How IBM’s 104-Qubit Processor Simulated Particle Production

The experiment in brief

Roland C. Farrell, Nikita A. Zemlevskiy, Marc Illa, and John Preskill used IBM’s 104-qubit ibm_marrakesh processor to carry out a digital, real-time simulation of a collision in one-dimensional Ising field theory. They prepared two separated light-particle wavepackets moving toward each other, evolved the system through the collision, and measured its local energy density after the particles had passed. The hardware runs used up to 5,589 two-qubit gates to reach post-collision dynamics. 7

The central observation was evidence for inelastic particle production: rather than simply producing two outgoing light particles, the collision was consistent with a final state containing one light particle and a slower, heavier excitation. 7

How the particle collision was simulated

The team encoded a discretized Ising field theory on the processor and implemented its real-time evolution with a Trotterized quantum circuit. In practical terms, the calculation proceeded in three stages:

1. Prepare incoming particles. Two localized light-particle wavepackets were initialized far apart with opposite momenta.
2. Evolve through the impact. Trotter steps approximated the theory’s time evolution as the wavepackets approached, overlapped, and moved apart.
3. Measure the outgoing state. Local energy-density measurements tracked the distribution of energy after the collision. 7

This is a scattering experiment in a simplified quantum field theory—not a literal collision of physical particles inside the chip. Its value is that quantum hardware directly represents the many-body quantum dynamics that make real-time scattering difficult to calculate in general.

Why preparing the wavepackets was the key challenge

A scattering calculation is only as useful as its incoming state. The researchers first created a tailored ( |W(k_0)\rangle )-like state, designed to set a wavepacket’s spatial envelope, momentum content, and quantum numbers. Because that starting state can include unwanted multiparticle components, they then used symmetry-preserving variational circuits to lower its energy toward the target one-particle eigenstate. 7

The large-lattice Ising wavepacket circuits were designed with matrix-product-state simulations before being used as initial states on the 104-qubit processor. 7

The W-state advantage

Conventional preparation methods can become deeper as a wavepacket spreads across more lattice sites. The team instead adapted W-state preparation techniques using mid-circuit measurement and classical feed-forward:

• Mid-circuit measurements produce classical outcomes before the quantum program finishes.
• Feed-forward uses those outcomes to choose later conditional operations.
• Together, they enable an ancilla-free preparation protocol whose circuit depth is independent of wavepacket size and spatial dimension, apart from dependence on the physical correlation length. 7

The authors describe this scaling as a superexponential improvement over earlier approaches with expensive classical preprocessing or circuit depths that grow polynomially with wavepacket size. 27

How energy-density skewness revealed a heavy particle

The experiment did not identify outgoing particles by taking a conventional detector image. Instead, it used the shape of the measured energy-density profile.

For a comparatively elastic outcome, two outgoing light particles should give a more symmetric energy distribution after the collision. In the inelastic channel, a heavy particle moves more slowly, creating a detectable asymmetry in the post-collision profile. 7

The researchers quantified that asymmetry with skewness, a statistical measure of whether a distribution has a longer or heavier tail on one side. A nonzero skewness that matched the expected heavy-particle trajectory served as the signature of the inelastic process. 7

That matters because the signal supports more than energy being redistributed. It is consistent with the incoming light particles’ kinetic energy being converted into a final state with a new, heavier excitation plus a light particle. 7

Why this matters for quantum field theory

Particle production is a defining feature of high-energy collisions. In the long term, researchers want to predict how collisions in strongly coupled quantum field theories produce particles and collective behavior—problems related to hadron production and ultimately quantum chromodynamics (QCD). Generic real-time quantum dynamics are especially challenging for classical methods, making them a natural target for future quantum computers. 7

The wavepacket-preparation approach is not limited to the Ising demonstration. The authors also constructed or demonstrated it for one-dimensional scalar field theory, the Schwinger model, and two-dimensional Ising field theory. 27

What the result does—and does not—show

This was an important algorithmic and hardware demonstration, but its scope should be stated precisely.

• It modeled one-dimensional Ising field theory, not QCD, realistic three-dimensional hadron collisions, or collider jets. 7
• It found evidence for an inelastic channel, rather than delivering a systematically precise, fault-tolerant scattering prediction. 7
• It relied on noisy quantum hardware and error-mitigation techniques. The work therefore differs fundamentally from a fully error-corrected quantum calculation. 78
• Reaching realistic gauge theories remains difficult. Higher-dimensional encodings, long real-time evolution, scalable state preparation, and precise extraction of many-particle observables will require substantially more capable quantum systems.

The immediate achievement is narrower but concrete: a quantum processor followed a prepared two-particle state through a collision and produced a measurable signature consistent with a heavier outgoing particle. That makes scalable scattering-state preparation and real-time particle-production studies more credible targets for future quantum hardware. 7