搜索

x
中国物理学会期刊

外部激励作用下的多稳态忆阻FitzHugh-Nagumo神经元电路的动力学调控

Dynamic Regulation of a Multistable Memristive FitzHugh-Nagumo Neuronal Circuit Under External Excitation

PDF
导出引用
  • 为揭示外部激励作用下忆阻记忆效应对神经元复杂放电行为及其初值调控机制的影响,本文以经典FitzHugh-Nagumo(FHN)神经元电路为基础,采用压控型多稳态忆阻器替代传统电路中的隧道二极管,构建了一种多稳态忆阻FHN神经元电路,并结合哈密顿能量函数分析了此电路的能量调控特性。数值分析结果表明,该电路能够产生簇发振荡、尖峰放电及混沌振荡等丰富动力学行为,其中低频驱动条件下更易诱发明显的簇发放电模态。进一步地,通过多种仿真手段揭示了参数对动力学演化过程的调制作用,揭示了复杂放电行为的形成机制。此外,研究发现该电路在忆阻内部状态方向上具有周期复制特征,改变忆阻内部状态初值可诱导多个共存吸引子及其周期性平移分布,表现出显著的多稳态特征。最后,基于PSpice平台完成了电路设计与仿真,验证了所建模型的物理可实现性。研究表明,外部激励与多稳态忆阻器的耦合作用显著增强了FHN神经元电路的动力学复杂性与放电行为可调控性,为忆阻神经元动力学分析及神经形态电路设计提供了新的理论参考。

     

    To reveal how memristive memory, external periodic excitation, and initial-state regulation jointly affect the complex dynamics of a FitzHugh-Nagumo (FHN) neuronal circuit, a multistable memristive FHN neuronal circuit is proposed by replacing the conventional tunnel diode in the classical FHN circuit with a voltage-controlled multistable memristor. First, the multistable memristor model is established from its port current relation and internal state equation. Its memristive and multistable properties are verified through pinched hysteresis loops, state-variable evolution, and responses under different initial conditions. Based on Kirchhoff’s laws, the circuit equations of the proposed neuron are then derived and transformed into an equivalent dimensionless model by voltage, current, and time scaling, which provides a unified framework for stability analysis and numerical simulation. To further clarify the energy regulation mechanism, a dimensionless Hamiltonian energy function is constructed by considering the electric-field energy of the capacitor, the magnetic-field energy of the inductor, and the equivalent electromagnetic energy of the memristive channel. Helmholtz decomposition is used to decompose the vector field into conservative and dissipative components, showing that the introduced multistable memristor provides a state-dependent energy regulation mechanism for the FHN circuit. The effects of external excitation frequency, excitation amplitude, memristive feedback strength, and DC bias on neuronal firing dynamics are systematically investigated using equilibrium distribution, Jacobian eigenvalue analysis, two-dimensional dynamical maps, bifurcation diagrams, Lyapunov exponents, and phase trajectories. Numerical results show that the proposed circuit can generate various firing patterns, including periodic oscillations, bursting oscillations, spiking activity, weakly chaotic oscillations, and chaotic oscillations. In particular, low-frequency excitation promotes time-scale separation between the slow memristive state variable and the fast neuronal variables, thereby inducing pronounced bursting behavior. The bursting transition is jointly governed by stability switching induced by Hopf bifurcations and the unstable equilibrium skeleton shaped by fold bifurcations. Under high-frequency excitation, the system evolves from periodic firing to chaotic firing as the memristive nonlinear feedback strength varies. In addition, since the memristive internal state enters the system through periodic nonlinear functions, the vector field exhibits 2π translational invariance along the memristive state direction. This property enables regular translational displacement of trajectories when only the initial memristive state is changed, leading to multiple coexisting attractors with similar geometric structures but different spatial locations. Basin analysis further shows that the final attractor is mainly determined by the initial memristive state, and the well-defined basin boundaries indicate effective initial-state controllability. Finally, an equivalent PSpice circuit is designed to verify the circuit implementability of the proposed model. The circuit simulation results under different initial voltages of the memristive state capacitor agree well with the numerical results, confirming the initial-state-dependent coexistence and translational displacement of attractors. These results demonstrate that the multistable memristor not only enhances the diversity of firing patterns in the FHN neuronal circuit, but also endows the system with regular multistable coexistence that can be selected by the memristive initial state. The proposed model provides a physically realizable theoretical and circuit framework for studying complex firing regulation, multistable state control, and neuromorphic circuit design based on memristive neurons.

     

    目录

    /

    返回文章
    返回