Quantum Computer Recreates Particle Collisions That Turn Energy Into Matter

In the first instants after two high-energy particles slam into each other, something remarkable can happen: the kinetic energy of the collision transforms into entirely new particles that did not exist before the impact. This process, known as inelastic particle production, is one of the most fundamental consequences of quantum field theory and underlies what happens in particle colliders and in the hot, dense universe moments after the Big Bang. Yet simulating such out-of-equilibrium collisions on ordinary computers is extraordinarily difficult, because the entanglement and complexity of the quantum states involved grow explosively with system size and time. Now, a team of researchers has reported evidence of inelastic particle production in a quantum field theory using a digital quantum processor, running a simulation on 104 qubits and opening a practical route toward studying real-time scattering dynamics that have long been beyond the reach of classical computation.

The work, published in Nature Physics by Roland C. Farrell and John Preskill of the California Institute of Technology along with Nikita A. Zemlevskiy and Marc Illa of the University of Washington, focuses on the one-dimensional Ising field theory, a simplified but physically rich model of interacting particles. The team collided two wavepackets, each carrying the lightest particle in the theory, and watched what emerged after the crash. By measuring the skewness of the energy density left behind in the collision region, they identified an inelastic component of the outgoing radiation: a final state containing one light particle alongside one heavier particle, a configuration that could only arise if collision energy had been converted into new matter. The experiment used up to 5,589 two-qubit gates to access the post-collision dynamics, making it one of the deepest quantum simulations of scattering reported to date.

The central technical hurdle in simulating scattering on a quantum computer is not simply evolving the system forward in time, which Trotterized digital circuits can handle, but preparing physically sensible initial states: two well-localized wavepackets moving toward each other with well-defined momentum. Naive preparation methods are prohibitively expensive because they scale polynomially with the spatial volume occupied by the wavepacket, which grows rapidly as the packets spread and as lattice sizes increase. The new work circumvents this bottleneck with a quantum algorithm built around W states, multipartite entangled states in which a single excitation is coherently distributed across many qubits. By extending established protocols for efficiently creating W states, the researchers could lay down a single-particle skeleton for the wavepackets and then shape it into physically accurate moving packets with far shallower circuit structure than previous approaches.

The key innovation lies in the use of mid-circuit measurements combined with feed-forward operations that are conditioned on the outcomes of those measurements. This dynamical-circuit technique allows long-range entanglement to be established in ways that purely unitary, fixed-depth circuits cannot easily achieve. As a result, the depth of the wavepacket-preparation circuit becomes independent of the wavepacket’s spatial volume, rather than scaling polynomially as in earlier methods. For a simulation aiming at utility-scale devices, where every additional gate layer accumulates noise, this constant-depth behavior is transformative. It means the simulation can in principle be scaled to wavepackets spanning hundreds of lattice sites without the preparation cost exploding, effectively decoupling the quality of the initial state from the size of the quantum computer’s register.

Quantitatively, the construction proceeds in stages. A W-state-like superposition across the lattice sites encodes a single particle delocalized with a chosen central momentum, forming a seed state from which a genuine interacting wavepacket is grown. The researchers then apply layers of symmetry-preserving energy minimization, using an adaptive variational algorithm related to ADAPT-VQE, to transform the seed into a low-energy wavepacket consistent with the interacting theory. Classical tensor-network simulations validated the approach before it ever touched hardware: comparing prepared wavepackets on small and large lattices, and checking energy densities and infidelities against exact results, the team confirmed that circuits of manageable depth produced states faithful to the target physics. This careful classical benchmarking, performed with matrix product state methods, provided the reference curves against which the quantum hardware results were ultimately judged.

