Definite integrals

This is about the basics of evaluating definite integrals, and surprisingly how they can be used to define functions.

Contents

Evaluation

Suppose F(x)F(x) is an antiderivative of f(x)f(x), in this course the definite integral is evaluated using

x1x2f(x)dx=F(x2)F(x1)\int_{x_1}^{x_2} f(x)\d x = F(x_2) - F(x_1)

It can be equivalently notated as

x1x2f(x)dx=F(x)x1x2\int_{x_1}^{x_2} f(x)\d x = F(x)\Big\vert_{x_1}^{x_2}

And in cases to be absolutely clear, this can also be written as

x1x2f(x)dx=F(x)x=x1x=x2\int_{x_1}^{x_2} f(x)\d x = F(x)\Big\vert_{x = x_1}^{x = x_2}

x1x_1 and x2x_2 are the limits of integration.

Connecting back to indefinite integrals, we customarily set the +C+C to 00, but it does not matter as by taking the difference, it cancels.

Example: Evaluate

e211xdx\int_{- \e^2}^{-1}\frac 1x \d x

e211xdx=lnxe21=ln1lne2=ln1lne2=02=2\begin{align*}\int_{- \e^2}^{-1}\frac 1x \d x &= \ln\lvert x\rvert \Big\vert_{- \e^2}^{-1} \\ &= \ln\lvert -1\rvert - \ln\lvert - \e^2\rvert \\ &= \ln 1 - \ln \e^2 \\ &= 0 - 2 = -2 \qed\end{align*}

Where differentiation is about instantaneous change, definite integration is about the accumulation of change.

Definite integrals are defined as the signed area between a curve and the xx-axis. In this course, it is sufficient to use antiderivatives to find the area under a continuous function.

Properties

These properties assume the function is defined for all xx within the limits of integration, or is discontinuous (HL) at a finite number of points.

x1x2f(x)dx=x2x1f(x)dx\int_{x_1}^{x_2} f(x)\d x = -\int_{x_2}^{x_1} f(x)\d x
x1x2f(x)dx+x2x3f(x)dx=x1x3f(x)dx\int_{x_1}^{x_2} f(x)\d x + \int_{x_2}^{x_3} f(x)\d x = \int_{x_1}^{x_3} f(x)\d x

The properties from indefinite integrals also carry over.

x1x2f(x)+g(x)dx=x1x2f(x)dx+x1x2g(x)dx\int_{x_1}^{x_2} f(x) + g(x)\d x = \int_{x_1}^{x_2} f(x) \d x + \int_{x_1}^{x_2} g(x) \d x
x1x2kf(x)dx=kx1x2f(x)dx\int_{x_1}^{x_2} k f(x)\d x = k\int_{x_1}^{x_2} f(x)\d x

With F(x)F(x) being an antiderivative of f(x)f(x),

x1x2f(ax+b)dx=1aax1+bax2+bf(x)dx=1aF(ax2+b)1aF(ax1+b)\int_{x_1}^{x_2} f(ax+b)\d x = \frac 1a\int_{ax_1+b}^{ax_2+b} f(x)\d x = \frac 1a F(ax_2+b) - \frac 1a F(ax_1+b)

The last property comes directly from substituting in x1x_1 and x2x_2 into the chosen antiderivative. This as a uu-substitution.

Example: Find

01(4x+2)3dx\int_0^1 (4x+2)^3 \d x

Here we need to apply both integration of xnx^n and f(ax+b)f(ax + b).

The antiderivative of x3x^3 is 14x4+C\frac{1}{4}x^4 + C

Then the f(4x+2)f(4x + 2) means we need to divide by another 44.

01(4x+2)3dx=116(4x+2)401=124(6424)=341=80\begin{align*} \int_0^1 (4x+2)^3 \d x &= \frac {1}{16}(4x+2)^4 \Big\vert_0^1 \\ &= \frac{1}{2^4}(6^4 - 2^4) \\ &= 3^4 - 1 \\ &= 80 \qed \end{align*}

Alternatively we could have factored out a 232^3 early on

01(4x+2)3dx=801(2x+1)3dx=8124[(2x+1)4]01=811=80\begin{align*} \int_0^1 (4x+2)^3 \d x &= 8\int_0^1 (2x+1)^3 \d x \\ &= 8\cdot\frac{1}{2\cdot 4}\left[(2x+1)^4\right]_0^1 \\ &= 81 - 1 \\ &= 80 \qed \end{align*}

which in retrospect is easier, but not by a lot.

