Question about total variation, positive variation and negative variation

1.7k Views Asked by At

I have a question about the following problem:

Let $f:[a,b] \to \mathbb{R}$ be a function of bounded variation with $f(a) = 0$ and $f_1$, $f_2$ be two increasing function such that $f_1(a) = 0$ and $f_2(a)=0$. and $f=f_1-f_2$. Define positive and negative variation $$TV^{+}(f)_a^x = \frac{TV(f)_a^x + f(x)}{2},TV^{-}(f)_a^x = \frac{TV(f)_a^x - f(x)}{2} $$ Then by Jordan decomposition we know $$f = TV^+(f)_a^x - TV^-(f)_a^x$$ where two terms are increasing. Show this is a decomposition of $f$ as the difference of two increasing functions in the lowest terms ($TV^+(f) \leq f_1$ and $TV^-(f) \leq f_2$)

I first of all don't get what the question is telling us to show. So prove that for all increasing function positive and negative variation is the smallest?

Thank you!

1

There are 1 best solutions below

2
On

The question wants you to show that ${TV}^{+}(f)\leq f_1$ and ${TV}^{-}(f)\leq f_2$ as I think you've surmised :)
What that means in English is that any function of bounded variation can be decomposed into the difference of two increasing functions and that the smallest${}^{1}$ pair of such functions are the positive and negative variations of $f$ itself. This is useful in that it prescribes a way to find a decomposition of $f$ into the difference of increasing functions so that each time doesn't require a special effort (and proof).

${}^{1}$Normally for 'smallest' we'd say $|{TV}^{±}(f)|\leq f_i$ where $i=1,2$ respectively, but since we're talking about strictly non-negative functions we can drop the absolute values.

EDIT: To show that ${TV}^+(f)\leq f_1$ as well:
First of all note that it's enough to show that ${TV}^+(f)\leq f_1$, as ${TV}^-(f)\leq f_2$ is an immediate consequence (just compare the two decompositions of $f$). Next we need a technical lemma to help us:
Lemma: if there exists an increasing function $f_1:[a,b]\rightarrow \mathbb{R}$ and a function $f$ such that $|f(y)-f(x)|\leq f_1(y)-f_1(x)$ for all $[x,y] \subseteq [a,b]$ then $f$ is of bounded variation and, in particular, ${TV}(f)^y_x \leq f_1(y)-f_1(x)$.

Proof: ${TV}(f)^b_a =: \sup_{P\in \cal{P}} \{ \sum_{i=1}^{N}\left|f(x_i)-f(x_{i-1})\right| \} \leq \sup_{P\in \cal{P}} \{ \sum_{i=1}^{N}\left|f_1(x_i)-f_1(x_{i-1})\right| \} \leq f_1(b)-f_1(a)$ since the middle sum telescopes because $f_1$ is a non-negative increasing function.

Finally then: let $f=f_1 - f_2$ be the decomposition into non-negative increasing functions and let $a\leq x < y \leq b$. Then: $$\left|f(y)-f(x)\right| = \left|(f_1(y)-f_1(x)) - (f_2(y)-f_2(x)) \right| $$ $(f_2(y)-f_2(x)) \geq 0$ because $f_2$ is non-negative and increasing so it can only make the RHS smaller. Hence $$\left|f(y)-f(x)\right| \leq f_1(y)-f_1(x) $$ where we've dropped the absolute value because this RHS is also non-negative. We know that ${TV}(f)^y_x = {TV}^+(f)^y_x + {TV}^-(f)^y_x$ and that the RHS here is also two positive quantities. Putting this altogether: $$ {TV}^+(f)^y_x \leq {TV}^+(f)^y_x + {TV}^-(f)^y_x = {TV}(f)^y_x \leq f_1(y)-f_1(x) \leq f_1(y) $$ where the second-to-last inequality uses the technical lemma we proved.