One-dimensional [continuous](/page/Continuity) dynamical systems occupy a special place in bifurcation theory. The phase space is a subset of $\mathbb{R}$, and this imposes a severe [topological](/page/Topology) constraint on the dynamics: solutions cannot oscillate, and no limit cycles can exist. As a consequence, the long-time behavior of any trajectory is governed entirely by the equilibrium structure of the vector field. The central question of bifurcation theory is then to understand how this equilibrium structure changes as a real parameter $\mu$ is varied. We consider throughout a smooth one-parameter family of scalar ordinary differential equations. The state space is $\mathbb{R}$ and the parameter space is $\mathbb{R}$. The system is governed by: \begin{align*} \dot{y} = f(y, \mu), \quad y \in \mathbb{R},\quad \mu \in \mathbb{R} \end{align*} where $f \in C^\infty(\mathbb{R} \times \mathbb{R}, \mathbb{R})$. A point $y^* \in \mathbb{R}$ is called an **equilibrium** of the system at parameter value $\mu$ if $f(y^*, \mu) = 0$. Throughout this page, we focus on the behavior near a distinguished equilibrium. We assume $f(0, 0) = 0$, so the origin is an equilibrium at $\mu = 0$, and we analyze the local dynamics near $(y, \mu) = (0, 0)$. ## Formal Definitions We establish the stability concepts and algebraic invariants that govern the bifurcation analysis. [definition: Stability of an Equilibrium] Let $y^*$ be an equilibrium of $\dot{y} = f(y, \mu)$ at a fixed $\mu$. The equilibrium is: - **Stable** (or hyperbolic attracting) if $\partial_y f(y^*, \mu) < 0$. - **Unstable** (or hyperbolic repelling) if $\partial_y f(y^*, \mu) > 0$. - **Non-hyperbolic** if $\partial_y f(y^*, \mu) = 0$. A non-hyperbolic equilibrium $y^*$ is called **semi-stable** if $f$ does not change sign on either side of $y^*$: that is, if $y^*$ is attracting from one side and repelling from the other. [/definition] [definition: Lyapunov Coefficients] Consider the Taylor expansion of $f(y, 0)$ at the equilibrium $y = 0$: \begin{align*} f(y, 0) = \sum_{j=0}^{k} l_j\, y^j + O(|y|^{k+1}), \quad l_j := \frac{1}{j!}\frac{\partial^j f}{\partial y^j}\bigg|_{(0,0)}. \end{align*} The scalars $l_j$ are called the **Lyapunov coefficients** of the equilibrium. In the bifurcation setting, we assume $l_0 = 0$ (so that the origin is indeed an equilibrium at $\mu = 0$) and $l_1 = 0$ (so that the linearization is degenerate). For $k \ge 2$, the **first non-zero Lyapunov coefficient** is $l_k$, where $k = \min\{j \ge 2 : l_j \neq 0\}$. [/definition] The Lyapunov coefficients are the primary algebraic invariants that determine the local bifurcation type. In the non-degenerate case $l_1 \neq 0$, the equilibrium is hyperbolic and persists smoothly under parameter perturbation by the [Implicit Function Theorem](/page/Implicit%20Function%20Theorem). It is precisely the vanishing of $l_1$ that forces a bifurcation to occur. [definition: Forward-Invariant Interval] An open interval $I \subset \mathbb{R}$ is **forward-invariant** for $\dot{y} = f(y, \mu)$ at a fixed $\mu$ if every solution $y(t)$ with $y(0) \in I$ satisfies $y(t) \in I$ for all $t \ge 0$ for which the solution exists. This holds if and only if $f$ points inward at the [boundary](/page/Boundary) of $I$: for a bounded interval $I = (a, b)$, we require $f(a, \mu) \ge 0$ and $f(b, \mu) \le 0$. [/definition] ## Key Results ### Qualitative Behavior on Intervals The one-dimensionality of the phase space yields a complete classification of long-time behavior. This stands in sharp contrast to higher-dimensional systems, where trajectories may wind around limit cycles or exhibit chaotic motion. [quotetheorem:236] The Interval Dichotomy reflects the fundamental rigidity of one-dimensional flows: the absence of topological room for recurrent non-equilibrium behavior forces all bounded trajectories to converge. In higher dimensions, the Poincaré-Bendixson theorem provides a partial analogue (allowing limit cycles as $\omega$-[limits](/page/Limit) in $\mathbb{R}^2$), but in one dimension the conclusion is absolute. ### Counting Bifurcating Equilibria Before analyzing the precise geometry of specific bifurcations, we establish a general upper bound on how many equilibria can emerge from a degenerate equilibrium as $\mu$ is varied. This bound is determined entirely by the order of vanishing of $f$ at the origin. [quotetheorem:237] [remark: Sharpness and Parity] The bound $k$ is sharp: in the generic case, exactly $k$ equilibria (counted with multiplicity) are involved in the bifurcation. The parity of $k$ plays a decisive role. If $k$ is odd, the leading term $l_k y^k$ changes sign between $y > 0$ and $y < 0$, so the [intermediate value theorem](/theorems/629) guarantees at least one equilibrium persists for every $\mu$ near zero. If $k$ is even, the leading term $l_k y^2$ does not change sign, and the equilibrium at the origin may annihilate entirely — giving birth to either zero or two nearby equilibria as $\mu$ varies. [/remark] ### The Saddle-Node Bifurcation The most generic degenerate case occurs when $l_1 = 0$ and $l_2 \neq 0$. This is the **saddle-node bifurcation** (also called a fold bifurcation). By the Lyapunov Bound, at most two equilibria can be born near the origin. The following theorem gives a complete and precise description of the three distinct parameter regimes. [quotetheorem:238] [remark: Bifurcation Diagram] The bifurcation diagram for the saddle-node is a parabola in the $(y, \mu)$-plane, given by the curve $\mu = -l_2 y^2 + \text{h.o.t.}$ The vertex of the parabola at the origin represents the semi-stable equilibrium at $\mu = 0$. On one side of the parabola (when $\mu l_2 < 0$) there are two equilibria; on the other (when $\mu l_2 > 0$) there are none. [illustration:saddle-node-bifurcation-diagram] [/remark] ## Examples [example: Canonical Normal Form] The simplest system exhibiting the saddle-node bifurcation is the **canonical normal form**: \begin{align*} \dot{y} = \mu - y^2. \end{align*} Here $f(y, \mu) = \mu - y^2$, so $l_0 = 0$, $l_1 = 0$, and $l_2 = \frac{1}{2!}(-2) = -1$. The transversality condition $\partial_\mu f(0,0) = 1 \neq 0$ holds, so all hypotheses of the Saddle-Node Bifurcation Theorem are satisfied. **Equilibria:** Setting $f(y,\mu) = 0$ gives $y^2 = \mu$. This has no real solutions when $\mu < 0$, a unique solution $y = 0$ when $\mu = 0$, and two solutions $y = \pm\sqrt{\mu}$ when $\mu > 0$. **Consistency with the theorem:** Since $l_2 = -1 < 0$, the condition $\mu l_2 < 0$ corresponds to $\mu > 0$, precisely the regime where two equilibria exist. **Stability:** We compute $\partial_y f = -2y$. At $O_+ = +\sqrt{\mu}$, this gives $-2\sqrt{\mu} < 0$, so $O_+$ is **stable**. At $O_- = -\sqrt{\mu}$, this gives $+2\sqrt{\mu} > 0$, so $O_-$ is **unstable**. At $\mu = 0$, the single equilibrium $y = 0$ has $\partial_y f = 0$, confirming it is semi-stable (trajectories approach from above and repel from below). [illustration:saddle-node-vector-field] [illustration:saddle-node-phase-portraits] [/example] [example: Cusp Bifurcation — Higher Order Degeneracy] When $l_1 = l_2 = 0$ but $l_3 \neq 0$, the Lyapunov Bound predicts at most three bifurcating equilibria. The canonical form in this case is: \begin{align*} \dot{y} = \mu + y^3. \end{align*} Here $f(y, \mu) = \mu + y^3$, giving $l_1 = l_2 = 0$ and $l_3 = \frac{1}{3!}(6) = 1$. **Equilibria:** Setting $f = 0$ gives $y^3 = -\mu$, i.e., $y = -\mu^{1/3}$. For every $\mu \in \mathbb{R}$ there is exactly one real equilibrium. No collision or bifurcation of equilibria occurs; instead, the unique equilibrium at $\mu = 0$ is non-hyperbolic (since $\partial_y f(0,0) = 0$) but the equilibrium persists for all $\mu$. **Stability:** $\partial_y f = 3y^2 \ge 0$ with equality only at $y = 0$. Thus the equilibrium is non-hyperbolic at $\mu = 0$ and semi-stable (it repels from both sides near $\mu = 0$). For $\mu \neq 0$, $y = -\mu^{1/3} \neq 0$, and $\partial_y f = 3(-\mu^{1/3})^2 = 3\mu^{2/3} > 0$, so the equilibrium is always **unstable** away from the origin. This example illustrates that the parity of $k$ matters: for $k$ odd, one equilibrium is guaranteed to persist, and the bound of $k$ bifurcating equilibria is not always achieved. [illustration:cubic-normal-form-bifurcation] [/example] ## Problems [problem] Consider the system: \begin{align*} \dot{y} = \mu + 2y^2 - y^4. \end{align*} 1. Compute the Lyapunov coefficients $l_0, l_1, l_2$ at $(y, \mu) = (0,0)$ and verify that the hypotheses of the Saddle-Node Bifurcation Theorem are satisfied. 