Quantum simulation of materials and quantum matter: what quantum computers have done, and what classical computers matched
Quantum simulation means using a controllable quantum system to compute how another quantum system behaves: a magnet, electrons in a crystal, the nuclear spins of a molecule. It was Feynman's original reason for building quantum computers, and it is where today's machines come closest to doing something classical computers cannot. The strongest claim, Google's Quantum Echoes measurement on Willow, still has no classical match after extensive attempts. Three others, IBM's utility experiment, Quantinuum's square-lattice magnetism simulation and Q-CTRL's Fermi-Hubbard run, were reproduced by improved classical methods; D-Wave's spin-glass simulation is contested; and the newest experiments on superconducting pairing and driven magnets have no classical reply yet. None of them has yet designed a material. Below is each claim, how it was checked, and what would have to happen before quantum simulation changes materials science.
Updated
The short answer
- What quantum simulation is
- Using qubits or atoms to reproduce the behaviour of another quantum system, so its dynamics or phases can be measured. Feynman proposed it[1] and Lloyd showed a quantum computer can simulate any local quantum system efficiently[2].
- Still standing
- Google's Quantum Echoes measurement on Willow: no classical method has reproduced it[13][15].
- Matched by classical methods
- IBM's kicked Ising experiment[5], Quantinuum's square-lattice magnetism simulation[19] and Q-CTRL's Fermi-Hubbard run[24].
- Contested or open
- D-Wave's spin-glass dynamics[8][9], and newer experiments on driven magnets and superconducting pairing that no classical group has yet answered[25][20].
- Materials designed so far
- None. Every result is a measurement on a model Hamiltonian or a small molecule. Designing a material needs accurate answers for realistic models, which most estimates place on error-corrected machines.
What quantum simulation is, and why it was the first idea
In a 1982 paper Richard Feynman pointed out that a classical computer needs memory that grows exponentially with the number of particles to keep track of a quantum system, and asked whether a computer built from quantum parts could simulate nature without that cost[1]. Seth Lloyd answered yes in 1996: a quantum computer can be programmed to simulate any quantum system with local interactions, with a cost that grows only polynomially with its size and the simulated time[2]. The state of spins has amplitudes. At fifty spins that is complex numbers, about bytes at double precision, and every added spin doubles it; a quantum simulator holds the same state in fifty qubits.
The exponential wall is the worst case. Classical physicists have spent decades on methods that avoid it when the physics allows: tensor networks store states with little entanglement compactly, quantum Monte Carlo samples states without a sign problem, and Pauli or operator propagation tracks only the few terms an observable actually needs. Every claim below is, in the end, a test of whether a quantum device found a corner of physics none of those methods can reach.
This page is about the dynamics and phases of quantum matter: spin models, Hubbard models of electrons in a lattice, magnetism, superconducting correlations and how isolated systems thermalize. Computing the electronic energy of a specific molecule, for catalysts, drugs and batteries, is a related but different problem with its own algorithms and evidence, covered in quantum computing for chemistry.
Digital and analog quantum simulation
Two ways to make one quantum system imitate another. Many machines do one; a few do both. How the qubit types themselves differ is in superconducting vs trapped-ion vs neutral-atom qubits.
| Aspect | Digital simulation | Analog simulation |
|---|---|---|
| How it works | The evolution is cut into small time steps (Trotter steps), each built from gates | The hardware's own interactions are tuned to match the model, and the system evolves continuously |
| Which models | Any model that can be written as qubit operators, including fermions | Models close to the hardware's native interaction, such as Ising or XY spins for Rydberg atoms |
| Main error | Gate errors that add up with the number of steps, plus the Trotter error of the time step | Imperfect calibration of the interactions, and no gate-level way to correct it |
| Typical machines | IBM Heron, Google Willow, Quantinuum H2 and Helios | QuEra Aquila and Pasqal Orion (Rydberg atoms), D-Wave annealers |
| Example below | Quantinuum's Ising and Hubbard simulations[18][20] | Spin liquids and Ising transitions in atom arrays[29][31] |
Google has run both modes on one superconducting chip: analog evolution of an XY magnet with digital state preparation[17]. Daley and colleagues argue that analog devices already give a first practical advantage for specialised problems, while digital ones open the full range of applications but need fault-tolerant hardware[3].
