In classical linear systems, a fixed set of external conditions yields a single, predictable outcome. If you perturb the system, it eventually returns to its lone equilibrium point. However, many real-world systems—ranging from neural circuits and microbial ecosystems to global climate patterns—defy this intuition. Under identical environmental parameters, these systems can settle into entirely different long-term behaviors.
This phenomenon is known as multistability, and the individual stable states the system can adopt are called coexisting attractors. Understanding why multistability occurs requires shifting our perspective from simple linear cause-and-effect to the geometric landscape of nonlinear dynamics.
The Geometry of Phase Space and Basins of Attraction
To understand why a system can possess multiple stable states simultaneously, we can visualize its behavior using a phase space—a geometric map where every point represents a specific state of the system. In a multistable system, this space is carved into distinct regions called basins of attraction. Each basin acts like a valley in a physical landscape: any initial state within that valley will naturally roll down and settle into the attractor at the bottom.
As Yiran Feng demonstrates in the context of secure communications (A New Four-Dimensional Chaotic System and its Circuit Implementation), the final state of a high-dimensional chaotic system is determined solely by its initial conditions when parameters are held constant. This sensitivity makes the system highly unpredictable to an outside observer, which is a distinct advantage for cryptographic applications but a challenge for predicting natural systems.
The boundaries separating these basins are rarely clean, straight lines. In discrete multi-species ecosystems, for example, Da Wang et al. showed that the boundary between long-term species survival and total extinction can exhibit intricate, three-dimensional fractal characteristics (Fractals analysis and control for a kind of three-species ecosystem with symmetrical coupled predatory behavior). Because of this topological complexity, a minute variation in initial population densities can completely flip the system's ultimate fate from stable coexistence to collapse.
Initial State A Initial State B
│ │
▼ ▼
┌─────────────────┐ ┌─────────────────┐
│ Basin of High │ │ Basin of Low │
│ Moisture │ │ Moisture │
└────────┬────────┘ └────────┬────────┘
│ │
▼ ▼
┌─────────────────┐ ┌─────────────────┐
│ Attractor 1 │ │ Attractor 2 │
│ (Lush Monsoon) │ │ (Arid Drought) │
└─────────────────┘ └─────────────────┘
In macro-scale systems like global weather, these basins represent distinct climate regimes. Davide Faranda et al. modeled binary climate regimes using a double-well potential landscape (El Niño and droughts in Southeast Asia: A stochastic-chaotic modeling approach). They found that regional drought conditions represent a distinct, lower-dimensional basin of attraction within the chaotic climate system. Entering a drought state involves a fundamental shift in atmospheric complexity, making the pattern self-sustaining and difficult to exit.
Shifting States: Bifurcations and Hysteresis
A multistable system can transition between coexisting attractors in two ways: through a massive external shock that physically pushes the system's state across a boundary into an adjacent basin, or through a gradual change in external parameters that reshapes the underlying landscape itself.
When internal forces or external parameters are modulated, the system can undergo a bifurcation—a threshold point where an existing attractor destabilizes or vanishes entirely, forcing the system to jump to an alternative stable state. Aaron J. Gruber et al. illustrated this mechanism in neurobiology (Dopamine Induced Bistability Enhances Signal Processing in Spiny Neurons). By modulating the scaling factors of opposing internal ionic currents, dopamine induces a saddle-node bifurcation in striatal spiny neurons. This transition creates a bistable regime, allowing the neurons to act as sophisticated filters that lock into an "up" state to amplify time-correlated signals while suppressing background noise in the "down" state.
A defining hallmark of these transitions is hysteresis: the path taken to exit a state is not the same as the path taken to enter it.
State Value
▲
│ ┌───────────────► Attractor 2 (High Cholesterol / Arid Cloud)
│ │ ▲
│ │ │ Bifurcation Point (Tipping Point)
│ │ │
│ ──────┴───────────────┘
│ ▲
│ │ Bifurcation Point
│ │
│ └─────────────────────► Attractor 1 (Low Cholesterol / Thick Cloud)
│
└──────────────────────────────────────────────────────────────────► Parameter
Consider the dynamical model of lipoprotein metabolism developed by Elias August et al. (A Dynamical Model of Lipoprotein Metabolism). The system can settle into either a low-cholesterol or a high-cholesterol steady state. Once an individual's metabolism is pushed into the high-cholesterol attractor, hysteresis dictates that simply reducing the initial triggering behavior will not automatically return the system to the low-cholesterol state; the underlying parameters must be altered drastically to destroy the high-cholesterol attractor entirely.
