This is a proof verification post. I'm attaching the problem and my solution to it. I believe that my solution is too long. Please check it it is correct and also let me know if there are methods by which I can shorten the proof a bit. Thank you.

The Problem : Let $\left(\Omega,\mathcal{F},\mu\right)$ be a measure space and $f,g:\Omega \to \mathbb{R}$ be Borel functions, i.e. $f^{-1}\left(\mathcal{B}\right) \subset \mathcal{F},~g^{-1}\left(\mathcal{B}\right) \subset \mathcal{F},$ where $\mathcal{B}$ denotes the Borel-$\sigma$-field of $\mathbb{R}$.

Show that, if $\int_{\Omega} f d\mu$ exists and $c \in \mathbb{R},$ then $\int_{\Omega} \left(cf\right) d\mu$ exists and $$\int_{\Omega} \left(cf\right) d\mu=c\int_{\Omega} f d\mu$$

My Solution : To show existence, it is enough to show that $cf$ is Borel. We have to show that $\left(cf\right)^{-1}\left(\mathcal{B}\right) \subset \mathcal{F}$. Let us choose (arbitrarily) $B \in \mathcal{B}$. We want to show that $\left(cf\right)^{-1}\left(B\right) \in \mathcal{F}$.

Note that, $\left(cf\right)^{-1}\left(B\right)=\left\{\omega \in \Omega : \left(cf\right)\left(\omega\right) \in B\right\}=\left\{\omega \in \Omega : cf\left(\omega\right) \in B\right\}$

Suppose $c=0$. Either we have $0 \in B,$ or $0 \notin B$. Thus, $\left\{\omega \in \Omega : cf\left(\omega\right) \in B\right\}=\left\{\omega \in \Omega : 0 \in B\right\}=\phi \text{ or } \Omega,$ and hence, $\left(cf\right)^{-1}\left(B\right) \in \mathcal{F}$.

Now suppose $c \neq 0$. Define, $\frac{B}{c}:=\left\{\frac{b}{c} : b \in B\right\}$. Then, \begin{align*} \left(cf\right)^{-1}\left(B\right)&=\left\{\omega \in \Omega : cf\left(\omega\right) \in B\right\}\\ &=\left\{\omega \in \Omega :\exists b \in B, \text{ such that } cf\left(\omega\right)=b\right\}\\ &=\left\{\omega \in \Omega :\exists b \in B, \text{ such that } f\left(\omega\right)=\frac{b}{c}\right\}\\ &=\left\{\omega \in \Omega : f\left(\omega\right) \in \frac{B}{c}\right\}\\ &=f^{-1}\left(\frac{B}{c}\right) \end{align*} If we can show that $\frac{B}{c} \in \mathcal{B}$, we are done.

Now, $B \in \mathcal{B} \implies \exists \text{ a sequence } \left(A_n\right)_{n=1}^{\infty} \text{ of open intervals in } \mathbb{R} \text{ such that } B=\bigcup_{n=1}^{\infty}A_n$

Claim : $\frac{B}{c}=\bigcup_{n=1}^{\infty}\frac{A_n}{c}$. Now,

\begin{align*} a \in \frac{B}{c} &\implies \exists b \in B, \text{ such that } a=\frac{b}{c}\\ &\implies \exists n \in \mathbb{N} \setminus \left\{0\right\}, \text{ and } b \in A_n, \text{ such that } a=\frac{b}{c}\\ &\implies \exists n \in \mathbb{N} \setminus \left\{0\right\}, \text{ such that } a=\frac{A_n}{c}\\ &\implies a \in \bigcup_{n=1}^{\infty} \frac{A_n}{c} \end{align*}

Thus, $$\frac{B}{c} \subset \bigcup_{n=1}^{\infty} \frac{A_n}{c}$$ To show the other side,

\begin{align*} a \in \bigcup_{n=1}^{\infty} \frac{A_n}{c} &\implies \exists n \in \mathbb{N} \setminus \left\{0\right\}, \text{ such that } a=\frac{A_n}{c}\\ &\implies \exists n \in \mathbb{N} \setminus \left\{0\right\}, \text{ and } b \in A_n, \text{ such that } a=\frac{b}{c}\\ &\implies \exists b \in B, \text{ such that } a=\frac{b}{c}\\ &\implies a \in \frac{B}{c} \end{align*} Thus, $$\bigcup_{n=1}^{\infty} \frac{A_n}{c} \subset \frac{B}{c}$$ Hence we have, $$\frac{B}{c}=\bigcup_{n=1}^{\infty} \frac{A_n}{c}$$

The crucial (although trivial) thing is that for all $n \in \mathbb{N} \setminus \left\{0\right\}, A_n \text{ is an open interval in } \mathbb{R} \implies \frac{A_n}{c} \text{ is an open interval in } \mathbb{R}$. And hence, $\frac{B}{c} \in \mathcal{B}$. This completes the existence part.