Every major beyond-classical claim in quantum simulation, and the classical reply
Status as of November 2026. "Matched" means a classical method later reproduced the result as well or better, in comparable or less time. "Standing" means no published classical method has. The full list of advantage claims, simulation and otherwise, is on the quantum advantage page.
| Year | Team and hardware | What was simulated | How it was checked | Classical status |
|---|---|---|---|---|
| 2023 | IBM, Eagle (superconducting, 127 qubits[4]) | Magnetisation of a kicked Ising model | Error mitigation, checked against exact results where they exist | Matched within months by tensor networks and sparse Pauli dynamics[5][7] |
| 2025 | Google, analog-digital simulator (superconducting, 69 qubits[17]) | Thermalization and phase transition of a 2D XY magnet | Beyond classical only in its cross-entropy benchmark; the physics was compared with theory | No beyond-classical claim for the physics |
| 2025 | D-Wave, Advantage2 (quantum annealer)[8] | Quench dynamics of spin glasses in 2D, 3D and infinite dimensions | Agreement with Schrödinger-equation solutions at sizes classical methods reach | Contested: tensor networks and variational Monte Carlo reproduce part of it[9][10]; D-Wave says the hardest cases stand[11] |
| 2025 | Google, Willow (65 qubits[14] for the hardest data) | Second-order out-of-time-order correlators (Quantum Echoes) | Nine classical algorithms tried[14]; the value can in principle be reproduced on another quantum device | Standing[15] |
| 2025 to 2026 | Quantinuum, H2 (trapped ion, 56 qubits[18]) | Thermalization of a digitized 2D quantum Ising model | Compared with MPS, PEPS, sparse Pauli and neural-network simulations, none converged in the hardest regimes | Matched on the square lattice by an MPS heuristic on a single A100 GPU[19] |
| 2025 to 2026 | Quantinuum, Helios (trapped ion, up to 90 qubits[20] used) | Superconducting pairing correlations in Fermi-Hubbard models | Exact results on a benchmark lattice; the authors call one regime likely hard for classical methods | Open: no quantified classical comparison, no classical reply yet |
| 2025 to 2026 | Phasecraft and Google, Willow (72 qubits[21]) | Dynamics of a 2D Fermi-Hubbard model | Exact results where feasible, tensor-network and operator-propagation comparisons elsewhere | Beyond exact state vectors only; approximate methods agree in some regimes and differ in others |
| 2026 | Q-CTRL with IBM (superconducting, 120 qubits[22]) | Dynamics of a 1D Fermi-Hubbard model | Agreement with a tensor network where it converged | Matched: reproduced classically faster than the quantum run[24] |
| 2026 | Qedma and partners, IBM Heron r3 (superconducting, up to 74 qubits[25]) | Prethermal oscillations of a periodically driven Ising magnet | Several error-mitigation estimators; selected runs repeated on Quantinuum H2 and Helios | Open: tensor networks did not converge, Pauli-path results depended on truncation |
IBM's utility experiment and the classical replies
In 2023 IBM simulated a kicked Ising model, a lattice of spins flipped by periodic pulses, on its 127-qubit[4] Eagle processor, used error mitigation to remove the effect of noise, and reported accurate expectation values at a scale beyond brute-force classical computation[4]. It was a deliberate step from random sampling to expectation values, the kind of number physics needs.
Classical replies came within weeks, using two of the methods this page keeps returning to. Tindall, Fishman, Stoudenmire and Sels built a tensor network in the shape of the chip and contracted it with belief propagation, getting results significantly more accurate and precise than the quantum processor[5]. Begušić and Chan used sparse Pauli dynamics, a form of Pauli propagation that follows the observable backwards through the circuit and drops small terms, and ran it on a single core of a laptop[6], faster than the experiment. A later paper with Gray converged several such methods orders of magnitude faster than the experiment[7]. The lesson was specific: the circuit's correlations stayed tree-like enough for a method built for that structure to follow them[5].
