Hodgkin–Huxley Excitability Model

• The excitability model is a quantitative framework that employs voltage-dependent ion channel gating to simulate action potential generation and propagation in neurons.
• It integrates classical Hodgkin–Huxley kinetics with bifurcation and energy-based threshold analyses to distinguish spiking regimes.
• Hybrid and stochastic extensions enrich the model by incorporating data-driven, mesoscale, and homeostatic dynamics to capture realistic neuronal variability.

The excitability model of Hodgkin–Huxley (HH) defines a quantitative and biophysically interpretable framework for understanding how voltage-dependent ion channel dynamics underlie the generation and propagation of action potentials in excitable cells. The standard HH formulation treats individual channel populations—predominantly Na⁺ and K⁺—as sets of macroscopic, voltage- and time-dependent conductances gated by empirically defined kinetic variables. Recent work has both expanded and hybridized the classical paradigm, embedding it within stochastic, mesoscale, electromechanical, homeostatic, and data-driven architectures.

1. Classical Hodgkin–Huxley Formalism: Core Equations and Kinetics

The HH model specifies the temporal evolution of the membrane potential $V(t)$ in response to an externally injected current $I_\mathrm{ext}(t)$ through the equation

$$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$

where $C_m$ is membrane capacitance, $\bar{g}_{\mathrm{Na}}$, $\bar{g}_K$, $g_L$ are maximal conductances for sodium, potassium, and leakage, $E_{\mathrm{Na}}, E_K, E_L$ are their Nernst potentials, and $m$, $h$, $I_\mathrm{ext}(t)$0 are gating variables for sodium activation, sodium inactivation, and potassium activation, respectively.

The gating variables evolve according to

$I_\mathrm{ext}(t)$1

for $I_\mathrm{ext}(t)$2, with rate functions empirically derived from voltage-clamp data, e.g. (units: $I_\mathrm{ext}(t)$3 in mV): $I_\mathrm{ext}(t)$4 This deterministic, nine-parameter system robustly reproduces all-or-none action potentials, threshold phenomena, refractoriness, and repetitive firing (Estienne, 2023).

2. Modern Bifurcation, Threshold, and Excitability Mechanics

2.1 Type I/II Excitability, Bifurcations, and Periodic Orbits

Periodic firing and underlying bifurcations are explicit in the HH model. The saddle–node of limit cycles at $I_\mathrm{ext}(t)$5 underlies type I excitability: arbitrarily low-frequency oscillations near threshold. Type II excitability emerges via a subcritical Hopf bifurcation at $I_\mathrm{ext}(t)$6, where oscillations appear at nonzero frequency. Beyond the Hopf point, unstable periodic orbits and additional bifurcations (period-doubling, homoclinic) structure the transition from quiescence to repetitive spiking (Balti et al., 2015).

2.2 Energy-Based Threshold: Dissipativity and Required Supply

A recent approach formalizes threshold as a local maximum of the required supply $I_\mathrm{ext}(t)$7, defined as the minimal external energy required to drive the system from rest to a target state $I_\mathrm{ext}(t)$8 (Sepulchre et al., 2 Apr 2025): $I_\mathrm{ext}(t)$9 In the HH framework, the energy-based threshold unambiguously separates subthreshold and suprathreshold responses and generalizes classical definitions, offering robust, input–output-based criteria independent of specific voltage, current, or channel state (Sepulchre et al., 2 Apr 2025).

3. Dimensional Reduction, Robustness, and Homeostasis

Although sensitive to parameter variations, the functional output of the HH model is largely determined by two dimensions: the instantaneous “structural” conductance ratio ($$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$0) and the “kinetic” recovery rate ($$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$1) (Ori et al., 2018). Specifically,

$$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$2

$$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$3

Models collapse, in $$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$4 space, into three regimes: non-excitable, single-spike excitable, and oscillatory (pacemaking). Slow inactivation of Na⁺ channels is modeled as a dynamic reduction of $$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$5, providing automatic, local homeostatic stabilization of excitability (Ori et al., 2018).

