Skip to content

Mathematics · Ch 11 — Definite Integration

Definite Integral as Limit of a Sum

11.1

Definite Integral as Limit of a Sum

A definite integral of a function f(x)f(x) on an interval [a,b][a,b] is written ∫abf(x) dx\int_a^b f(x)\,dx, read as "the integral from aa to bb of f(x)f(x) with respect to xx." Here a<ba<b are real numbers and ff is defined (and, for this introduction, continuous and non-negative) on [a,b][a,b].

Geometric meaning. ∫abf(x) dx\int_a^b f(x)\,dx is defined as the area of the plane region bounded above by the curve y=f(x)y=f(x), below by the X-axis, and on the sides by the vertical lines x=ax=a and x=bx=b. If g(x)g(x) is any primitive (antiderivative) of f(x)f(x) — that is, g′(x)=f(x)g'(x)=f(x) — then this area turns out to equal g(b)−g(a)g(b)-g(a). The rest of this section explains why that is true, by building the area up from thin strips and taking a limit.

Partitioning the interval. Divide [a,b][a,b] into nn equal sub-intervals using the points

a=x0<x1<x2<⋯<xn−1<xn=b.a=x_0<x_1<x_2<\dots<x_{n-1}<x_n=b.

Each sub-interval [xr,xr+1][x_r,x_{r+1}] determines a thin vertical strip of the region under the curve. The width of every strip is the same, h=b−anh=\dfrac{b-a}{n}, so that xr=a+rhx_r=a+rh and, since xn=bx_n=b, we always have nh=b−anh=b-a.

Approximating a strip by a rectangle. Figure 4.1 shows one such strip. Its true area (bounded above by the curved arc of y=f(x)y=f(x)) is approximated by the area of a rectangle with the same base [xr,xr+1][x_r,x_{r+1}] and height f(tr)f(t_r), where trt_r is some point strictly between xrx_r and xr+1x_{r+1} chosen by the Mean Value Theorem (explained next). This rectangle's area is (xr+1−xr)⋅f(tr)(x_{r+1}-x_r)\cdot f(t_r).

Figure 4.1Fig. 4.1 — the definite integral as a limit of a sum: the area under y = f(x) from x = a (point A) to x = b (point B) is sliced into thin vertical strips, and one strip is approximated by the rectangle M_r M_{r+1} Q P of base [x_r, x_{r+1}] and height f(t_r).
Fig. 4.1 — Fig. 4.1 — the definite integral as a limit of a sum: the area under y = f(x) from x = a (point A) to x = b (point B) is sliced into thin vertical strips, and one strip is approximated by the rectangle M_r M_{r+1} Q P of base [x_r, x_{r+1}] and height f(t_r).

Drawn by us to help you understand the concept clearly, and verified to make sure it's accurate. For exams, practice from your textbook's own diagram.

What this figure shows. Shows the curve y = f(x) drawn over the interval [a, b] on the X-axis, with the region between the curve, the X-axis, and the ordinates x = a and x = b divided by the partition points x0 = a, x1, x2, ..., xn = b into n thin vertical strips. One representative strip is marked with corner points labelled Mr, Mr+1, Q, P, showing how its curved top is approximated by the flat top of a rectangle of width (xr+1 - xr) and height f(tr) for some tr between xr and xr+1, which is the building block used to derive the definite integral as a limit of a Riemann-type sum.

4.1: Area under y = f(x) approximated by strips.

Bringing in the Mean Value Theorem. If g(x)g(x) is a primitive of f(x)f(x), the Mean Value Theorem applied to gg on [xr,xr+1][x_r,x_{r+1}] says there exists trt_r with xr<tr<xr+1x_r<t_r<x_{r+1} such that

g(xr+1)−g(xr)=(xr+1−xr)⋅g′(tr)=(xr+1−xr)⋅f(tr),g(x_{r+1})-g(x_r) = (x_{r+1}-x_r)\cdot g'(t_r) = (x_{r+1}-x_r)\cdot f(t_r),

because g′=fg'=f. So the exact increase in gg across the strip equals exactly the approximate rectangle area built from f(tr)f(t_r) — the approximation is not a coincidence, it is forced by the Mean Value Theorem.

