Sub-Riemannian geometry of measurement based quantum computation

The computational power of quantum phases of matter with symmetry can be accessed through local measurements, but what is the most efficient way of doing so? In this work, we show that minimizing operational resources in measurement-based quantum computation on subsystem symmetric resource states amounts to solving a sub-Riemannian geodesic problem between the identity and the target logical unitary. This reveals a geometric structure underlying MBQC and offers a principled route to optimize quantum processing in computational phases.

💡 Research Summary

The paper establishes a deep connection between measurement‑based quantum computation (MBQC) on symmetry‑protected (or symmetry‑enriched) phases of quantum matter and sub‑Riemannian geometry. In such computational phases, the set of logical gates that can be implemented by local measurements is not arbitrary; it is fixed by the underlying symmetry and consists of a small, finite collection of Pauli‑type generators. This restriction naturally defines a control‑subspace that is a classic object of sub‑Riemannian geometry, where only motions along certain “horizontal” directions are allowed.

A central physical quantity is the computational order parameter σ (0 ≤ σ ≤ 1). In one‑dimensional cluster phases σ coincides with the familiar string‑order parameter; in two‑dimensional subsystem‑symmetry phases it plays an analogous role. σ = 1 corresponds to an ideal resource state (e.g., the exact cluster state) where logical operations are noiseless, while σ = 0 describes a trivial phase that cannot support any computation. For intermediate σ the logical channels are noisy, and the error scales with σ as previously shown for a single small rotation: ϵ ≤ α²/(N·σ²)·(1 − σ²)⁻¹, where α is the rotation angle and N the number of times the rotation is split.

The authors generalize this result from a single rotation to an arbitrary target unitary U. They introduce the notion of a sub‑Riemannian geodesic (or Carnot‑Carathéodory distance) d_CC(e,U) between the identity e and U, defined as the minimal length of a horizontal curve C(t) whose tangent lies in the span of the allowed generators. The length L(C) = ∫₀¹‖c(t)‖dt, with c(t) the control vector, is interpreted as the “runtime” of the implementation.

The main theorem (Theorem 1) states that if a target unitary U belongs to the Lie group generated by the allowed generators, then it can be realized on the given resource state with an error bounded by

  ϵ ≤ (1/N)·(1/σ² − 1)·d_CC(e,U)² + O(N⁻²).

Thus the total error factorizes into three intuitive contributions: (i) the inverse number of measurement steps 1/N, (ii) a factor (1/σ² − 1) that quantifies the quality of the resource state, and (iii) the square of the sub‑Riemannian distance, which captures the intrinsic geometric complexity of the desired unitary within the constrained control space.

The proof proceeds in three stages. First, a local error analysis shows that each measurement contributes ϵ_i ≈ (1/σ² − 1)·α_i²/2, where α_i is the effective rotation angle after accounting for σ. Second, using the triangle inequality, the total error is bounded by the sum of the local errors. Third, the authors invoke the Lie‑Trotter‑Suzuki product formula to approximate the optimal horizontal curve (the geodesic) by a finite sequence of small rotations, thereby constructing an explicit measurement protocol that achieves the bound. Numerical simulations for X‑axis Pauli rotations (σ = 0.9, N = 500) confirm that the bound is tight for the diamond norm; other norms (trace, Frobenius) give looser bounds, as expected.

Conceptually, the work overturns the view that MBQC is purely discrete and lacks a notion of “time” or continuous optimization. By recognizing that the symmetry‑restricted gate set defines a sub‑Riemannian structure, the authors import the powerful toolbox of optimal control and geometric analysis into MBQC. Practically, the result suggests two complementary routes to improve MBQC performance: (a) material engineering to increase σ (e.g., preparing states deeper in the computational phase), and (b) algorithmic design that follows sub‑Riemannian geodesics rather than naïvely decomposing circuits into the native MBQC gate set. The latter can reduce the number of required measurements dramatically, especially for unitaries that are “far” from the identity in the sub‑Riemannian metric.

The paper also discusses the implications for higher‑dimensional phases and subsystem symmetries, where the same formalism applies but the Lie algebra generated by the allowed generators may be larger, potentially reducing d_CC(e,U) for many useful unitaries. The authors note that while the upper bound is proven, establishing a matching lower bound for the diamond norm remains an open problem.

In summary, the authors provide a rigorous geometric framework that translates the problem of resource‑optimal MBQC into a sub‑Riemannian geodesic problem. This bridges discrete measurement‑driven computation with continuous optimal‑control theory, offering a principled pathway to design more efficient quantum algorithms on symmetry‑protected quantum matter.