A numerical approach for solving boundary value problems (BVPs) for ordinary differential equations by converting the BVP into one or more initial value problems (IVPs) and iteratively adjusting unknown initial conditions until the solution meets the boundary conditions.

Principle

Exploit continuous dependence of IVP solutions on initial data: treat the unknown initial values as parameters, integrate the IVP, evaluate the mismatch at the boundary, and use root-finding or optimization to adjust parameters until the mismatch vanishes.

Demonstration

Second-order BVP y'' = f(x,y,y'), y(a)=α, y(b)=β: pick a trial initial slope s, solve IVP with y(a)=α, y'(a)=s to obtain y_s(b); then find s such that y_s(b)=β using a secant or Newton method.

Misapplication

Using naive single shooting when the IVP is highly sensitive (ill-conditioned) to initial data, when multiple solutions exist (leading to convergence to unintended roots), or when the problem is stiff or yields blow-up before the boundary — situations where multiple-shooting, continuation, or collocation are safer.

Consequence

Transforms BVP solvers into standard IVP integrators plus nonlinear solvers, often simple to implement and efficient for well-behaved problems; yields good accuracy when the parameter-to-boundary map is smooth and well-conditioned.

Reversal

Direct discretization methods (finite difference, collocation, spectral/Galerkin) solve the BVP as a global system and avoid parameter shooting; multiple shooting splits the interval to reduce sensitivity.

Boundary

Intended for ODE boundary value problems on finite intervals; not directly applicable to PDE boundary problems without dimensional reduction, and effectiveness depends on stability and conditioning of the IVP and the chosen root-finding method.

Semantic Tension

Tension between the simplicity of shooting and the robustness of global discretization: shooting treats the problem as parameter-fitting for IVPs, whereas collocation/Galerkin approximate the entire solution space; numerical analysts must choose based on conditioning and nonlinearity.

Synthesis

The shooting method reduces a BVP to parameter-dependent IVPs and a root-finding task: it is practical and efficient for many smooth, well-conditioned ODE BVPs but must be replaced by multiple shooting or global discretization when sensitivity, stiffness, multiplicity, or blow-up compromise single-shot approaches.