Presentation Information
[23p-22B-12]Spin cluster glass-based Ising machine with stochastic resonance effect
〇Zhiqiang Liao1, Kaijie Ma1, Hiroyasu Yamahara1, Munetoshi Seki1, Hitoshi Tabata1 (1.Univ. of Tokyo)
Keywords:
Ising machine,Spin cluster glass,Stochastic resonance
Gain-dissipative Ising machines (GIM) has the potential to address NP-hard problems with high efficiency. To get rid of local minimum, GIM needs the help of noise to the increase escape rate. However, the inevitable noise in real-world physical systems can lead to a decrease in the performance of GIM.
To enhance the performance of GIM within noisy environments, we conducted testing in this study by utilizing Lu3Fe4Co0.5Si0.5O12 (LFCS) as the nonlinear component of the Ising machine. LFCS is a material derived from the Lu3Fe5O12 (LuIG) doped by Co2+ and nonmagnetic Si4+, known for its exhibiting of spin glass behavior. Spin glasses display memory effects, where the material retains information about its past states. This effect can be achieved when the material is cooled below a certain temperature called the spin freezing temperature. In the experimental setup, the personal computer is employed for implementing programmable spin interactions, while the magnetic field is applied as feedback signal, enabling LFCS to perform nonlinear mappings on the input. In this study, upon completing the mapping of the magnetic field onto the sample, the recorded normalized magnetization is considered as the original spin amplitude.
As the testing results, we found the LFCS-based GIM reaches its ground state most rapidly around 170K. At higher temperatures like 220K and 300K, LFCS and LuIG exhibit similar properties. However, at lower temperatures such as 110K and 10K, the LFCS-based GIM experiences a gradual decrease in speed due to enhanced spin freezing effects. This trend of initially increasing and then decreasing performance with rising temperatures is known as the stochastic resonance effect. This implies that around 170K, the LFCS-based GIM effectively utilizes noise energy, achieving optimal performance.
To enhance the performance of GIM within noisy environments, we conducted testing in this study by utilizing Lu3Fe4Co0.5Si0.5O12 (LFCS) as the nonlinear component of the Ising machine. LFCS is a material derived from the Lu3Fe5O12 (LuIG) doped by Co2+ and nonmagnetic Si4+, known for its exhibiting of spin glass behavior. Spin glasses display memory effects, where the material retains information about its past states. This effect can be achieved when the material is cooled below a certain temperature called the spin freezing temperature. In the experimental setup, the personal computer is employed for implementing programmable spin interactions, while the magnetic field is applied as feedback signal, enabling LFCS to perform nonlinear mappings on the input. In this study, upon completing the mapping of the magnetic field onto the sample, the recorded normalized magnetization is considered as the original spin amplitude.
As the testing results, we found the LFCS-based GIM reaches its ground state most rapidly around 170K. At higher temperatures like 220K and 300K, LFCS and LuIG exhibit similar properties. However, at lower temperatures such as 110K and 10K, the LFCS-based GIM experiences a gradual decrease in speed due to enhanced spin freezing effects. This trend of initially increasing and then decreasing performance with rising temperatures is known as the stochastic resonance effect. This implies that around 170K, the LFCS-based GIM effectively utilizes noise energy, achieving optimal performance.