A transform is an operator, not a magic change of letters
Objective: Why is s not replaced by t after the integral is evaluated?
An integral transform compares a function with a family of kernel shapes and records the resulting coefficients. The transformed variable labels which kernel member was used; it is not the original input renamed. Before using a transform table, name the original variable, transformed variable, kernel, integration or summation region, normalization, units, and convergence assumptions. The exact claim for this chapter is: If the integral exists for f and g, then T(af+bg)=aTf+bTg for scalars a,b.
Here is the object in precise language: For a kernel K(s,t), an integral transform is (Tf)(s)=∫_D K(s,t)f(t)dt whenever the integral exists; t is integrated out, while s indexes the output. A transform is a new representation of the same declared object, not permission to ignore domain, endpoint, regularity, or inverse-existence conditions.
Use the concrete problem “Let K(s,t)=e^{-st} on t≥0. Compute T(2e^{-t}+3e^{-2t})(s) for real s>−1.” to make every symbol observable. Rebuild the three steps, check the result in the original representation, and then test this boundary: Linearity does not guarantee existence: two individually divergent integrals cannot be separated and manipulated as though both were finite.
If the integral exists for f and g, then T(af+bg)=aTf+bTg for scalars a,b.
Insert af(t)+bg(t) into the defining integral.
Use linearity of integration to separate the two terms and move constants outside.
Recognize the two remaining integrals as Tf and Tg on their common output domain.
Let K(s,t)=e^{-st} on t≥0. Compute T(2e^{-t}+3e^{-2t})(s) for real s>−1.
- Combine exponents: the two integrands are 2e^{-(s+1)t} and 3e^{-(s+2)t}.
- For a>0, ∫₀∞e^{-at}dt=1/a; s>−1 makes both a=s+1 and a=s+2 positive.
- Apply linearity to obtain 2/(s+1)+3/(s+2).
Result: Tf(s)=2/(s+1)+3/(s+2) for real s>−1.