B.3. THE YOUNG INTEGRAL 381

Also let P be a dissection 0 = t0 < t1 < · · ·< tr = T .Claim: Then there exists ti such that ω (ti−1, ti+1)≤ 5

r ω (0,T ).Proof of claim: If r = 2 so there are only two intervals in the dissection,

∥F∥pp,[0,T ] ≤ ∥F∥

pp,[0,T ]+∥Z∥

pp,[0,T ] = ω (0,T )<

52

ω (0,T ) .

So assume r > 2. Then how many disjoint intervals [ti−1, ti+1] are there for either i odd or ieven? Certainly fewer than r and at least as many as r

2 . Then we have

∑i even

ω (ti−1, ti+1)+ ∑i odd

ω (ti−1, ti+1)≤ 2ω (0,T )

Now if it is not true that there exists such an i, then for each i,ω (ti−1, ti+1)>5r ω (0,T ) .

52

ω (0,T ) =5r

ω (0,T )r2≤ ∑

i even

5r

ω (0,T )< ∑i even

ω (ti−1, ti+1)≤ ω (0,T )

52

ω (0,T ) =5r

ω (0,T )r2≤ ∑

i odd

5r

ω (0,T )< ∑i odd

ω (ti−1, ti+1)≤ ω (0,T )

Hence 5ω (0,T )< 2ω (0,T ) which is absurd. Therefore, there exists such a ti. This showsthe claim.

Next points of P are deleted beginning with ti where ω (ti−1, ti+1)≤ 5r ω (0,T ) .∫

PZdF−

∫P\{ti}

ZdF = Z (ti−1)(F (ti)−F (ti−1))

+Z (ti)(F (ti+1)−F (ti))−Z (ti−1)(F (ti+1)−F (ti−1))

This equals Z (ti−1)F (ti)+Z (ti)F (ti+1)−Z (ti)F (ti)−Z (ti−1)F (ti+1)

= Z (ti)(F (ti+1)−F (ti))+Z (ti−1)(F (ti)−F (ti+1))

= (Z (ti)−Z (ti−1))(F (ti+1)−F (ti))

It follows that∣∣∣∣∫P

ZdF−∫

P\{ti}ZdF

∣∣∣∣≤ |(Z (ti)−Z (ti−1))| |F (ti+1)−F (ti)|

≤ ω (ti−1, ti)1/q

ω (ti, ti+1)1/p ≤ ω (ti−1, ti+1)

1/p+1/q ≤(

5r

ω (0,T ))1/p+1/q

In particular, this yields∣∣∣∣∫P(Y −Y (0))dF−

∫P\{ti}

(Y −Y (0))dF∣∣∣∣≤ (5

rω (0,T )

)1/p+1/q

=

(5r

(∥F∥p

p,[0,T ]+∥Y∥qq,[0,T ]

))1/p+1/q

B.3. THE YOUNG INTEGRAL381Also let Y be a dissection 0 = t9 <t) <-::<t-=T.Claim: Then there exists ¢; such that @ (t;-1, 11) < 2@ (0,7).Proof of claim: If r = 2 so there are only two intervals in the dissection5Pp P P _ =FIP or) SFU 9 F IZM 9.7) = O0.7) < 500.7).So assume r > 2. Then how many disjoint intervals [t;_; ,t;:1] are there for either i odd or ieven? Certainly fewer than r and at least as many as 5. Then we haveYY? @(t-1.tii)+ VY @(G-1,ti41) < 2@ (0,7)ieven ioddNow if it is not true that there exists such an i, then for each i, @ (4-1,t:41) > 2@(0,T)-~0@(0,T)=-@ o(0,T) O (ti-1, ti <@(0,T50(0,T) < Y 200.7) < Y ol-stisr) <@(.7)ieven ievenmT>09(0,7) =" > (0,7)a5<y —@(0,T) < y O (ti-1,ti41) < @(0,T)iodd 7 i oddHence 5@ (0,7) < 2@(0,T) which is absurd. Therefore, there exists such a t;. This showsthe claim.Next points of Y are deleted beginning with ¢; where @ (t;-1,ti+1) < 2 (0,7)| [ ZdF — Do uyZ@ = Z(t) (F (ti) — F (t-1))+Z (ti) (F (tit) — F (Gi) —Z (t-1) (F (G41) — F (G-1))This equals Z (t;,1) F (t;) +Z (ti) F (tin1) — Z(G) F (G) —Z (G1) F (tit)Z (ti) (F (+1) — F (ti) +Z (G-1) F (i) — F (tit)(Z (ti) —Z (t-1)) (F (+1) — F (t))It follows that/ zar— | zarP| <|(Z(tt) —Z(t-r))|F (tent) —F (0)2 ANI}5 1/p+1/q< 0 (t;-1,t1)'!4 0 (titist)/? < @(t1,tig1) PT < (200.7)rIn particular, this yieldsP\ti}[, (Y —¥ (0)) dF — (YY (0) AF] < (Soon)PDP ,5 1/p+1/q_ (; (\IF I? on+l¥llon))