Let $f:[0,\infty)\to\mathbb{R}$ and $R>0$. How does Fubini's theorem imply $$\int_0^r\frac 1{\sigma^{n-1}}\int_0^\sigma\rho^{n-1}f(\rho)\;d\rho\;d\sigma\color{red}{=\int_0^r\frac{r^{2-n}-\rho^{2-n}}{2-n}\rho^{n-1}f(\rho)\;d\rho}\;\;\;\text{for }r\in (0,R)$$ (I don't understand why the left part is equal to the $\color{red}{\text{red}}$ part)?

I know Fubini's theorem in the following version: Let

  • $(\Omega_i,\mathcal{A}_i,\mu_i)$ be a $\sigma$-finite measure space
  • $g:\Omega_1\times\Omega_2\to\overline{\mathbb{R}}$ be measurable with respect to $\mathcal{A}_1\otimes\mathcal{A}_2$ and nonnegative or $(\mu_1\otimes\mu_2)$-integrable

Then, $$G_1:\Omega_1\to\mathbb{R}\;,\;\;\;\omega_1\mapsto\int g(\omega_1,\cdot)\;d\mu_2$$ and $$G_2:\Omega_2\to\mathbb{R}\;,\;\;\;\omega_2\mapsto\int g(\cdot,\omega_2)\;d\mu_1$$ are $\mathcal{A}_2$- and $\mathcal{A}_1$-measurable, respectively, and it holds $$\int g\;d(\mu_1\otimes\mu_2)=\int G_1\;d\mu_1=\int G_2\;d\mu_2$$

What is $g$ in the given scenario?


Here is a hint : consider the measure spaces $\Omega_1=\,]0,r[\,=\Omega_2$ (with Lebesgue measure) and the function $$g(\rho,\sigma)=\frac1{\sigma^{n-1}}\mathbf 1_{]0, \rho[}(\sigma) \rho^{n-1}f(\rho)=\frac1{\sigma^{n-1}}\mathbf 1_{]\sigma, r[}(\rho) \rho^{n-1}f(\rho)\, .$$


You just permute the two integrals (looking at the endpoints gives you $0 \leq \rho \leq \sigma \leq r$), obtaining $\int \limits _o ^r \rho ^{n-1} f(\rho) \int \limits _\rho ^r \frac 1 {\sigma ^{n-1}} \mathbb{d}\sigma \mathbb{d}\rho$, and the inner integral is easy to compute.

(More detailed: you pass from $\int (\int \mathbb{d}\rho) \mathbb{d}\sigma$ to $\int \mathbb{d}(\rho, \sigma)$ (that's the integral on the product space) and then to $\int (\int \mathbb{d}\sigma) \mathbb{d}\rho$. So, you apply Fubini twice, the first time "backwards".)


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.