48

BRUCE K. DRIVER

4. The Riemann Integral In this short chapter, the Riemann integral for Banach space valued functions is dened and developed. Our exposition will be brief, since the Lebesgue integral and the Bochner Lebesgue integral will subsume the content of this chapter. The following simple Bounded Linear Transformation theorem will often be used here and in the sequel to dene linear transformations. Theorem 4.1 (B. L. T. Theorem). Suppose that Z is a normed space, X is a Banach space, and S Z is a dense linear subspace of Z. If T : S X is a bounded linear transformation (i.e. there exists C < such that kT z k C kz k L(Z, X ) and this for all z S ), then T has a unique extension to an element T extension still satises z C kz k for all z S . T Exercise 4.1. Prove Theorem 4.1. For the remainder of the chapter, let [a, b] be a xed compact interval and X be a Banach space. The collection S = S ([a, b], X ) of step functions, f : [a, b] X, consists of those functions f which may be written in the form (4.1) f (t) = x0 1[a,t1 ] (t) +
n 1 X i=1

xi 1(ti ,ti+1 ] (t),

where {a = t0 < t1 < < tn = b} is a partition of [a, b] and xi X. For f as in Eq. (4.1), let (4.2) I (f )
n 1 X i=0

(ti+1 ti )xi X.

Exercise 4.2. Show that I (f ) is well dened, independent of how f is represented as a step function. (Hint: show that adding a point to a partition of [a, b] does not change the right side of Eq. (4.2).) Also verify that I : S X is a linear operator. Proposition 4.2 (Riemann Integral). The linear function I : S X extends from S (the closure of the step functions uniquely to a continuous linear operator I inside of ([a, b], X )) to X and this operator satises, (f )k (b a) kf k for all f S . (4.3) kI

Furthermore, C ([a, b], X ) S as (4.4)

(f ) may be computed ([a, b], X ) and for f , I

n 1 X i=0

(f ) = lim I

| |0

f (c i )(ti+1 ti )

where {a = t0 < t1 < < tn = b} denotes a partition of [a, b], | | = max {|ti+1 ti | : i = 0, . . . , n 1} is the mesh size of and c i may be chosen arbitrarily inside [ti , ti+1 ]. Proof. Taking the norm of Eq. (4.2) and using the triangle inequality shows, (4.5) kI (f )k
n 1 X i=0 n 1 X i=0

(ti+1 ti )kxi k

(ti+1 ti )kf k (b a)kf k .

ANALYSIS TOOLS W ITH APPLICATIONS

49

satisfying Eq. (4.3) is a consequence of Theorem 4.1. The existence of I For f C ([a, b], X ), {a = t0 < t1 < < tn = b} a partition of [a, b], and c i [ti , ti+1 ] for i = 0, 1, 2 . . . , n 1, let f (t) f (c0 )0 1[t0 ,t1 ] (t) + Pn1
n 1 X i=1

f (c i )1(ti ,ti+1 ] (t).

Then I (f ) = i=0 f (c i )(ti+1 ti ) so to nish the proof of Eq. (4.4) and that , it suces to observe that lim||0 kf f k = 0 because f is C ([a, b], X ) S uniformly continuous on [a, b]. such that limn kf fn k = 0, then for a < b, If fn S and f S then 1[, ] fn S and limn 1[, ] f 1[, ] fn = 0. This shows 1[, ] f S . whenever f S and a b we will write denote I (1[, ] f ) by Notation 4.3. For f S R R f (t) dt or [, ] f (t)dt. Also following the usual convention, if a b, we will let Z Z (1[,] f ) = f (t) dt = I f (t) dt.

The next Lemma, whose proof is left to the reader (Exercise 4.4) contains some of the many familiar properties of the Riemann integral. ([a, b], X ) and , , [a, b], the Riemann integral satises: Lemma 4.4. For f S R (1) f (t) dt ( ) sup {kf (t)k : t } . R R R (2) f (t) dt = f (t) dt + f (t) dt. Rt (3) The function G(t) := a f ( )d is continuous on [a, b]. ([a, b], Y ) and (4) If Y is another Banach space and T L(X, Y ), then T f S ! Z Z f (t)dt = T f (t)dt. T

([a, b], R) and f g, then (6) If f, g S Z Z b f (t)dt

a

([a, b], R) and (5) The function t kf (t)kX is in S Z Z b b f (t) dt kf (t)k dt. a a
b

g (t)dt.

(4.6)

Theorem 4.5 (Baby Fubini Theorem). Let a, b, c, d R and f (s, t) X be a continuous function of (s, t) for s between a and b and t between c and d. Then the Rb Rd maps t a f (s, t)ds X and s c f (s, t)dt are continuous and # # Z "Z Z "Z
d b b d

f (s, t)ds dt =

f (s, t)dt ds.

