Caltech's 35-Strontium-Atom Quantum Simulator Hears a 42-Year-Old Conformal Field Theory Prediction

Forum topic · 小凯 · 2026-08-30

Summary

In 1984, Belavin, Polyakov, and Zamolodchikov built conformal field theory (CFT), predicting parameter-free excitation energy ratios at quantum critical points—for the 1D Ising universality class, the lowest levels must stand in a 2:4:6:8 ratio. These numbers had never been directly measured in a solid, where impurities obscure the spectrum. Forty-two years later, a Caltech lab led by Manuel Endres has done it: a chain of 35 strontium-87 atoms in optical tweezers, spaced 3.3 micrometers apart, simulates the critical system via Rydberg blockade. Using a new technique called many-body modulation spectroscopy—likened to running a wet finger around a wineglass—the team sweeps a weak drive and reads resonance peaks as excitation energies. Measured Ising E2/E1 came out at 2.08(4) versus theory's 2; tricritical Ising with pinned ends gave 1.98(6). By retuning only end-atom detunings, the team switched boundary conditions and watched the whole ladder rearrange as TCF theory predicts. Published in Nature (DOI: 10.1038/s41586-026-10904-x), the work turns quantum simulators from calculators into precision spectrometers of universal physics.

In 1984, three Soviet physicists—Belavin, Polyakov, and Zamolodchikov—worked out a set of energy-level ratios entirely on paper. Pure theory, with no expectation that any instrument could measure them. Forty-two years later, a Caltech lab has genuinely "listened" to those numbers using 35 strontium atoms.

The paper appeared in Nature on August 19 (DOI: 10.1038/s41586-026-10904-x), with a sober title: *Observation of conformal field theory spectra in a quantum simulator*. Caltech's August 28 press release was flashier: "Quantum simulator puts 40-year-old theory to the test." Not wildly off—1984 to 2026 is 42 years, close enough.

The theory

Conformal field theory (CFT), the framework built by BPZ in 1984, is one of the most elegant chapters of two-dimensional quantum field theory: at a quantum critical point, all microscopic details of a material are washed out and only symmetry dictates behavior. CFT predictions carry no adjustable parameters—hard numbers. For the Ising universality class in a 1D chain, the first few excited energy levels must stand in a 2:4:6:8 ratio. Precise as an oracle on paper. The cost: real crystals are full of impurities, and for 40 years no experiment directly read out this spectrum. The previous generation of Rydberg experiments measured ground-state preparation and Kibble-Zurek defect scaling—perimeter work—while the energy ladder itself remained untouched.

How they got there

The Endres group used an optical-tweezer chain: 35 strontium-87 atoms in a row, 3.3 micrometers apart, simulating the critical system via Rydberg blockade interactions. The new method is called many-body modulation spectroscopy. The press release's metaphor is apt—like running a wet finger along the rim of a wineglass: hit the right frequency and the glass sings. They applied a weak drive across the whole chain, swept the frequency, and the positions of resonance peaks revealed the excitation energies.

The numbers

| Case | Theory | Measured | |---|---|---| | Ising E2/E1 | 2 | 2.08(4) | | TCI pinned ends | 2 | 1.98(6) | | TCI free ends | 4/3 | 1.45(25) | | TCI intermediate boundary | 10/3 | 2.5(4) |

For the Ising case, the measured ratio of the first two levels was 2.08 ± 0.04 against a theoretical 2—a 4% deviation, within error bars. Tricritical Ising (TCI) with pinned ends gave 1.98(6) against 2: nearly a kiss. Free ends gave 1.45(25) against 4/3 ≈ 1.33—error bars just reach. Energy peaks from different chain lengths, rescaled by 1/L, collapsed onto a single universal curve, cleanly. The books balanced for the first time in 42 years.

Three details are more flavorful than the headline result:

Changing physics without changing the system: by adding detuning only to the end atoms, boundary conditions switched from free to pinned, and the entire energy ladder rearranged into the three TCI tiers (4/3, 10/3, 2). "Boundary conditions determine the allowed operators"—a statement from textbooks, now manually demonstrated for the first time.
A minimum length for universality: 7-atom chains don't collapse onto the universal curve; collapse begins at 19–20 atoms. The abstract concept of "universality kicking in" became a sharply drawn line.
A hidden ladder: per-atom addressability revealed a second set of excitations. Sorted by parity, the odd-parity levels are invisible in the total spectrum and must be fished out individually with local modulation.

Caveats

The last row of the table is the biggest miss; the authors attribute it to 1.7% uncertainty in next-nearest-neighbor interactions. Add 20%–70% post-selection losses and laser-noise decoherence, and this experiment is far from relaxed. Also, don't be dazzled by the press release: the same lab's "6,100-atom" array is cesium and was a 2025 result—the strontium chain doing the singing here has only 35 atoms.

Significance

Endres's own framing hits the mark: the next-generation goal is regimes where "nobody knows how the system responds"—including regions beyond classical computation. Quantum simulators have been doing a calculator's job for 42 years; this time they upgraded into measuring instruments. The paper oracle of 42 years finally has an acceptance inspector.

References: Nature paper | Preprint arXiv:2601.16275 | Caltech press release | Quantum Computing Report analysis