122 CHAPTER 5. LINE INTEGRALS AND CURVES

and so γh ∈C1 ([a,b]) . Next is a fundamental estimate.

Lemma 5.2.4 V (γh, [a,b])≤V (γ, [a,b]) .

Proof: Let a = t0 < t1 < · · ·< tn = b. Then using the definition of γh and changing thevariables to make all integrals over [0,2h] ,

n

∑j=1

∣∣γh (t j)− γh(t j−1

)∣∣= n

∑j=1

∣∣∣∣∣[

12h∫ 2h

0 γ(s−2h+ t j +

2hb−a (t j−a)

)−γ(s−2h+ t j−1 +

2hb−a

(t j−1−a

)) ]∣∣∣∣∣≤ 1

2h

∫ 2h

0

n

∑j=1

∣∣∣∣ γ(s−2h+ t j +

2hb−a (t j−a)

)−γ(s−2h+ t j−1 +

2hb−a

(t j−1−a

)) ∣∣∣∣ds

For a given s∈ [0,2h] , the points, s−2h+ t j +2h

b−a (t j−a) for j = 1, · · · ,n form an increas-ing list of points in the interval [a−2h,b+2h] and so the integrand is bounded above byV (γ, [a−2h,b+2h]) =V (γ, [a,b]) . It follows

n

∑j=1

∣∣γh (t j)− γh(t j−1

)∣∣≤V (γ, [a,b]) so V (γh, [a,b])≤V (γ, [a,b]) ■

With this lemma the proof of Theorem 5.2.3 can be completed without too much trou-ble. By uniform continuity of γ, if h is small enough, say h < δ 1, then for all t ∈ [a,b] ,

|γ (t)− γh (t)| ≤1

2h

∫ t+ 2h(b−a) (t−a)

−2h+t+ 2h(b−a) (t−a)

|γ (s)− γ (t)|ds

<1

2h

∫ t+ 2h(b−a) (t−a)

−2h+t+ 2h(b−a) (t−a)

εds = ε (5.14)

This proves 5.11. It remains to verify the approximation of the integrals.Let P = {t0, ..., tn}

Sh (P) ≡n

∑k=1

f(γh (τk)) · (γh (tk)− γh (tk−1)) (5.15)

S (P) ≡n

∑k=1

f(γ (τk)) · (γ (tk)− γ (tk−1)) (5.16)

From estimates of Theorem 5.1.5 and the fact that the total variation of γh is no morethan that of γ , there exists δ 2 such that if ∥P∥< δ 2, then∣∣∣∣∫

γ

f ·dγ (t)−S (P)∣∣∣∣< ε

3,

∣∣∣∣∫γh

f ·dγh (t)−Sh (P)∣∣∣∣< ε

3(5.17)

Then consider |S (P)−Sh (P)| where 0 < h < min(δ 1,δ 2). For such a fixed P, choose hsmall enough in 5.15, 5.16 that |S (P)−Sh (P)|< ε

3 . Then for this h and P,∣∣∣∣∫γ

f ·dγ (t)−∫

γh

f ·dγh (t)∣∣∣∣

≤∣∣∣∣∫

γ

f ·dγ (t)−S (P)∣∣∣∣+ |S (P)−Sh (P)|+

∣∣∣∣∫γh

f ·dγh (t)−Sh (P)∣∣∣∣

< ε/3+ ε/3+ ε/3 = ε

122 CHAPTER 5. LINE INTEGRALS AND CURVESand so y, € C! ({a,b]). Next is a fundamental estimate.Lemma 5.2.4 V (y,,[a,b]) <V(%[a,5]).Proof: Let a=t <t; <---<t, =b. Then using the definition of y, and changing thevariables to make all integrals over [0, 2h] ,oH 0 (s—2h+t;+ 24 (t;—a))(82h +t)1 + pq (4-1-4)Me:1% (t7) — Yp (ti-1) | -¥2h nN<a &For a given s € [0,2], the points, s—2h+1;+ 74 (t; —a) for j= 1,--- ,n form an increas-ing list of points in the interval [a —2h,b+ sr ‘and so the integrand j is s bounded above byV (y, [a —2h, b+ 2h|) = V(¥, |a, b]) . It follows1J¥(s—2h+4)+ po (ya)Yom. it pa BA ty. 1—a)) “YE lnled)— malo] SV Clad) s0V (lesb) < Vlad)With this lemma the proof of Theorem 5.2.3 can be completed without too much trou-ble. By uniform continuity of y, if # is small enough, say h < 61, then for all t € [a,b],®—n0| < 5 fo Ins) —710|as2htt+ ey (ta)t+ Bra (t—a)< it =€E 5.14Glos,This proves 5.11. It remains to verify the approximation of the integrals.Let P = {to,...,tn}Sn(P) = Y.£( yy (te) - (a (te) ~ Yo (te-1)) (5.15)k=1s(P) = Yk(y(te)) (y(t) —Y(te-1)) (5.16)ilinFrom estimates of Theorem 5.1.5 and the fact that the total variation of Y;, is no morethan that of y, there exists 52 such that if ||P|| < 62, then-dy(t)ft dy, (t)— 54(P) <5 (5.17)Then consider si) —S,(P)| where 0 < h < min(6), 62). For such a fixed P, choose hsmall enough in 5.15, 5.16 that |S(P) — S;, (P)| < §. Then for this h and P,f-dy(t)— |] f-dy,(t)Y Vh[tars] +15@)-s)1+|/ tani") —Si(P)< €/3+6€/3+8/3=€<