Research collaboration — universities, semi-states, EU programmes

Why a sovereign machine changes the collaboration model

Most Irish quantum research today runs on remote cloud queues — IBM Quantum, AWS Braket, Azure Quantum. That is fine for algorithm development and for small-circuit experiments, but it has three structural limits for serious work. First, queue depth: a researcher running a variational eigensolver on a 27-qubit shared device can wait hours between iterations, which kills the inner optimisation loop. Second, calibration drift: when you do not control the cryostat schedule, you do not know which T1/T2 regime your shots actually landed in, so reproducing a result six weeks later is hard. Third, data sovereignty: pre-publication chemistry Hamiltonians and grid-topology data for Irish semi-states should not be sitting in a US-jurisdiction queue.

Ireland Quantum 100 is being built in Co. Tipperary specifically to close those three gaps for Irish quantum research. A 100-physical-qubit superconducting transmon machine on a heavy-hex lattice, sitting in a dilution refrigerator at sub-15 mK, with a published calibration cadence and an Irish-jurisdiction data plane. That is the substrate the collaboration model is built on.

What universities actually get at the gate level

For a research group, the useful unit is not "access to a quantum computer" — it is gate fidelity, coherence time, connectivity, and shot rate. The IQ100 system targets the engineering envelope you would expect from a current-generation transmon device: single-qubit gates in the tens of nanoseconds, two-qubit echoed cross-resonance or tunable-coupler gates in the low hundreds of nanoseconds, T1 and T2 in the order of one to two hundred microseconds on calibrated qubits, and readout in the high-90s percent on a good day. Heavy-hex connectivity means each qubit has degree two or three, which constrains compilation but also reduces frequency-collision and crosstalk problems compared to a denser lattice.

For a Ph.D. student writing in OpenQASM 3 or driving the device through Qiskit, PennyLane, or Cirq, what this means in practice is that you can plan a SWAP-aware transpilation pass, budget your circuit depth against the two-qubit gate error, and know roughly how many shots you need before the noise eats the signal. We are publishing the calibration data — qubit-by-qubit T1, T2, single- and two-qubit gate errors, readout assignment matrices — on a documented cadence so groups can cite a specific calibration epoch in a paper and another group can reproduce against the same epoch later.

The collaboration tiers

University research access

The intent for Irish university groups is straightforward: dedicated reservation windows on the device, not best-effort cloud queue. A group running a UCCSD ansatz on a small molecular Hamiltonian, or a QAOA instance on a graph problem, gets a block of wall-clock time on a known calibration. That lets them close the variational loop tightly, run noise-extrapolation experiments without queue jitter, and treat the device as an instrument rather than a service. We expect the priority cohort to come from chemistry, condensed-matter, and computer-science departments working on quantum error correction, variational methods, and quantum machine learning.

Semi-state and applied research

Ireland's semi-states have problems that map well to early fault-tolerant and noisy intermediate-scale quantum work: grid optimisation under increasing renewable penetration, demand-response scheduling, water-network flow optimisation, transport routing. None of these need a million qubits to be useful — many of them are interesting at the 50–100 logical-equivalent scale once you push circuits through error mitigation. The collaboration model here is joint problem definition: a semi-state brings the operational data and the loss function, a university group brings the algorithm, and IQ100 provides the hardware and the compilation stack.

EU programmes and the wider European stack

The European Quantum Flagship, the EuroHPC quantum-integration programme, and the Digital Europe quantum-computing pillar are all funding work that needs hardware access. An Irish-resident 100-qubit superconducting machine is a credible node in that landscape, particularly for consortia that want their hardware partner inside the EU jurisdiction. We are designing the access stack to be compatible with the European federated-quantum-computing model: standard authentication, OpenQASM 3 ingress, results returnable through Qiskit Runtime-style primitives so that existing Horizon Europe codebases do not need a rewrite to retarget.

Climate chemistry as the anchor workload

The priority cohort is climate science, and the technical reason is that the chemistry is genuinely hard for classical methods and genuinely tractable for near-term quantum hardware. Carbon-capture amines and metal-organic frameworks, perovskite photovoltaic candidates, lithium-sulphur and sodium-ion battery electrolyte chemistry, nitrogenase-style catalysts for ammonia synthesis — these are all problems where the strongly-correlated electron structure makes density-functional theory unreliable and full configuration interaction intractable beyond about 20 spin-orbitals.

The mapping is well-understood: Jordan-Wigner or Bravyi-Kitaev encoding of the second-quantised Hamiltonian, active-space reduction down to the chemically relevant orbitals, then a hardware-efficient or unitary-coupled-cluster ansatz run as a VQE on the device. With 100 physical qubits and aggressive error mitigation — zero-noise extrapolation, probabilistic error cancellation, measurement-error mitigation — useful active spaces of 30 to 50 spin-orbitals come into reach. That is not a million-molecule screen, but it is enough to settle disputed transition-state energies on candidate molecules that the classical screen has already shortlisted.

Timeline and what access looks like by quarter

The delivery is twelve months. Q3 2026 is site fit-out in Co. Tipperary — power, cooling, vibration isolation, the dilution refrigerator support frame. Q4 2026 is the cryostat install and cooldown to base temperature. Q1 2027 is first-light: single-qubit characterisation, T1/T2 mapping, single-qubit gate calibration across the lattice. Q2 2027 is multi-qubit operation — two-qubit gate calibration, randomised benchmarking, simultaneous-gate crosstalk characterisation — and the opening of the first customer and research-collaborator access windows. Surface-code experiments on small patches are a 2027–2028 roadmap item, not a launch claim.

What we are doing about this

If you lead a research group, a semi-state R&D function, or an EU consortium