Jordan decomposition lemma (bounded variation)

A function of bounded variation is the difference of two increasing functions
Jordan decomposition lemma (bounded variation)

Let f:[a,b]Rf:[a,b]\to\mathbb{R} be a function of . For x[a,b]x\in[a,b], write Vax(f)=sup{j=1nf(tj)f(tj1): a=t0<t1<<tn=x} V_a^x(f)=\sup\left\{\sum_{j=1}^n |f(t_j)-f(t_{j-1})|:\ a=t_0<t_1<\cdots<t_n=x\right\} for the of ff on [a,x][a,x].

Jordan decomposition lemma: If ff has bounded variation on [a,b][a,b], then there exist functions g,h:[a,b]Rg,h:[a,b]\to\mathbb{R} such that f=gh. f=g-h. One explicit choice is h(x)=Vax(f),g(x)=f(x)+Vax(f). h(x)=V_a^x(f),\qquad g(x)=f(x)+V_a^x(f). Then hh is increasing, and gg is increasing as well.

This lemma is fundamental for the : bounded-variation integrators behave like differences of increasing (hence “measure-like”) functions.

Proof sketch: The map xVax(f)x\mapsto V_a^x(f) is increasing by definition. For ax<yba\le x<y\le b, the variation estimate implies Vay(f)Vax(f)f(y)f(x)(f(y)f(x)), V_a^y(f)-V_a^x(f)\ge |f(y)-f(x)|\ge -(f(y)-f(x)), so g(y)g(x)=(f(y)f(x))+(Vay(f)Vax(f))0, g(y)-g(x)=(f(y)-f(x))+\bigl(V_a^y(f)-V_a^x(f)\bigr)\ge 0, hence gg is increasing. Finally, f=ghf=g-h holds by construction.