Canonical: https://knowlpedia.com/differential-equations/smooth-parameter-dependence/

# Smooth dependence of ODE solutions on parameters

Smooth coefficients and initial data yield solutions smooth in parameters on a common local domain.

Suppose \(F(t,y,\lambda)\) is [\(C^k\)](https://knowlpedia.com/real-analysis/class-ck-map/), \(k\ge1\), near \((t_0,y_0,\lambda_0)\). The local solution of \(y'=F(t,y,\lambda)\) depends \(C^k\) on time, initial state, initial time and \(\lambda\), on a common sufficiently small neighborhood. Smooth parameterized initial data can be composed with this solution map.

## Variational equation

For one parameter and fixed initial time, \(z=\partial_\lambda y\) satisfies
\[
z'=D_yF(t,y,\lambda)z+\partial_\lambda F(t,y,\lambda),
\qquad z(t_0)=\partial_\lambda y_0.
\]
The [Duhamel formula](https://knowlpedia.com/differential-equations/duhamel-formula/) controls this derivative. Further differentiation gives linear equations in the highest parameter derivative, with sources involving lower derivatives. Uniform estimates over a longer interval require common bounds and a common domain; pointwise existence for each parameter does not supply them.

## References

- [Gerald Teschl, Ordinary Differential Equations and Dynamical Systems, Theorem 2.11](https://www.mat.univie.ac.at/~gerald/ftp/book-ode/ode.pdf).