The hardware itself was IBM’s ibm_marrakesh processor, on which the team ran scattering circuits across 104 qubits. The collision dynamics were implemented through second-order Trotterized time evolution, with the lattice divided into regions so that two incoming wavepackets could approach, collide, and separate. To extract meaningful signals from a noisy device running circuits of this depth, the team deployed a battery of error mitigation techniques. Pauli twirling and randomized compiling converted coherent errors into stochastic ones that are easier to model; twirled readout error mitigation corrected measurement biases; and dynamical decoupling suppressed idling decoherence. Zero-noise extrapolation style analyses, along with careful signal-strength estimation of local observables, allowed the raw device data to be pushed toward the true expectation values of the ideal circuit.

The decisive observable was the skewness of the measured energy density. In a purely elastic collision, where the incoming particles simply bounce off or pass through one another, the post-collision energy distribution retains a symmetric character, splitting into two outgoing lumps of equal energy. When inelastic production occurs, energy is shared asymmetrically between final states of different particle content, and the energy density develops a measurable skewness. By computing the third moment of the energy density distribution across a range of energy cutoffs, the researchers extracted a quantity that distinguishes elastic from inelastic outcomes. Their measurements, on both the quantum processor and in high-accuracy tensor-network simulations, revealed the signature of a final state with one light and one heavy particle, evidence that the collision had genuinely converted kinetic energy into new particle mass.

A striking feature of the study is the way low-energy and high-energy collisions were contrasted. At low collision energies, the dynamics are approximately elastic and integrable, so the outgoing radiation mirrors the incoming packets. At higher energies, the inelastic channel opens, and the skewness signal emerges clearly. The team also demonstrated the generality of their wavepacket-preparation machinery by constructing circuits for other theories: one-dimensional scalar field theory, the Schwinger model of quantum electrodynamics in one dimension, and the two-dimensional Ising field theory. This toolbox character matters because the ultimate ambition of the field is to simulate quantum chromodynamics, the theory of quarks and gluons, where inelastic processes like hadron production and jet formation are the bread and butter of collider physics.

The broader context is a decade-long effort to bring quantum computing to high-energy and nuclear physics. Since the early proposals for quantum computation of scattering in scalar field theories, and the first few-qubit demonstrations of real-time lattice gauge dynamics in 2016, the community has progressed through simulations of the Schwinger model on roughly one hundred qubits, tensor-network-validated hadron dynamics, and cold-atom and Rydberg simulators probing string breaking and gauge invariance. The present result adds a crucial ingredient to that program: an efficient, scalable way to prepare the scattering states that all such simulations require. Combined with continuing advances in quantum error correction, gate fidelity, and error mitigation, the constant-depth W-state approach suggests that detailed, real-time studies of particle production may move from proof-of-principle demonstrations toward genuine computational tools.

There remain caveats and frontiers. The simulated theory is one-dimensional, the wavepackets span a finite lattice, and error mitigation rather than full error correction carried the weight of noise suppression. Extending to nonintegrable higher-dimensional theories, resolving differential cross sections rather than energy-density signatures, and achieving the precision needed to confront experimental collider data are formidable challenges ahead. Yet the demonstration that 104 qubits and thousands of gates can witness energy becoming matter in a quantum field theory marks a milestone in the convergence of quantum information science and fundamental physics. As the authors and their collaborators in this growing field emphasize, the quantum computers being built today may ultimately be the only calculational instruments capable of answering certain questions about the real-time, out-of-equilibrium behavior of nature’s most fundamental constituents.

Subject of Research: Digital quantum simulation of inelastic scattering and particle production in quantum field theories using W states on a quantum processor

Article Title: Digital quantum simulations of scattering in quantum field theories using W states

Article References: Farrell, R. C., Zemlevskiy, N. A., Illa, M., & Preskill, J. (2026). Digital quantum simulations of scattering in quantum field theories using W states. Nature Physics. https://doi.org/10.1038/s41567-026-03436-8

Image Credits: AI Generated

DOI: 10.1038/s41567-026-03436-8

Keywords: quantum simulation, quantum field theory, particle scattering, inelastic particle production, W states, Ising field theory, digital quantum computers, wavepacket preparation, mid-circuit measurements, error mitigation, Schwinger model, Nature Physics