Integration with IMPT — how the quantum machine reinforces the carbon stack

Why the quantum machine sits next to the carbon stack

IMPT.io operates a software stack that audits, scores and re-bundles voluntary carbon offsets. The hard problem in that stack is not blockchain or accounting — it is chemistry and earth-system uncertainty. How much CO₂ does a given mineralisation pathway actually fix per kilogram of feedstock? How stable is a soil-carbon claim under realistic microbial dynamics? How does a direct-air-capture sorbent degrade after ten thousand cycles? Each of these questions reduces, eventually, to a many-body electronic-structure problem or a coupled stochastic PDE that classical HPC solves only approximately and at enormous cost.

Ireland Quantum 100 — the sovereign 100-physical-qubit superconducting transmon machine being commissioned in Co. Tipperary across the next twelve months — is being built explicitly to attack the chemistry layer underneath that stack. This page sets out the engineering position on how the two systems connect.

The chemistry workloads that actually move the needle

For a 100-qubit heavy-hex transmon device with realistic two-qubit gate fidelities in the 99.0–99.5% range pre-error-correction, the workloads worth running are not full-protein simulation and not Shor at scale. They are tightly-scoped electronic-structure problems on active spaces of roughly 30–60 spin-orbitals, where Variational Quantum Eigensolver (VQE) and Quantum Phase Estimation (QPE) variants give a credible advantage path against CCSD(T) and DMRG once noise is properly handled.

The candidate carbon-relevant problems we are queueing for first-light science include:

• Amine and amino-acid CO₂ binding energetics. Reaction-energy surfaces for monoethanolamine, piperazine and 2-amino-2-methyl-1-propanol carbamate formation, where classical DFT struggles with dispersion and proton-transfer transition states.
• Metal-organic framework active sites. Open-shell transition-metal centres (Cu, Fe, Mn) in MOF-74-class frameworks where multireference character defeats single-reference methods.
• Mineralisation kinetics. Olivine and basalt dissolution intermediates, including Mg²⁺ and Fe²⁺ coordination shells under varying pH.
• Photocatalytic water-splitting and CO₂-reduction intermediates on TiO₂, BiVO₄ and Cu-based catalysts — directly relevant to synthetic-fuel offset pathways.

None of these is a marketing claim. Each is a problem with a published quantum-chemistry track record on smaller devices (IBM, Quantinuum, IQM), where the bottleneck has been qubit count and circuit depth rather than algorithmic novelty.

The integration architecture between IMPT and Ireland Quantum 100

The two systems are deliberately decoupled at the runtime layer. The quantum machine exposes an OpenQASM 3 endpoint with Qiskit, PennyLane and Cirq transpiler support; jobs are submitted, queued and returned as classical result objects. IMPT's offset-scoring pipeline does not call the quantum backend in any synchronous path — that would be reckless given current cycle times and queue depths.

Instead, the integration runs as a chemistry knowledge-base feed:

• Quantum-chemistry runs produce ground-state energies, reaction enthalpies, activation barriers and dipole moments for specific catalyst/sorbent candidates.
• Results are written to a versioned chemistry KB with full provenance: ansatz used (UCCSD, hardware-efficient, ADAPT-VQE), active space, qubit mapping, error-mitigation strategy (zero-noise extrapolation, probabilistic error cancellation, readout twirling), and shot count.
• IMPT's offset-evaluation models read from that KB to refine per-tonne durability and additionality scores for projects whose underlying chemistry has been simulated.

This is the operationally honest version of "quantum carbon offset" — the chain of evidence from a transmon at sub-15 mK in Tipperary to a tonne of CO₂ on an offset registry is several layers deep, and we do not want a marketing arrow that pretends otherwise.

Error-correction roadmap and what it means for delivery dates

The hardware path is the standard one. Sub-15 mK base-plate temperature in a dilution refrigerator, fixed-frequency transmons on a heavy-hex lattice, dispersive readout via Purcell-filtered resonators, and a control stack capable of arbitrary-waveform pulses with sub-nanosecond timing. The 12-month delivery window breaks down as: Q3 2026 site fit-out and shielding; Q4 2026 cryostat install and wiring; Q1 2027 first-light single-qubit characterisation (T1, T2, gate fidelity, readout assignment); Q2 2027 multi-qubit benchmarking and the first external-user access cohort.

Surface-code error correction is on the roadmap, not the delivery. A 100-physical-qubit machine is below the threshold for a useful logical qubit at distance-5 surface code (which wants on the order of 49 data qubits plus ancillas per logical qubit, at error rates well below 1%). The honest framing for the first eighteen months of operation is noisy intermediate-scale quantum (NISQ) chemistry with aggressive error mitigation, with logical-qubit demonstrations targeted as a research milestone rather than a service line.

That distinction matters for IMPT integration. The chemistry problems queued for the NISQ phase are deliberately chosen to be tolerant of finite shot noise and bias-mitigatable error — single-point energy differences with cancellation of correlated errors, rather than absolute energies that demand chemical accuracy in one shot.

Climate cohort priority and queue policy

The Ireland climate stack rationale is straightforward: the machine is being built in Ireland, with Irish capital, and the first cohort of users is climate-science workloads. That is a deliberate procurement choice, not a slogan. Carbon-capture chemistry, photovoltaic materials discovery, battery cathode and electrolyte modelling, climate-relevant protein folding (e.g. carbonic anhydrase variants for industrial CCS), and grid-optimisation problems framed as QAOA instances are all in the priority lane.

The queue policy for the first