On-chain offset verification meets quantum chemistry

Carbon offsets have a credibility problem, and the fix is not another registry. The fix is being able to prove, computationally, that a tonne of CO₂ a project claims to remove or avoid actually corresponds to chemistry, biology, or physics that survives independent simulation. That is where quantum chemistry meets the blockchain layer — not as a marketing pairing, but as two computational substrates that, used together, can put real physics behind a credit and an immutable record behind the proof.

Why offset verification keeps failing the technical test

Most offset methodologies in active use are spreadsheet exercises wrapped in PDF. A direct-air-capture project models its sorbent regeneration energy from vendor literature. A reforestation project uses allometric equations from the 1990s. A blue-carbon project assumes sediment burial rates from one peer-reviewed paper applied to a coastline it never measured. The registry signs off. The credit lists. The buyer retires it.

The breakage is not malice — it is that the underlying chemistry, soil dynamics, or ocean alkalinity is genuinely hard to model classically. Density functional theory at useful accuracy for a novel amine sorbent costs you weeks on a CPU cluster, and even then DFT functionals systematically under- or over-bind certain bonds. So projects ship with assumed numbers and registries trust the assumptions. When journalists pull the thread later, the credits unravel.

The honest engineering answer is: if you cannot simulate the binding energy, the regeneration cycle, or the catalyst turnover at chemical accuracy, you cannot verify the credit. You can only believe it.

Where quantum changes the chemistry budget

Superconducting transmon qubits, operated in a dilution refrigerator at sub-15 mK, are not a general-purpose accelerator. They are very specifically good at a class of problems where the wavefunction is the answer — strongly correlated electronic structure, transition-metal catalysis, and certain conformational searches in protein folding. These are exactly the problems that sit at the bottom of every offset methodology a buyer is asked to trust.

The interesting near-term workload is the variational quantum eigensolver, or VQE, running against an active space carved out of the molecule of interest. You leave the inactive electrons to a classical Hartree-Fock or DFT scaffold, you map the active orbitals to qubits via Jordan-Wigner or Bravyi-Kitaev, and you let a parameterised circuit hunt the ground-state energy. With a 100-qubit transmon machine on a heavy-hex topology, you have enough room to handle active spaces that genuinely matter for industrial sorbents and small catalysts — not the hydrogen-molecule demos of a decade ago.

None of this requires fault tolerance to be useful. It does require honest noise characterisation, error mitigation — zero-noise extrapolation, probabilistic error cancellation, readout calibration — and a willingness to say when a result is not converged. The surface-code roadmap matters for the decade ahead; the chemistry that anchors today's offset claims can be attacked with what we are building now.

What "on-chain" actually has to mean

The phrase on chain quantum gets used loosely. Let me be precise about what is and is not useful.

Putting a full quantum circuit execution log on a public chain is pointless — the data is enormous, the chain is the wrong storage layer, and nobody will replay it. What you want on chain is a verification artefact: a content-addressed hash of the circuit, the transpilation target, the calibration data of the device on the day, the measured expectation values, and the post-processed energy. That artefact is small, it is signed, and it lets a third party — months or years later — fetch the underlying data from cold storage and confirm the run actually happened on the hardware claimed.

The blockchain layer is doing one job here: ordering and immutability. It is not doing the chemistry. It is making it impossible for an offset issuer to quietly swap out the simulation that justified the credit. Combined with attestation from the quantum facility itself — the device serial, the fridge temperature log, the qubit T1 and T2 coherence times during the run — you get a verification chain that an auditor can actually walk.

This is the workable definition of quantum offset verification: the credit references a hash, the hash resolves to a chemistry computation, the computation references a specific quantum device run, and the device run references calibration data the operator cannot retroactively edit.

A worked example: amine sorbent regeneration

Take a direct-air-capture operator using a proprietary amine sorbent. The credit they want to issue depends on three numbers: how much CO₂ binds per cycle, how much energy it costs to regenerate, and how many cycles before the sorbent degrades. The first two are quantum-chemistry questions. The third is materials-degradation, which is harder, but the first two are tractable now.

The workflow looks like this. The operator's chemistry team builds the sorbent–CO₂ adduct geometry classically. They identify the active space — typically the nitrogen lone pair, the C=O π system, and the relevant σ framework, maybe twenty to forty spin orbitals depending on the amine. They submit a job to the quantum facility specifying the Hamiltonian, the ansatz (UCCSD or a hardware-efficient variant), the optimiser, and the error mitigation strategy. The facility runs it, returns the energy with uncertainty bands, and writes a signed manifest.

The manifest goes on chain. The credit issuer references it. An auditor — academic, regulator, or buyer's technical team — can pull the manifest, fetch the raw counts from the facility's archive, re-run the classical post-processing, and confirm the energy. If the operator later changes their methodology, the chain shows it. If a new functional or a deeper active space changes the answer, the chain shows that too. Credits become re-verifiable rather than re-believed.

This is what we mean when we talk about a quantum carbon credit: not a credit issued by a quantum computer, but a credit whose physical claim is anchored in a quantum simulation that anyone can re-fetch and re-check.

Architecture: where the pieces actually live

The deployment we are building in Tipperary is sovereign — the cryostat, the control electronics, and the data plane sit in one facility under Irish jurisdiction. That matters for offset verification because the calibration data and the raw shot counts cannot be useful as audit evidence if they live in a black-box cloud account in another regulatory regime. The full Ireland Quantum 100 programme is built around the principle that the operator can show the auditor the fridge.

The software stack is deliberately conventional: OpenQASM 3 as the circuit interchange format, Qiskit and PennyLane as the front-end SDKs, Cirq supported for teams already invested in it. Nothing about the verification model requires exotic tooling. A chemistry team that already runs PySCF or OpenFermion classically can extend their workflow into the active-space-on-quantum pattern without rewriting their pipeline.

The on-chain side is where teams over-engineer. You do not need a bespoke L1. You need a public chain with predictable finality, a smart contract that ingests signed manifests, and an off-chain content store — IPFS, S3 with object lock, or an institutional archive — for the bulk data. The chain holds the hash and the signature. Everything else lives where storage is cheap.

What this changes for IMPT and for the offset market

IMPT's offset stack today aggregates supplier projects across reforestation, blue carbon, biochar, mineralisation, and engineered removal. The supplier-quality work is already serious, but the technical due-diligence ceiling is bounded by what classical chemistry can verify. As the Tipperary facility comes online through 2027, candidate suppliers in the engineered-removal and mineralisation tranches become re-evaluable at chemical accuracy. That is not a marketing claim — it is a workflow change. Suppliers whose numbers survive a VQE re-check on their actual sorbent or actual mineral pathway get treated differently from suppliers whose numbers came from a vendor PDF.

For the broader market, blockchain quantum verification is not going to replace registries — registries do legal and contractual work that computation cannot. What it can do is force registries to upgrade their evidence standards. Once a few credit classes exist with verifiable on-chain quantum-chemistry backing, the credits without any anchored physics start trading at a discount, and that discount is the market doing the work that auditors have struggled to do.

Where to start this week

If you run an offset project with engineered-removal or mineralisation chemistry at its core, do one thing this week: write down the three to five numbers your credit actually depends on, and next to each, write down where the number came from. If any of them trace back to "vendor literature" or "assumed from analogous compound", you have identified the workloads that need re-anchoring. Bring that list to a chemistry team that can scope an active-space calculation. The hardware to run it at chemical accuracy is being built — the question is whether your methodology is ready to be verified when it arrives.