1: \begin{abstract}

2: The Variational Quantum Eigensolver (VQE) is a promising candidate for

3: quantum applications on near-term Noisy Intermediate-Scale Quantum (NISQ) computers.

4: Despite a lot of empirical studies and recent progress in theoretical understanding of VQE's optimization landscape,

5: the convergence for optimizing VQE is far less understood. 

6: We provide the first rigorous analysis of the convergence of VQEs in the

7: over-parameterization regime.

8: By connecting the training dynamics with the

9: Riemannian Gradient Flow on the unit-sphere, 

10: we establish a threshold on the sufficient number of parameters

11: for efficient convergence, which depends polynomially on the system dimension and the spectral ratio, a property of the

12: problem Hamiltonian, and could be resilient to gradient noise to some extent.

13: We further illustrate that this over-parameterization threshold could be vastly reduced for specific VQE instances by establishing an ansatz-dependent threshold paralleling our main result.

14: We showcase that our ansatz-dependent threshold could serve as a proxy of the trainability of different VQE ansatzes without performing empirical experiments, which hence leads to a principled way of evaluating ansatz design.  

15: Finally, we conclude with a comprehensive empirical study that supports our theoretical findings. 

16: \end{abstract}

17: