Long-Horizon Simulation Challenges
Why simulating dynamics for long times is fundamentally hard, and what limits both classical and quantum approaches.
The problem of long times
Many questions in physics, plasma confinement over many turbulence correlation times, chemical reactions over long timescales, transport in materials, require evolving a system for a long total time t. Both classical and quantum methods face growing cost as t increases, though for different reasons.
The quantum cost of time
For quantum simulation, gate count grows at least linearly in t; the optimal Hamiltonian-simulation algorithms scale as O(t + log(1/epsilon)), and this linear-in-t term is provably unavoidable, a no-fast-forwarding theorem. You cannot generically simulate long evolution in sublinear time, because doing so would let you solve problems known to be hard. Long horizons therefore mean many gates and long coherence.
Error accumulation
- On noisy hardware, errors compound with circuit depth, capping usable t.
- Trotter error accumulates across steps unless step size shrinks.
- Fault tolerance overhead grows with total gate count, hence with t.
- State-preparation and measurement errors add on top.
The classical comparison
Classically, entanglement often grows with time, so tensor-network methods that work at short times become intractable once entanglement saturates the bond dimension. Chaotic and turbulent systems are the hardest, because small errors amplify. This is precisely where a quantum advantage might appear, but only if the quantum machine can maintain coherence long enough.
The special obstacle for chaotic and nonlinear systems
Plasma turbulence and other chaotic dynamics combine long horizons with nonlinearity and exponential sensitivity to initial conditions. Quantum simulation handles linear dynamics well but nonlinear systems only indirectly, and long horizons magnify every source of error. This combination is why long-time kinetic plasma simulation is among the most demanding targets and remains, for now, out of reach for both paradigms at full fidelity. Realistic progress will likely come from linearized subproblems and hybrid classical-quantum schemes rather than brute-force long evolution.