Dairong Chen,Andrew D. Kent,Dries Sels,Flaviano Morone
Abstract:We present a method to approximately solve general instances of combinatorial optimization problems using the physical dynamics of 3d rotors obeying Landau-Lifshitz-Gilbert dynamics. Conventional techniques to solve discrete optimization problems that use simple continuous relaxation of the objective function followed by gradient descent minimization are inherently unable to avoid local optima, thus producing poor-quality solutions. Our method considers the physical dynamics of macrospins capable of escaping from local minima, thus facilitating the discovery of high-quality, nearly optimal solutions, as supported by extensive numerical simulations on a prototypical quadratic unconstrained binary optimization (QUBO) problem. Our method produces solutions that compare favorably with those obtained using state-of-the-art minimization algorithms (such as simulated annealing) while offering the advantage of being physically realizable by means of arrays of stochastic magnetic tunnel junction devices.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to approximately solve combinatorial optimization problems by using the stochastic Landau - Lifshitz - Gilbert (sLLG) dynamical system. Specifically, the paper proposes a new energy minimization framework, using the dynamics of physical systems to solve combinatorial optimization problems, especially the Quadratic Unconstrained Binary Optimization (QUBO) problem.
### Problem Background
Traditional methods for solving discrete optimization problems usually adopt simple continuous relaxation techniques, that is, after the objective function is made continuous, the gradient descent method is used for minimization. However, these methods are prone to fall into local optimal solutions, thus producing solutions of poor quality. To overcome this problem, this paper proposes a dynamic method based on macrospins, which can escape from local minima and are more likely to find high - quality approximate optimal solutions.
### Solution
The core idea of the paper is to use the dynamic behavior of 3D rotors (i.e., macrospins) to simulate the solution process of optimization problems. The specific steps are as follows:
1. **Introducing the sLLG dynamical system**: Each magnetic moment \(\vec{m}_i\) evolves according to the Landau - Lifshitz - Gilbert equation:
\[
\frac{d\vec{m}_i}{dt}=\gamma' \vec{m}_i\times\vec{H}_{\text{eff},i}-\alpha_i\frac{\vec{m}_i\times d\vec{m}_i}{dt}
\]
where \(\gamma'\) is the gyromagnetic ratio, \(\alpha_i\) is the damping constant, and \(\vec{H}_{\text{eff},i}\) is the effective field of the \(i\) - th magnetic moment.
2. **Introducing thermal fluctuations**: To simulate thermal effects, a thermal noise term \(\vec{H}_{\text{th},i}\) is added to the effective field, and its correlation is:
\[
\langle\vec{H}_{\text{th},i,\mu}(t)\vec{H}_{\text{th},i,\nu}(t')\rangle = C_i\delta_{\mu\nu}\delta(t - t')
\]
where
\[
C_i=\frac{2\alpha_i k_B T}{\mu_0^2 m_i\gamma_i(1+\alpha_i^2)}
\]
3. **Applying to the QUBO problem**: The above - mentioned dynamical system is applied to the Sherrington - Kirkpatrick (SK) model, which is a typical QUBO problem. The Hamiltonian of the SK model is:
\[
H_{\text{SK}}(s)=-\sum_{\langle i,j\rangle}J_{ij}\sqrt{n}s_is_j
\]
where \(s_i = \pm1\) represents Ising spins, and \(J_{ij}\) is the exchange interaction coefficient.
4. **Mapping to the LLG model**: By mapping \(J_{ij}\) to the exchange interaction term in the LLG model and setting the external magnetic field to zero, the ground state of the SK model can be mapped to the ground state of the LLG system.
### Experimental Results
The paper verifies the effectiveness of this method through numerical simulation. Compared with the traditional Glauber dynamics, the LLG method can obtain lower energy values under different system sizes, close to the theoretically optimal solution. Specifically, as shown in Figure 3 and Figure 4, when the results of the LLG method are extrapolated to an infinite - sized system, the energy value is very close to the exact solution.
### Conclusion
The optimization method based on sLLG dynamics proposed in this paper not only has theoretical advantages but also can be implemented experimentally, for example, by using an array of magnetic tunnel junction devices. This provides a way to solve complex combinatorial optimization problems.