Silicon photovoltaics took roughly seventy years to climb from a curiosity in a Bell Labs notebook to the cheapest source of bulk electricity on the planet. Perovskites — the lead-halide and tin-halide hybrids that have dominated PV research papers for the last fifteen years — won't get seventy years. The grid wants them now. The chemistry, however, is a mess: hundreds of thousands of plausible compositions, each with a defect landscape and a degradation pathway that classical density-functional theory either approximates poorly or doesn't touch at all. This is exactly the workload a 100-qubit superconducting machine in Co. Tipperary is being built for.
Why perovskites are a quantum-shaped problem
A methylammonium lead iodide unit cell is small. The interesting physics isn't. You have spin-orbit coupling from the heavy lead atom, dynamic disorder from the rotating organic cation, polaron formation that couples the electronic structure to lattice phonons, and a defect chemistry where a single iodine vacancy can dominate the device's open-circuit voltage. None of those phenomena live cleanly inside a mean-field approximation.
Classical methods get you to the door. Hybrid functionals like HSE06 give reasonable bandgaps once you tune the mixing parameter — which you tune, of course, by fitting to the answer you're trying to predict. GW and Bethe-Salpeter give better excitonic properties at a cost that scales as the sixth or seventh power of system size. For a single static cell at zero kelvin, fine. For a mixed-cation, mixed-halide composition with thermal disorder and a grain-boundary defect, you've already lost the afternoon and most of the cluster.
Quantum chemistry on a quantum computer doesn't make this trivial — anyone telling you it does is selling something — but it does change which approximations you're forced to make. The strongly correlated subspace, typically the frontier orbitals around the bandgap, can in principle be solved on hardware. The rest stays classical. This is the active-space embedding picture that's emerged across the field, and it's the picture our machine is being commissioned around.
The hardware that has to do the work
Ireland Quantum 100 is a 100-physical-qubit superconducting transmon machine. Transmons because the fabrication is mature, the gate set is well-understood, and the field's tooling — Qiskit, PennyLane, Cirq, OpenQASM 3 — assumes them as the default. The chip lives at the bottom of a dilution refrigerator below 15 millikelvin, which is colder than interstellar space and considerably more annoying to maintain.
The topology is heavy-hex. That choice trades raw connectivity for syndrome-extraction simplicity: when you eventually layer a surface code on top, the heavy-hex layout maps cleanly to the stabiliser measurements, and the data/measure qubit assignments don't fight the chip geometry. For chemistry workloads in the near term, before logical qubits are routine, you live inside the noise. That means short, hardware-aware circuits — variational ansätze that respect the coupling map, error mitigation via zero-noise extrapolation and probabilistic error cancellation, and a brutally honest accounting of what the noise floor is doing to your expectation values.
None of this is hypothetical architecture choice. It's what the field has converged on, because every other choice is worse for the same money.
From a perovskite formula to a quantum circuit
The pipeline for photovoltaic material discovery on near-term quantum hardware looks roughly like this, and it is worth being concrete about each step because the gap between "we ran something on a quantum computer" and "we learned something about a solar cell" is where most projects quietly die.
- Geometry and classical pre-processing. Build the unit cell, relax it with a classical DFT code, identify the orbitals that matter — typically the lead 6p, the halide np, and any defect states pinned in the gap.
- Active-space selection. Project the full Hamiltonian down to that frontier subspace. For a clean bulk cell you might get away with eight to twelve spatial orbitals. For a defect or interface you want more, and you immediately bump into the qubit budget.
- Fermion-to-qubit mapping. Jordan-Wigner is the textbook choice; Bravyi-Kitaev and parity mappings give better locality on hardware that has a constrained coupling map. Heavy-hex pushes you toward mappings that don't demand all-to-all SWAP networks.
- Ansatz and solver. Hardware-efficient ansätze for raw VQE; ADAPT-VQE when you can afford the gradient measurements; quantum Krylov and quantum subspace expansion methods when you need excited states, which for photovoltaics you always do — the bandgap is an excited-state question.
- Error mitigation and post-processing. Symmetry verification, readout-error mitigation, ZNE. Then back to the classical workflow to feed the corrected energies into a transport or device-level model.
The honest framing: a near-term machine is not going to simulate a full perovskite solar cell. It's going to give you better numbers for the small, hard piece in the middle that classical methods botch — the strongly correlated active space around the gap and around the defects. Everything else stays where it already lives.
What "next-gen solar" actually means in this context
The phrase next-gen solar gets used for tandem cells, for thin-film flexible modules, for building-integrated PV, for any number of things. For our purposes it means three concrete material families where quantum simulation has the best shot at telling us something classical methods can't.
First, mixed-cation lead-halide perovskites with controlled disorder. The state of the art device stack uses formamidinium-caesium mixes with bromide-iodide tuning, and the question of how composition controls phase stability under operating temperature is fundamentally a free-energy question with strong correlation contributions.
Second, lead-free alternatives. Tin-halide perovskites, double perovskites with bismuth or antimony, chalcogenide perovskites. The motivation is regulatory and environmental — getting lead out of a roof-deployed module — but the materials are less well-characterised, the spin-orbit coupling is different, and the defect chemistry is wide open.
Third, perovskite-silicon tandems. The classical modelling here is reasonable for the silicon bottom cell and the recombination layer; the quantum-relevant piece is the perovskite top cell's behaviour at the interface, where strain and band alignment are sensitive to electronic-structure details that survive the active-space reduction.
Across all three, the workload pattern is the same: thousands of candidate compositions and defect configurations, each one a small but classically painful electronic-structure problem. That's a screening problem, and screening is where quantum-classical hybrid pipelines have a credible near-term story. More on the broader workload mix lives on the climate workloads page.
What we're not promising
A 100-qubit transmon machine, even a well-calibrated one, does not give you fault-tolerant chemistry. The error rates are what they are; the circuit depths you can run before decoherence eats the signal are what they are. Any vendor — including us — telling you they'll be solving production-grade material-discovery problems on bare physical qubits in 2027 is overselling.
What it does give you, and this is the genuinely useful part, is a real platform to develop the algorithms, the embedding workflows, the mitigation stacks, and the classical-quantum handshake that the fault-tolerant era will run on. Groups that wait for logical qubits to start writing perovskite-quantum codes will find themselves three years behind groups that started on noisy hardware. The software stack is where the moat is, and the software stack is built by running real workloads on real machines, badly at first.
The other thing it gives us, specifically, is sovereignty. An Irish climate-research group that wants to run a perovskite screening campaign today is queueing on someone else's hardware in someone else's jurisdiction under someone else's terms of service. That is not a tenable position for the next twenty years of climate-relevant compute.
Where to start this week
If you're a materials group, a PV company, or a chemistry PhD with an interest in quantum solar work: install Qiskit and PennyLane locally, take a small perovskite cluster you already understand classically — a single octahedron, a defect dimer, something with eight to twelve relevant orbitals — and run it through an active-space VQE on a simulator with a realistic noise model. Get the embedding pipeline working end-to-end on a problem where you already know the answer. When the Tipperary machine comes online for external workloads, the groups that have done this exercise will be the ones who get useful science out of the first cohort. If you want to be in that cohort, the Ireland Quantum 100 page is where the access process will be published as it firms up.