194
Part Five · Convergence

Simulating tunneling in enzymes on a quantum device

Chapter 164 described hydrogen tunnelling in enzymes. Can a quantum computer simulate it? The idea is to store the particle’s position on a grid, write the grid index in binary across qubits, and evolve the wavefunction with gates. Four qubits give a 16-point grid, enough to see a particle tunnel through a barrier. This chapter runs that simulation in the book’s simulator, compares it with the exact answer, and asks when doing it on a quantum computer would ever be worthwhile.

On the plate: a particle tunnelling in a double well, on 16 grid points.

Grid encoding

Discretize position into 2n2^n points and encode point jj as the basis state ∣j⟩|j\rangle of nn qubits: amplitudes on a grid are amplitudes of a quantum state. The potential energy V(x)V(x) is diagonal in this basis, so e−iVΔte^{-iV\Delta t} is a set of phase rotations computed from the binary digits. The kinetic energy is diagonal in momentum, reached with the quantum Fourier transform (Chapter 112). So one Trotter step (Chapter 118) is: phases for VV, QFT, phases for p2/2mp^2/2m, inverse QFT. This is the split-operator method of Chapter 31, run on qubits. Stephen Wiesner and Christof Zalka proposed it in 1996–98. Established

Run it

Barrier height 2.5 and Trotter steps per time unit 4. The particle starts in the left well; the plate plots the probability of finding it on the right, from exact diagonalization of the 16 × 16 Hamiltonian (white) and from the Trotterized circuit (amber). The probability oscillates at the tunnelling splitting, which shrinks exponentially as the barrier rises. Too few Trotter steps and the amber curve drifts away; more steps converge to the exact answer, at proportionally more gates. Established

The critique

A single proton in one dimension is trivial classically. Realistic enzyme tunnelling involves the proton coupled to many protein and solvent modes; classical methods handle that well too, through ring-polymer and path-integral molecular dynamics, multiconfigurational time-dependent Hartree (MCTDH) for tens of coupled modes, and semiclassical instanton theory. A quantum computer would only help if very many coupled quantum degrees of freedom needed exact treatment at once, and the grid for dd dimensions needs d×nd\times n qubits, cheap in qubits but costly in gates and precision. No current study shows an advantage for enzyme tunnelling. Speculative as an advantage; toy demonstrations work.

What you now have

Binary grid encoding of wavefunctions, diagonal potential phases, the QFT-based kinetic step, a complete 4-qubit tunnelling simulation checked against exact dynamics, Trotter convergence, and the classical methods that make an advantage unlikely.

· · ·

Three questions to carry with you.

How is the kinetic energy applied on qubits?tap
QFT to momentum space, diagonal phases, inverse QFT.
What sets the tunnelling oscillation period?tap
The energy splitting between the lowest two states, which shrinks exponentially with barrier height.
Which classical methods compete?tap
Path-integral MD, MCTDH and instanton theory.

And one to work out. How many grid points do 10 qubits encode?

points =

Sources: S. Wiesner, arXiv:quant-ph/9603028 (1996). C. Zalka, Proc. R. Soc. A 454, 313 (1998). H.-D. Meyer, U. Manthe and L. S. Cederbaum, Chem. Phys. Lett. 165, 73 (1990) (MCTDH). Checked by the plate: exact eigen-decomposition of the 16-point Hamiltonian and an explicit split-step circuit using the discrete Fourier transform.

Next · Chapter 195
Simulating photosynthetic energy transfer as an open quantum system

Quantum hardware has already been built to imitate a leaf's energy transfer, and noise helped transport.

Turn the page