Exchange Interaction

The exchange interaction originates from the requirement that the total wavefunction of two identical fermions be antisymmetric under particle exchange, a consequence of the Pauli exclusion principle. When two electrons are confined in adjacent quantum dots and the tunnel barrier between them is lowered, their wavefunctions overlap and the antisymmetry constraint forces a correlation between their spin and spatial degrees of freedom. The effective spin Hamiltonian describing this interaction is the Heisenberg exchange Hamiltonian H = J * (S1 . S2), where J is the exchange coupling energy and S1, S2 are the spin operators of the two electrons. The sign and magnitude of J depend on the orbital overlaps and can be tuned over many orders of magnitude by adjusting the voltage on electrostatic gates that control the tunnel barrier height.

Time-evolving under the Heisenberg Hamiltonian with a controlled J(t) pulse generates two-qubit gates. Integrating J(t) over time to a value of h/4 produces the sqrt(SWAP) gate, and two sequential sqrt(SWAP) operations produce a full SWAP. Combined with single-qubit rotations, sqrt(SWAP) is a universal two-qubit gate. Alternatively, sequences of exchange pulses and single-qubit operations can synthesize CNOT gates. The Loss-DiVincenzo proposal in 1998 first outlined this scheme for spin qubits in quantum dots, and it remains the leading two-qubit gate mechanism for silicon and GaAs spin qubit platforms. Gate times are typically in the nanosecond to microsecond range, much faster than the qubit coherence times in the best silicon devices, giving exchange-based gates a favorable duty cycle.

The fidelity of exchange gates is limited primarily by charge noise, the dominant noise source in semiconductor quantum dots. Because J depends sensitively on the tunnel barrier height, which is set by gate voltages, any fluctuation in those voltages directly modulates J and introduces phase errors on the qubits. Operating at sweet spots where dJ/dV = 0, using dynamical decoupling sequences, and encoding logical qubits in decoherence-free subspaces are all strategies employed to mitigate this coupling. Silicon spin qubits benefit from a lower density of charge noise compared to GaAs because silicon can be isotopically purified to remove magnetic 29Si nuclei, eliminating hyperfine-driven dephasing and pushing T2* times into the millisecond range.

Comparing exchange-based gates to other leading platforms highlights distinct trade-offs. Superconducting qubit two-qubit gates, such as the cross-resonance gate used by IBM, are driven by microwave-frequency control electronics and achieve gate times of 100-400 ns with fidelities above 99%, but require millikelvin cryogenic cooling and are sensitive to flux noise. Trapped-ion platforms use the Molmer-Sorensen gate, driven by laser or microwave pulses that excite a shared motional mode, achieving gate fidelities above 99.9% with gate times in the range of microseconds to milliseconds. Exchange-based spin qubit gates occupy a middle ground: they are fast, operate at millikelvin temperatures alongside superconducting control circuitry, and benefit from the mature semiconductor fabrication industry, which raises hopes for large-scale integration that may be harder to achieve with ion traps or superconducting circuits.