Relationship between spin-glass three-dimensional (3D) Ising model and traveling salesman problems

TL;DR

This paper explores the deep mathematical relationship between 3D spin-glass Ising models and 3D traveling salesman problems, revealing shared complexity bounds and implications for algorithm development.

Contribution

It establishes a mapping between the spin-glass 3D Ising model and the TSP, identifying common lower bounds and complexity characteristics of their absolute minimum core models.

Findings

Existence of an absolute minimum core (AMC) model in both systems.

The AMC model equals the difference between specific 2D and 3D models.

Lower bounds of complexity are subexponential and superpolynomial.

Abstract

In this work, the relationship between a spin-glass three-dimensional (3D) Ising model with the lattice size N = mnl and the traveling salesman problem (TSP) in a 3D lattice is studied. In particular, the mathematical structures of the two systems are investigated in details. In both the hard problems, the nontrivial topological structures, the non-planarity graphs, the nonlocalities and/or the long-range spin entanglements exist, while randomness presents, which make the computation very complicated. It is found that an absolute minimum core (AMC) model exists not only in the spin-glass 3D Ising model but also in the 3D TSP for determining the lower bound of their computational complexities, which can be mapped each other. It is verified that the spin-glass AMC model equals to the difference between a two-level (l = 2) grid spin-glass 3D Ising model and a spin-glass 2D Ising model, which is NP-complete. Furthermore, according to the mapping between the spin-glass 3D Ising model and the TSP, it is proven that the AMC model for the TSP identifies to the difference between a two-level (l = 2) grid TSP model and a 2D TSP model, which is NP-complete also. The AMC models in both models are proven to be at the border between the NP-complete problems and the NP-intermediate problems. Because the lower bound of the computational complexity of the spin-glass 3D Ising model is the computational complexity by brute force search of the AMC model, the lower bound of the computational complexity of the TSP in a 3D lattice is the computational complexity by brute force search of the AMC model for the TSP. All of them are in subexponential and superpolynomial. The present work provides some implications on numerical algorithms for the NP-complete problems, for instance, one cannot develop a polynomial algorithm, but may develop a subexponential algorithm for the 3D spin-glass problem or TSP.