Now to show that, $$\int_{\Omega} \left(cf\right) d\mu=c\int_{\Omega} f d\mu \tag{1}$$

If $c=0,$ $(1)$ holds trivially.

$$\text{By definition, }L.H.S.=\int_{\Omega} \left(cf\right) d\mu=\int_{\Omega} \left(cf\right)^+ d\mu-\int_{\Omega} \left(cf\right)^- d\mu$$ $$\int_{\Omega} \left(cf\right)^+ d\mu=\sup\left\{\int_{\Omega}\psi d\mu : \psi \text{ is a simple function}, 0 \leq \psi \leq \left(cf\right)^+\right\}$$ Suppose, $c>0$. Then, $\left(cf\right)^+ = \max\left\{cf,0\right\} = c \max\left\{f,0\right\} = cf^+$

\begin{align*} \int_{\Omega} \left(cf\right)^+ d\mu&=\sup\left\{\int_{\Omega}\psi d\mu : \psi \text{ is a simple function}, 0 \leq \psi \leq cf^+\right\}\\ &=\sup\left\{\int_{\Omega}\psi d\mu : \psi \text{ is a simple function}, 0 \leq \frac{\psi}{c} \leq f^+\right\}\\ &=\sup\left\{\int_{\Omega}c\cdot\frac{\psi}{c} d\mu : \frac{\psi}{c} \text{ is a simple function}, 0 \leq \frac{\psi}{c} \leq f^+\right\}\\ &\left[\psi \text{ is a simple function } \implies \frac{\psi}{c} \text{ is a simple function } \right]\\ &=\sup\left\{\int_{\Omega}c\psi^* d\mu : \psi^* \text{ is a simple function}, 0 \leq \psi^* \leq f^+\right\}\\ &=\sup\left\{c\int_{\Omega}\psi^* d\mu : \psi^* \text{ is a simple function}, 0 \leq \psi^* \leq f^+\right\}\\ &=c\sup\left\{\int_{\Omega}\psi^* d\mu : \psi^* \text{ is a simple function}, 0 \leq \psi^* \leq f^+\right\}\\ &=c\int_{\Omega} f^+ d\mu \end{align*}

In the same way, $\int_{\Omega} \left(cf\right)^- d\mu = c\int_{\Omega} f^- d\mu,$ and hence,

$$\int_{\Omega} \left(cf\right) d\mu=\int_{\Omega} \left(cf\right)^+ d\mu-\int_{\Omega} \left(cf\right)^- d\mu=c\int_{\Omega} f^+ d\mu-c\int_{\Omega} f^- d\mu=c\int_{\Omega} f d\mu$$

Similarly, if $c<0,$ we have $\left(cf\right)^+ = \max\left\{cf,0\right\} = c\min\left\{f,0\right\} = -c\max\left\{-f,0\right\} = -cf^-,$ and hence, $\int_{\Omega} \left(cf\right)^+ d\mu = -c\int_{\Omega} f^- d\mu$.

Also in this case, $\left(cf\right)^- = \max\left\{-cf,0\right\} = -c\max\left\{f,0\right\} = -cf^+,$ and hence, $\int_{\Omega} \left(cf\right)^- d\mu = -c\int_{\Omega} f^+ d\mu$.

Thus we have $$\int_{\Omega} \left(cf\right) d\mu=\int_{\Omega} \left(cf\right)^+ d\mu-\int_{\Omega} \left(cf\right)^- d\mu=-c\int_{\Omega} f^- d\mu+c\int_{\Omega} f^+ d\mu=c\int_{\Omega} f d\mu$$

Hence, in all the cases, $(1)$ holds. Hence the proof.


Seems like you are doing a lot of extra work. How about recalling that the integral of a simple function $\psi(x) = \sum \limits_{i=1}^n a_i \chi_{A_i} \left( x \right)$ is $\sum \limits_{i=1}^n a_i \mu(A_i) $. In this case the proof is trivial. Then you can wrap up by e.g. recalling that any measurable $f$ is a pointwise limit of a monotone sequence of simple functions and then applying monotone convergence theorem.

  • $\begingroup$ Of course, this depends on what you are allowed to use. $\endgroup$ – Matias Heikkilä Aug 5 '17 at 8:27
  • $\begingroup$ Thanks for the response. I'm only beginning and I've not studied MCT yet. But it's reassuring to know that I don't have to do such long proofs using only fundamental definitions in future. $\endgroup$ – pkxop Aug 5 '17 at 9:21

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.