String Breaking Observed in a 13-Ion Quantum Simulator

Pull two particles of opposite electric charge far apart and the energy between them grows, then saturates: the electromagnetic force reaches a finite maximum potential and the two simply sit there. In the strong force there is no such ceiling. The energy between a quark and an antiquark rises indefinitely with distance, stretching a flux tube of gluon field that binds them together — a string. Eventually that stored energy is converted into a new particle-antiparticle pair, the new charges screen the original ones, and the string snaps. That process, string breaking, is thought to govern how matter forms in heavy-ion collisions, and it is stubbornly hard to compute classically in real time.

A team led by Christopher Monroe at Duke encoded a stripped-down version of this physics into 13 trapped ions and watched string breaking happen. The surprising part is not that they saw pairs appear. It is where they appeared. The textbook Schwinger picture predicts particle-antiparticle creation happening uniformly through the bulk of the field. In this experiment the new pairs showed up preferentially at the edges of the string, near the static boundary charges, and the timescale barely responded to changes in tension or coupling strength. The team published it in Nature Physics on September 23, 2026.

Key takeaways:

• This is an analogue quantum simulation of a (1+1)-dimensional ℤ₂ lattice gauge theory. It is not QCD, and it is not a particle collider.
• The Schwinger mechanism is exponentially sensitive to tension and coupling. The observed mechanism was not, which is why the authors call it a new, non-conventional route.
• The experimental innovation was site-by-site control of the local field, using a second array of tightly focused laser beams. That is what let them emulate static charges without a hard boundary.
• The system never reached thermal equilibrium during the observation window, in any parameter regime tested.

Why string breaking is hard to simulate classically

Once the energy stored in a string exceeds roughly twice the mass of the lightest particle-antiparticle pair the theory can produce, that energy can convert into a new pair. Because those new charges partially neutralise the original ones, you end up with two screening pairs rather than one connected pair, and the string breaks. String fragmentation is conjectured to be the mechanism behind hadronisation in heavy-ion collisions and the cooling of the universe after the Big Bang. Conjectured is doing real work in that sentence.

The conventional expectation is the Schwinger mechanism: pair creation begins through a process attributed to Schwinger, occurring spatially uniformly through the bulk of the field. Its signature is extreme sensitivity — the creation rate and characteristic timescale are exponentially enhanced by the string tension and the coupling strength. Push the coupling down and the process should slow to a crawl.

What the Duke experiment found was the opposite behaviour, in a different place.

The system: 13 ions, one spatial dimension

The published details matter here, because the popular summaries compress several distinctions into one.

The mapping is elegant. Fermions live on sites and ℤ₂ gauge fields live on the links between them. Applying Gauss’s law eliminates the fermions and leaves a purely bosonic dual chain: charges become spin kinks, and strings become domains of down-spins. The physics the experiment aims at is then expressed in the behaviour of a spin chain that a trapped-ion machine is genuinely good at simulating.

The implemented Hamiltonian is deceptively simple:

H = -Σ J_ij σz_i σz_j  -  Σ h_i σz_i  -  Σ g σx_i

The first term is the string tension coupling, the second is the local effective electric field, and the third is a transverse fluctuation term.

The control problem, and the trick that solved it

Two laser arrays did the work, and the second one is the actual novelty.

• The first array used bichromatic beams with collective addressing via the Mølmer–Sørensen protocol to set the long-range spin coupling and the uniform transverse field, with the coupling profile set by a decay factor of about 0.78.
• The second array used tightly focused, individually addressed beams on carrier Raman transitions to impose a site-dependent longitudinal field at every ion simultaneously. That is what lets the experiment emulate semi-infinite virtual spin chains on both sides, which encode the static probe charges and the static string, and which remove the artefacts a hard physical boundary would introduce.

In other words: a real string in this simulation needs static charges at both ends, but the machine has finite ions. Rather than accept a boundary, the team simulated the surroundings.

What was observed

Two experiments ran. Measurements were averaged over 300 repetitions, and a classical simulation of the same Hamiltonian was run alongside as a cross-check.

A single charge: spreading and trapping

With zero tension, a single charge spreads ballistically across the lattice until it reaches the static boundaries, at a velocity consistent with the predicted maximum. As tension increases, the behaviour inverts: the charge stops spreading and instead performs coherent, localised Bloch-like oscillations about its starting position, the signature of Wannier-Stark localisation. The predicted spatial amplitude and temporal period both match the data, and the agreement holds even at strongly non-perturbative coupling.