2. Identify the parameter regime (in terms of $\mu$) in which two equilibria exist near the origin, and determine their stability. 3. Using the Interval Dichotomy, describe the long-time behavior of any solution starting in a bounded forward-invariant interval containing the two equilibria, in the parameter regime found in part (2). [/problem] [solution] **Part 1: Computing the Lyapunov Coefficients and Verifying Hypotheses.** We have $f(y, \mu) = \mu + 2y^2 - y^4$. Computing partial [derivatives](/page/Derivative) at $(0,0)$: \begin{align*} l_0 &= f(0, 0) = 0, \\ l_1 &= \frac{\partial f}{\partial y}\bigg|_{(0,0)} = \left(4y - 4y^3\right)\bigg|_{y=0} = 0, \\ l_2 &= \frac{1}{2!}\frac{\partial^2 f}{\partial y^2}\bigg|_{(0,0)} = \frac{1}{2}\left(4 - 12y^2\right)\bigg|_{y=0} = 2. \end{align*} So $l_2 = 2 \neq 0$, confirming (SN2). We also check (SN3): $\partial_\mu f(0,0) = 1 \neq 0$. All hypotheses of the Saddle-Node Bifurcation Theorem hold. **Part 2: Existence and Stability of Equilibria.** Since $l_2 = 2 > 0$, the condition $\mu l_2 < 0$ reduces to $\mu < 0$. By the Saddle-Node Bifurcation Theorem, for $\mu < 0$ (and $|\mu|$ small), two equilibria exist near the origin: \begin{align*} O_{\pm}: \quad y = \pm\sqrt{\frac{-\mu}{l_2}} + O(|\mu|^{1/2}) = \pm\sqrt{\frac{-\mu}{2}} + O(|\mu|^{1/2}). \end{align*} To determine stability, we compute $\partial_y f = 4y - 4y^3$. For small $|y|$, the cubic term is negligible, so: \begin{align*} \partial_y f(O_\pm, \mu) \approx 4 \cdot \left(\pm\sqrt{\frac{-\mu}{2}}\right) = \pm 4\sqrt{\frac{-\mu}{2}}. \end{align*} Since $\mu < 0$, the quantity $\sqrt{-\mu/2} > 0$. Therefore: \begin{align*} \partial_y f(O_+, \mu) \approx +4\sqrt{-\mu/2} > 0 \implies O_+ \text{ is **unstable**},\\ \partial_y f(O_-, \mu) \approx -4\sqrt{-\mu/2} < 0 \implies O_- \text{ is **stable**}. \end{align*} **Part 3: Long-Time Behavior via the Interval Dichotomy.** Fix $\mu < 0$ small and choose a bounded interval $I$ containing both $O_-$ and $O_+$ that is forward-invariant (such an interval exists because $f(y, \mu) \approx -y^4 < 0$ for $|y|$ large, so we can find $b > O_+$ with $f(b, \mu) < 0$ and $a < O_-$ with $f(a, \mu) > 0$). By the Interval Dichotomy, any solution $y(t)$ with $y(0) \in I$ converges to an equilibrium in $\overline{I}$. The equilibria in $I$ are $O_-$ (stable) and $O_+$ (unstable). Trajectories starting between $O_-$ and $O_+$ (i.e., $y(0) \in (O_-, O_+)$) satisfy $f > 0$ and hence increase, converging to $O_+$ from below. Trajectories starting to the right of $O_+$ satisfy $f < 0$ for $y$ slightly above $O_+$ (up to the next zero of $f$) and decrease back toward $O_+$. The stable equilibrium $O_-$ attracts all trajectories starting to its left within $I$. In summary, the basin of attraction of $O_-$ is $\{y \in I : y < O_-\}$, the basin of attraction of $O_+$ (from above) consists of trajectories slightly above $O_+$, and trajectories between the two equilibria are repelled from $O_-$ and attracted to $O_+$, consistent with the Interval Dichotomy. [/solution] ## References 1. Kuznetsov, Y.A., *Elements of Applied Bifurcation Theory*, 4th ed., Springer (2004). 2. Strogatz, S.H., *Nonlinear Dynamics and Chaos*, 2nd ed., Westview Press (2015). 3. Perko, L., *Differential Equations and Dynamical Systems*, 3rd ed., Springer (2001). 4. Guckenheimer, J. and Holmes, P., *Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields*, Springer (1983).