• $\lambda$ denote the Lebesgue measure on $\mathbb R^2$
  • $\tilde\Delta,\Delta\subseteq\mathbb R^2$ be nonempty and compact
  • $F:\tilde\Delta\to\Delta$ with $$F(\tilde x)=A\tilde x+b\;\;\;\text{for all }\tilde x\in\tilde\Delta\tag1$$ for some invertible $A\in\mathbb R^{d\times d}$ and $b\in\mathbb R^d$

If $v\in L^1(\Delta)$, we obtain $$\int_\Delta v\:{\rm d}\lambda=\int_{\tilde\Delta}\left(v\circ F'\right)\left|\det F'\right|\:{\rm d}\lambda\tag2$$ (where $F'=A$, but that's not important in the sequel) via integration by substitution. Now, if $v$ admits a weak gradient $\nabla v\in L^1(\Delta,\mathbb R^d)$, I want to express $$\int_\Delta\nabla v\:{\rm d}\lambda\tag3$$ in the same way in terms of an integral over $\tilde\Delta$.

How do we need to apply integration by substitution for $(3)$?

I was quite sure that the solution is $$\int_\Delta\nabla v\:{\rm d}\lambda=\int_{\tilde\Delta}\left(\nabla v\circ F\right)\left|\det F'\right|\:{\rm d}\lambda\tag4\;.$$ However, I've read (for example, in Remark 5.45 of a lecture note) that we have $$\int_\Delta\nabla v\:{\rm d}\lambda=\int_{\tilde\Delta}\nabla\left(v\circ F\right)\left|\det F'\right|\:{\rm d}\lambda\tag5$$ instead.

The crucial point is that in $(4)$ we take the weak gradient of $v$ and evaluate it at $F(\tilde x)$ whereas we take the weak gradient of $v\circ F$ (which has to be calculated by the chain rule) and evaluate it at $\tilde x$ in $(5)$.

Actually, $(5)$ doesn't make sense to me. The function we're considering in the integration by substitution rule should be $x\mapsto\nabla v(x)$. I guess there is something I'm missing which prevents us from using the conventional integration by substitution rule and hence $(5)$ is the result of some modified rule for integrands involving the gradient.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.