A quenched string: where the pairs appeared

Here a classical string was stretched between two static charges, then both the transverse field and the tension were abruptly increased from their ground-state values. The result:

• Charge pairs form near the edges of the string, close to the static boundary charges — not uniformly through the bulk
• At low or vanishing tension those pairs perform coherent oscillations confined to the edges
• As tension increases, the pairs propagate and spread inward from the edges into the bulk
• The onset time and oscillation period of edge pair creation stay consistent across the whole range of tension and coupling scanned, down to a coupling ratio of 0.75

That last point is the discriminator. If pair creation were governed by the Schwinger mechanism, reducing the coupling would suppress it dramatically. It did not.

The authors describe the mechanism as edge-facilitated string breaking. Perturbative analysis supports a simple picture: the abrupt quench creates a quantum superposition of a single pair located near either edge, and as tension rises an equipotential channel opens in configuration space that allows the pair to quantum-diffuse from the edges into the interior. The genuine Schwinger bulk process, being suppressed faster than any power of the coupling, is not observable on this experiment’s timescale at all.

A separate observable told the same story: the size distribution of the largest empty vacuum bubble started concentrated at small sizes and developed a second peak that travels to larger sizes whenever tension is non-zero — the signature of a string that is actually breaking rather than a system settling into equilibrium.

What this is not

This section matters more than any headline number, because the gaps are easy to overstate.

• Not QCD. The system is a ℤ₂ gauge theory in one spatial dimension, not SU(3) Yang-Mills in three. The paper states plainly that a quantum chromodynamics simulator remains beyond current capabilities.
• Not a collider. There are no beams and no scattering. The probe charges are emulated by virtual static spin chains.
• Not a digital quantum computation. No quantum gates, no error-corrected logical qubits, no Trotter circuit.
• Not an equilibrium measurement. The system remained out of equilibrium for the entire observation window in every regime tested. There is no thermal Schwinger result here.
• Not a claim of quantum advantage. At 13 sites this is not yet a regime where classical simulation is out of reach, and the classical cross-check was run deliberately.
• No transverse degrees of freedom. One spatial dimension means no transverse string fluctuations and none of the rope-like behaviour that would require two.

Where the field is heading

This result sits on a clear trajectory, and it is worth seeing the whole line rather than a single point.

Two patterns stand out. First, dimensionality: this field moved from one spatial dimension to two within about a year. Second, and less discussed, the mode flipped. The one-dimensional result is analogue; the two-dimensional trapped-ion results are digital.

There is also a structural reason the jump to two dimensions is not automatic. Plaquette, or magnetic, terms are what give the gauge field its own dynamics and permit photon-like propagation. With the plaquette term switched off, a two-dimensional model collapses back to one-dimensional physics. The 2026 Quantinuum results report genuine two-dimensional string behaviour only with that term enabled — which is a sharper statement than “they ran on more qubits.”

The stated next steps are fully dynamical strings with a dynamical surrounding environment, two-dimensional ion crystals, neutral-atom arrays, and non-Abelian gauge theories.

Every one of those directions multiplies the required qubit count, which is where error rates rather than physics tend to set the limit. Post-selecting quantum shots with a GNN is one measured attempt to buy accuracy back on noisy circuits; it is not required for the 13-ion result above, but it is the kind of technique that decides how far the next generation of these simulations can actually run.

Why it is worth your attention

String breaking is central to how matter forms in high-energy collisions, but it is too hard to compute classically to watch in real time. How hard is a moving target rather than a fixed cost, and when a classical computer can still beat a noisy quantum circuit sets out where that line currently sits. What the Duke team did was encode a stripped-down version of a confining gauge theory into a string of 13 ions, watch the process occur, and find that it does not follow the textbook route. The physics is real, the result is peer-reviewed and open access, and the caveats above are stated by the authors themselves rather than bolted on afterwards.

References

• De, Lerose, De Luo, Surace, Schuckert, Bennewitz, Ware, Morong, Collins, Davoudi, Gorshkov, Katz and Monroe, Observation of string-breaking dynamics in a quantum simulator, Nature Physics (2026), DOI 10.1038/s41567-026-03422-0
• arXiv:2410.13815 — preprint, submitted October 2024
• Quantum Device Simulates Matter Popping into Existence — Duke Pratt School of Engineering, September 23, 2026
• arXiv:2604.07435 and arXiv:2604.07436 — 2+1D lattice gauge simulations on Quantinuum H2
• arXiv:2409.17142 — 2+1D ℤ₂ string dynamics on a 72-qubit Google Sycamore processor

Related Tools

Try these companion tools on DevStackIO Tools:

• scientific calculator — perform computations for quantum physics simulations.