D-Wave's spin glasses: the contested annealer claim
In 2025 D-Wave and collaborators reported in Science that its annealers simulate the quench dynamics of spin glasses, disordered magnets swept through a phase transition, in two-, three- and infinite-dimensional lattices, in close agreement with the Schrödinger equation, and that leading tensor-network and neural-network methods could not reach the same accuracy in reasonable time[8]. D-Wave states that matching its largest simulations with matrix product states would take nearly a million years[12] on the Frontier supercomputer.
The Flatiron Institute group that matched IBM's experiment then simulated the same two- and three-dimensional models with lattice-shaped tensor networks and belief propagation, at modest cost, published in Science in 2026[9]. Mauron and Carleo used time-dependent variational Monte Carlo on up to 128 spins[10] of a diamond lattice with correlation errors below 7%[10]. D-Wave answers that neither work attempted the infinite-dimensional lattices, the low-precision ensembles where correlations grow fastest, the longest times or the full set of observables[11][12]. Both sides agree the classical work narrowed the claim; whether anything beyond classical reach remains depends on the cases not yet simulated. An annealer is an analog simulator of one model, the transverse-field Ising model, which is why this result belongs to quantum simulation and not to optimisation.
Google: Quantum Echoes, molecules by NMR, and an analog-digital simulator
Quantum Echoes. In October 2025 Google measured second-order out-of-time-order correlators (OTOCs) on its 103-qubit[13] Willow processor. The circuit runs a many-body evolution forward, perturbs one qubit, runs it backward and measures how the disturbance spread, which is sensitive to interference between many paths through the system[13]. The hardest data used 65 qubits[14] and took about 2 hours[14], against an estimated 3.2 years[14] per data point on Frontier, a gap of about 13,000 times[14]. Google spent about 10 person-years[14] trying to beat it with nine[14] classical algorithms. A 2026 study showed that belief-propagation tensor networks, which matched IBM's and part of D-Wave's experiments, cannot feasibly simulate it either, because the circuits generate too much entanglement to compress[15]. It is the strongest simulation result standing. Its circuits are random, so the advantage is on a physics quantity, not yet on a system someone needed measured.
The NMR follow-up. The same team measured OTOCs in a nuclear magnetic resonance experiment on two real molecules, toluene and a dimethylbiphenyl, dissolved in a liquid crystal, and used them to estimate a hydrogen-hydrogen distance and a dihedral angle with accuracy similar to independent spectroscopic measurements[16]. The quantum computer's role was to simulate the molecules' OTOCs, with a root-mean-square error of 0.05[16]. Google says plainly that this demonstration is not yet beyond classical[14]. It is a protocol for when the molecules get larger, not an advantage today.
Thermalization on an analog-digital simulator. In a 2025 Nature paper Google ran a 69-qubit[17] chip in both modes: gates to prepare states, then continuous analog evolution of a two-dimensional XY magnet. The team observed signatures of the classical Kosterlitz-Thouless transition and deviations from Kibble-Zurek scaling during thermalization[17]. The beyond-classical part of that paper is a cross-entropy benchmark of the device; the physics measurements are not presented as out of classical reach.
Quantinuum: digital magnetism and superconducting pairing
Digital magnetism on H2. Quantinuum simulated the digitized dynamics of a two-dimensional quantum Ising model on its 56-qubit[18] H2 machines, with circuits of over 2,000 two-qubit gates[18] and a native partial-entangler fidelity of 99.94%[18]. It watched the spins thermalize on a square lattice and, reprogrammed, on a frustrated triangular one. The authors tried matrix product states, belief-propagation PEPS, sparse Pauli dynamics and neural-network states, and none gave converged results in the hardest regimes[18]. A Google team then proposed a heuristic that rescales low-bond-dimension MPS results by their fidelity, simulated the same 7 by 8[19] periodic lattice with bond dimension 4,096[19] on a single A100 GPU[19], and reported excellent agreement with the hardware. The square-lattice result is therefore matched; the triangular-lattice runs have not, as far as we found, been reproduced classically.
