Integration by Parts

Resources

Main Idea

For an integration involving two functions, we can interchange differentiation operators between them, as long as we account for boundary conditions.

Definition

ΩuDvdΩ=ΩvDudΩ+ΓuvdΓ

This is an application of the Divergence Theorem (Gauss's Theorem), which is

ΩDwdΩ=ΓwdΓ,

and we take w=uv, apply the product rule for D, and split the integral based on the summation in the integrand. Often this is presented with, D=, but as seen in Weak SINDy, the differential operator D can be more arbitrary (but must still be first order in this formula).

For a derivative Dα, multi-indices α and β with order |α| and |β|, if

[Dβv]Γ=0,β{β:βj1αj}

then,

ΩuDαvdΩ=(1)|α|ΩvDαudΩ.

In other words, if v and a sufficient number of its derivatives are 0 along the boundary, then the derivative operator is easily interchanged, up to the difference of the sign.

Summation by Parts

For a linear Finite Difference scheme Δ,

i=mnviΔui=vn+1un+1vmumi=mnΔviui.

One can take for instance Δyi=yi+1yi and show this by expanding the left side, adding a final 0=vn+1un+1vn+1un+1, grouping the like ui terms, and factoring these terms. The technique is similar to that of telescoping series. This is also known as Abel's Transformation.