2
$\begingroup$

Having a semester of abstract measure theory behind me, it is still hard to solve actual exercises. Recently I stumbled upon a very interesting one, which I have not encountered before. Consider the emasure space $(\mathbb{R},\mathcal{B}(\mathbb{R}),dx)$ where $\mathcal{B}(\mathbb{R})$ denotes the usual Borel $\sigma$-algebra and $dx$ the Lebesgue measure. For a constant $a > 0$ define the measure $\mu$ on $\mathcal{B}(\mathbb{R})$ by $$\mu(E) := \int_E\chi_{[0,\infty)}(x) e^{-ax} dx$$

Now the task is to calculate $$\int_{\mathbb{R}}x^2 d\mu$$

My first (and natural) idea was to approximate $x^2$ by simple functions and then find a monotone sequence $\varphi_n$ which converges pointwise to $x^2$. I mean this how one calculates the integral by definition. I think this gets kind of complicated if one does not find the right simple functions. Has anyone a better idea?

  • 2
    Why not just spell it out as $\int_{\bf R} x^2\chi_{[0,\infty)}(x)e^{-ax}\,\textrm{d}x$?2017-02-08
  • 0
    @tomasz Actually, that was my very first idea. The problem is, that I did not (and also not yet) see the legitimation why I can do this.2017-02-08
  • 0
    $\mu$ has a density with respect to $dx$, namely $f(x) = \chi_{[0, \infty)}(x) e^{-ax}$.2017-02-08
  • 0
    @user251257 I think I do not fully understand. Have you any references about this which I can read?2017-02-08
  • 2
    If $\mu(E)=\int_E g dx$, then $$\int f d\mu=\int fg dx.$$ You can prove this just like you suggested, it is trivially true for characteristic functions of measurable sets, hence also for simple functions by linearity of integral. For arbitrary measurable function $f$, take sequence of simple functions $f_n \uparrow f$ and use monotone convergence theorem.2017-02-08
  • 0
    @TheGeekGreek: The legitimation is simple: it trivally works for simple functions, and by definition of the Lebesgue integral, that is sufficient.2017-02-08

1 Answers 1

0

Thanks to tomasz, user251257 and Zoran Loncarevic I present my solution, which solves the stated question as a corollary.

Proposition. Let $(X,\mathcal{A},\mu)$ be a measure space and $f \in \mathcal{M}$. Then the function $\nu: \mathcal{A} \to [0,\infty]$ defined by $$\nu(A) := \int_A f d\mu$$ is a measure on $\mathcal{A}$.

For calculating an integral with respect to above measure, there is a nice proposition.

Proposition. Let $(X,\mathcal{A},\mu)$ be a measure space and $f,g \in \mathcal{M}$. With the terminology established above it holds that $$\int_X g d\nu = \int_X fg d\mu$$

Proof. Assume $g := \sum_{i = 1}^n a_i \chi_{A_i} \in \Sigma^+$. Then $$\int_X g d\nu = \sum_{i = 1}^n a_i \nu(A_i) = \sum_{i = 1}^n a_i \int_{A_i} f d\mu = \int_X f\left(\sum_{i = 1}^n a_i\chi_{A_i}\right) d\mu = \int_X fg d\mu$$

Now let $g \in \mathcal{M}^+$. Then we find a sequence $(\varphi_n)_{n \in \mathbb{N}}$ of functions in $\Sigma^+$ such that $\varphi_n \nearrow g$ pointwise. Hence $$\int_X g d\nu = \lim_{n \to\infty} \int_X \varphi_n d\nu = \lim_{n \to \infty} \int_X f\varphi_n d\mu = \int_X fg d\mu$$

by monotone convergence since $f\varphi_n \nearrow fg$. Lastly let $g \in \mathcal{M}$. Write $g = g^+ - g^-$. Thus $$\int_X g d\nu = \int_Xg^+ d\nu - \int_X g^- d\nu = \int_X fg^+ d\mu - \int_X fg^- d\mu = \int_X fg d\mu$$

Corollary. We have $$\int_{- \infty}^\infty x^2 d\mu = \frac{2}{a^3}$$

Proof. By above proposition we have that $$\int_{- \infty}^\infty x^2 d\mu = \int_0^\infty x^2 e^{-ax}dx$$ Twice partial integration yields $$\begin{align}\int_0^\infty x^2 e^{-ax}dx &= -\frac{1}{a}x^2e^{-ax}\Big\vert_0^\infty + \frac{2}{a}\int_0^\infty xe^{-ax}dx\\ &= \frac{2}{a}\int_0^\infty xe^{-ax}dx\\ &= \frac{2}{a}\left[-\frac{1}{a}xe^{-ax}\Big\vert_0^\infty + \frac{1}{a}\int_0^\infty e^{-ax}dx \right]\\ &= -\frac{2}{a^3}e^{-ax}\Big\vert_0^\infty\\ &= \frac{2}{a^3} \end{align}$$

  • 0
    So what's the answer to the question?2017-02-08
  • 0
    @zhw. Updated my answer.2017-02-09