Superconducting pairing on Helios. On Helios, which has 98[20] effectively all-to-all connected qubits, Quantinuum measured pairing correlations, the signature of superconductivity, in three Fermi-Hubbard settings: pairing driven by an electromagnetic field on a half-filled square lattice, d-wave pairing in a doped checkerboard model, and s-wave pairing in a bilayer model relevant to nickelate superconductors. The largest runs simulated up to 72 orbitals[20] with up to 90 qubits[20] and 3,439 two-qubit gates[20]. The authors call one regime likely difficult to access classically, but the preprint does not quantify a classical comparison and makes no advantage claim. It shows the measurement can be done reliably, which is a precondition for using it on open questions about superconductors.
Analog neutral-atom simulators: spin liquids and phase transitions
Atoms held in optical tweezers and excited to Rydberg states interact strongly enough that an array of them evolves as a quantum spin model. Groups at Harvard and MIT began with a 51-atom[27] simulator in 2017 and were later joined by the company QuEra. In 2021 they mapped phases of a two-dimensional Ising-type model on arrays of up to 256[28] atoms, and placed 219[29] atoms on the links of a kagome lattice to detect the signatures of a quantum spin liquid of the toric-code type, a phase with topological order and no local magnetism[29]. In 2025 they followed how ordered domains coarsen after crossing a quantum phase transition[30].
Antoine Browaeys and Thierry Lahaye's group at Institut d'Optique, Lahaye being a co-founder of Pasqal[33], simulated the antiferromagnetic transverse-field Ising model on up to 196 atoms[31], with good agreement with numerical calculations up to around 100 particles[31], the size those calculations could reach. A later experiment with a Pasqal co-author realised a dipolar XY magnet and observed long-range order that only the dipolar interaction allows[32], on arrays of up to 100[33] atoms.
These experiments are science, not advantage claims. They reach sizes where exact simulation stops and produce phases that are hard to prepare and measure any other way, but the quantities they report can often be estimated by classical approximations, and the authors check them against such methods where possible. Their strength is breadth of physics, not a certified speed-up.
Claims from 2026
Q-CTRL's Fermi-Hubbard simulation. Q-CTRL simulated a one-dimensional Fermi-Hubbard model with up to 120 qubits[22] of an IBM processor and reported a wall-clock run time nearly three orders of magnitude[22] faster than its tensor-network baseline. Rausch and colleagues exploited the model's symmetries on four NVIDIA H200 GPUs[23] and cut the gap to about 36 times[23]. Ouyang, Chi and Chan then reproduced all 7,260[24] measured observable trajectories with a transverse tensor-network contraction at bond dimension 32[24], faster and more accurately than the hardware. One-dimensional dynamics is where tensor networks are strongest.
Driven magnets on IBM Heron. Qedma and partners, using its error-mitigation software on an IBM Heron processor, measured long-lived prethermal oscillations of a periodically driven Ising magnet on up to 74 qubits[25] of a heavy-hex lattice. Leading tensor-network simulations failed to converge and sparse Pauli-path simulations depended on their truncation, even with extensive runs on GPUs and the Fugaku supercomputer; selected runs were repeated on Quantinuum hardware as a cross-check[25]. It is open, and the kind of candidate the public Quantum Advantage Tracker exists to test: experiments post their observables with error bars and anyone can submit a classical method[34].
Molecules by NMR on H2. A team using Quantinuum's H2 simulated the proton NMR spectrum of a diphosphane chosen as a classically challenging benchmark, through Trotterized evolution of an effective 21-spin[26] model, and the spectrum agreed with classical reference calculations[26]. Like Google's NMR work, it shows a route from spin simulation to a measurement chemists use, at sizes classical methods still check.
The Hubbard model on Willow. Phasecraft and Google simulated the dynamics of the two-dimensional Fermi-Hubbard model on lattices up to 6 by 6[21] with 72 qubits[21], observing magnetic polarons and stripe dynamics. The scale is beyond exact state-vector simulation; approximate classical methods agree in some regimes and not in others, and the authors say it is unclear which is more accurate where no exact answer exists[21].
