the convolution theorem
Here is one of the most useful trades in all of transform mathematics. In the time world, blending two signals by convolution is a heavy integral. In the s-world, that same blend becomes plain multiplication. The convolution theorem is the bridge between the two — and it almost always means you do the hard work in whichever world makes it easy.
The theorem states: L{f * g} = F(s) G(s). The Laplace transform of a convolution is just the product of the separate transforms. Reading it forward turns a convolution integral into a quick multiplication. But the reverse reading is the everyday hero: when you finish an ODE and your answer in the s-domain is a PRODUCT F(s) G(s) that you cannot easily split, you do not have to force partial fractions. You invert each factor separately to get f(t) and g(t), then write the answer as the single convolution integral (f * g)(t) = integral from 0 to t of f(tau) g(t - tau) d tau. The product in s is guaranteed to come from that convolution in time.
This is why convolution and the Laplace transform are inseparable. It gives a clean closed form for the response of a linear system to an arbitrary input (input convolved with impulse response), it rescues inversions that partial fractions cannot handle (such as a forcing with no nice transform), and it is the seed of Duhamel's principle. Whenever you see a product of two transforms, think convolution.
To invert 1/(s^2 (s^2 + 1)), read it as (1/s^2)·(1/(s^2 + 1)); these invert to t and sin(t), so the answer is the convolution t * sin(t) = integral from 0 to t of tau · sin(t - tau) d tau = t - sin(t).
A product of two transforms inverts to a single convolution integral — no partial fractions needed.
The theorem applies to a true convolution f * g, not to an ordinary product: L{f(t) g(t)} is NOT F(s) G(s). Mistaking pointwise multiplication for convolution is the most common error here.