Setting up the Riemann sum. Let f(x) be a real-valued, bounded function on the closed interval [a,b], a<b. Unlike in a first geometric picture, f(x) need not keep the same sign throughout [a,b] — it may take both positive and negative values.
Partition [a,b] into n subintervals [x0,x1],[x1,x2],…,[xn−1,xn] with
a=x0<x1<x2<⋯<xn−1<xn=b.
In each subinterval [xi−1,xi], i=1,…,n, choose an arbitrary point ξi with xi−1≤ξi≤xi, and form the sum
∑i=1nf(ξi)(xi−xi−1)=f(ξ1)(x1−x0)+f(ξ2)(x2−x1)+⋯+f(ξn)(xn−xn−1).…(1)
This is called a Riemann sum of f(x) for the given partition. Because ξi can be any point of [xi−1,xi], there are infinitely many different Riemann sums for the same partition.
Definition of the Riemann (definite) integral. If, as n→∞ with max(xi−xi−1)→0 (the widest subinterval shrinking to zero), the sum in (1) tends to a single finite value A — regardless of how the partition and the points ξi were chosen — then A is called the definite integral of f with respect to x on [a,b], also called the Riemann integral, denoted
∫abf(x)dx,
read "the integral of f(x) with respect to x from a to b". When a=b, ∫aaf(x)dx=0.
This chapter works with f(x) continuous on [a,b]; the Riemann integral also exists more generally for bounded, piecewise-continuous f. The integration variable is a dummy variable: ∫abf(x)dx=∫abf(u)du — it can be renamed freely.
Three standard choices of the evaluation point ξi (all yield the same limit for continuous f, but give different finite-n approximations):
- Left-end rule (ξi=xi−1): ∫abf(x)dx=n→∞limi=1∑nf(xi−1)(xi−xi−1).
- Right-end rule (ξi=xi): ∫abf(x)dx=n→∞limi=1∑nf(xi)(xi−xi−1).
- Mid-point rule (ξi=21(xi−1+xi)): ∫abf(x)dx=n→∞limi=1∑nf(2xi−1+xi)(xi−xi−1).
Exercise 9.1 uses each rule without taking the limit — i.e. computes just one finite Riemann sum S=∑f(ξi)Δx from a given 5-point partition of [1,1.5] — which is only an approximate value of the true integral (worked in the text as Example 9.1, estimating ∫00.5x2dx via all three rules on 5 equal subintervals of width h=0.1).
Remarks connecting sign and area (used throughout §9.8):
- If ∫abf(x)dx exists, then F(x)=∫axf(u)du is a well-defined function on [a,b].
- If f(x)≥0 throughout [a,b], ∫abf(x)dx equals the geometric area between y=f(x), the x-axis, and x=a,x=b. …