Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Again I am stuck with this problem in Rudin:

Assume that $|f|_{r}<\infty$ for some $r<0$. Prove that $$\lim_{p\rightarrow 0}|f|_{p}=e^{\int_{X}\log |f|d\mu}$$

Let $r<p<0$ and we have $K=\frac{r}{p}>1$. Denote its conjugate by $K'$. Then we have $$\int f^{p}d\mu=\int (f^{p})*1d\mu\le (\int [f^{p}]^{\frac{r}{p}})^{\frac{p}{r}}$$ since by assumption $\mu(X)=1$. So in particular we have $$|f|_{p}\le |f|_{r}<\infty$$ Since $|f|_{p}$ is monotonely decreasing with $p\rightarrow 0$, it must have a limit.

We now apply Jensen's inequality, which gives us $$\log^{\int_{X}Fd\mu}\ge \int_{X}\log[F]d\mu$$ Here $F=f^{p}$. So we have $$ \int_{X}f^{p}d\mu\ge (e^{\int_{X}\log[f]d\mu})^{p} $$ taking the $p$-th root on both sides we conclude that $$|f|_{p}\ge e^{\int \log|f|d\mu}$$

But then I got totally stuck. It is worth pointing that Jensen's inequality is only an equality when $f^{p}=c$ is a constant. Therefore $f$ has to be a constant as well.

share|cite|improve this question
This has already been asked on the site. See, for example,… – Martin Argerami Dec 25 '12 at 6:34
.......:( sorry! I am not good at searching... – Bombyx mori Dec 25 '12 at 6:35
But there is a difference; here the $p$ is approaching from negative instead of positive. – Bombyx mori Dec 25 '12 at 6:35
No worries. It actually took me several minutes to find the duplicate, and that was because I vaguely remembered having answered a similar question. – Martin Argerami Dec 25 '12 at 6:36
I see. In your arguments you seem to use Hölder's inequality for negative exponents: what is your source for that? – Martin Argerami Dec 25 '12 at 6:39
up vote 2 down vote accepted

Note that for $u>0$, $$ k(u)=u\log(u)-u+1\ge0\tag{1} $$ This is because $k'(u)=\log(u)$ and $k''(u)=1/u$ show that $k(u)$ has a minimum at $u=1$.

Now, by L'Hospital, we have $$ \lim_{p\to0}\frac{x^p-1}p=\log(x)\tag{2} $$ Furthermore, applying $(1)$ yields $$ \begin{align} \frac{\mathrm{d}}{\mathrm{d}p}\frac{x^p-1}p &=\frac{p\log(x)x^p-x^p+1}{p^2}\\ &=\frac{k(x^p)}{p^2}\\ &\ge0\tag{3} \end{align} $$ Thus, by Dominated Convergence, we have $$ \begin{align} \log\left(\lim_{p\to0}\left(\int_X|f(x)|^p\,\mathrm{d}\mu\right)^{1/p}\right) &=\lim_{p\to0}\frac1p\log\left(\int_X|f(x)|^p\,\mathrm{d}\mu\right)\\ &=\lim_{p\to0}\frac1p\log\left(1+p\int_X\frac{|f(x)|^p-1}{p}\,\mathrm{d}\mu\right)\\ &=\lim_{p\to0}\frac1p\log\left(1+p\int_X\log|f(x)|\,\mathrm{d}\mu\right)\\ &=\int_X\log|f(x)|\,\mathrm{d}\mu\tag{4} \end{align} $$ Therefore, $$ \lim_{p\to0}\left(\int_X|f(x)|^p\,\mathrm{d}\mu\right)^{1/p} =\exp\left(\int_X\log|f(x)|\,\mathrm{d}\mu\right)\tag{5} $$

share|cite|improve this answer
Thank you so much for the help! – Bombyx mori Dec 25 '12 at 12:35
You're welcome. Each time I prove this theorem, I see some new underlying idea that makes it simpler. – robjohn Dec 25 '12 at 12:41

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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