Coupling-Induced Multistability
Coupling-induced multistability is the phenomenon in which a dynamical system, composed of interacting subsystems that are each monostable in isolation, exhibits multiple stable states — attractors — solely as a consequence of the coupling between them. The individual components have no capacity for multistability; the system as a whole does, and this capacity is an emergent property of the interaction topology.
The phenomenon is distinct from parametric multistability, in which a single system exhibits multiple attractors because its parameters place it in a bistable regime. In coupling-induced multistability, the parameters of the individual systems and the coupling strength can be fixed, and yet the coupled system will support multiple attractors that the uncoupled systems cannot. The multistability is a property of the whole, not of the parts.
Mechanism
The simplest example is two coupled bistable oscillators. Each oscillator, in isolation, has a single stable limit cycle. When coupled, the system of two oscillators can exhibit multiple stable phase-locked states — in-phase, anti-phase, and various asymmetric phase relations — depending on the coupling function and the initial conditions. The phase space of the coupled system contains attractors that do not exist in the phase space of either oscillator alone.
More generally, coupling-induced multistability arises whenever the coupling function introduces nonlinear terms that create new fixed points or limit cycles in the joint phase space. In neural networks, synaptic coupling with state-dependent plasticity rules can create multiple stable activity patterns from neurons that are each tonically active in isolation. In gene regulatory networks, the cooperative binding of transcription factors can create multiple stable expression states from genes that are each constitutively expressed when uncoupled.
Relation to Emergence and Systems Theory
Coupling-induced multistability is a paradigmatic case of emergence: a property of the whole (multiple attractors) that is not present in the parts and cannot be predicted from the properties of the parts in isolation. It is also a paradigmatic case of downward causation: the system's occupation of a particular attractor constrains the behavior of the individual components, which would otherwise explore their full phase space.
The phenomenon has direct implications for control theory. A system with coupling-induced multistability cannot be controlled by acting on individual components alone, because the control input must be strong enough to push the system across a separatrix — the boundary between basins of attraction. Local control strategies fail because the attractor structure is global. This is why many biological and social systems resist piecemeal intervention: the system's stable states are collective properties, and changing them requires collective action.
Open Questions
- What are the necessary and sufficient topological conditions for coupling-induced multistability? Not all coupled systems exhibit it; the graph structure and the coupling function must satisfy specific constraints.
- How does coupling-induced multistability relate to symmetry breaking in physical systems? Both produce multiple stable states from symmetric initial conditions, but the mechanisms differ.
- Can coupling-induced multistability be exploited for memory and computation? The multiple attractors of a coupled system can serve as discrete memory states, and transitions between them can serve as computational operations.
Coupling-induced multistability is the refutation of reductionism in dynamical form. It demonstrates that you cannot understand what a system can do by understanding what its parts can do. The parts, in isolation, are simple. The whole, in interaction, is not. The difference is not complexity; it is organization.