2
$\begingroup$

Note: We define the signum function, $\text{sgn}$ by $\text{sgn}(x) = 1$ if $x \geq 0$, and $-1$ otherwise.

Suppose $f: [0, \infty) \to \mathbb R$ is continuous and of locally bounded variation, with $f(0) = 0$. Is it true that the integral

$$\int_0^t f \, d\, \text{sgn(f)} := f(t) \, \text{sgn}(f(t)) - \int_0^t \text{sgn} (f(s)) \, df(s)$$

$$= \lvert f(t) \rvert - \int_0^t \text{sgn} (f(s)) \, df(s)$$ vanishes for all $t \geq 0?$

Remark: Note that the RHS is a definition for the LHS! The integral on the RHS is to be interpreted as a regular Lebesgue Stiltjes integral.

$\endgroup$

1 Answer 1

2
$\begingroup$

$\newcommand\sgn{\operatorname{sgn}}$Yes, this is true. Indeed, fix any real $t>0$. Let \begin{equation} S_+:=\{s\in(0,t)\colon f(s)\ge0\},\quad S_-:=\{s\in(0,t)\colon f(s)<0\}. \end{equation} Then \begin{equation} I:=\int_0^t\sgn(f(s))\,df(s)=I_+ - I_-, \end{equation} where \begin{equation} I_-:=\int_{S_-}df(s), \end{equation} \begin{equation} I_+:=\int_{S_+}df(s)=\int_0^1 df(s)-\int_{S_-}df(s)=f(t)-I_-, \end{equation}
so that \begin{equation} I=f(t) - 2I_-. \end{equation} Since $f$ is continuous, the set $S_-$ is open, so that $S_-$ is the union of at most countably many disjoint nonempty open intervals. If the right endpoint of one of those intervals is $t$, let us write that interval as $(a,t)$, with $a\in[0,t)$. If the right endpoint of none of those intervals is $t$, let us use the empty interval $(a,t)$ with $a=t$. So,
\begin{equation} S_-=(a,t)\cup\bigcup_{j\in J}(a_j,b_j), \end{equation} where $J$ is an at most countable set, $0\le a_j<b_j\le a\le t$ and $b_j<t$ for all $j\in J$ and the intervals $(a_j,b_j)$ are pairwise disjoint.

If $f(t)>0$, then $a=t$ by the continuity of $f$, and hence $\int_{(a,t)}df(s)=0=f(t)\,1(f(t)\le0)$.

If $f(t)\le0$ and $a=t$, then $f(t)=0$ by the continuity of $f$, and hence $\int_{(a,t)}df(s)=0=f(t)\,1(f(t)\le0)$.

If $f(t)\le0$ and $a<t$, then $f(a)=0$ by the continuity of $f$, and hence $\int_{(a,t)}df(s)=f(t)-f(a)=f(t)=f(t)\,1(f(t)\le0)$.

So, in any case, \begin{equation} \int_{(a,t)}df(s)=f(t)\,1(f(t)\le0). \end{equation}

Also, by the continuity of $f$, we have $\int_{(a_j,b_j)}df(s)=f(b_j)-f(a_j)=0-0$ for all $j\in J$.

So, \begin{equation} I_-=\int_{(a,t)}df(s)+\sum_{j\in J}\int_{(a_j,b_j)}df(s)=f(t)\,1(f(t)\le0). \end{equation} Thus, \begin{equation} I=f(t) - 2I_-=f(t) - 2f(t)\,1(f(t)\le0)=|f(t)|, \end{equation} as claimed.

$\endgroup$
4
  • $\begingroup$ Ah, may I ask how the Stiltjes integral over $S_+$, and $S_-$ is defined? I have not seen partially defined Lebesgue Stiltjes integrals before. $\endgroup$
    – Nate River
    Commented Oct 24, 2022 at 14:44
  • $\begingroup$ @NateRiver : This is the Lebesgue integral with respect to the Lebesgue--Stieltjes measure. $\endgroup$ Commented Oct 24, 2022 at 14:57
  • $\begingroup$ Ahh, right that makes sense. $\endgroup$
    – Nate River
    Commented Oct 24, 2022 at 15:02
  • $\begingroup$ @NateRiver : You are welcome! $\endgroup$ Commented Oct 24, 2022 at 15:08

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .