3

Suppose $F$ is an increasing function on $[a, b]$. (a) Prove that we can write $$ F=F_{A}+F_{C}+F_{J} $$ where each of the functions $F_{A}, F_{C}$, and $F_{J}$ is increasing and:

(i) $F_{A}$ is absolutely continuous.

(ii) $F_{C}$ is continuous, but $F_{C}^{\prime}(x)=0$ for a.e. $x$.

(iii) $F_{j}$ is a jump function.

If we let $$ j_{n}(x)=\left\{\begin{array}{cl} 0 & \text { if } x<x_{n} \\ \theta_{n} & \text { if } x=x_{n} \\ 1 & \text { if } x>x_{n} \end{array}\right. $$ then we define the jump function associated to $F$ by $$ J_{F}(x)=\sum_{n=1}^{\infty} \alpha_{n} j_{n}(x) . $$

I am kind of sure that I have to use the property that an increasing function can have only countably many jump discontinuity. Given an increasing order of discontinuities $\{x_n\}$ of $F$ i.e., $x_n\le x_{n+1}$ and $f$ has discontinuities at $x_n$ then we can represent the function $$ F(x)=f_n(x); x_n\le x\le x_{n+1} $$ and $f_n$ is continuous on $(x_n,x_{n+1})$.

Now we can define a function $$G(x)=\left\{\begin{array}{cl} g_1(x):=f_1(x) & \text { if } a=x_1\leq x<x_{2} \\ g_{n}(x) & \text { if } x_n\le x\le x_{n+1} \end{array}\right. $$ where $g_n(x)$ is a function defined on $(x_n,x_{n+1})$ having the same slope of $f_n(x)$ on $(x_n,x_{n+1})$ so that $G(x)$ is continuous basically I am removing the jump discontinuities of $F$. So we can write $F(x)=G(x)+J(x)$ where $$J(x)=f_n(\frac{x_n+ x_{n+1}}{2})-g_n(\frac{x_n+ x_{n+1}}{2}); x_n\le x\le x_{n+1} $$

then $J(x)$ is a step function. I am wondering if I can at all transform this thing into the solution I want. Let me know if there is any other way. Please help!

Ri-Li
  • 9,038
  • 1
    Hm, this is essentially Lebesgue's decomposition theorem using that "strictly increasing functions" are basically the same as "measures on [a,b]" and "jump discontinuity" corresponds to "point mass", "absolutely continuous" refers to, uh, "absolutely continuous" (but with measures this time), and the second condition refers to measures singular to Lebesgue measure with no point masses. Would a proof translating this into measure theory and using that theorem be useful? – Milo Brandt May 02 '21 at 02:09
  • I think so. If you can do that. – Ri-Li May 02 '21 at 15:51

0 Answers0