50

BRUCE K. DRIVER

Proof. With out loss of generality we may assume a < b and c < d. By uniform continuity of f, Exercise 3.15,
ctd

sup kf (s, t) f (s0 , t)k 0 as s s0

d c

and so by Lemma 4.4 Z f (s, t)dt Rd

c

f (s0 , t)dt as s s0

showing the continuity of s similarly. Now let

f (s, t)dt. The other continuity assertion is proved

= {a s0 < s1 < < sm = b} and 0 = {c t0 < t1 < < tn = d} be partitions of [a, b] and [c, d] respectively. For s [a, b] let s = si if s (si , si+1 ] and i 1 and s = s0 = a if s [s0 , s1 ]. Dene t0 for t [c, d] analogously. Then # # Z b "Z d Z b Z b "Z d 0 f (s, t)dt ds = f (s, t )dt ds + 0 (s)ds
a c a

= where
0

"Z

d c

f (s , t0 )dt ds + ,0 + Z

b 0 (s)ds

(s) = Z
b

and ,0 =

d c

c d

f (s, t)dt

f (s, t0 )dt #

"Z

{f (s, t0 ) f (s , t0 )} dt ds.

The uniform continuity of f and the estimates Z d kf (s, t) f (s, t0 )k dt sup k 0 (s)k sup
s[a,b] s[a,b] c

(d c) sup {kf (s, t) f (s, t0 )k : (s, t) Q}

d

and k,0 k Z
b

"Z

kf (s, t0 ) f (s , t0 )k dt ds

(b a)(d c) sup {kf (s, t) f (s, t0 )k : (s, t) Q} allow us to conclude that # Z Z b "Z d f (s, t)dt ds
a c b a

"Z

f (s , t0 )dt ds 0 as | | + | 0 | 0.

By symmetry (or an analogous argument), # # Z d "Z b Z d "Z b f (s, t)ds dt f (s , t0 )ds dt 0 as | | + | 0 | 0.

c a c a

ANALYSIS TOOLS W ITH APPLICATIONS

51

This completes the proof since # Z b "Z d f (s , t0 )dt ds =

a c

0i<m,0j<n

"Z

X
a

f (si , tj )(si+1 si )(tj +1 tj ) #

f (s , t0 )ds dt.

4.1. The Fundamental Theorem of Calculus. Our next goal is to show that our Riemann integral interacts well with dierentiation, namely the fundamental theorem of calculus holds. Before doing this we will need a couple of basic denitions and results. Denition 4.6. Let (a, b) R. A function f : (a, b) X is dierentiable at )f (t) t (a, b) i L := limh0 f (t+hh exists in X. The limit L, if it exists, will be df denoted by f(t) or dt (t). We also say that f C 1 ((a, b) X ) if f is dierentiable at all points t (a, b) and f C ((a, b) X ). Proposition 4.7. Suppose that f : [a, b] X is a continuous function such that f(t) exists and is equal to zero for t (a, b). Then f is constant. Proof. Let > 0 and (a, b) be given. (We will later let 0 and a.) By the denition of the derivative, for all (a, b) there exists > 0 such that (4.7) kf (t) f ( )k = f (t) f ( ) f( )(t ) |t | if |t | < . Let (4.8) A = {t [, b] : kf (t) f ()k (t )} and t0 be the least upper bound for A. Eq. (4.7) with = shows t0 > and a simple continuity argument shows t0 A, i.e. (4.9) kf (t0 ) f ()k (t0 ) For the sake of contradiction, suppose that t0 < b. By Eqs. (4.7) and (4.9), kf (t) f ()k kf (t) f (t0 )k + kf (t0 ) f ()k (t0 ) + (t t0 ) = (t ) for 0 t t0 < t0 which violates the denition of t0 being an upper bound. Thus we have shown Eq. (4.8) holds for all t [, b]. Since > 0 and > a were arbitrary we may conclude, using the continuity of f, that kf (t) f (a)k = 0 for all t [a, b]. Remark 4.8. The usual real variable proof of Proposition 4.7 makes use Rolles theorem which in turn uses the extreme value theorem. This latter theorem is not available to vector valued functions. However with the aid of the Hahn Banach Theorem 18.16 and Lemma 4.4, it is possible to reduce the proof of Proposition 4.7 and the proof of the Fundamental Theorem of Calculus 4.9 to the real valued case, see Exercise 18.12. Theorem 4.9 (Fundamental Theorem of Calculus). Suppose that f C ([a, b], X ), Then Rt d (1) dt f ( ) d = f (t) for all t (a, b). a

52

BRUCE K. DRIVER

(2) Now assume that F C ([a, b], X ), F is continuously dierentiable on (a, b), extends to a continuous function on [a, b] which is still denoted by and F . Then F Z b (t) dt = F (b) F (a). F
a

