, static planet positions and simplified, constant fractional measurement errors. Actual precision improves as a source brightens during magnification. Kane argues that the enormous signal deficit nevertheless prevents a realistic noise model from overturning the result. Other distances are explored, but the study does not integrate detection efficiency across a complete Galactic population.

Roman’s transit observations offer a different opportunity. Kane cites simulations by Tamburo and colleagues predicting about 1,350 small transiting planets around mid-M and ultracool dwarfs—not 1,350 complete TRAPPIST-1 analogs. Such detections would predominantly sample nearer foreground stars, rather than the same distances as the representative microlensing configuration.

The demographic lesson is therefore specific: failing to find compact terrestrial systems through microlensing would not show that they are rare. Combining complementary detection methods is necessary to distinguish an absent population from one a survey cannot see.