Fundamental Limitations of QAOA on Constrained Problems and a Route to Exponential Enhancement

TL;DR

The paper identifies a fundamental feasibility bottleneck for generic QAOA on permutation-constrained problems, showing that shallow circuits cannot concentrate significant probability mass on feasible solutions within the one-hot manifold. It introduces CE–QAOA, a constraint-enhanced kernel that operates inside the feasible subspace with a block-XY mixer, and proves an exponential separation in feasible mass between CE–QAOA and generic QAOA, scalable as exp(Θ(n^2)). The authors develop a cohesive toolkit—Walsh–Fourier analysis, Kostka-like Krawtchouk bounds, angle-averaging, and light-cone arguments—to bound the generic envelope and establish parameter-transfer amplification from generic to constrained variants. These results support a design principle that embedding feasibility into the ansatz can transform global coordination problems into locally solvable ones, with practical implications for routing and scheduling NP-hard problems. The work also outlines a clear path for extending the harmonic-analytic framework to quantify CE–QAOA’s behavior at deeper depths and under broader constraint families.

Abstract

We study fundamental limitations of the generic Quantum Approximate Optimization Algorithm (QAOA) on constrained problems where valid solutions form a low dimensional manifold inside the Boolean hypercube, and we present a provable route to exponential improvements via constraint embedding. Focusing on permutation constrained objectives, we show that the standard generic QAOA ansatz, with a transverse field mixer and diagonal r local cost, faces an intrinsic feasibility bottleneck: even after angle optimization, circuits whose depth grows at most linearly with n cannot raise the total probability mass on the feasible manifold much above the uniform baseline suppressed by the size of the full Hilber space. Against this envelope we introduce a minimal constraint enhanced kernel (CE QAOA) that operates directly inside a product one hot subspace and mixes with a block local XY Hamiltonian. For permutation constrained problems, we prove an angle robust, depth matched exponential enhancement where the ratio between the feasible mass from CE QAOA and generic QAOA grows exponentially in $n^2$ for all depths up to a linear fraction of n, under a mild polynomial growth condition on the interaction hypergraph. Thanks to the problem algorithm co design in the kernel construction, the techniques and guarantees extend beyond permutations to a broad class of NP-Hard constrained optimization problems.

Figures (3)

• Figure 1: Constraint-Enhanced QAOA dominates generic QAOA at depth $p{=}1$. For medium-size TSP instances from QOptLibOsaba2024Qoptlib, the unconstrained ansatz spreads its amplitude almost entirely outside the permutation subspace ($n!/2^{n^{2}}\!\to\!0$), whereas the block-encoded circuit concentrates an $\Omega(n^{-m})$ fraction of its probability on valid tours.
• Figure 2: Empirical versus theoretical feasibility mass after parameter optimization on noiseless simulations at depth p = 1 for the three QOPLib benchmark instances wi3, wi4, and wi5 Osaba2024Qoptlib. Each instance was executed with 500,000 shots so that Generic QAOA returns at least one feasible bit-string
• Figure 3: Empirical versus theoretical feasibility gap with parameter transfer. Noiseless simulations at depth $p=1$ for the three Qoplib benchmarksOsaba2024Qoptlibwi3, wi4, and wi5, each executed with 500 000 shots so that Generic QAOA returns at least one feasible bit-string. Black dots: measured ratios $P^{(\mathrm{block})}_{1}/P^{(\mathrm{generic})}_{1}$; green dashed line: analytic lower bound $2^{n^{2}}/n^{\,n}$ from Theorem \ref{['thm:param-transfer']} with $n=3,4,5$. Every empirical point lies on or above the bound, visually confirming the predicted exponential advantage of parameter transfer from Generic to Constraint-Enhanced QAOA

Theorems & Definitions (53)

• Definition 1: CE--QAOA kernel
• Lemma 2: One-block twirl
• proof
• Lemma 3: Blockwise (tensor) twirl to $I_D/D$
• proof
• Lemma 4: Conjugation invariance
• proof
• Lemma 5: Average success probability under blockwise twirl
• proof
• Proposition 6: Existence of a blockwise relabeling achieving the baseline