Agriculture and photosynthesis — synthetic chlorophyll on a quantum machine

Why photosynthesis is a quantum problem, not a classical one

Photosynthesis runs at energy-transfer efficiencies that classical chemistry cannot fully explain. The Fenna–Matthews–Olson (FMO) complex in green sulphur bacteria, and the analogous light-harvesting complexes in higher plants, route excitons from antenna chlorophylls to reaction centres with quantum yields above 95%. The mechanism involves vibronic coupling between electronic and vibrational states, with coherence lifetimes on the order of hundreds of femtoseconds at biological temperatures. Modelling this on a classical machine means truncating the Hilbert space, approximating the bath, and accepting that you cannot resolve the very effects that make the process efficient.

For quantum agriculture work, this matters because every percentage point of light-harvesting efficiency translates directly into yield per hectare per unit of solar flux. A faithful simulation of the chlorophyll-a/b binding protein complex (LHCII), with its 14 chlorophylls and 4 carotenoids per monomer, is a problem that scales badly: the electronic Hamiltonian alone, even truncated to single-excitation manifolds, pushes 10⁴–10⁵ basis states once you include the vibrational modes that drive the energy transfer. That is squarely in the regime where a 100-qubit superconducting transmon machine, properly programmed, has something to say.

What a 100-qubit transmon machine can actually do here

Let's be honest about what 100 physical qubits — pre-error-correction — can and cannot do. With a heavy-hex topology, two-qubit gate fidelities in the 99.0–99.5% range, and circuit depths bounded by T₂ coherence (typical transmon T₂ sits in the 100–300 µs band), you are not going to run a full FMO simulation under fault tolerance in 2027. You are going to run useful, falsifiable approximations.

The realistic 2027 workload set for quantum photosynthesis on Ireland Quantum 100 looks like this:

• Variational Quantum Eigensolver (VQE) on reduced active spaces of synthetic chlorophyll analogues — typically 8–20 qubits for the electronic structure, with the rest of the register absorbed into ansatz depth and error-mitigation overhead.
• Quantum subspace expansion for excited-state energies, which is what you actually need to design a chlorophyll variant: ground-state energy alone tells you nothing about absorption spectra.
• Trotterised dynamics of exciton transport on small chromophore networks (4–8 sites), with engineered dephasing to study environment-assisted transport.
• Quantum imaginary-time evolution for thermal-state preparation at biological temperatures, which classical methods handle badly when coherence is non-negligible.

All of this is expressible in OpenQASM 3 and runs through Qiskit or PennyLane. Nothing exotic — the engineering challenge is in the error-mitigation stack (zero-noise extrapolation, probabilistic error cancellation, Clifford data regression) and in the classical co-processor that drives the VQE outer loop.

Synthetic chlorophyll: the design target

Natural chlorophyll-a absorbs strongly in the blue (~430 nm) and red (~662 nm), with a green gap that is, in agronomic terms, wasted photons. Roughly half the photosynthetically active radiation reaching a leaf canopy is reflected or transmitted rather than absorbed by the dominant pigment. Synthetic biology quantum work is interested in chlorin and bacteriochlorin scaffolds with shifted Q-band absorption — pulling the red edge out to 720–750 nm, or filling the green gap with engineered carotenoid-chlorophyll hybrids.

The design problem is: given a candidate macrocycle with substituent modifications at the meso and β-pyrrolic positions, what is the absorption spectrum, the excited-state lifetime, and the redox potential of the lowest singlet? Classical TDDFT gets you the spectrum to maybe 0.2–0.3 eV accuracy, which is not good enough to discriminate between candidates that differ by 20–30 nm in absorption maximum. Multireference methods (CASPT2, NEVPT2) do better but choke on the active spaces required for full porphyrin-class molecules. A quantum-classical hybrid where the strongly correlated π-system is handled on the QPU, embedded in a classical mean-field environment, is the path that actually scales.

Crop genetics and the rubisco problem

Photosynthesis is rate-limited not by light harvesting in most C3 crops but by RuBisCO — ribulose-1,5-bisphosphate carboxylase/oxygenase — which fixes CO₂ and oxygen indiscriminately, with the oxygenation reaction producing the wasteful photorespiration pathway. Crop genetics quantum workloads are interested in two angles here: simulating the active-site chemistry of natural and engineered RuBisCO variants to improve carboxylation specificity, and exploring carbon-concentrating mechanisms that can be transplanted from C4 plants and cyanobacteria into C3 staples.

Both reduce, computationally, to transition-state energy calculations on metalloenzyme active sites — the Mg²⁺ coordination in RuBisCO, the carbamylated lysine, and the substrate enediol. These are exactly the multireference, transition-metal-containing systems where classical quantum chemistry is least reliable and where a QPU has the cleanest theoretical advantage. Active-space sizes of 20–40 spatial orbitals are tractable on Ireland Quantum 100 with appropriate compression schemes (qubit tapering, point-group symmetries, locality-aware encodings).

The engineering path on Ireland Quantum 100

The hardware timeline is fixed. Site fit-out at Annerpark begins Q3 2026, the dilution refrigerator goes in during Q4 2026 with base-temperature operation below 15 mK, single-qubit characterisation runs through Q1 2027, and multi-qubit access opens to the climate-priority cohort in Q2 2027. Agriculture and photosynthesis workloads sit inside that cohort.

The software stack that lands with the machine supports OpenQASM 3, Qiskit Runtime-compatible primitives, PennyLane for differentiable quantum circuits, and Cirq for users who prefer it. The chemistry-