A Trotter simulation you can check exactly, with eigenkit
import numpy as np
from eigenkit import Circuit, all_z
from eigenkit.core.blocks import TrotterBlock
# A transverse-field Ising chain: H = -J sum Z_i Z_(i+1) - h sum X_i
n, J, h, t = 8, 1.0, 1.0, 2.0
terms = []
for i in range(n - 1):
terms.append((J, "I" * i + "ZZ" + "I" * (n - i - 2)))
for i in range(n):
terms.append((h, "I" * i + "X" + "I" * (n - i - 1)))
# TrotterBlock builds exp(+iAt), so A = -H gives the physical exp(-iHt).
def trotter_magnetisation(steps):
circuit = Circuit([TrotterBlock(terms, time=t, steps=steps, order=2)], observables=all_z(n))
return float(np.mean(circuit(np.zeros(0)))) # start in |00000000>, all spins up
# The exact answer: build the 256 x 256 matrix and diagonalise it.
paulis = {"I": np.eye(2), "X": np.array([[0, 1], [1, 0]]), "Z": np.diag([1.0, -1.0])}
def matrix(label):
m = np.array([[1.0]])
for ch in label:
m = np.kron(m, paulis[ch])
return m
H = -sum(c * matrix(p) for c, p in terms)
energies, vectors = np.linalg.eigh(H)
psi0 = np.zeros(2**n); psi0[0] = 1.0
psi_t = vectors @ (np.exp(-1j * energies * t) * (vectors.conj().T @ psi0))
Zavg = sum(matrix("I" * i + "Z" + "I" * (n - i - 1)) for i in range(n)) / n
exact = float(np.real(psi_t.conj() @ Zavg @ psi_t))
print(f"exact magnetisation at t={t}: {exact:+.4f}") # +0.0195
for steps in (2, 5, 10, 20):
m = trotter_magnetisation(steps)
print(f"{steps:>2} Trotter steps: {m:+.4f} error {abs(m - exact):.1e}")
# 2 steps: error 2.2e-01; 5: 1.1e-03; 10: 3.9e-04; 20: 1.1e-04
# The exact method stores 2**n complex amplitudes, 16 bytes each.
for qubits in (8, 30, 50, 100):
print(f"{qubits:>3} qubits: {16 * 2.0**qubits:.1e} bytes") # 4.1e+03, 1.7e+10, 1.8e+16, 2.0e+31This is the digital simulation of the experiments above in miniature: Trotter steps built from gates, compared with the exact answer. At eight spins the exact answer is a 256-dimensional matrix; at fifty the state vector needs petabytes, which is why the experiments are checked against tensor networks and Pauli propagation instead. The Trotter error shrinks as the steps get smaller, but each step costs gates, and on hardware gate errors grow with every step. Runs on a bare pip install eigenkit.
From a model to a material: what would have to happen
None of the experiments on this page designed a material, and none claimed to. They measure dynamics or correlations of model Hamiltonians, simplified descriptions such as the Ising or Hubbard model whose parameters are chosen, not derived from a real crystal. Turning simulation into materials design takes three more steps, each a research programme of its own[35].
Accuracy on a model that matters. The two-dimensional Fermi-Hubbard model is the standard target, because it is the starting point for modelling many strongly correlated materials, high-temperature superconductors among them[20]. Fault-tolerant estimates for its ground-state energy are modest by the standards of the field: classically intractable instances with a few hundred thousand physical qubits[36] at a physical error rate of one part per thousand[36], around one million Toffoli gates[37] with better Trotter bounds, and a crossover with the best tensor-network methods within runtimes of hours[38] on a few hundred thousand physical qubits[38]. Yoshioka and colleagues conclude that condensed-matter models are likely the earliest platform for a practical advantage, ahead of the usual chemistry candidates[38].
From the model to the real material. A real material has many orbitals per atom, long-range interactions and lattice vibrations. Reducing it to a lattice model a quantum computer can handle, and knowing that the reduction kept the physics, is a classical step, and a better quantum solver does not remove its errors[35].
Closing the loop with experiment. A simulation helps design a material only when it predicts something measurable, a transition temperature, a magnetic order, a spectrum, that guides what to synthesise next. The NMR work by Google and Quantinuum is the first step in that direction: quantum simulations compared directly with laboratory spectra[16][26]. For the electronic energies that decide catalysts and batteries, the route is phase estimation on error-corrected machines, covered in quantum computing for chemistry.
