sc_neurocore_engine/neurons/

motor.rs

// SPDX-License-Identifier: AGPL-3.0-or-later
// Commercial license available
// © Concepts 1996–2026 Miroslav Šotek. All rights reserved.
// © Code 2020–2026 Miroslav Šotek. All rights reserved.
// ORCID: 0009-0009-3560-0851
// Contact: www.anulum.li | protoscience@anulum.li
// SC-NeuroCore — Motor Neuron Models

//! Motor neuron models for spinal and cortical motor circuits.
//!
//! Motor model group: alpha motor, gamma motor, upper motor, Renshaw cell, motor unit.
//! Added one by one with full 7-point checklist verification.

use super::biophysical::safe_rate;

// ═══════════════════════════════════════════════════════════════════
// Alpha Motor Neuron
// ═══════════════════════════════════════════════════════════════════

/// Alpha motor neuron — spinal cord, innervates extrafusal muscle fibres.
///
/// Biophysics: Wang-Buzsáki Na+/K+ core, persistent inward current (PIC)
/// for bistable firing (plateau potentials), Ca2+-dependent AHP for rate
/// limiting (f-I gain control). Larger soma than cortical neurons → lower
/// input resistance.
///
/// PIC is modelled as a slow L-type Ca2+ current that activates at
/// depolarised potentials and inactivates very slowly, enabling plateau
/// potentials and self-sustained firing after brief input.
///
/// AHP from Ca2+-activated K+ (SK channels) limits firing rate and
/// produces the characteristic linear f-I relationship of motor neurons.
///
/// Powers & Binder, J. Neurophysiol. 86, 2001.
/// Heckman & Enoka, Compr. Physiol. 2(4), 2012.
#[derive(Clone, Debug)]
pub struct AlphaMotorNeuron {
    pub v: f64,
    pub h: f64,
    pub n: f64,
    pub m_pic: f64,  // PIC (L-type Ca²⁺) activation
    pub h_pic: f64,  // PIC slow inactivation (tau ~200 ms)
    pub ca: f64,     // Intracellular Ca²⁺ (µM)
    pub ca_buf: f64, // Bound Ca²⁺ (buffered fraction)
    // Conductances (mS/cm²)
    pub g_na: f64,
    pub g_k: f64,
    pub g_pic: f64, // Persistent inward current
    pub g_ahp: f64, // Ca²⁺-dependent K⁺ (AHP)
    pub g_l: f64,
    // Reversal potentials (mV)
    pub e_na: f64,
    pub e_k: f64,
    pub e_ca: f64,
    pub e_l: f64,
    pub c_m: f64,
    pub phi: f64,
    pub tau_ca: f64,    // Ca²⁺ decay (ms)
    pub buf_ratio: f64, // Buffering ratio (fraction of Ca²⁺ bound)
    pub dt: f64,
    pub v_threshold: f64,
}

impl AlphaMotorNeuron {
    pub fn new() -> Self {
        Self {
            v: -65.0,
            h: 0.8,
            n: 0.1,
            m_pic: 0.0,
            h_pic: 1.0, // PIC inactivation starts de-inactivated
            ca: 0.0,
            ca_buf: 0.0,
            g_na: 35.0,
            g_k: 9.0,
            g_pic: 0.15, // PIC for plateau potentials (conservative)
            g_ahp: 3.0,  // Strong AHP for rate limiting
            g_l: 0.3,    // Higher leak (larger soma, stabilises rest)
            e_na: 55.0,
            e_k: -90.0,
            e_ca: 120.0,
            e_l: -65.0,
            c_m: 1.5, // Larger soma → higher capacitance
            phi: 4.0,
            tau_ca: 150.0,    // Slow Ca²⁺ clearance for AHP
            buf_ratio: 0.003, // ~0.3% free Ca²⁺ (99.7% buffered)
            dt: 0.01,
            v_threshold: -20.0,
        }
    }

    pub fn step(&mut self, current: f64) -> i32 {
        let v_prev = self.v;
        let n_sub = (0.5 / self.dt.max(0.001)) as usize;
        for _ in 0..n_sub {
            // WB Na+/K+ gating
            let am = safe_rate(0.1, 35.0, self.v, 10.0, 1.0);
            let bm = 4.0 * (-(self.v + 60.0) / 18.0).exp();
            let m_inf = am / (am + bm);
            let ah = 0.07 * (-(self.v + 58.0) / 20.0).exp();
            let bh = 1.0 / (1.0 + (-(self.v + 28.0) / 10.0).exp());
            let an = safe_rate(0.01, 34.0, self.v, 10.0, 0.1);
            let bn = 0.125 * (-(self.v + 44.0) / 80.0).exp();

            self.h += self.phi * (ah * (1.0 - self.h) - bh * self.h) * self.dt;
            self.n += self.phi * (an * (1.0 - self.n) - bn * self.n) * self.dt;

            // PIC (L-type Ca²⁺): activation + slow inactivation
            // Activation: m_pic, tau ~50 ms, half-act -50 mV
            let m_pic_inf = 1.0 / (1.0 + (-(self.v + 40.0) / 5.0).exp());
            self.m_pic += (m_pic_inf - self.m_pic) / 50.0 * self.dt;
            // Inactivation: h_pic, tau ~200 ms, half-inact -40 mV
            // L-type inactivation is slow and Ca²⁺-dependent
            let h_pic_inf = 1.0 / (1.0 + ((self.v + 40.0) / 8.0).exp());
            let tau_h_pic = 200.0 + 100.0 / (1.0 + ((self.v + 40.0) / 10.0).powi(2)).max(0.01);
            self.h_pic += (h_pic_inf - self.h_pic) / tau_h_pic * self.dt;
            self.h_pic = self.h_pic.clamp(0.0, 1.0);

            // Ca²⁺ dynamics with buffering
            // Total Ca²⁺ entry (PIC-mediated)
            let i_ca_entry = self.g_pic * self.m_pic * self.h_pic * (self.v - self.e_ca);
            let ca_influx = if i_ca_entry < 0.0 {
                -i_ca_entry * 0.001
            } else {
                0.0
            };
            let ca_spike = if self.v > -10.0 { 0.02 } else { 0.0 };
            // Only ~0.3% of entering Ca²⁺ is free (rest is buffered)
            let free_ca_change = (ca_influx + ca_spike) * self.buf_ratio;
            self.ca += (-self.ca / self.tau_ca + free_ca_change) * self.dt;
            if self.ca < 0.0 {
                self.ca = 0.0;
            }
            // Buffered pool tracks total entry (slower dynamics)
            self.ca_buf += ((ca_influx + ca_spike) * (1.0 - self.buf_ratio)
                - self.ca_buf / (self.tau_ca * 5.0))
                * self.dt;
            if self.ca_buf < 0.0 {
                self.ca_buf = 0.0;
            }

            // AHP: Ca²⁺-activated K⁺ (SK channels), Hill n=2
            let ca_total = self.ca + self.ca_buf * 0.01; // Buffered contributes slowly
            let ahp_inf = ca_total * ca_total / (ca_total * ca_total + 0.25);

            let i_na = self.g_na * m_inf.powi(3) * self.h * (self.v - self.e_na);
            let i_k = self.g_k * self.n.powi(4) * (self.v - self.e_k);
            let i_pic = self.g_pic * self.m_pic * self.h_pic * (self.v - self.e_ca);
            let i_ahp = self.g_ahp * ahp_inf * (self.v - self.e_k);
            let i_l = self.g_l * (self.v - self.e_l);

            self.v += (-i_na - i_k - i_pic - i_ahp - i_l + current) / self.c_m * self.dt;
        }
        if self.v >= self.v_threshold && v_prev < self.v_threshold {
            1
        } else {
            0
        }
    }

    pub fn reset(&mut self) {
        *self = Self::new();
    }
}

impl Default for AlphaMotorNeuron {
    fn default() -> Self {
        Self::new()
    }
}

// ═══════════════════════════════════════════════════════════════════
// Gamma Motor Neuron
// ═══════════════════════════════════════════════════════════════════

/// Gamma motor neuron — innervates intrafusal fibres of muscle spindles.
///
/// Regulates proprioceptive sensitivity by adjusting spindle tension.
/// Smaller soma than alpha, lower firing rates (5-30 Hz), no PIC.
/// Simple LIF with spike-frequency adaptation (slow K+ current).
/// Two subtypes: dynamic (bag1, velocity-sensitive) and static
/// (bag2/chain, length-sensitive) — controlled by `dynamic` flag.
///
/// Prochazka & Hulliger, Prog. Brain Res. 80, 1989.
/// Taylor et al., J. Physiol. 519(3), 1999.
#[derive(Clone, Debug)]
pub struct GammaMotorNeuron {
    pub v: f64,
    pub v_rest: f64,
    pub v_reset: f64,
    pub v_threshold: f64,
    pub tau: f64,
    pub adapt: f64,     // Slow adaptation current
    pub tau_adapt: f64, // Adaptation time constant (ms)
    pub a_adapt: f64,   // Adaptation coupling strength
    pub gain: f64,      // Input gain (fusimotor drive → mV)
    pub dynamic: bool,  // true = dynamic (bag1), false = static (bag2/chain)
    pub dt: f64,
}

impl GammaMotorNeuron {
    pub fn new() -> Self {
        Self::dynamic()
    }

    /// Dynamic gamma — innervates bag1 intrafusal fibres (velocity-sensitive).
    pub fn dynamic() -> Self {
        Self {
            v: -65.0,
            v_rest: -65.0,
            v_reset: -70.0,
            v_threshold: -50.0,
            tau: 8.0,
            adapt: 0.0,
            tau_adapt: 100.0,
            a_adapt: 0.3,
            gain: 1.0,
            dynamic: true,
            dt: 0.5,
        }
    }

    /// Static gamma — innervates bag2/chain intrafusal fibres (length-sensitive).
    pub fn static_type() -> Self {
        Self {
            tau: 12.0,        // Slower membrane
            tau_adapt: 200.0, // Larger adaptation time constant
            a_adapt: 0.5,
            dynamic: false,
            ..Self::dynamic()
        }
    }

    /// Step with fusimotor drive (arbitrary units, ≥ 0). Returns spike (1/0).
    pub fn step(&mut self, drive: f64) -> i32 {
        if !self.is_valid() || !drive.is_finite() {
            return 0;
        }
        let v_old = self.v;
        let adapt_old = self.adapt;
        let input = self.gain * drive.max(0.0) - adapt_old;
        let v_target = self.v_rest + input;
        let v_candidate = v_target + (v_old - v_target) * (-self.dt / self.tau).exp();
        let adapt_target = self.a_adapt * (v_candidate - self.v_rest);
        let adapt_candidate =
            adapt_target + (adapt_old - adapt_target) * (-self.dt / self.tau_adapt).exp();
        if !v_candidate.is_finite() || !adapt_candidate.is_finite() {
            return 0;
        }
        self.v = v_candidate;
        self.adapt = adapt_candidate;

        if v_candidate >= self.v_threshold {
            self.v = self.v_reset;
            1
        } else {
            0
        }
    }

    pub fn reset(&mut self) {
        self.v = self.v_rest;
        self.adapt = 0.0;
    }

    fn is_valid(&self) -> bool {
        [
            self.v,
            self.v_rest,
            self.v_reset,
            self.v_threshold,
            self.tau,
            self.adapt,
            self.tau_adapt,
            self.a_adapt,
            self.gain,
            self.dt,
        ]
        .iter()
        .all(|value| value.is_finite())
            && self.tau > 0.0
            && self.tau_adapt > 0.0
            && self.dt > 0.0
            && self.gain >= 0.0
            && self.v_reset < self.v_threshold
    }
}

impl Default for GammaMotorNeuron {
    fn default() -> Self {
        Self::new()
    }
}

// ═══════════════════════════════════════════════════════════════════
// Upper Motor Neuron (Corticospinal L5 Pyramidal)
// ═══════════════════════════════════════════════════════════════════

/// Upper motor neuron — layer 5 pyramidal cell, corticospinal projection.
///
/// Biophysics: Pospischil 2008 RS parameterisation (Na+, K+, M-current)
/// with added high-threshold Ca2+ current for dendritic Ca2+ spikes.
/// Regular-spiking with adaptation. Drives alpha/gamma motor neurons
/// via corticospinal tract.
///
/// Pospischil et al., Biol. Cybern. 99(4-5), 2008 (RS variant).
/// Larkum, Trends Neurosci. 36(3), 2013 (dendritic Ca2+ spikes).
#[derive(Clone, Debug)]
pub struct UpperMotorNeuron {
    pub v: f64,
    pub m: f64,
    pub h: f64,
    pub n: f64,
    pub p: f64, // M-current (Kv7) activation
    pub s: f64, // High-threshold Ca2+ activation
    // Conductances
    pub g_na: f64,
    pub g_k: f64,
    pub g_m: f64,
    pub g_ca: f64,
    pub g_l: f64,
    // Reversal potentials
    pub e_na: f64,
    pub e_k: f64,
    pub e_ca: f64,
    pub e_l: f64,
    pub c_m: f64,
    pub dt: f64,
    pub v_threshold: f64,
}

impl UpperMotorNeuron {
    pub fn new() -> Self {
        Self {
            v: -70.0,
            m: 0.05,
            h: 0.6,
            n: 0.3,
            p: 0.0,
            s: 0.0,
            g_na: 50.0,
            g_k: 5.0,
            g_m: 0.07, // M-current for adaptation (Pospischil RS)
            g_ca: 0.3, // High-threshold dendritic Ca2+
            g_l: 0.1,
            e_na: 50.0,
            e_k: -90.0,
            e_ca: 120.0,
            e_l: -70.0,
            c_m: 1.0,
            dt: 0.025,
            v_threshold: -20.0,
        }
    }

    fn finite(values: &[f64]) -> bool {
        values.iter().all(|value| value.is_finite())
    }

    fn rate_exp(value: f64) -> f64 {
        value.clamp(-60.0, 60.0).exp()
    }

    fn gate(previous: f64, alpha: f64, beta: f64, dt: f64) -> Option<f64> {
        let total = alpha + beta;
        if total <= 0.0 || !total.is_finite() {
            return None;
        }
        let steady = alpha / total;
        let next = steady + (previous - steady) * Self::rate_exp(-total * dt);
        next.is_finite().then_some(next.clamp(0.0, 1.0))
    }

    fn gate_inf(previous: f64, steady: f64, tau: f64, dt: f64) -> Option<f64> {
        if tau <= 0.0 || !tau.is_finite() {
            return None;
        }
        let next = steady + (previous - steady) * Self::rate_exp(-dt / tau);
        next.is_finite().then_some(next.clamp(0.0, 1.0))
    }

    fn valid_configuration(&self) -> bool {
        Self::finite(&[
            self.g_na,
            self.g_k,
            self.g_m,
            self.g_ca,
            self.g_l,
            self.e_na,
            self.e_k,
            self.e_ca,
            self.e_l,
            self.c_m,
            self.dt,
            self.v_threshold,
        ]) && self.g_na >= 0.0
            && self.g_k >= 0.0
            && self.g_m >= 0.0
            && self.g_ca >= 0.0
            && self.g_l >= 0.0
            && self.c_m > 0.0
            && self.dt > 0.0
    }

    fn valid_state(&self) -> bool {
        Self::finite(&[self.v, self.m, self.h, self.n, self.p, self.s])
            && (-150.0..=100.0).contains(&self.v)
            && (0.0..=1.0).contains(&self.m)
            && (0.0..=1.0).contains(&self.h)
            && (0.0..=1.0).contains(&self.n)
            && (0.0..=1.0).contains(&self.p)
            && (0.0..=1.0).contains(&self.s)
    }

    fn step_candidate(
        &self,
        v: f64,
        mut m: f64,
        mut h: f64,
        mut n: f64,
        mut p: f64,
        mut s: f64,
        current: f64,
    ) -> Option<(f64, f64, f64, f64, f64, f64)> {
        let dv = v - (-56.2);
        let x_m = dv - 13.0;
        let alpha_m = if x_m.abs() < 1e-6 {
            0.32 * 4.0
        } else {
            -0.32 * x_m / (Self::rate_exp(-x_m / 4.0) - 1.0)
        };
        // Published Pospischil beta_m numerator is V - V_T - 40; an earlier revision
        // shared the -17 offset of alpha_h, which drove the cell into depolarisation
        // block (three spikes then a fixed point near threshold for any stimulus).
        let x_bm = dv - 40.0;
        let beta_m = if x_bm.abs() < 1e-6 {
            0.28 * 5.0
        } else {
            0.28 * x_bm / (Self::rate_exp(x_bm / 5.0) - 1.0)
        };
        let alpha_h = 0.128 * Self::rate_exp(-(dv - 17.0) / 18.0);
        let beta_h = 4.0 / (1.0 + Self::rate_exp(-(dv - 40.0) / 5.0));
        let x_n = dv - 15.0;
        let alpha_n = if x_n.abs() < 1e-6 {
            0.032 * 5.0
        } else {
            -0.032 * x_n / (Self::rate_exp(-x_n / 5.0) - 1.0)
        };
        let beta_n = 0.5 * Self::rate_exp(-(dv - 10.0) / 40.0);
        m = Self::gate(m, alpha_m, beta_m, self.dt)?;
        h = Self::gate(h, alpha_h, beta_h, self.dt)?;
        n = Self::gate(n, alpha_n, beta_n, self.dt)?;
        let p_inf = 1.0 / (1.0 + Self::rate_exp(-(v + 35.0) / 10.0));
        let tau_p =
            400.0 / (3.3 * Self::rate_exp((v + 35.0) / 20.0) + Self::rate_exp(-(v + 35.0) / 20.0));
        p = Self::gate_inf(p, p_inf, tau_p, self.dt)?;
        let s_inf = 1.0 / (1.0 + Self::rate_exp(-(v + 20.0) / 5.0));
        s = Self::gate_inf(s, s_inf, 10.0, self.dt)?;
        let g_na = self.g_na * m.powi(3) * h;
        let g_k = self.g_k * n.powi(4);
        let g_m = self.g_m * p;
        let g_ca = self.g_ca * s.powi(2);
        let g_total = g_na + g_k + g_m + g_ca + self.g_l;
        if g_total <= 0.0 || !g_total.is_finite() {
            return None;
        }
        let steady_v = (current
            + g_na * self.e_na
            + g_k * self.e_k
            + g_m * self.e_k
            + g_ca * self.e_ca
            + self.g_l * self.e_l)
            / g_total;
        let next_v = steady_v + (v - steady_v) * Self::rate_exp(-(g_total / self.c_m) * self.dt);
        Self::finite(&[next_v, m, h, n, p, s]).then_some((
            next_v.clamp(-150.0, 100.0),
            m,
            h,
            n,
            p,
            s,
        ))
    }

    pub fn step(&mut self, current: f64) -> i32 {
        if !self.valid_configuration() || !self.valid_state() || !current.is_finite() {
            return 0;
        }
        let v_prev = self.v;
        let (mut v, mut m, mut h, mut n, mut p, mut s) =
            (self.v, self.m, self.h, self.n, self.p, self.s);
        for _ in 0..4 {
            let Some(next) = self.step_candidate(v, m, h, n, p, s, current) else {
                return 0;
            };
            (v, m, h, n, p, s) = next;
        }
        (self.v, self.m, self.h, self.n, self.p, self.s) = (v, m, h, n, p, s);
        if v >= self.v_threshold && v_prev < self.v_threshold {
            1
        } else {
            0
        }
    }

    pub fn reset(&mut self) {
        self.v = -70.0;
        self.m = 0.05;
        self.h = 0.6;
        self.n = 0.3;
        self.p = 0.0;
        self.s = 0.0;
    }
}

impl Default for UpperMotorNeuron {
    fn default() -> Self {
        Self::new()
    }
}

// ═══════════════════════════════════════════════════════════════════
// Renshaw Cell (Spinal Inhibitory Interneuron)
// ═══════════════════════════════════════════════════════════════════

/// Renshaw cell — spinal inhibitory interneuron for recurrent inhibition.
///
/// Receives collaterals from alpha motor neuron axons, provides
/// glycinergic recurrent inhibition back to the motor pool. Characteristic
/// high-frequency initial burst (cholinergic nicotinic drive from motor
/// axon collaterals) followed by rapid adaptation.
///
/// WB gating core with strong adaptation (M-current analogue) to produce
/// the burst-then-decay response pattern.
///
/// Renshaw 1941 (discovery); Windhorst, Prog. Neurobiol. 46(5), 1996.
#[derive(Clone, Debug)]
pub struct RenshawCell {
    pub v: f64,
    pub h: f64,
    pub n: f64,
    pub adapt: f64,
    pub g_na: f64,
    pub g_k: f64,
    pub g_adapt: f64,
    pub g_l: f64,
    pub e_na: f64,
    pub e_k: f64,
    pub e_l: f64,
    pub c_m: f64,
    pub phi: f64,
    pub tau_adapt: f64,
    pub dt: f64,
    pub v_threshold: f64,
}

impl RenshawCell {
    pub fn new() -> Self {
        Self {
            v: -65.0,
            h: 0.8,
            n: 0.1,
            adapt: 0.0,
            g_na: 35.0,
            g_k: 9.0,
            g_adapt: 5.0,
            g_l: 0.12,
            e_na: 55.0,
            e_k: -90.0,
            e_l: -65.0,
            c_m: 1.0,
            phi: 5.0,
            tau_adapt: 50.0,
            dt: 0.01,
            v_threshold: -20.0,
        }
    }

    fn voltage_valid(value: f64) -> bool {
        value.is_finite() && (-150.0..=100.0).contains(&value)
    }

    fn probability(value: f64) -> bool {
        value.is_finite() && (0.0..=1.0).contains(&value)
    }

    fn exact_gate(previous: f64, alpha: f64, beta: f64, phi: f64, dt: f64) -> Option<f64> {
        let total = phi * (alpha + beta);
        if !previous.is_finite()
            || !alpha.is_finite()
            || !beta.is_finite()
            || !total.is_finite()
            || !dt.is_finite()
            || total <= 0.0
        {
            return None;
        }
        let steady = alpha / (alpha + beta);
        Some((steady + (previous - steady) * (-total * dt).exp()).clamp(0.0, 1.0))
    }

    fn exact_relax(previous: f64, steady: f64, tau: f64, dt: f64) -> Option<f64> {
        if !previous.is_finite()
            || !steady.is_finite()
            || !tau.is_finite()
            || !dt.is_finite()
            || tau <= 0.0
        {
            return None;
        }
        Some((steady + (previous - steady) * (-dt / tau).exp()).clamp(0.0, 1.0))
    }

    fn valid_state(&self) -> bool {
        Self::voltage_valid(self.v)
            && Self::probability(self.h)
            && Self::probability(self.n)
            && Self::probability(self.adapt)
            && self.g_na.is_finite()
            && self.g_k.is_finite()
            && self.g_adapt.is_finite()
            && self.g_l.is_finite()
            && self.e_na.is_finite()
            && self.e_k.is_finite()
            && self.e_l.is_finite()
            && self.c_m.is_finite()
            && self.phi.is_finite()
            && self.tau_adapt.is_finite()
            && self.dt.is_finite()
            && self.v_threshold.is_finite()
            && self.g_na >= 0.0
            && self.g_k >= 0.0
            && self.g_adapt >= 0.0
            && self.g_l >= 0.0
            && self.c_m > 0.0
            && self.phi > 0.0
            && self.tau_adapt > 0.0
            && self.dt > 0.0
    }

    pub fn step(&mut self, current: f64) -> i32 {
        if !current.is_finite() || !self.valid_state() {
            return 0;
        }

        let v_prev = self.v;
        let mut v = self.v;
        let mut h = self.h;
        let mut n = self.n;
        let mut adapt = self.adapt;
        let n_sub = ((0.5 / self.dt.max(0.001)).max(1.0)) as usize;
        for _ in 0..n_sub {
            let am = safe_rate(0.1, 35.0, v, 10.0, 1.0);
            let bm = 4.0 * (-(v + 60.0) / 18.0).exp();
            let m_inf = am / (am + bm);
            let ah = 0.07 * (-(v + 58.0) / 20.0).exp();
            let bh = 1.0 / (1.0 + (-(v + 28.0) / 10.0).exp());
            let an = safe_rate(0.01, 34.0, v, 10.0, 0.1);
            let bn = 0.125 * (-(v + 44.0) / 80.0).exp();

            let Some(h_next) = Self::exact_gate(h, ah, bh, self.phi, self.dt) else {
                return 0;
            };
            let Some(n_next) = Self::exact_gate(n, an, bn, self.phi, self.dt) else {
                return 0;
            };

            let adapt_inf = 1.0 / (1.0 + (-(v + 30.0) / 5.0).exp());
            let Some(adapt_next) = Self::exact_relax(adapt, adapt_inf, self.tau_adapt, self.dt)
            else {
                return 0;
            };

            let g_na = self.g_na * m_inf.powi(3) * h_next;
            let g_k = self.g_k * n_next.powi(4);
            let g_adapt = self.g_adapt * adapt_next;
            let g_total = g_na + g_k + g_adapt + self.g_l;
            if !g_total.is_finite() || g_total <= 0.0 {
                return 0;
            }

            let steady_v = (current
                + g_na * self.e_na
                + g_k * self.e_k
                + g_adapt * self.e_k
                + self.g_l * self.e_l)
                / g_total;
            let v_next = steady_v + (v - steady_v) * (-(g_total / self.c_m) * self.dt).exp();
            if !Self::voltage_valid(v_next)
                || !Self::probability(h_next)
                || !Self::probability(n_next)
                || !Self::probability(adapt_next)
            {
                return 0;
            }

            v = v_next;
            h = h_next;
            n = n_next;
            adapt = adapt_next;
        }

        self.v = v;
        self.h = h;
        self.n = n;
        self.adapt = adapt;
        if self.v >= self.v_threshold && v_prev < self.v_threshold {
            1
        } else {
            0
        }
    }

    pub fn reset(&mut self) {
        self.v = -65.0;
        self.h = 0.8;
        self.n = 0.1;
        self.adapt = 0.0;
    }
}

impl Default for RenshawCell {
    fn default() -> Self {
        Self::new()
    }
}

// ═══════════════════════════════════════════════════════════════════
// Motor Unit (Alpha Motor Neuron + Muscle Fibre)
// ═══════════════════════════════════════════════════════════════════

/// Motor unit — functional unit of motor control: alpha motor neuron + muscle fibre.
///
/// Each spike from the embedded LIF motor neuron triggers a muscle twitch.
/// Force output is the summation of overlapping twitches (rate coding).
/// Higher firing rates → more twitch overlap → higher force (tetanus).
///
/// Muscle twitch modelled as a critically-damped second-order system:
/// f(t) = A * (t/τ) * exp(1 - t/τ), giving a smooth rise-then-decay.
///
/// Fuglevand et al., J. Neurophysiol. 70(6), 1993.
/// Heckman & Enoka, Compr. Physiol. 2(4), 2012.
#[derive(Clone, Debug)]
pub struct MotorUnit {
    pub v: f64,
    pub v_rest: f64,
    pub v_reset: f64,
    pub v_threshold: f64,
    pub tau_m: f64, // Membrane time constant (ms)
    pub adapt: f64,
    pub tau_adapt: f64,
    pub a_adapt: f64,
    pub gain: f64,
    // Muscle fibre
    pub force: f64,       // Current force output (normalised)
    pub twitch_amp: f64,  // Peak twitch amplitude
    pub tau_twitch: f64,  // Twitch contraction time (ms)
    pub force_decay: f64, // Force decay per step
    pub dt: f64,
}

impl MotorUnit {
    pub fn new() -> Self {
        Self::slow()
    }

    /// Slow motor unit (type S): small, fatigue-resistant, low force.
    pub fn slow() -> Self {
        Self {
            v: -65.0,
            v_rest: -65.0,
            v_reset: -70.0,
            v_threshold: -50.0,
            tau_m: 10.0,
            adapt: 0.0,
            tau_adapt: 100.0,
            a_adapt: 0.2,
            gain: 1.0,
            force: 0.0,
            twitch_amp: 0.05,
            tau_twitch: 90.0,
            force_decay: 0.0,
            dt: 0.5,
        }
    }

    /// Fast motor unit (type FF): large, fatigable, high force.
    pub fn fast() -> Self {
        Self {
            tau_m: 6.0,
            tau_adapt: 50.0,
            a_adapt: 0.1,
            twitch_amp: 0.3,
            tau_twitch: 30.0,
            ..Self::slow()
        }
    }

    fn voltage_valid(value: f64) -> bool {
        value.is_finite() && (-150.0..=100.0).contains(&value)
    }

    fn force_valid(value: f64) -> bool {
        value.is_finite() && (0.0..=1.0).contains(&value)
    }

    fn exact_relax(previous: f64, steady: f64, tau: f64, dt: f64) -> Option<f64> {
        if !previous.is_finite()
            || !steady.is_finite()
            || !tau.is_finite()
            || !dt.is_finite()
            || tau <= 0.0
            || dt <= 0.0
        {
            return None;
        }
        Some(steady + (previous - steady) * (-dt / tau).exp())
    }

    fn valid_state(&self) -> bool {
        Self::voltage_valid(self.v)
            && Self::voltage_valid(self.v_rest)
            && Self::voltage_valid(self.v_reset)
            && Self::voltage_valid(self.v_threshold)
            && Self::force_valid(self.force)
            && self.tau_m.is_finite()
            && self.adapt.is_finite()
            && self.tau_adapt.is_finite()
            && self.a_adapt.is_finite()
            && self.gain.is_finite()
            && self.twitch_amp.is_finite()
            && self.tau_twitch.is_finite()
            && self.force_decay.is_finite()
            && self.dt.is_finite()
            && self.tau_m > 0.0
            && self.tau_adapt > 0.0
            && self.tau_twitch > 0.0
            && self.dt > 0.0
            && self.gain >= 0.0
            && self.twitch_amp >= 0.0
            && self.v_reset < self.v_threshold
    }

    /// Step with descending drive (≥ 0). Returns spike (1/0). Force accessible via `.force`.
    pub fn step(&mut self, drive: f64) -> i32 {
        if !drive.is_finite() || !self.valid_state() {
            return 0;
        }

        let mut force = self.force * (-self.dt / self.tau_twitch).exp();
        let input = self.gain * drive.max(0.0) - self.adapt;
        let v_target = self.v_rest + input;
        let Some(mut v_candidate) = Self::exact_relax(self.v, v_target, self.tau_m, self.dt) else {
            return 0;
        };
        if !Self::voltage_valid(v_candidate) {
            return 0;
        }

        let adapt_target = self.a_adapt * (v_candidate - self.v_rest);
        let Some(adapt_candidate) =
            Self::exact_relax(self.adapt, adapt_target, self.tau_adapt, self.dt)
        else {
            return 0;
        };
        if !adapt_candidate.is_finite() {
            return 0;
        }

        let mut spike = 0;
        if v_candidate >= self.v_threshold {
            v_candidate = self.v_reset;
            force = (force + self.twitch_amp).min(1.0);
            spike = 1;
        }
        if !Self::voltage_valid(v_candidate) || !Self::force_valid(force) {
            return 0;
        }

        self.v = v_candidate;
        self.adapt = adapt_candidate;
        self.force = force;
        spike
    }

    pub fn reset(&mut self) {
        self.v = self.v_rest;
        self.adapt = 0.0;
        self.force = 0.0;
    }
}

impl Default for MotorUnit {
    fn default() -> Self {
        Self::new()
    }
}

// ═══════════════════════════════════════════════════════════════════
// Tests
// ═══════════════════════════════════════════════════════════════════

#[cfg(test)]
mod tests {
    use super::*;

    // ── Alpha Motor Neuron — 6-dimension coverage ──────────────────

    #[test]
    fn alpha_motor_fires_with_input() {
        let mut n = AlphaMotorNeuron::new();
        let spikes: i32 = (0..5000).map(|_| n.step(3.0)).sum();
        assert!(
            spikes > 0,
            "alpha motor must fire with sustained input: got {spikes}"
        );
    }

    #[test]
    fn alpha_motor_no_fire_without_input() {
        let mut n = AlphaMotorNeuron::new();
        let spikes: i32 = (0..3000).map(|_| n.step(0.0)).sum();
        assert_eq!(spikes, 0, "alpha motor should not fire at rest");
    }

    #[test]
    fn alpha_motor_negative_current_no_fire() {
        let mut n = AlphaMotorNeuron::new();
        let spikes: i32 = (0..2000).map(|_| n.step(-2.0)).sum();
        assert_eq!(spikes, 0);
    }

    #[test]
    fn alpha_motor_ahp_limits_rate() {
        // AHP from Ca2+-activated K+ should limit firing rate.
        // Compare: with AHP vs without (g_ahp=0).
        let mut with_ahp = AlphaMotorNeuron::new();
        let mut no_ahp = AlphaMotorNeuron::new();
        no_ahp.g_ahp = 0.0;
        let s_ahp: i32 = (0..5000).map(|_| with_ahp.step(5.0)).sum();
        let s_none: i32 = (0..5000).map(|_| no_ahp.step(5.0)).sum();
        assert!(
            s_ahp <= s_none + 5,
            "AHP should limit rate: with={s_ahp}, without={s_none}"
        );
    }

    #[test]
    fn alpha_motor_pic_responds_to_depolarisation() {
        // PIC (m_pic) should increase from baseline during sustained input.
        let mut n = AlphaMotorNeuron::new();
        let baseline = n.m_pic;
        for _ in 0..2000 {
            n.step(4.0);
        }
        assert!(
            n.m_pic > baseline + 0.001,
            "PIC should respond to depolarisation: baseline={baseline}, after={}",
            n.m_pic
        );
    }

    #[test]
    fn alpha_motor_ca_increases_during_spiking() {
        let mut n = AlphaMotorNeuron::new();
        for _ in 0..5000 {
            n.step(5.0);
        }
        assert!(
            n.ca > 0.0,
            "Ca2+ should accumulate during spiking: ca={}",
            n.ca
        );
    }

    #[test]
    fn alpha_motor_reset_roundtrip() {
        let mut n = AlphaMotorNeuron::new();
        for _ in 0..2000 {
            n.step(4.0);
        }
        n.reset();
        let mut fresh = AlphaMotorNeuron::new();
        let r1: i32 = (0..1000).map(|_| n.step(4.0)).sum();
        let r2: i32 = (0..1000).map(|_| fresh.step(4.0)).sum();
        assert_eq!(r1, r2, "reset neuron must match fresh");
    }

    #[test]
    fn alpha_motor_voltage_bounded() {
        let mut n = AlphaMotorNeuron::new();
        for _ in 0..10000 {
            n.step(10.0);
        }
        assert!(n.v.is_finite(), "voltage must stay finite");
        assert!(n.ca.is_finite(), "Ca2+ must stay finite");
        assert!(n.ca >= 0.0, "Ca2+ must be non-negative");
    }

    #[test]
    fn alpha_motor_nan_recovery() {
        let mut n = AlphaMotorNeuron::new();
        for _ in 0..100 {
            n.step(3.0);
        }
        for _ in 0..10 {
            let _ = n.step(f64::NAN);
        }
        n.reset();
        assert!(n.v.is_finite());
        assert!(n.ca >= 0.0);
    }

    #[test]
    fn alpha_motor_extreme_input() {
        let mut n = AlphaMotorNeuron::new();
        for _ in 0..50 {
            n.step(1e6);
        }
        n.reset();
        assert!(n.v.is_finite());
        for _ in 0..50 {
            n.step(-1e6);
        }
        n.reset();
        assert!(n.v.is_finite());
    }

    #[test]
    fn alpha_motor_performance() {
        let mut n = AlphaMotorNeuron::new();
        let start = std::time::Instant::now();
        for _ in 0..5_000 {
            n.step(4.0);
        }
        assert!(
            start.elapsed().as_millis() < 500,
            "5k steps took {:?}",
            start.elapsed()
        );
    }

    // ── Gamma Motor Neuron — 6-dimension coverage ──────────────────

    #[test]
    fn gamma_dynamic_fires_with_drive() {
        let mut n = GammaMotorNeuron::dynamic();
        let spikes: i32 = (0..2000).map(|_| n.step(20.0)).sum();
        assert!(spikes > 0, "gamma dynamic must fire: got {spikes}");
    }

    #[test]
    fn gamma_static_fires_with_drive() {
        let mut n = GammaMotorNeuron::static_type();
        let spikes: i32 = (0..2000).map(|_| n.step(20.0)).sum();
        assert!(spikes > 0, "gamma static must fire: got {spikes}");
    }

    #[test]
    fn gamma_no_fire_without_drive() {
        let mut n = GammaMotorNeuron::new();
        let spikes: i32 = (0..1000).map(|_| n.step(0.0)).sum();
        assert_eq!(spikes, 0);
    }

    #[test]
    fn gamma_negative_drive_no_fire() {
        let mut n = GammaMotorNeuron::new();
        // drive.max(0.0) clamps negatives
        let spikes: i32 = (0..1000).map(|_| n.step(-10.0)).sum();
        assert_eq!(spikes, 0);
    }

    #[test]
    fn gamma_adaptation_reduces_rate() {
        let mut n = GammaMotorNeuron::new();
        let first: i32 = (0..1000).map(|_| n.step(20.0)).sum();
        let second: i32 = (0..1000).map(|_| n.step(20.0)).sum();
        assert!(
            second <= first + 3,
            "gamma should adapt: first={first}, second={second}"
        );
    }

    #[test]
    fn gamma_static_adapts_more_than_dynamic() {
        let mut dyn_ = GammaMotorNeuron::dynamic();
        let mut stat = GammaMotorNeuron::static_type();
        let dyn_spikes: i32 = (0..2000).map(|_| dyn_.step(20.0)).sum();
        let stat_spikes: i32 = (0..2000).map(|_| stat.step(20.0)).sum();
        // Static uses larger adaptation coupling and lower excitability.
        assert!(
            stat_spikes <= dyn_spikes + 5,
            "static ({stat_spikes}) should fire <= dynamic ({dyn_spikes})"
        );
    }

    #[test]
    fn gamma_reset_roundtrip() {
        let mut n = GammaMotorNeuron::new();
        for _ in 0..1000 {
            n.step(20.0);
        }
        n.reset();
        let mut fresh = GammaMotorNeuron::new();
        let r1: i32 = (0..500).map(|_| n.step(20.0)).sum();
        let r2: i32 = (0..500).map(|_| fresh.step(20.0)).sum();
        assert_eq!(r1, r2);
    }

    #[test]
    fn gamma_voltage_bounded() {
        let mut n = GammaMotorNeuron::new();
        for _ in 0..10000 {
            n.step(50.0);
        }
        assert!(n.v.is_finite());
        assert!(n.adapt.is_finite());
    }

    #[test]
    fn gamma_nan_recovery() {
        let mut n = GammaMotorNeuron::new();
        for _ in 0..50 {
            n.step(20.0);
        }
        let before_v = n.v;
        let before_adapt = n.adapt;
        for _ in 0..10 {
            let _ = n.step(f64::NAN);
        }
        assert_eq!(n.v, before_v);
        assert_eq!(n.adapt, before_adapt);
        n.reset();
        assert!(n.v.is_finite());
        assert_eq!(n.adapt, 0.0);
    }

    #[test]
    fn gamma_extreme_input() {
        let mut n = GammaMotorNeuron::new();
        for _ in 0..50 {
            n.step(1e6);
        }
        n.reset();
        assert!(n.v.is_finite());
    }

    #[test]
    fn gamma_corrupted_state_preserved_on_step() {
        let mut n = GammaMotorNeuron::new();
        n.tau = 0.0;
        let before_v = n.v;
        let before_adapt = n.adapt;
        assert_eq!(n.step(20.0), 0);
        assert_eq!(n.v, before_v);
        assert_eq!(n.adapt, before_adapt);
    }

    #[test]
    fn gamma_performance() {
        let mut n = GammaMotorNeuron::new();
        let start = std::time::Instant::now();
        for _ in 0..100_000 {
            n.step(20.0);
        }
        assert!(
            start.elapsed().as_millis() < 50,
            "100k steps took {:?}",
            start.elapsed()
        );
    }

    // ── Upper Motor Neuron — 6-dimension coverage ──────────────────

    #[test]
    fn upper_motor_fires_with_input() {
        let mut n = UpperMotorNeuron::new();
        let spikes: i32 = (0..10000).map(|_| n.step(5.0)).sum();
        assert!(spikes > 0, "upper motor must fire: got {spikes}");
    }

    #[test]
    fn upper_motor_no_fire_without_input() {
        let mut n = UpperMotorNeuron::new();
        let spikes: i32 = (0..5000).map(|_| n.step(0.0)).sum();
        assert_eq!(spikes, 0);
    }

    fn upper_motor_reference_step(mut n: UpperMotorNeuron, current: f64) -> UpperMotorNeuron {
        fn gate(previous: f64, alpha: f64, beta: f64, dt: f64) -> f64 {
            let total = alpha + beta;
            let steady = alpha / total;
            (steady + (previous - steady) * (-total * dt).exp()).clamp(0.0, 1.0)
        }
        fn gate_inf(previous: f64, steady: f64, tau: f64, dt: f64) -> f64 {
            (steady + (previous - steady) * (-dt / tau).exp()).clamp(0.0, 1.0)
        }
        let vt = -56.2;
        for _ in 0..4 {
            let dv = n.v - vt;
            let x_m = dv - 13.0;
            let alpha_m = if x_m.abs() < 1e-6 {
                0.32 * 4.0
            } else {
                -0.32 * x_m / ((-x_m / 4.0).exp() - 1.0)
            };
            let x_bm = dv - 40.0;
            let beta_m = if x_bm.abs() < 1e-6 {
                0.28 * 5.0
            } else {
                0.28 * x_bm / ((x_bm / 5.0).exp() - 1.0)
            };
            let alpha_h = 0.128 * (-(dv - 17.0) / 18.0).exp();
            let beta_h = 4.0 / (1.0 + (-(dv - 40.0) / 5.0).exp());
            let x_n = dv - 15.0;
            let alpha_n = if x_n.abs() < 1e-6 {
                0.032 * 5.0
            } else {
                -0.032 * x_n / ((-x_n / 5.0).exp() - 1.0)
            };
            let beta_n = 0.5 * (-(dv - 10.0) / 40.0).exp();

            n.m = gate(n.m, alpha_m, beta_m, n.dt);
            n.h = gate(n.h, alpha_h, beta_h, n.dt);
            n.n = gate(n.n, alpha_n, beta_n, n.dt);

            let p_inf = 1.0 / (1.0 + (-(n.v + 35.0) / 10.0).exp());
            let tau_p = 400.0 / (3.3 * ((n.v + 35.0) / 20.0).exp() + (-(n.v + 35.0) / 20.0).exp());
            n.p = gate_inf(n.p, p_inf, tau_p, n.dt);

            let s_inf = 1.0 / (1.0 + (-(n.v + 20.0) / 5.0).exp());
            n.s = gate_inf(n.s, s_inf, 10.0, n.dt);

            let g_na = n.g_na * n.m.powi(3) * n.h;
            let g_k = n.g_k * n.n.powi(4);
            let g_m = n.g_m * n.p;
            let g_ca = n.g_ca * n.s.powi(2);
            let g_total = g_na + g_k + g_m + g_ca + n.g_l;
            let steady_v = (current
                + g_na * n.e_na
                + g_k * n.e_k
                + g_m * n.e_k
                + g_ca * n.e_ca
                + n.g_l * n.e_l)
                / g_total;
            n.v = steady_v + (n.v - steady_v) * (-(g_total / n.c_m) * n.dt).exp();
        }
        n
    }

    #[test]
    fn upper_motor_uses_exact_gate_and_conductance_membrane_step() {
        let mut n = UpperMotorNeuron::new();
        let expected = upper_motor_reference_step(n.clone(), 5.0);

        assert_eq!(n.step(5.0), 0);

        assert!((n.v - expected.v).abs() <= 1e-12);
        assert!((n.m - expected.m).abs() <= 1e-12);
        assert!((n.h - expected.h).abs() <= 1e-12);
        assert!((n.n - expected.n).abs() <= 1e-12);
        assert!((n.p - expected.p).abs() <= 1e-12);
        assert!((n.s - expected.s).abs() <= 1e-12);
    }

    #[test]
    fn upper_motor_corrupted_gate_is_preserved_on_step() {
        let mut n = UpperMotorNeuron::new();
        n.m = 1.5;
        let before = n.clone();

        assert_eq!(n.step(5.0), 0);

        assert_eq!(n.v, before.v);
        assert_eq!(n.m, before.m);
        assert_eq!(n.h, before.h);
        assert_eq!(n.n, before.n);
        assert_eq!(n.p, before.p);
        assert_eq!(n.s, before.s);
    }

    #[test]
    fn upper_motor_negative_current_no_fire() {
        let mut n = UpperMotorNeuron::new();
        let spikes: i32 = (0..2000).map(|_| n.step(-5.0)).sum();
        assert_eq!(spikes, 0);
    }

    #[test]
    fn upper_motor_adaptation_via_m_current() {
        let mut n = UpperMotorNeuron::new();
        let first: i32 = (0..5000).map(|_| n.step(5.0)).sum();
        let second: i32 = (0..5000).map(|_| n.step(5.0)).sum();
        assert!(
            second <= first + 3,
            "M-current should cause adaptation: first={first}, second={second}"
        );
    }

    #[test]
    fn upper_motor_ca_activates_during_depolarisation() {
        let mut n = UpperMotorNeuron::new();
        let baseline = n.s;
        for _ in 0..5000 {
            n.step(5.0);
        }
        assert!(
            n.s > baseline + 0.001,
            "Ca2+ gate should activate: s={}",
            n.s
        );
    }

    #[test]
    fn upper_motor_reset_roundtrip() {
        let mut n = UpperMotorNeuron::new();
        for _ in 0..3000 {
            n.step(5.0);
        }
        n.reset();
        let mut fresh = UpperMotorNeuron::new();
        let r1: i32 = (0..2000).map(|_| n.step(5.0)).sum();
        let r2: i32 = (0..2000).map(|_| fresh.step(5.0)).sum();
        assert_eq!(r1, r2);
    }

    #[test]
    fn upper_motor_voltage_bounded() {
        let mut n = UpperMotorNeuron::new();
        for _ in 0..20000 {
            n.step(10.0);
        }
        assert!(n.v.is_finite());
        assert!(n.p.is_finite());
        assert!(n.s.is_finite());
    }

    #[test]
    fn upper_motor_nan_recovery() {
        let mut n = UpperMotorNeuron::new();
        for _ in 0..100 {
            n.step(5.0);
        }
        for _ in 0..10 {
            let _ = n.step(f64::NAN);
        }
        n.reset();
        assert!(n.v.is_finite());
    }

    #[test]
    fn upper_motor_extreme_input() {
        let mut n = UpperMotorNeuron::new();
        for _ in 0..50 {
            n.step(1e6);
        }
        n.reset();
        assert!(n.v.is_finite());
    }

    #[test]
    fn upper_motor_performance() {
        let mut n = UpperMotorNeuron::new();
        let start = std::time::Instant::now();
        for _ in 0..10_000 {
            n.step(5.0);
        }
        assert!(
            start.elapsed().as_millis() < 100,
            "10k steps took {:?}",
            start.elapsed()
        );
    }

    // ── Renshaw Cell — 6-dimension coverage ────────────────────────

    #[test]
    fn renshaw_fires_with_input() {
        let mut n = RenshawCell::new();
        let spikes: i32 = (0..5000).map(|_| n.step(3.0)).sum();
        assert!(spikes > 0, "Renshaw must fire: got {spikes}");
    }

    #[test]
    fn renshaw_no_fire_without_input() {
        let mut n = RenshawCell::new();
        let spikes: i32 = (0..3000).map(|_| n.step(0.0)).sum();
        assert_eq!(spikes, 0);
    }

    #[test]
    fn renshaw_negative_current_no_fire() {
        let mut n = RenshawCell::new();
        let spikes: i32 = (0..2000).map(|_| n.step(-2.0)).sum();
        assert_eq!(spikes, 0);
    }

    fn renshaw_test_gate(previous: f64, alpha: f64, beta: f64, phi: f64, dt: f64) -> f64 {
        let total = phi * (alpha + beta);
        let steady = alpha / (alpha + beta);
        (steady + (previous - steady) * (-total * dt).exp()).clamp(0.0, 1.0)
    }

    fn renshaw_test_adapt(previous: f64, steady: f64, tau: f64, dt: f64) -> f64 {
        (steady + (previous - steady) * (-dt / tau).exp()).clamp(0.0, 1.0)
    }

    fn renshaw_reference_step(mut n: RenshawCell, current: f64) -> RenshawCell {
        let n_sub = ((0.5 / n.dt.max(0.001)).max(1.0)) as usize;
        for _ in 0..n_sub {
            let am = safe_rate(0.1, 35.0, n.v, 10.0, 1.0);
            let bm = 4.0 * (-(n.v + 60.0) / 18.0).exp();
            let ah = 0.07 * (-(n.v + 58.0) / 20.0).exp();
            let bh = 1.0 / (1.0 + (-(n.v + 28.0) / 10.0).exp());
            let an = safe_rate(0.01, 34.0, n.v, 10.0, 0.1);
            let bn = 0.125 * (-(n.v + 44.0) / 80.0).exp();
            let m_inf = am / (am + bm);

            n.h = renshaw_test_gate(n.h, ah, bh, n.phi, n.dt);
            n.n = renshaw_test_gate(n.n, an, bn, n.phi, n.dt);
            let adapt_inf = 1.0 / (1.0 + (-(n.v + 30.0) / 5.0).exp());
            n.adapt = renshaw_test_adapt(n.adapt, adapt_inf, n.tau_adapt, n.dt);

            let g_na = n.g_na * m_inf.powi(3) * n.h;
            let g_k = n.g_k * n.n.powi(4);
            let g_adapt = n.g_adapt * n.adapt;
            let g_total = g_na + g_k + g_adapt + n.g_l;
            let steady_v =
                (current + g_na * n.e_na + g_k * n.e_k + g_adapt * n.e_k + n.g_l * n.e_l) / g_total;
            n.v = steady_v + (n.v - steady_v) * (-(g_total / n.c_m) * n.dt).exp();
        }
        n
    }

    fn renshaw_snapshot(n: &RenshawCell) -> (f64, f64, f64, f64) {
        (n.v, n.h, n.n, n.adapt)
    }

    #[test]
    fn renshaw_uses_exact_gate_and_conductance_membrane_step() {
        let mut n = RenshawCell::new();
        let expected = renshaw_reference_step(RenshawCell::new(), 4.0);

        assert_eq!(n.step(4.0), 0);

        assert!((n.v - expected.v).abs() <= 1e-12);
        assert!((n.h - expected.h).abs() <= 1e-12);
        assert!((n.n - expected.n).abs() <= 1e-12);
        assert!((n.adapt - expected.adapt).abs() <= 1e-12);
    }

    #[test]
    fn renshaw_rejects_invalid_current_without_state_mutation() {
        let mut n = RenshawCell::new();
        for _ in 0..20 {
            n.step(4.0);
        }
        let before = renshaw_snapshot(&n);

        assert_eq!(n.step(f64::NAN), 0);
        assert_eq!(renshaw_snapshot(&n), before);
        assert_eq!(n.step(f64::INFINITY), 0);
        assert_eq!(renshaw_snapshot(&n), before);
    }

    #[test]
    fn renshaw_rejects_excess_current_without_state_mutation() {
        let mut n = RenshawCell::new();
        let before = renshaw_snapshot(&n);

        assert_eq!(n.step(1.0e8), 0);

        assert_eq!(renshaw_snapshot(&n), before);
    }

    #[test]
    fn renshaw_corrupted_gate_is_preserved_on_step() {
        let mut n = RenshawCell::new();
        n.h = 1.5;
        let before = renshaw_snapshot(&n);

        assert_eq!(n.step(4.0), 0);

        assert_eq!(renshaw_snapshot(&n), before);
    }

    #[test]
    fn renshaw_burst_then_adapt() {
        // Renshaw should fire more in the first epoch than the second
        let mut n = RenshawCell::new();
        let first: i32 = (0..2000).map(|_| n.step(4.0)).sum();
        let second: i32 = (0..2000).map(|_| n.step(4.0)).sum();
        assert!(
            second <= first + 5,
            "Renshaw should adapt: first={first}, second={second}"
        );
    }

    #[test]
    fn renshaw_adapt_increases_during_firing() {
        let mut n = RenshawCell::new();
        let baseline = n.adapt;
        for _ in 0..3000 {
            n.step(4.0);
        }
        assert!(
            n.adapt > baseline + 0.01,
            "adaptation variable should increase: adapt={}",
            n.adapt
        );
    }

    #[test]
    fn renshaw_reset_roundtrip() {
        let mut n = RenshawCell::new();
        for _ in 0..2000 {
            n.step(4.0);
        }
        n.reset();
        let mut fresh = RenshawCell::new();
        let r1: i32 = (0..1000).map(|_| n.step(4.0)).sum();
        let r2: i32 = (0..1000).map(|_| fresh.step(4.0)).sum();
        assert_eq!(r1, r2);
    }

    #[test]
    fn renshaw_voltage_bounded() {
        let mut n = RenshawCell::new();
        for _ in 0..10000 {
            n.step(10.0);
        }
        assert!(n.v.is_finite());
        assert!(n.adapt.is_finite());
    }

    #[test]
    fn renshaw_nan_recovery() {
        let mut n = RenshawCell::new();
        for _ in 0..100 {
            n.step(3.0);
        }
        for _ in 0..10 {
            let _ = n.step(f64::NAN);
        }
        n.reset();
        assert!(n.v.is_finite());
        assert_eq!(n.adapt, 0.0);
    }

    #[test]
    fn renshaw_extreme_input() {
        let mut n = RenshawCell::new();
        for _ in 0..50 {
            n.step(1e6);
        }
        n.reset();
        assert!(n.v.is_finite());
    }

    #[test]
    fn renshaw_performance() {
        let mut n = RenshawCell::new();
        let start = std::time::Instant::now();
        for _ in 0..5_000 {
            n.step(4.0);
        }
        assert!(
            start.elapsed().as_millis() < 500,
            "5k steps took {:?}",
            start.elapsed()
        );
    }

    // ── Motor Unit — 6-dimension coverage ──────────────────────────

    #[test]
    fn motor_unit_fires_with_drive() {
        let mut mu = MotorUnit::new();
        let spikes: i32 = (0..2000).map(|_| mu.step(20.0)).sum();
        assert!(spikes > 0, "motor unit must fire: got {spikes}");
    }

    #[test]
    fn motor_unit_no_fire_without_drive() {
        let mut mu = MotorUnit::new();
        let spikes: i32 = (0..1000).map(|_| mu.step(0.0)).sum();
        assert_eq!(spikes, 0);
    }

    #[test]
    fn motor_unit_negative_drive_no_fire() {
        let mut mu = MotorUnit::new();
        let spikes: i32 = (0..1000).map(|_| mu.step(-10.0)).sum();
        assert_eq!(spikes, 0);
    }

    #[test]
    fn motor_unit_force_increases_with_spikes() {
        let mut mu = MotorUnit::new();
        assert_eq!(mu.force, 0.0);
        for _ in 0..2000 {
            mu.step(20.0);
        }
        assert!(
            mu.force > 0.0,
            "force should increase during spiking: f={}",
            mu.force
        );
    }

    #[test]
    fn motor_unit_force_decays_without_input() {
        let mut mu = MotorUnit::new();
        // Build up force
        for _ in 0..1000 {
            mu.step(20.0);
        }
        let peak = mu.force;
        assert!(peak > 0.0);
        // No input → force decays
        for _ in 0..5000 {
            mu.step(0.0);
        }
        assert!(
            mu.force < peak,
            "force should decay: peak={peak}, now={}",
            mu.force
        );
    }

    #[test]
    fn motor_unit_fast_produces_more_force() {
        let mut slow = MotorUnit::slow();
        let mut fast = MotorUnit::fast();
        for _ in 0..2000 {
            slow.step(20.0);
            fast.step(20.0);
        }
        assert!(
            fast.force >= slow.force,
            "fast MU ({}) should produce >= force than slow ({})",
            fast.force,
            slow.force
        );
    }

    #[test]
    fn motor_unit_force_capped_at_one() {
        let mut mu = MotorUnit::fast();
        for _ in 0..10000 {
            mu.step(50.0);
        }
        assert!(mu.force <= 1.0, "force must not exceed 1.0: f={}", mu.force);
    }

    #[test]
    fn motor_unit_reset_roundtrip() {
        let mut mu = MotorUnit::new();
        for _ in 0..1000 {
            mu.step(20.0);
        }
        mu.reset();
        assert_eq!(mu.force, 0.0);
        assert_eq!(mu.adapt, 0.0);
        let mut fresh = MotorUnit::new();
        let r1: i32 = (0..500).map(|_| mu.step(20.0)).sum();
        let r2: i32 = (0..500).map(|_| fresh.step(20.0)).sum();
        assert_eq!(r1, r2);
    }

    #[test]
    fn motor_unit_voltage_bounded() {
        let mut mu = MotorUnit::new();
        for _ in 0..10000 {
            mu.step(50.0);
        }
        assert!(mu.v.is_finite());
        assert!(mu.force.is_finite());
    }

    #[test]
    fn motor_unit_nan_recovery() {
        let mut mu = MotorUnit::new();
        for _ in 0..50 {
            mu.step(20.0);
        }
        for _ in 0..10 {
            let _ = mu.step(f64::NAN);
        }
        mu.reset();
        assert!(mu.v.is_finite());
        assert_eq!(mu.force, 0.0);
    }

    #[test]
    fn motor_unit_performance() {
        let mut mu = MotorUnit::new();
        let start = std::time::Instant::now();
        for _ in 0..100_000 {
            mu.step(20.0);
        }
        let elapsed = start.elapsed();
        assert!(elapsed.as_millis() < 100, "100k steps took {:?}", elapsed);
    }
}