Definitions. A finite subset P={x0,x1,…,xnP}⊂[a,b] is a partition of the segment [a,b] if and only if a=x0<x1<…,xnP−1<xnP=b. The set of all partitions of the segment [a,b] is denoted by Part[a,b]. If f:[a,b]→R is a function, then the variation of f(x) on the partition P is the sum V(f;P)=i=1∑nP∣f(xi)−f(xi−1)∣. A quantity Vab(f):=sup{V(f;P):P∈Part[a,b]} is called by the total variation of the function f(x) on the segment [a,b]. The function f(x) is a function of bounded variation on the segment [a,b], if and only if Vab(f)<+∞.
Theorem 1. If f(x) is a function of bounded variation on the segment [a,b], then f(x) is bounded on [a,b].
Proof. Suppose that f(x) is unbounded on [a,b]. Then for all k∈N there exists tk∈[a,b] such that ∣f(tk)∣→+∞, as k→+∞ . Consider the partition Pk={a,tk,b}∈Part[a,b].
V(f;Pk)=∣f(tk)−f(a)∣+∣f(tk)−f(b)∣≥2∣f(tk)∣−∣f(a)∣−∣f(b)∣→+∞
Therefore, Vab(f)=sup{V(f;P):P∈Part[a,b]}=+∞ and f(x) is not a function of bounded variation on [a,b]. The received contradiction proves the theorem.
Theorem 2. If c∈(a,b) then Vab(f)=Vac(f)+Vcb(f).
Proof. Let P1={x0,x1,…,xn1}∈Part[a,c], P2={y0,y1,…,yn2}∈Part[c,b]. Put P=P1∪P2∈Part[a,b]. Then
V(f;P)=i=1∑n1∣f(xi)−f(xi−1)∣+j=1∑n2∣f(yj)−f(yj−1)∣=V(f;P1)+V(f;P2)
Hence
(*) Vab(f)=sup{V(f;P):P∈Part[a,b]}≥
sup{V(f;P1)+V(f;P2)∣P1∈Part[a,c],P2∈Part[c,b]}=
=Vac(f)+Vcb(f)
By now, let P∈Part[a,b] be arbitrary. Put P′=P∪{c}.
If c∈P, then P′=P andV(f;P′)=V(f;P).
If c∈/P then c∈(xk,xk+1) for some k and, hence, V(f;P′)−V(f;P)=∣f(c)−f(xk)∣+∣f(xk+1)−f(c)∣−∣f(xk+1)−f(xk)∣≥0 by the triangle inequality.
Put LP=P′∩[a,c]∈Part[a,c], RP=P′∩[c,b]∈Part[c,b]. Then P′=LP∪RP and LP∩RP={c}. For such partitions it was proved above that V(f;P′)=V(f;LP)+V(f;RP)
Therefore
V(f;P)≤V(f;P′)=V(f;LP)+V(f;RP)≤Vac(f)+Vcb(f)
Taking the supremum by P , we obtain
Vab(f)≤Vac(f)+Vcb(f)
Combining this with (*), we get Vab(f)=Vac(f)+Vcb(f).
Q.E.D.
Comments