How to read the next quantum simulation claim
Five questions that sorted every result above.
What model, and does anyone need its answer?
A random circuit is a test of the machine. A named model with physical parameters, such as a Hubbard lattice at a given doping, is closer to science. A real material or molecule compared with a laboratory measurement is closer still.
Which classical methods were tried, and by whom?
The authors' own baselines are the weakest test. For dynamics the challengers are tensor networks with belief propagation, Pauli propagation, MPS heuristics and variational Monte Carlo. IBM's, Quantinuum's and Q-CTRL's results each fell to a method their authors had not used[5][19][24].
Is the geometry hard for classical methods?
One-dimensional chains and short times are where tensor networks are strongest. Two- and three-dimensional lattices, frustrated geometries, all-to-all couplings and long times are harder. A claim on a 1D model deserves more scepticism than one on a frustrated 2D lattice.
Run it on a real QPU with EigenOrbit Routera Trotter circuit like the one above on a real QPUIn eigenkit: pip install "eigenkit[router]", then provider=RouterConfig(device=...) (docs).
Questions
What is quantum simulation?
Quantum simulation is using a controllable quantum system, such as qubits in a quantum computer or atoms in a laser array, to reproduce how another quantum system behaves, so that its dynamics or phases can be measured. Richard Feynman proposed it because classical computers need memory that grows exponentially with the number of particles to do the same job exactly.
What is the difference between analog and digital quantum simulation?
A digital simulation breaks the evolution into short time steps built from quantum gates, so it can simulate any model that can be written in qubit operators. An analog simulation tunes the hardware's own interactions to match the model and lets the system evolve continuously, which reaches larger sizes but only for models close to the hardware's native physics.
Has quantum simulation beaten classical computers?
Possibly, for one result so far: Google's Quantum Echoes measurement on Willow has no classical match after extensive attempts. IBM's kicked Ising experiment, Quantinuum's square-lattice magnetism simulation and Q-CTRL's Fermi-Hubbard run were later reproduced by improved classical methods, D-Wave's spin-glass result is contested, and the newest experiments on superconducting pairing and driven magnets are still waiting for a classical reply.
Can quantum computers simulate materials today?
They simulate simplified models of materials, such as Ising, XY and Fermi-Hubbard models, at sizes beyond exact classical simulation, and measure properties like thermalization and superconducting pairing. None has yet designed or predicted a new material. That needs accurate answers on realistic models, which most resource estimates place on error-corrected quantum computers.
What is the difference between quantum simulation and quantum chemistry on a quantum computer?
Quantum simulation usually means the dynamics and phases of many interacting spins or electrons in a lattice, the physics of magnets and superconductors. Quantum chemistry means computing the electronic energy of specific molecules, for reaction mechanisms, catalysts and drugs. They share methods, and the line blurs for molecular spectroscopy such as NMR.
Is a quantum simulator the same as a simulator of a quantum computer?
No. In physics a quantum simulator is quantum hardware that imitates another quantum system. Software that runs quantum circuits on a classical computer is also called a quantum simulator, but it is a classical program whose cost grows exponentially with the number of qubits for general circuits.
Sources
- R. P. Feynman. Simulating physics with computers. International Journal of Theoretical Physics 21, 467–488, 1982.
- S. Lloyd. Universal quantum simulators. Science 273, 1073–1078, 1996.
- A. J. Daley, I. Bloch, C. Kokail, S. Flannigan, N. Pearson, M. Troyer, P. Zoller. Practical quantum advantage in quantum simulation. Nature 607, 667–676, 2022.
- Y. Kim, A. Eddins, S. Anand, K. X. Wei, E. van den Berg, S. Rosenblatt, H. Nayfeh, Y. Wu, M. Zaletel, K. Temme, A. Kandala. Evidence for the utility of quantum computing before fault tolerance. Nature 618, 500–505, 2023.
- J. Tindall, M. Fishman, E. M. Stoudenmire, D. Sels. Efficient tensor network simulation of IBM's Eagle kicked Ising experiment. PRX Quantum 5, 010308, 2024. arXiv:2306.14887.
