Quantum Optimization Landscape Analysis

---
name: quantum-optimization-landscape-analysis
description: "Quantum optimization landscape analysis methodology for NISQ-era variational algorithms. Covers VQE/QAOA performance analysis, expressibility-noise barren plateaus, feedback-guided methods (FALQON), and physics-informed circuit co-design for strongly correlated many-body systems. Activation: quantum optimization, VQE, QAOA, barren plateau, FALQON, feedback quantum algorithm, strongly correlated systems, many-body phase transitions, variational quantum algorithm, quantum criticality, expressibility, noise-induced plateau"
license: Complete terms in LICENSE.txt
metadata:
  arxiv_id: "2606.03147"
  published: "2026-06-02"
  authors: "G. E. L. Pexe, L. A. M. Rattighieri, P. M. Prado, A. R. Fritsch, F. F. Fanchini"
  tags: ["quantum-optimization", "VQE", "QAOA", "barren-plateau", "FALQON", "many-body", "NISQ", "phase-transition"]
---

# Quantum Optimization Landscape Analysis

Analyze and select quantum optimization algorithms for strongly correlated many-body systems in the NISQ era.

## Context

Classical simulation of strongly correlated systems is intractable due to exponential Hilbert space growth and fermionic sign problem. Variational quantum algorithms (VQE, QAOA) offer a path but face operational bottlenecks.

## Core Methodology

### 1. Algorithm Selection Framework

| Algorithm | Type | Gradient | Robustness | Best For |
|-----------|------|----------|------------|----------|
| VQE | Variational | Gradient-based | Low (barren plateaus) | Ground state energy |
| QAOA | Variational | Gradient-based | Low (expressibility limits) | Combinatorial optimization |
| FALQON | Feedback-based | Deterministic | High (gradient-free) | Phase transitions, critical points |

### 2. Barren Plateau Diagnosis

Two distinct failure modes:
- **Expressibility-induced**: Over-parameterized circuits flatten gradients → reduce ansatz depth, use problem-informed initialization
- **Noise-induced**: Hardware noise suppresses signal-to-noise ratio → use error mitigation, shallow circuits

### 3. Feedback-Guided Advantage

FALQON and similar deterministic feedback methods:
- Avoid gradient computation entirely
- Navigate energy landscapes via geometric feedback
- Provide more robust trajectories through noisy optimization spaces
- Scale better with circuit depth than gradient-based methods

### 4. Physics-Informed Circuit Co-Design

Key pattern: combine domain physics knowledge with circuit architecture:
- Encode symmetries directly in ansatz (reduces search space)
- Use physical intuition for initial parameter values
- Design measurement strategies aligned with observables of interest

## Application Domains

- Deconfined Quantum Criticality
- Strange metals
- Many-Body Localization
- Topological phase transitions
- Quantum spin liquids

## Implementation Steps

1. **Problem classification**: Identify if target system exhibits strong correlation (fermionic sign problem present)
2. **Hardware assessment**: Determine available qubits, coherence times, gate fidelities
3. **Algorithm selection**: Prefer feedback-guided (FALQON) over gradient-based for deep circuits on noisy hardware
4. **Circuit design**: Build physics-informed ansatz encoding known symmetries
5. **Validation**: Compare against classical baselines (DMRG, tensor networks) for small system sizes

## Pitfalls

- **Gradient-based methods on NISQ**: Barren plateaus make VQE/QAOA unreliable beyond ~10-15 qubits without mitigation
- **Pure simulation comparison**: Classical tensor networks often outperform NISQ variational algorithms for 1D/2D systems
- **Ignoring noise structure**: Noise is not uniform across qubits — map logical to physical qubits considering device topology
- **Over-parameterized ansatz**: More parameters ≠ better results; leads to expressibility-induced barren plateaus

## Verification

- Test on known systems (e.g., transverse-field Ising model) where exact solutions exist
- Verify energy convergence matches analytical predictions at small system sizes
- Compare feedback-guided vs gradient-based trajectories on same problem instance