For a linear time‑invariant (LTI), causal system, Duhamel's integral expresses the time‑domain response y(t) to an arbitrary input f(t) as the convolution of the system's impulse response h(t) with the input: y(t) = ∫_0^t h(t − τ) f(τ) dτ (plus any homogeneous solution accounting for initial conditions). It operationalizes superposition by integrating the contributions of infinitesimal impulses over past time.

Principle

Because an LTI system's response to a shifted impulse is the shifted impulse response and responses superpose, the total response to an arbitrary forcing is the time convolution of forcing with impulse response; Duhamel's integral is the causal convolution implementing this principle in the time domain.

Demonstration

Illustrative scenario → A damped single‑degree‑of‑freedom oscillator has impulse response h(t) (measurable or derivable). Given a measured force history f(t) applied from t = 0 onward in a temperature- and parameter‑stable system, compute displacement y(t) for t ≥ 0 by evaluating y(t) = ∫_0^t h(t − τ) f(τ) dτ and adding the homogeneous solution if nonzero initial displacement/velocity exist.

Misapplication

Using Duhamel's integral for nonlinear, time‑varying, or noncausal systems; the semantic error is assuming linear superposition and time invariance where they do not hold, which produces incorrect responses because impulse responses and convolutional superposition are not valid outside LTI causality.

Consequence

Provides a direct time‑domain method to compute system responses from impulse responses or modal contributions, underpinning time‑domain simulation, control design, structural response to transient loads and signal processing; it also yields the same information as transfer‑function methods but highlights causality and temporal assembly of effects.

Reversal

For linear but time‑varying systems, the kernel becomes a two‑time Green's function G(t, τ) and the response is y(t) = ∫ G(t, τ) f(τ) dτ; for nonlinear systems one must resort to methods such as Volterra series, numerical integration of governing equations, or incremental linearization — Duhamel's simple convolution no longer applies.

Boundary

Clearly within: causal, linear time‑invariant systems with known impulse response and inputs defined for t ≥ 0. Boundary case: systems that are approximately LTI over a limited operating range where convolution gives useful approximate predictions. Clearly outside: strongly nonlinear dynamics, parametric time‑varying systems, or systems with memory kernels that violate superposition.

Semantic Tension

Duhamel's time‑domain convolutional formulation competes with frequency‑domain transfer‑function methods: both are mathematically equivalent for LTI systems, but time‑domain convolution emphasizes causality and transient assembly while frequency methods emphasize spectral characteristics and steady‑state behavior.

Synthesis

Duhamel's integral is the causal time‑domain implementation of LTI superposition: it reconstructs a system's response by integrating impulse contributions over past inputs; its reliability depends entirely on linearity, time invariance and causality, and it generalizes to Green's functions for time‑varying problems.