Continuous Dependence on Parameters

Solutions of ODEs depend continuously (and often smoothly) on initial conditions, parameters, and even the right-hand side. This structural stability is fundamental for applications and perturbation theory.

Statement

Theorem8.6Continuous dependence on initial conditions and parameters

Consider the parametrized IVP:

$\mathbf{x}' = \mathbf{F}(t, \mathbf{x}, \boldsymbol{\mu}), \quad \mathbf{x}(t_0) = \boldsymbol{\xi},$

where $\mathbf{F}$ is continuous in all variables and Lipschitz in $\mathbf{x}$ uniformly in $(t, \boldsymbol{\mu})$ . Let $\mathbf{x}(t; \boldsymbol{\xi}, \boldsymbol{\mu})$ denote the unique solution.

Then the map $(\boldsymbol{\xi}, \boldsymbol{\mu}) \mapsto \mathbf{x}(\cdot\,; \boldsymbol{\xi}, \boldsymbol{\mu})$ is continuous. That is, if $\boldsymbol{\xi}_n \to \boldsymbol{\xi}_0$ and $\boldsymbol{\mu}_n \to \boldsymbol{\mu}_0$ , then $\mathbf{x}(t; \boldsymbol{\xi}_n, \boldsymbol{\mu}_n) \to \mathbf{x}(t; \boldsymbol{\xi}_0, \boldsymbol{\mu}_0)$ uniformly on compact subsets of the maximal interval.

Differentiable Dependence

Theorem8.7Differentiability with respect to initial conditions

If $\mathbf{F}(t, \mathbf{x})$ is $C^1$ in $(t, \mathbf{x})$ , then the solution $\mathbf{x}(t; t_0, \boldsymbol{\xi})$ is $C^1$ in $\boldsymbol{\xi}$ , and the matrix $\Phi(t) = \partial \mathbf{x} / \partial \boldsymbol{\xi}$ satisfies the variational equation:

$\Phi'(t) = \frac{\partial \mathbf{F}}{\partial \mathbf{x}}\big(t, \mathbf{x}(t)\big) \Phi(t), \quad \Phi(t_0) = I_n.$

This is a linear matrix ODE whose coefficient depends on the base solution $\mathbf{x}(t)$ .

ExampleVariational equation for a scalar ODE

For $x' = -x^3$ , $x(0) = \xi$ , the solution is $x(t; \xi) = \xi / \sqrt{1 + 2\xi^2 t}$ .

The variational equation is $\phi' = -3x(t)^2 \phi$ , $\phi(0) = 1$ , giving:

$\phi(t) = \frac{\partial x}{\partial \xi} = \frac{1}{(1 + 2\xi^2 t)^{3/2}}.$

This quantifies how perturbations in the initial condition decay: nearby trajectories converge, reflecting asymptotic stability.

The Flow Map

Definition8.6Flow of a differential equation

For the autonomous system $\mathbf{x}' = \mathbf{F}(\mathbf{x})$ , the flow is the map $\varphi_t: \mathbb{R}^n \to \mathbb{R}^n$ defined by $\varphi_t(\mathbf{x}_0) = \mathbf{x}(t; \mathbf{x}_0)$ . The flow satisfies:

$\varphi_0 = \mathrm{id}$ (identity map).
$\varphi_{t+s} = \varphi_t \circ \varphi_s$ (group property).
$\varphi_t$ is a $C^k$ diffeomorphism if $\mathbf{F}$ is $C^k$ .

The map $(t, \mathbf{x}_0) \mapsto \varphi_t(\mathbf{x}_0)$ is $C^k$ jointly in both variables.

ExampleFlow of a linear system

For $\mathbf{x}' = A\mathbf{x}$ , the flow is $\varphi_t(\mathbf{x}_0) = e^{tA}\mathbf{x}_0$ . The Jacobian is $D\varphi_t = e^{tA}$ , and the variational equation $\Phi' = A\Phi$ has solution $\Phi(t) = e^{tA}$ .

For $A = \begin{pmatrix}0 & 1 \\ -1 & 0\end{pmatrix}$ , $e^{tA} = \begin{pmatrix}\cos t & \sin t \\ -\sin t & \cos t\end{pmatrix}$ , giving rotation flow.

RemarkSmooth dependence and structural stability

If $\mathbf{F}$ is $C^k$ ( $k \geq 1$ ), then the solution depends in a $C^k$ manner on all data (initial conditions, parameters, time). This has practical consequences:

Numerical stability: small errors in initial conditions produce proportionally small errors in the solution (on bounded intervals).
Bifurcation theory: smooth parameter dependence enables the study of qualitative changes as parameters vary.
Sensitivity analysis: the variational equation quantifies the response of solutions to perturbations.