Video · + 1 min read

The technique of differentiation under the integral sign, popularly known as Feynman's trick, is a way to impress your friends and colleagues in dire circumstances:

Often, integrals, like ∫0∞x2e−xdx\int_0^\infty x^2 e^{-x} dx, live naturally not in isolation, but in families. When you learn about differential equations, you see that a function can arise in more than one way, e.g. as explicit formula y=exy = e^x vs. as differential equation y′=yy' = y. There lies the philosophy behind Feynman's method: an integral may have a useful alternative description, perhaps as a derivative of another family of integrals, or as a double integral (integral of a family of integrals).

Finding that alternative description (in terms of families) is more of an art than an exact science. Here, ∫0∞x2e−xdx\int_0^\infty x^2 e^{-x} dx fits in the family I(n,t)=∫0∞xne−txdx,I(n,t) = \int_0^\infty x^n e^{-tx} dx, as the member I(2,1)I(2,1). The derivative of I(n,t)I(n,t) with respect to tt is ddtI(n,t)=−∫0∞xn+1e−txdx,\frac{d}{dt}I(n,t) = -\int_0^\infty x^{n+1} e^{-tx} dx, a.k.a. just −I(n+1,t)-I(n+1,t)! Thus I(2,t)I(2,t) is just the second derivative of I(0,t)I(0,t)!

On the other hand, when n=0n=0, we easily evaluate the integral as long as t>0t>0. So, I(2,t)=I′′(0,t)=2t3,I(2,t) = I''(0,t) = \frac{2}{t^3}, and we find the integral we started with, ∫0∞x2e−xdx=I(2,1)=2.\int_0^\infty x^2 e^{-x} dx = I(2,1) = \boxed{2}.

Type a concept, like “slope” or “photosynthesis”.

↑ ↓ to choose · Enter to openSearch explanations and problems