This exact mathematical behavior underpins physical systems as well. François Alouges et al. proved that certain external magnetic or electric field strengths allow for the coexistence of multiple stable defect configurations in liquid crystals (The Saturn Ring Effect in Nematic Liquid Crystals with External Field: Effective Energy and Hysteresis). The transition point between a dipole configuration and a "Saturn ring" configuration differs depending on whether the external field strength is increasing or decreasing.
From Two States to Infinity: Extreme and Marginal Stability
Multistability is not limited to binary choices. Some systems exhibit extreme multistability, where an infinite number of stable states coexist for the exact same parameter settings.
How can a system sustain infinite stable configurations without collapsing into chaos? Jürgen Kurths et al. demonstrated a clear physical mechanism for this using symmetrically coupled pendulum clocks (Extreme multistability in symmetrically coupled clocks). When a system of coupled oscillators organizes into independent clusters that become mathematically uncoupled from one another, it can maintain an infinite variety of stable phase shifts between those clusters indefinitely.
In highly complex, disordered systems—such as massive ecosystems with hundreds of interacting species—this phenomenon manifests as marginal stability. Giulio Biroli et al. showed that when ecosystems are driven by strong, heterogeneous interactions, they naturally settle into a critical state poised at the edge of chaos (Marginally stable equilibria in critical ecosystems). This state is characterized by a vast multiplicity of coexisting stable equilibria, allowing the ecosystem to adapt dynamically to environmental variations by shifting between nearby stable configurations.
Multistability Across Disciplines
Because multistability is a fundamental property of nonlinear mathematics, the same underlying equations explain phenomena across vastly different fields:
- Microbial Ecology: The human microbiome does not possess a single, uniform healthy state. Peng Yang et al. used ensemble clustering to show that healthy human habitats are characterized by multiple stable structural configurations of microbial communities (Microbial community pattern detection in human body habitats via ensemble clustering framework). However, this multi-stability also means that external shocks can be dangerous. Stein et al. demonstrated that perturbations like antibiotics can inadvertently push the gut microbiome out of a healthy equilibrium and into an alternative stable state that is highly persistent and susceptible to pathogen colonization.
- Climate and Meteorology: On a macro scale, Norbert Marwan et al. showed that the South American monsoon system exhibits bistability driven by a positive feedback loop involving atmospheric moisture (A deforestation-induced tipping point for the South American monsoon system). Beyond a certain threshold of deforestation, the high-precipitation attractor disappears, triggering a tipping point into a permanent, low-precipitation state. On a micro-meteorological scale, A. Berner et al. used cloud-resolving models to confirm that marine stratocumulus clouds exhibit atmospheric bistability (Marine boundary layer cloud regimes and POC formation in a CRM coupled to a bulk aerosol scheme). They can coexist side-by-side indefinitely as either a thick, polluted closed-cell state or a thin, drizzling open-cell state due to self-stabilizing feedback loops.
- Robotics and Engineering: Multistability is also a vital consideration in mechanical design. Jorge Cham demonstrated that when realistic gravity and damping are incorporated into hybrid dynamical models of vertical hopping robots, multiple steady-state trajectories coexist for a single stride period (Dynamic Stability of Open-Loop Hopping). This multi-stability explains why physical robots experience sudden, unpredictable jumps between different hopping modes. Conversely, engineers can intentionally harness this property; Mrunmayi Mungekar et al. utilized emergent multistability in large, kirigami-inspired planar structures to design soft deployable robots capable of maintaining multiple stable shapes without requiring continuous power (Design of Bistable Soft Deployable Structures via a Kirigami‐Inspired Planar Fabrication Approach).
By mapping complex systems as geometric landscapes of coexisting attractors, researchers can move past simple linear predictions. Whether trying to prevent a climate tipping point, cure a metabolic imbalance, or secure a cryptographic network, the key lies in understanding the boundaries of these hidden basins of attraction and the forces that govern the transitions between them.
Key Foundational Work
Effects of nutrient-poor and anaerobic conditions on maintenance of R100-1 and RSF2124 plasmids in model populations of Escherichia coli. (1989) — contributed 2 concept elements to this direction.
Bacterial Wall Attachment in a Flow Reactor (2002) — contributed 1 concept elements to this direction.
Dopamine Induced Bistability Enhances Signal Processing in Spiny Neurons (2002) — contributed 1 concept elements to this direction.