4. Extended and Hybrid Models

4.1 Data-Driven Hybrid Hodgkin–Huxley Models

Estienne’s hybrid HH-ANN model replaces each empirical rate function $$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$6 with a trainable two-layer neural network (ANN), optimized via backpropagation to match experimentally observed voltage traces under known stimuli:

• Each rate function net has two layers, each with a single neuron.
• First layer activation: (log-)sigmoid (monotonicity); second: ReLU (non-negativity).
• Only six nets (one for each $$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$7, $$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$8).
• Training on only two suprathreshold pulses (with augmentation) suffices to recover thresholds, spike shape, refractoriness, and frequency–current curves with fidelity (errors: waveform amplitude/duration $$C_m \frac{dV}{dt} = - \left[ \bar{g}_{\mathrm{Na}} m^3 h (V - E_{\mathrm{Na}}) + \bar{g}_K n^4 (V - E_K) + g_L (V - E_L) \right] + I_\mathrm{ext}(t)$$9, f–I curve error $C_m$020%) (Estienne, 2023).

4.2 Mesoscopic and Stochastic Extensions

Recent mesoscale HH reductions describe collective excitability of fields, with state variables such as sodium kinetic-energy density $C_m$1 and excitability $C_m$2, and treat firing rate as a dynamic redistribution of energy. These models reproduce wave propagation, oscillation spectra, and damped temporal/spatial responses in neural tissue (Qin et al., 2022).

Stochastic HH models introduce rigorously validated multiplicative noise into gate kinetics: $C_m$3 ensuring invariance of gating variables in $C_m$4 (both Itô and Stratonovich sense), and reproducing channel-noise-induced firing, spike-time jitter, and physiologically realistic subthreshold fluctuations (Cresson et al., 2012).

5. Biophysical, Homeostatic, and Multiscale Generalizations

Subsequent models have incorporated additional physical mechanisms:

• Electrodiffusion: A “primitive” model treats Na⁺ spike activation/deactivation as an electrodiffusive phenomenon without gating variables, reproducing HH-like spikes solely via Nernst–Planck and Poisson equations in a growing hemispherical volume (Patel et al., 2024).
• Ionic Homeostasis: Extensions with dynamic intra/extracellular ion concentrations and pumps (e.g., Na⁺/K⁺-ATPase), as well as bath or glial buffer coupling, reveal new classes of bistable (pathological free-energy-starvation, FES) and excitable (spreading depression, seizure-like) behaviors via low-dimensional bifurcation structures, e.g., in the “potassium gain/loss” parameter (Hübel et al., 2014, Hübel et al., 2013).
• Quantum Biophysics: Selectivity-filter gating models (BS model) introduce quantum corrections to HH sodium conductance by modulating $C_m$5, where $C_m$6 encodes filter entanglement; these corrections slightly sharpen spike onset and preserve all standard excitability features (Moradi et al., 2014).

6. Mechanistic, Spatio-Temporal, and Network-Scale Formulations

Alternative mechanistic models replace empirical gating by ODEs grounded in first-principles conductance kinetics, predicting saddle-node thresholds and all-or-nothing responses with reduced parameter sets (Deng, 2015). Memristive circuit models generalize HH–type excitability to spatio-temporal domains, unifying positive/negative feedback and threshold mechanisms from single neurons to population-level neural fields, and reproducing both SNIC bifurcations and Amari-type spatial bumps (Burger et al., 28 May 2025).

7. Integrative and Extended Theoretical Frameworks

Modern theoretical developments place HH excitability within input–output dissipativity theory and multivariable singular perturbation; the mesoscopic, energy-based, and multiscale approaches enable quantitative, robust, and physiologically meaningful descriptions of neuronal excitability. These frameworks offer interpretability, data-driven adaptability, and extensibility to include engineering, network, and clinical perspectives, supporting future research in channelopathy, neural coding, and complex systems neuroscience (Estienne, 2023, Sepulchre et al., 2 Apr 2025, Ori et al., 2018, Qin et al., 2022).