Summing all strips. Adding the rectangle areas for r=0,1,…,n−1r=0,1,\dots,n-1 gives an approximating sum SnS_n for the whole region:

Sn=∑r=0n−1(xr+1−xr)⋅f(tr)=∑r=0n−1[g(xr+1)−g(xr)].S_n=\sum_{r=0}^{n-1}(x_{r+1}-x_r)\cdot f(t_r)=\sum_{r=0}^{n-1}\big[g(x_{r+1})-g(x_r)\big].

The right-hand sum telescopes — every interior g(xr)g(x_r) cancels once as a subtrahend and once as a minuend — leaving only the first and last terms:

Sn=g(xn)−g(x0)=g(b)−g(a).S_n = g(x_n)-g(x_0) = g(b)-g(a).

This is true for every nn, so in particular

g(b)−g(a)=lim⁡n→∞Sn=lim⁡n→∞∑r=0n−1(xr+1−xr)f(tr).g(b)-g(a) = \lim_{n\to\infty} S_n = \lim_{n\to\infty}\sum_{r=0}^{n-1}(x_{r+1}-x_r)f(t_r).

This limit — the limit, as the strips get infinitely thin, of the sum of thin-rectangle areas — is defined to be the definite integral:

∫abf(x) dx=lim⁡n→∞Sn=g(b)−g(a).\int_a^b f(x)\,dx = \lim_{n\to\infty} S_n = g(b)-g(a).

The limit-of-sum working formula. In practice trt_r is taken to be the right end-point a+rha+rh of each strip (a legitimate choice since the Mean Value Theorem only guarantees some point in the strip, and for a continuous ff the limit is the same regardless of which point in each strip is used), so the formula actually used to evaluate an integral from first principles is

∫abf(x) dx=lim⁡n→∞∑r=1nh⋅f(a+rh),h=b−an.\int_a^b f(x)\,dx=\lim_{n\to\infty}\sum_{r=1}^{n}h\cdot f(a+rh), \qquad h=\frac{b-a}{n}.

The word "integrate" literally means "to find the sum of" — reflecting exactly this construction. This technique underlies not just areas but also arc lengths, volumes of revolution, and similar "sum of infinitely many infinitesimal pieces" constructions met later.

Worked Example 1 — ∫12(2x+5) dx\int_1^2(2x+5)\,dx. Here f(x)=2x+5f(x)=2x+5, a=1, b=2a=1,\ b=2, so h=1nh=\frac{1}{n} and nh=1nh=1. Then f(a+rh)=f(1+rh)=2(1+rh)+5=7+2rhf(a+rh)=f(1+rh)=2(1+rh)+5=7+2rh. The sum becomes

∑r=1nh(7+2rh)=7h∑r=1n1+2h2∑r=1nr=7hn+2h2⋅n(n+1)2=7(hn)+(hn)2(1+1n).\sum_{r=1}^n h(7+2rh)=7h\sum_{r=1}^n 1+2h^2\sum_{r=1}^n r = 7hn+2h^2\cdot\frac{n(n+1)}{2}=7(hn)+ (hn)^2\Big(1+\frac1n\Big).

As n→∞n\to\infty, hn=1hn=1 stays fixed and 1/n→01/n\to0, so the limit is 7(1)+(1)2(1+0)=87(1)+(1)^2(1+0)=8. So ∫12(2x+5) dx=8\int_1^2(2x+5)\,dx=8.

Worked Example 2 — ∫237x dx\int_2^3 7^x\,dx. Here f(x)=7xf(x)=7^x, a=2,b=3a=2,b=3, h=1/nh=1/n. Then f(a+rh)=72+rh=72⋅7rhf(a+rh)=7^{2+rh}=7^2\cdot 7^{rh}, and the sum ∑r=1n7rh\sum_{r=1}^n 7^{rh} is a finite geometric series with common ratio 7h7^h and nn terms, summing to 7h(7h)n−17h−1=7h(7nh−1)7h−17^h\dfrac{(7^h)^n-1}{7^h-1}=\dfrac{7^h(7^{nh}-1)}{7^h-1}. Multiplying by h⋅72h\cdot7^2 and using nh=1nh=1, the sum is 49⋅7h(7−1)7h−1⋅h=49⋅6⋅h7h−1⋅7h49\cdot\dfrac{7^h(7-1)}{7^h-1}\cdot h= 49\cdot 6\cdot\dfrac{h}{7^h-1}\cdot 7^h. As h→0h\to0, 7h→17^h\to1 and 7h−1h→log⁡7\dfrac{7^h-1}{h}\to\log 7 (the defining limit of the derivative of 7x7^x at 00), so h7h−1→1log⁡7\dfrac{h}{7^h-1}\to\dfrac1{\log7}. The limit is therefore 49⋅6log⁡7=294log⁡7\dfrac{49\cdot6}{\log7}=\dfrac{294}{\log 7}.

