1: \begin{abstract}

2: Variational integrators are momentum-preserving and symplectic numerical methods used to propagate the evolution of Hamiltonian systems. 

3: In this paper, we introduce a new class of variational integrators that achieve fourth-order convergence despite having the same integration scheme as traditional second-order variational integrators.

4: The new class of integrators are created by replacing a dynamical system's Lagrangian in the variational integration algorithm with its surrogate Lagrangian. 

5: By incorporating the surrogate Lagrangian the propagation errors induced by variational integrators, up to a given order, are eliminated. 

6: Furthermore, no assumption on the Lagrangian's structure is made and, therefore, the proposed approach is applicable to a large range of dynamical systems.

7: In addition, surrogate variational integrators are also constructed for Hamiltonian systems subjected to holonomic constraints and external forces.  

8: Finally, the methodology is extended to derive higher-order surrogate variational integrators that achieve an arbitrary order of accuracy but retain second-order complexity in the integration scheme. 

9: Several numerical experiments are presented to demonstrate the efficacy of our approach.

10: \end{abstract}

11: