On the Contraction of Excitable Systems
Alessandro Cecconi, Michelangelo Bin, Lorenzo Marconi, Rodolphe Sepulchre · 2025
Read On the Contraction of Excitable Systems by Alessandro Cecconi, Michelangelo Bin, Lorenzo Marconi, Rodolphe Sepulchre (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-15477.txt.
Introduction
Alessandro Cecconi, Michelangelo Bin, Lorenzo Marconi, and Rodolphe Sepulchre study contraction in excitable systems, using the Hodgkin-Huxley model of a neuron as their main case. Contraction means nearby states move closer over time, so small differences fade rather than grow. They ask when that property holds, with and without incoming synaptic pulses, and what it means for spike timing.
Without input, the model contracts in the range that matters for real physiology. With pulse inputs, contraction survives when input events stay sparse enough, and it is lost when the firing rate grows too high. Spike timings prove reliable exactly in the contracting regime, linking a math property to trust in timing.
Catalog record
| CREATOR | Alessandro Cecconi, Michelangelo Bin, Lorenzo Marconi, Rodolphe Sepulchre |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-15477
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On the Contraction of Excitable Systems Alessandro Cecconi, Michelangelo Bin, Lorenzo Marconi, Rodolphe Sepulchre Abstract We analyze contraction in conductance-based excitable systems and link it to reliable spike timing. For a Hodgkin-Huxley neuron with synaptic input, we prove the unforced dynamics is incrementally exponentially stable on a compact physiological set. With impulsive inputs, contraction persists under an average dwell-time condition, and for periodic spike trains we derive an explicit lower bound on the inter-impulse period. In the high-rate limit the synapse is effectively held open, the conductance is nearly constant, and self-sustained limit-cycle oscillations can arise, so contraction no longer holds. This continuous-time view explains when spike times extracted by an event detector remain robust across trials, and simulations confirm both regimes. I Introduction Excitable systems, exemplified by conductance-based models, mix the continuous and the discrete. Their trajectories follow differential equations but consist of sequences of discrete events [ 1 ] . Building on this mixed description, neuromorphic control exploits excitable internal models to achieve event regulation, rather than trajectory regulation. Internal event generators can synchronize with external events despite a persistent mismatch between the corresponding continuous-time trajectories [ 2 , 3 ] . In neuroscience, reliability refers to the trial-to-trial reproducibility of spike times under the same input stimulation, with output spike voltages that lock to input synaptic currents despite variability in initial conditions and other perturbations [ 4 , 5 , 6 ] . The objective of this work is to show that an excitable behavior becomes contractive when regarded as an event generator, meaning that the input current keeps the state variables near an exponentially stable equilibrium except for brief spikes, that is, short and sharp excursions separated by refractory periods. In this setting, the continuous contraction property of the flow explains the discrete reliability of the observed events. Contraction theory formalizes robustness as incremental exponential convergence of trajectories driven by the same input [ 7 , 8 , 9 ] . In excitable dynamics contraction is input dependent [ 10 , 11 ] . Constant or effectively constant inputs can generate limit-cycle oscillations that are not contractive and that remain sensitive to phase perturbations [ 12 , 13 , 14 ] . In contrast, subthreshold inputs, or event inputs that trigger spikes without holding the system away from its stable rest, result in contracting behavior with trajectories that draw together and with spike times that align up to bounded timing variability. This input-selective viewpoint clarifies how the very same neuron can appear phase-sensitive in a tonic regime yet display highly reproducible timing when driven by sparse triggering signals. We formalize this viewpoint on a standard conductance-based formulation with gating dynamics and a first-order synaptic state. Physiological bounds on voltage and gates define a compact forward-invariant set on which the unforced Hodgkin–Huxley flow is incrementally exponentially stable. To connect the continuous flow with discrete observations we pair the model with an event detector that maps pre-synaptic spike trains to post-synaptic spike times. When the flow is contractive, different initial conditions under the same admissible input lead to trajectories that converge and to event sequences that coincide after a short transient. We then study impulsive synaptic drives and identify conditions under which contraction, and therefore event-level reliability, persists. For general spike trains, an average dwell-time requirement ensures impulses are sufficiently spaced so that the state returns near equilibrium between events. For periodic spike trains, this requirement becomes a lower bound on the inter-impulse period that depends on the contraction margin of the unforced flow. The opposite regime reveals the limits of reliability. At high input rates the synaptic state saturates, the effective conductance becomes nearly constant, and the neuron may enter self-sustained periodic firing on a limit cycle, which destroys global contractivity and makes timing sensitive to phase. Numerical experiments on a Hodgkin–Huxley neuron illustrate both behaviors predicted by the analysis. With adequately spaced inputs, trajectories contract and spike times lock tightly after a brief transient. With high-rate drives, tonic oscillations exhibit persistent phase offsets across trials and timing variability does not shrink. Together these findings provide a simple and testable account of when excitable systems are contractive at the level of events and why their spike timings can be robust in practice. II Modeling Excitable Systems We consider excitable systems modeled by the following conductance-based model \displaystyle\tau_{j}(v)\dot{x}_{j} \displaystyle=-x_{j}+\mu_{j}(v),\qquad j=1,\dots,m, (1) \displaystyle C\dot{v} \displaystyle=-\bar{g}_{\mathrm{leak}}(v-E_{\mathrm{leak}})-\sum_{k=1}^{n}i_{k}(x,v)+i_{s}, where v\in\mathbb{R} is the membrane voltage, and x_{j}\in\mathbb{R} are gating variables. The functions \mu_{j}:\mathbb{R}\to(0,1) and \tau_{j}:\mathbb{R}\to\mathbb{R}_{+} are the steady-state activation and the voltage-dependent time constant, respectively. These functions typically have a sigmoidal shape [ 13 , Sec. 1.8] . The quantities i_{k}(x,v) represent internal ionic currents. They have an Ohmic form i_{k}(x,v)=\bar{g}_{k}\,\varphi_{k}(x)\,(v-E_{k}),\qquad k=1,\dots,n, where \bar{g}_{k}>0 is the maximal conductance, E_{k}\in\mathbb{R} the corresponding reversal potential, and \varphi_{k}:\mathbb{R}^{m}\to(0,1) is an activation function determining the fraction of open channels. The variable i_{s} denotes an external synaptic current, which is modeled by \displaystyle\dot{s}=-\frac{1}{\tau_{s}}s+\alpha(1-s)v_{\text{pre}}(t),\quad i_{s}(t)=-\bar{g}_{s}s(v-E_{s}), (2) where \alpha\in(0,1] , \tau_{s}>0 is the synaptic time constant, \bar{g}_{s}>0 is the maximal synaptic conductance, E_{s}\in\mathbb{R} is the synaptic reversal potential, and v_{\text{pre}}(t) is the pre-synaptic voltage, the model input. The synapse is a one-port element of the same type as the internal currents. We model pre-synaptic spikes by a locally finite set of times. Let \mathbb{U}\coloneq\{\,I\subseteq[0,\infty):\forall T>0,\ \card(I\cap[0,T])<\infty\,\}. Each \mathcal{I}\in\mathbb{U} admits a unique strictly increasing sequence \mathcal{I}=\{t_{\ell}\}_{\ell=1}^{N} with N\in\mathbb{N} . With each \mathcal{I} we associate the spike-train signal v_{\text{pre}}(t)\coloneq\sum_{t_{\ell}\in\mathcal{I}}\delta(t-t_{\ell}), (3) where \delta is the Dirac distribution. If t_{\ell}=\ell T , then the spike train is periodic. This representation is standard in neuroscience and neuromorphic engineering [ 14 , 15 ] . The Hodgkin–Huxley model [ 16 ] is obtained as a particular case for m=3 gates modulating n=2 currents i_{\mathrm{Na}}=\bar{g}_{\mathrm{Na}}\,x_{1}^{3}\,x_{2}(v-E_{\text{Na}}),\qquad i_{\mathrm{K}}=\bar{g}_{\mathrm{K}}\,x_{3}^{4}(v-E_{\text{K}}), with i_{\mathrm{Na}} , and i_{\mathrm{K}} representing the sodium and potassium current, respectively. The equivalent circuit is shown in Figure 1 . Figure 1 : Circuit schematic of the conductance-based model ( 1 ) with the external synapse ( 2 ). In physiological conditions, neurons have a well-defined resting state, which corresponds to a unique stable equilibrium. Furthermore, the membrane voltage remains confined in an interval determined by the minimum and maximum of the reverse potentials, that is, in the case of the Hodgkin-Huxley, the Nernst potential of potassium and the Nernst potential of sodium, respectively [ 12 ] . We link such conditions to our model ( 1 )-( 2 ). With v_{\text{pre}}\equiv 0 , the unforced voltage dynamics in ( 1 ) can be rewritten as C\dot{v}=-G(x,s)\big(v-E(x,s)\big), (4) where G(x,s)\coloneq\bar{g}_{\mathrm{leak}}+\sum_{k=1}^{n}\bar{g}_{k}\varphi_{k}(x)+\bar{g}_{s}s\geq\bar{g}_{\text{leak}}>0, (5) and E(x,s)\coloneq\frac{\bar{g}_{\mathrm{leak}}E_{\mathrm{leak}}+\sum_{k}\bar{g}_{k}\varphi_{k}(x)E_{k}+\bar{g}_{s}s\,E_{s}}{\bar{g}_{\mathrm{leak}}+\sum_{k}\bar{g}_{k}\varphi_{k}(x)+\bar{g}_{s}s}. (6) Hence, E({x},s) is a convex combination of reversal potentials. In particular, let E_{\text{min}}\coloneq\min\{E_{\mathrm{leak}},E_{k},E_{s}\} , and E_{\text{max}}\coloneq\max\{E_{\mathrm{leak}},E_{k},E_{s}\} . Then, for every (x,v) , it holds E(x,s)\in[E_{\text{min}},E_{\text{max}}]. For instance, for the Hodgkin–Huxley model one has E({x},s)\in[E_{\text{K}},E_{\text{Na}}] , provided that E_{\text{s}}\in(E_{\text{K}},E_{\text{Na}}) . More in general, one can show the following. Lemma 1 . With \mathcal{S}=[0,1] , \mathcal{X}=[0,1]^{m} , \mathcal{V}=[E_{\min},E_{\max}] , define \mathcal{Z}\coloneq\mathcal{S}\times\mathcal{X}\times\mathcal{V} . Then, \mathcal{Z} is compact and forward invariant for ( 1 )–( 2 ) under ( 3 ). Proof. See Appendix. As for the rest state, we notice that an equilibrium (s^{*},x^{*},v^{*}) of ( 1 )–( 2 ) with v_{\text{pre}}\equiv 0 , necessarily satisfies s^{*}=0 , and the current balance \bar{g}_{\mathrm{leak}}(v^{*}-E_{\mathrm{leak}})+\sum_{k=1}^{n}\bar{g}_{k}\varphi_{k}(x^{*})(v^{*}-E_{k})=0. (7) Assuming, as customary, v^{*}=0 (which