A NOTE ON RIEMANN INTEGRABILITY

In this note we define Riemann integrabillty for real valued functions defined on a compact metric space accompanied by a finite Borel measure. If the measure of each open ball equals the measure of its corresponding closed ball, then a bounded function is Riemann integrable if and only if its set of points of discontinuity has measure zero. Let denote the algebra of sets generated by the open and closed subintervals of an interval [a,b]. A bounded real valued function f defined on [a,b] is Riemann integrable if for each positive , there exist two functions and that are linear combinations of characteristic functions of sets in { satisfying <. f <_ and fba , dmsba dm < where m denotes ordinary Lebesgue measure. Riemann integrability may be defined in an analagous way for real valued functions defined on a compact metric space K accompanied by a finite Borel measure. If we make a simple


Let
denote the algebra of sets generated by the open and closed subintervals of an interval [a,b].A bounded real valued function f defined on [a,b] is Riemann integrable if for each positive , there exist two functions and that are linear combinations of characteristic functions of sets in { satisfying <. f <_ and fb a , dm-sba dm < where m denotes ordinary Lebesgue measure.Riemann integrability may be defined in an analagous way for real valued functions defined on a compact metric space K accompanied by a finite Borel measure.If we make a simple G.A. BEER assumption about the balls of K, then the following famous theorem of Lebesgue extends: a bounded real valued function f defined on [a,b] is Riemann integrable if and only if the set of points at which f is not continuous has Lebesgue measure zero.
Suppose that K is a compact metric space and is a finite Bore1 measure on K.
Let Br(X {y-d(x,y) < r} and B--r(X ) {y" d(x,y) r} denote the open and closed balls of radius r about a point x in K.
Let denote the algebra generated by all such balls.Any element of is of the form where Aik is a ball or its complement and {m,n l,...,nm are positive integers.A step function is a linear combination of characteristic functions determined by elements of x.Hence a step function has the form is real and A. 6 Sinceis an algebra the diXA.where each di 1 1 {A i} can be taken to be pairwise disjoint.It is easy to see that if # and are step functions, then so are + , , inf {,}, and sup {,}.DEFINITION.A bounded real valued function f defined on K is Riemann integrable if for each positive there exist step functions and $ such that _< f <_ and d d < e.
Given a bounded real valued function f defined on K, the upper envelope h of f is the function defined by h(x) infs>oSUpy 6 Bs{x) f(Y) x K and the lower envelope g of f is defined by g(x) sup>0infy C B(x) f(Y) x C K It is well known that h is upper semicontinuous, g is lower semi- continuous, g(x) <_ f(x) <_ h(x) for each x, and (x) h(x) if and only if f is continuous at x (see Royden [1,p.49]).
THEOREM.Suppose (B r(x)) (B r(x)) for each x in K and for each positive r.A bounded real valued function f defined on K is Riemann integrable if and only if the set of points at which f is discontinuous has v-measure zero.
Proof.Let h be the upper envelope of f and g its lower envelope.
Let be any step function that exceeds f.Since each member of d can be expressed in the form depicted in (1), the condition on the balls of K implies that each member of { is the union of an open set and a set of u-measure zero.It follows that can be represented as n E a XA" where (i) A.3 is an open set for 1 S j <_ m (ii) u(Aj) 0 for m < j <_ n (iii) {AI,A 2 A n} partition K.
m Let x C [3 A.. Since is constant near x, there exists j=l J such that (x) > supy C B6(x) f(Y) so that (x) >. h(x).Hence, v{x" (x) < h(x)} O, and we have f dv > f h du.We now construct a decreasing sequence of step functions converging pointwise to h so that G.A. BEER inf { d" >_ f and is a step function} I h d.
Let N be a fixed positive integer.Let {Brl(Xl)'''" 'Brm (Xm) } be a cover of K by balls of radius at most I/N such that if y 6 Br. (x i), then h(y) < h(xi) + I/N.Now let 8N'K / R be the step function described by eN(X inf {h(xi) + I/N" x C Br.(Xi)}.Define @N to be 1 as above and let @N+p 0 N. Given any positive integer p, define eN+p be inf {eN+p,N+p_I}.Clearly, for each p N+p is a step function, > > h To establish the pointwise convergence suppose to and N+p N+p+l the contrary that for some x 0 in K and > 0 we have for each p N+p(XO) > h(xo) + 2e Pick n so large that 1/n < e.There exists a point x n such that d(xo,Xn) < 1/n and ,n(Xo) <_ h(Xn) + 1/n.Clearly, h(Xn) > h(xo) + e which violates the upper semicontinuity of ,h.Hence, {n } is the desired sequence.
Using the above technique we can show in the same manner that g d sup { d" f and is a step function}.The proof is now completed by observing the equivalence of the following statements; (i) f is Riemann integrable (ii) g d I h d (iii) f is continuous except at a set of points of -measure zero.
A simple example shows that the theorem need not hold if our condition on the balls of the metric space is omitted.Let K be the closed unit disc in the plane with the usual metric.If B is a Borel subset of K, 2 y2 define (B) to be I(B {(x,y)" x + i}) + 2 2 2 {Bf {(x,y)" x + y