- T. Begušić, G. K.-L. Chan. Fast classical simulation of evidence for the utility of quantum computing before fault tolerance. arXiv preprint, 2023. arXiv:2306.16372.
- T. Begušić, J. Gray, G. K.-L. Chan. Fast and converged classical simulations of evidence for the utility of quantum computing before fault tolerance. Science Advances 10, eadk4321, 2024. arXiv:2308.05077.
- A. D. King, A. Nocera, M. M. Rams, J. Dziarmaga, R. Wiersema et al. Beyond-classical computation in quantum simulation. Science 388, 199–204, 2025. arXiv:2403.00910.
- J. Tindall, A. Mello, M. Fishman, E. M. Stoudenmire, D. Sels. Dynamics of disordered quantum systems with two- and three-dimensional tensor networks. Science 392, 868–872, 2026. arXiv:2503.05693.
- L. Mauron, G. Carleo. Challenging the quantum advantage frontier with large-scale classical simulations of annealing dynamics. arXiv preprint, 2025. arXiv:2503.08247.
- A. D. King, A. Nocera, M. M. Rams, J. Dziarmaga, J. Raymond, N. Kaushal, A. W. Sandvik, G. Alvarez, J. Carrasquilla, M. Franz, M. H. Amin. Comment on: Dynamics of disordered quantum systems with two- and three-dimensional tensor networks. arXiv preprint, 2025. arXiv:2504.06283.
- D-Wave Quantum Inc. D-Wave's quantum supremacy result stands (press release, filed with the SEC as exhibit 99.1). D-Wave press release, 2026.
- Google Quantum AI and collaborators (D. A. Abanin et al.). Observation of constructive interference at the edge of quantum ergodicity. Nature 646, 825–830, 2025. arXiv:2506.10191.
- X. Mi, K. Kechedzhi. A verifiable quantum advantage. Google Research blog, 2025.
- P. Bermejo, B. Villalonga, B. Ware, G. Vidal, A. Szasz. Tensor networks with belief propagation cannot feasibly simulate Google's quantum echoes experiment. arXiv preprint, 2026. arXiv:2604.15427.
- C. Zhang, R. G. Cortiñas, A. H. Karamlou, N. Noll, J. Provazza, J. Bausch et al., T. E. O'Brien. Quantum computation of molecular geometry via many-body nuclear spin echoes. arXiv preprint, 2025. arXiv:2510.19550.
- T. I. Andersen, N. Astrakhantsev, A. H. Karamlou, J. Berndtsson et al., X. Mi. Thermalization and criticality on an analogue-digital quantum simulator. Nature 638, 79–85, 2025. arXiv:2405.17385.
- R. Haghshenas, E. Chertkov, M. Mills, W. Kadow, S.-H. Lin et al., M. Foss-Feig. Digital quantum magnetism on a trapped-ion quantum computer. Nature 653, 56–62, 2026. arXiv:2503.20870.
- S. Mandrà, B. Ware, N. Astrakhantsev, S. Isakov, B. Villalonga, T. Westerhout, K. Kechedzhi. A heuristic for matrix product state simulation of out-of-equilibrium dynamics of two-dimensional quantum spin systems. PRX Quantum 7, 033048, 2026. arXiv:2511.23438.
- E. Granet, S.-H. Lin, K. Hémery, R. Haghshenas, P. Andres-Martinez et al., H. Dreyer. Superconducting pairing correlations on a trapped-ion quantum computer. arXiv preprint, 2025. arXiv:2511.02125.
- F. Alam, J. L. Bosse, I. Čepaitė, A. Chapman, L. Clinton et al., P. Roushan. Programmable digital quantum simulation of 2D Fermi-Hubbard dynamics using 72 superconducting qubits. arXiv preprint, 2025. arXiv:2510.26845.
- G. S. Hartnett, K. S. Najafi, A. Khindanov, H. Liao, M. Schutzman, M. R. Hush, M. J. Biercuk, Y. Baum. Fast, accurate, high-resolution simulation of large-scale Fermi-Hubbard models on a digital quantum processor. arXiv preprint, 2026. arXiv:2605.04025.