Proof. Let h > 0 be a small number and consider Z t+h Z t Z t+h k f ( )d f ( )d f (t)hk = k (f ( ) f (t)) d k
a a

h (h), where (h) max [t,t+h] k(f ( ) f (t))k. Combining this with a similar computation when h < 0 shows, for all h R suciently small, that Z t Z t+h f ( )d f ( )d f (t)hk |h| (h), k
a a

t t+h

k(f ( ) f (t))k d

where now (h) max [t|h|,t+|h|] k(f ( ) f (t))k. By continuity of f at t, (h) 0 Rt d and hence dt f ( ) d exists and is equal to f (t). a Rt ( ) d F (t). Then G is continuous by For the second item, set G(t) a F (t) = 0 for all t (a, b) by item 1. An application of Proposition Lemma 4.4 and G Rb ( ) d F (b) = 4.7 shows G is a constant and in particular G(b) = G(a), i.e. a F F (a). Corollary 4.10 (Mean Value Inequality). Suppose that f : [a, b] X is a continuous function such that f(t) exists for t (a, b) and f extends to a continuous function on [a, b]. Then (4.10) kf (b) f (a)k Z
b

Rb Proof. By the fundamental theorem of calculus, f (b) f (a) = a f(t)dt and then by Lemma 4.4, Z Z Z b b b kf(t)kdt f (t)dt kf (b) f (a)k = f dt = (b a) f . a a a Proposition 4.11 (Equality of Mixed Partial Derivatives). Let Q = (a, b) (c, d) f (s, t), s f (s, t) and be an open rectangle in R2 and f C (Q, X ). Assume that t t s f (s, t) exists and are continuous for (s, t) Q, then s t f (s, t) exists for (s, t) Q and (4.11) f (s, t) = f (s, t) for (s, t) Q. s t t s

kf(t)kdt (b a) f

ANALYSIS TOOLS W ITH APPLICATIONS

53

Proof. Fix (s0 , t0 ) Q. By two applications of Theorem 4.9, Z s f (s, t) = f (st0 , t) + f (, t)d s0 Z s Z s Z t d d = f (s0 , t) + (4.12) f (, t0 )d + f (, ) s0 s0 t0 and then by Fubinis Theorem 4.5 we learn Z s Z t Z s d d f (s, t) = f (s0 , t) + f (, t0 )d + f (, ). s0 t0 s0 Dierentiating this equation in t and then in s (again using two more applications of Theorem 4.9) shows Eq. (4.11) holds. 4.2. Exercises. Exercise 4.3. Let ([a, b], X ) {f : [a, b] X : kf k supt[a,b] kf (t)k < }. Show that ( ([a, b], X ), k k ) is a complete Banach space. Exercise 4.4. Prove Lemma 4.4. ([a, b], Rn ) i fi Exercise 4.5. Using Lemma 4.4, show f = (f1 , . . . , fn ) S ([a, b], R) for i = 1, 2, . . . , n and S Z ! Z b Z b b f (t)dt = f1 (t)dt, . . . , fn (t)dt .
a a a

Exercise 4.6. Give another proof of Proposition 4.11 which does not use Fubinis Theorem 4.5 as follows. (1) By a simple translation argument we may assume (0, 0) Q and we are trying to prove Eq. (4.11) holds at (s, t) = (0, 0). (2) Let h(s, t) := t s f (s, t) and Z s Z t G(s, t) := d d h(, )
0 0

so that Eq. (4.12) states

f (s, t) = f (0, t) +

and dierentiating this equation at t = 0 shows (4.13) f (s, 0) = f (0, 0) + G(s, 0). t t t Now show using the denition of the derivative that Z s (4.14) dh(, 0). G(s, 0) = t 0 Hint: Consider Z G(s, t) t
s

s 0

f (, t0 )d + G(s, t)

dh(, 0) =

(3) Now dierentiate Eq. (4.13) in s using Theorem 4.9 to nish the proof.

t 0

d [h(, ) h(, 0)] .

54

BRUCE K. DRIVER

Exercise 4.7. Give another proof of Eq. (4.6) in Theorem 4.5 based on Proposition 4.11. To do this let t0 (c, d) and s0 (a, b) and dene Z t Z s d df (, ) G(s, t) :=
t0 s0

Show G satises the hypothesis of Proposition 4.11 which combined with two applications of the fundamental theorem of calculus implies G(s, t) = G(s, t) = f (s, t). t s s t Use two more applications of the fundamental theorem of calculus along with the observation that G = 0 if t = t0 or s = s0 to conclude Z s Z t Z s Z t (4.15) G(s, t) = d d d d f (, ). G(, ) = s0 t0 s0 t0

Finally let s = b and t = d in Eq. (4.15) and then let s0 a and t0 c to prove Eq. (4.6).