For small systems, optimal controls can often be found exactly. Their dynamics lives on a Lie group (unitaries) or a sphere (states), the optimal-control conditions of Pontryagin's maximum principle become geometric statements, and optimal trajectories turn out to be geodesics of an appropriate metric. Geometric control provides these exact solutions, the minimal times they imply, and the qualitative structures — bang-bang, singular and geodesic arcs — that numerical methods then rediscover in larger systems.

Time-optimal control

The most common geometric problem asks for the shortest time in which a target can be reached under constraints on the controls. Two regimes are particularly clean.

Bounded controls on a single spin or qubit. With a drift and control amplitudes limited by \(u_{\max}\), the maximum principle shows that optimal controls are concatenations of bang arcs (\(u = \pm u_{\max}\)), possibly singular arcs, and off periods. The complete “optimal synthesis” — the optimal control from every initial point — is known for spins on the rotation group [1]. Time-optimal single-qubit operations [2], population transfer in two- and three-level systems [3] and the minimal time for a driven two-level system have all been characterized. In the last case, for some initial states the optimal protocol switches abruptly from bang–off–bang to bang–bang as the bound on the coupling decreases [4].

Fast local controls, slow couplings. In many platforms — coupled nuclear spins, superconducting qubits — local operations on each qubit are fast, while entangling interactions come from a fixed, slower coupling in the drift. If local operations are effectively instantaneous, the minimal time to build a unitary is set by the coupling alone. The problem then reduces to a geometric one on a quotient of the unitary group, solved with a Cartan decomposition (the familiar \(U = k_1\,a\,k_2\), local–entangling–local, decomposition of two-qubit gates) [5]. For three coupled spins, time-optimal sequences are geodesics of a sub-Riemannian metric, i.e. the shortest paths when only some directions can be moved in freely. These sequences can be substantially faster than the usual decoupling-based constructions [6].

The quantum brachistochrone

A complementary viewpoint fixes the Hamiltonian rather than the controls: which Hamiltonian, within a set of allowed ones (for example with bounded energy spread), takes a given initial state to a given final state fastest? A variational principle analogous to the classical brachistochrone gives an equation for the optimal Hamiltonian [7]. For the simplest constraint, a fixed energy spread, the optimal evolution follows a geodesic in the space of states. This connects the brachistochrone to the quantum speed limits of free evolution; see Quantum speed limits.

Geometry of quantum computation

The same ideas apply to circuit complexity. Define a Riemannian metric on the unitary group that makes many-body directions “expensive” and local directions “cheap”. The minimal number of gates needed to build a unitary is then essentially the length of the shortest geodesic to it, which opens Riemannian-geometry tools to the analysis of quantum algorithms [8]. See Control complexity & resources.

With dissipation

Relaxation makes time-optimality even more important, since every microsecond costs coherence. It also changes the structure of solutions. For a dissipative spin-1/2, the time-optimal control contains singular arcs, and the predicted protocol has been implemented in NMR [9]; see Control of open quantum systems.

Why it matters

  • Benchmarks for numerics. Exact minimal times and protocols are reference solutions for numerical methods such as GRAPE, and for scanning control times numerically to locate the speed limit [10].
  • Structure. Bang-bang, singular and geodesic segments explain features of numerically optimized pulses, and suggest compact parametrizations.
  • Design principles. Results such as “the minimal gate time is set by the coupling” guide hardware and gate design directly; see Quantum gate synthesis and Spins — NMR, ESR & colour centres.