3
$\begingroup$

Let $\alpha$ be Stieltjes weight function (monotone increasing), and $\alpha = \alpha_1+\alpha_2$, where $\alpha_1,\alpha_2$ are also monotone increasing. suppose $f$ is Stieltjes-integrable on $[a,b]$

$$\int_a^bf d\alpha = \int_a^b f d\alpha_1 + \int_a^b f d\alpha_2$$

Under what conditions is this statement true? How can I prove it easily using basic knowledge of upper and lower Stieltjes integral? If it is not true all the time counter examples?

$\endgroup$
  • $\begingroup$ It should be sufficient that measure $\alpha(M)= \alpha_1(M) + \alpha_2(M)$. $\endgroup$ – tom Feb 7 '13 at 22:53
  • $\begingroup$ I didn't learn measure yet, just in the basic analysis course $\endgroup$ – mez Feb 7 '13 at 22:56
  • $\begingroup$ Try to show that $\alpha(M) = \text{inf}\{ \sum_i \alpha(I_i): M\subset \bigcup_i I_i\} = \text{inf}\{ \sum_i \alpha_1(I_i): M\subset \bigcup_i I_i\} + \text{inf}\{ \sum_i \alpha_2(I_i): M\subset \bigcup_i I_i\} = \alpha_1(M) +\alpha_2(M) $. Where $I$ are intervals of type $ (c,d] $. Plus functions $\alpha,\alpha_1, \alpha_2$ should be right continuous. $\alpha((c,d]) = \alpha(d) - \alpha(c) $ $\endgroup$ – tom Feb 7 '13 at 23:01
3
$\begingroup$

$\newcommand\uint{\overline{\int_a^b}}\newcommand\lint{\underline{\int_a^b}}$

Notations

Suppose $\alpha$ is monotone increasing on $[a,b]$, $P=\{\,x_0,x_1,\dotsc,x_n\,\}$ is a partition of $[a,b]$, where $a=x_0\le x_1\le\dotsb\le x_n=b$, $f$ is a bounded function on $[a,b]$, we'll use notations here:

\begin{align} &\Delta\alpha_k=\alpha(x_k)-\alpha(x_{k-1})\\\\ &M_k=\sup_{x_{k-1}\le x\le x_k}f(x) &m_k=\inf_{x_{k-1}\le x\le x_k}f(x)\\ &U(P,f,\alpha)=\sum_{k=1}^nM_k\Delta\alpha_k &L(P,f,\alpha)=\sum_{k=1}^nm_k\Delta\alpha_k\\ &\uint fd\alpha=\inf_P U(P,f,\alpha) &\lint fd\alpha=\sup_P U(P,f,\alpha) \end{align}

Stronger results

\begin{align} \uint fd(\alpha_1+\alpha_2)&=\uint fd\alpha_1+\uint fd\alpha_2\\ \lint fd(\alpha_1+\alpha_2)&=\lint fd\alpha_1+\lint fd\alpha_2 \end{align}

Proof

We'll only prove the first one. For each partition $P$, it's easier to check that $$U(P,f,\alpha_1+\alpha_2)=U(P,f,\alpha_1)+U(P,f,\alpha_2)$$ Take $\inf_P$ on both sides, we have $$\uint fd(\alpha_1+\alpha_2)\ge\uint fd\alpha_1+\uint fd\alpha_2$$

Next, for each $u_1>\uint fd\alpha_1$, $u_2>\uint fd\alpha_2$, there's some partitons $P_1$ and $P_2$ such that $U(P_1,f,\alpha_1)<u_1$ and $U(P_2,f,\alpha_2)<u_2$.

Consider $P=P_1\cup P_2$, we have $U(P,f,\alpha_1)\le U(P_1,f,\alpha_1)<u_1$ and $U(P,f,\alpha_2)<u_2$, then $$U(P,f,\alpha_1+\alpha_2)=U(P,f,\alpha_1)+U(P,f,\alpha_2)<u_1+u_2$$ therefore $$\uint fd(\alpha_1+\alpha_2)<u_1+u_2$$

Hence $$\uint fd(\alpha_1+\alpha_2)\le\uint fd\alpha_1+\uint fd\alpha_2$$

Proved.

$\endgroup$
  • $\begingroup$ Very good proof. thanks. $\endgroup$ – mez Feb 8 '13 at 10:00
  • $\begingroup$ Your proof partially finishes my question. I have further question, what if now we only assume that $f$ is stieltjes integrable with respect to $\alpha = \alpha_1+\alpha_2$, and only that $\alpha_1$,$\alpha_2$ are monotone increasing bounded functions, can we deduce $$\int_a^b f d\alpha = \int_a^b f d\alpha_1 +\int_a^b f d\alpha_2$$? We are not sure whether $$\overline{\int}_a^b f d\alpha_1 = \underline{\int}_a^b f d\alpha_1$$, is it true to begin with? $\endgroup$ – mez Feb 8 '13 at 10:43
  • $\begingroup$ @mezhang According to the definition, $\uint fd\alpha=\lint fd\alpha$ if and only if $f\in\mathcal R(\alpha)$, i.e. integrable (according to $\alpha$). $\endgroup$ – Yai0Phah Feb 9 '13 at 11:14
  • $\begingroup$ This I know, but how does $\overline{\int}_a^b f d(\alpha_1+\alpha_2) = \underline{\int}_a^b f d(\alpha_1+\alpha_2)$ imply $\overline{\int}_a^b f d\alpha_1 = \underline{\int}_a^b f d\alpha_1$? $\endgroup$ – mez Feb 9 '13 at 13:01
  • $\begingroup$ @mezhang Since $\uint fd\alpha_1+\uint fd\alpha_2=\int_a^bfd(\alpha_1+\alpha_2)=\lint fd\alpha_1+\lint fd\alpha_2$, and $\uint fd\alpha_k\ge\lint fd\alpha_k$, so $\uint fd\alpha_k=\lint fd\alpha_k$ for $k=1,2$. $\endgroup$ – Yai0Phah Feb 10 '13 at 5:58

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.