Worked Example 3 — ∫04(x−x2) dx\int_0^4 (x-x^2)\,dx. Here f(x)=x−x2f(x)=x-x^2, a=0,b=4a=0,b=4, h=4/nh=4/n, and f(rh)=rh−r2h2f(rh)=rh-r^2h^2. Splitting the sum,

∑r=1nh(rh−r2h2)=h2∑r−h3∑r2=h2n(n+1)2−h3n(n+1)(2n+1)6.\sum_{r=1}^n h(rh-r^2h^2)=h^2\sum r - h^3\sum r^2 = h^2\frac{n(n+1)}2 - h^3\frac{n(n+1)(2n+1)}6.

Writing each in terms of the fixed quantity nh=4nh=4: h2n(n+1)=(nh)2(1+1/n)→16h^2n(n+1)=(nh)^2(1+1/n)\to 16, and h3n(n+1)(2n+1)=(nh)3(1+1/n)(2+1/n)→64⋅2=128h^3n(n+1)(2n+1)=(nh)^3(1+1/n)(2+1/n)\to 64\cdot2=128. So the limit is 162−1286=8−643=−403\tfrac{16}{2}-\tfrac{128}{6}=8-\tfrac{64}{3}=-\tfrac{40}{3}.

Worked Example 4 — ∫0π/2sin⁡x dx\int_0^{\pi/2}\sin x\,dx. Here a=0, b=π/2, h=π/2na=0,\ b=\pi/2,\ h=\dfrac{\pi/2}{n}, so nh=π/2nh=\pi/2, and f(rh)=sin⁡(rh)f(rh)=\sin(rh). The sum ∑r=1nsin⁡(rh)\sum_{r=1}^n \sin(rh) is a standard trigonometric sum; multiplying it by 2sin⁡h22\sin\frac h2 turns every term into a telescoping difference of cosines via the identity 2sin⁡Asin⁡B=cos⁡(A−B)−cos⁡(A+B)2\sin A\sin B=\cos(A-B)-\cos(A+B), and collapses to

2sin⁡h2∑r=1nsin⁡(rh)=cos⁡h2−cos⁡(nh+h2)=cos⁡h2−cos⁡(π2+h2)=cos⁡h2+sin⁡h2,2\sin\tfrac h2\sum_{r=1}^n\sin(rh)=\cos\tfrac h2-\cos\Big(nh+\tfrac h2\Big)=\cos\tfrac h2-\cos\Big(\tfrac\pi2+\tfrac h2\Big)=\cos\tfrac h2+\sin\tfrac h2,

using nh=π/2nh=\pi/2 and cos⁡(π/2+θ)=−sin⁡θ\cos(\pi/2+\theta)=-\sin\theta. So ∑r=1nsin⁡(rh)=cos⁡h2+sin⁡h22sin⁡h2\displaystyle\sum_{r=1}^n\sin(rh)=\frac{\cos\frac h2+\sin\frac h2}{2\sin\frac h2}, and the original sum is

h∑r=1nsin⁡(rh)=(cos⁡h2+sin⁡h2)⋅h2sin⁡h2=(cos⁡h2+sin⁡h2)⋅h2sin⁡h2.h\sum_{r=1}^n\sin(rh)=\Big(\cos\tfrac h2+\sin\tfrac h2\Big)\cdot\frac{h}{2\sin\frac h2}=\Big(\cos\tfrac h2+\sin\tfrac h2\Big)\cdot\frac{\frac h2}{\sin\frac h2}.

As n→∞n\to\infty, h→0h\to0, so cos⁡h2→1\cos\frac h2\to1, sin⁡h2→0\sin\frac h2\to0, and h/2sin⁡(h/2)→1\dfrac{h/2}{\sin(h/2)}\to1 (the standard θ/sin⁡θ→1\theta/\sin\theta\to1 limit). The whole expression →(1+0)⋅1=1\to (1+0)\cdot1=1. Hence ∫0π/2sin⁡x dx=1\int_0^{\pi/2}\sin x\,dx=1.