- R. Rausch, S. Singh, S. S. Jahromi, A. Kshetrimayum, R. Orús. Pushing the classical frontier of 1D Fermi-Hubbard quench dynamics beyond current quantum simulations. arXiv preprint, 2026. arXiv:2606.04771.
- X.-Y. Ouyang, R. Chi, G. K.-L. Chan. Fast classical simulation of "Fast, accurate, high-resolution simulation of large-scale Fermi-Hubbard models on a digital quantum processor". arXiv preprint, 2026. arXiv:2608.13805.
- E. Leviatan, T. Watad, R. Perry, L. Broers, M. Z. Mullath, O. Alberton et al., N. H. Lindner. Resolving structure in prethermal Floquet dynamics with precision quantum computation. arXiv preprint, 2026. arXiv:2607.24937.
- P. Stadler, A. Owens, E. Granet, D. Muñoz Ramo, M. Marthaler. Large-scale NMR simulation on a trapped-ion quantum computer. arXiv preprint, 2026. arXiv:2609.17102.
- H. Bernien, S. Schwartz, A. Keesling, H. Levine, A. Omran et al., M. D. Lukin. Probing many-body dynamics on a 51-atom quantum simulator. Nature 551, 579–584, 2017. arXiv:1707.04344.
- S. Ebadi, T. T. Wang, H. Levine, A. Keesling, G. Semeghini et al., M. D. Lukin. Quantum phases of matter on a 256-atom programmable quantum simulator. Nature 595, 227–232, 2021. arXiv:2012.12281.
- G. Semeghini, H. Levine, A. Keesling, S. Ebadi, T. T. Wang et al., M. D. Lukin. Probing topological spin liquids on a programmable quantum simulator. Science 374, 1242–1247, 2021. arXiv:2104.04119.
- T. Manovitz, S. H. Li, S. Ebadi, R. Samajdar, A. A. Geim et al., M. D. Lukin. Quantum coarsening and collective dynamics on a programmable simulator. Nature 638, 86–92, 2025. arXiv:2407.03249.
- P. Scholl, M. Schuler, H. J. Williams, A. A. Eberharter, D. Barredo et al., A. Browaeys. Quantum simulation of 2D antiferromagnets with hundreds of Rydberg atoms. Nature 595, 233–238, 2021. arXiv:2012.12268.
- C. Chen, G. Bornet, M. Bintz, G. Emperauger, L. Leclerc et al., A. Browaeys. Continuous symmetry breaking in a two-dimensional Rydberg array. Nature 616, 691–695, 2023. arXiv:2207.12930.
- Pasqal. Simulating phases of matter in magnetic materials with qubits. Pasqal blog, 2023.
- Algorithmiq, BlueQubit, Flatiron Institute, IBM and contributors. Quantum Advantage Tracker. Community website, 2025.
- B. Bauer, S. Bravyi, M. Motta, G. K.-L. Chan. Quantum algorithms for quantum chemistry and quantum materials science. Chemical Reviews 120, 12685–12717, 2020. arXiv:2001.03685.
- I. D. Kivlichan, C. Gidney, D. W. Berry, N. Wiebe, J. McClean, W. Sun, Z. Jiang, N. Rubin, A. Fowler, A. Aspuru-Guzik, H. Neven, R. Babbush. Improved fault-tolerant quantum simulation of condensed-phase correlated electrons via Trotterization. Quantum 4, 296, 2020. arXiv:1902.10673.
- E. T. Campbell. Early fault-tolerant simulations of the Hubbard model. Quantum Science and Technology 7, 015007, 2021. arXiv:2012.09238.
- N. Yoshioka, T. Okubo, Y. Suzuki, Y. Koizumi, W. Mizukami. Hunting for quantum-classical crossover in condensed matter problems. npj Quantum Information 10, 45, 2024. arXiv:2210.14109.
Run it yourself with eigenkit
eigenkit is EigenOrbit's open-source Python library: write a circuit once, run it on any simulator or QPU, and test it against a classical baseline on your own data before you invest in it or rule it out.
$ pip install eigenkit