Variable of integration

A key difference between indefinite and definite integration is that indefinite integrals return families of functions, but definite integration return a value. Just like how in f(x)f(x), xx just means the input, the xx in definite integrals is very much a placeholder in the same way.

As such all of the following are equivalent

35f(x)dx\int_{-3}^{5} f(x)\d x
35f(w)dw\int_{-3}^{5} f(w)\d w
35f(τ)dτ\int_{-3}^{5} f(\tau)\d \tau

Here ww, xx, and τ\tau are called dummy variables or variables of integration.

Note to HL candidates: for uu-substitution, typically the limits of integration change. Here, they do not change because the substitutions are w=xw = x and τ=x\tau = x

Using definite integral to define a function

There is a way to specify the antiderivative function F(x)F(x) by expressing it as a definite integral.

F(x)=x1xf(τ)dτF(x) = \int_{x_1}^{x} f(\tau)\d \tau

On a calculator, the solver should be used; graphing takes too long for definite integrals.

Example: Simplify

f(x)=0xcos2θdθf(x) = \int_{0}^{x} \cos^2\theta\d \theta

Use double angle identity for cos2θ\cos 2\theta involving cos2θ\cos^2\theta.

cos2θ=2cos2θ112+12cos2θ=cos2θ\begin{align*}\cos 2\theta &= 2\cos^2\theta - 1 \\ \frac12 + \frac12\cos 2\theta &= \cos^2\theta \end{align*}

We find the antiderivative, using the +C=0+C = 0 trick

f(x)=0x12+12cos2θdθ=[12θ+14sin2θ]θ=0θ=x=12x+14sin2x(12(0)+14sin2(0))=12x+14sin2x\begin{align*}f(x) &= \int_0^x \frac12 + \frac12\cos 2\theta \d\theta \\ &= \left[\frac12\theta + \frac14\sin2\theta \right]_{\theta=0}^{\theta=x} \\ &= \frac12x + \frac14\sin2x - \left(\frac12(0) + \frac14\sin2(0)\right) \\ &= \frac12x + \frac14\sin2x \qed\end{align*}

In this particular question the second part entirely cancels. This should not assumed to be the case.

See an example on TI-84 Plus.

Derivation: z-scores of normal distribution

Derivation: Let

f(x)=2π0xet2dtf(x) = \frac{2}{\sqrt\pi} \int_0^x \e^{-t^2} \d t

and given the normal distribution probability density function

g(x)=1σ2πe(xμ)22σ2g(x) = \frac{1}{\sigma\sqrt{2\pi}} \e^{-\frac{(x-\mu)^2}{2\sigma^2}}

Show that

μμ+zσg(x)dx=12f(z2) for all zR\int_{\mu}^{\mu+z\sigma} g(x) \d x = \frac12f\left(\frac{z}{\sqrt2}\right) \text{ for all } z\in\mathbb R

I=1σ2πμμ+zσe(xμ)22σ2dx=1σ2πμμ+zσe(xμσ2)2dx=σ2σ2πμμσ2μ+zσμσ2et2dt=1π0z2et2dt=12(f(z2)f(0))=12f(z2)\begin{align*} I &= \frac{1}{\sigma\sqrt{2\pi}} \int_{\mu}^{\mu+z\sigma} \e^{-\frac{(x-\mu)^2}{2\sigma^2}} \d x \\ &= \frac{1}{\sigma\sqrt{2\pi}} \int_{\mu}^{\mu+z\sigma} \e^{-\left(\frac{x-\mu}{\sigma\sqrt2}\right)^2} \d x \\ &= \frac{\sigma\sqrt2}{\sigma\sqrt{2\pi}} \int_{\frac{\mu - \mu}{\sigma\sqrt2}}^{\frac{\mu + z\sigma - \mu}{\sigma\sqrt2}} \e^{-t^2} \d t \\ &= \frac{1}{\sqrt\pi} \int_{0}^{\frac{z}{\sqrt2}} \e^{-t^2} \d t \\ &= \frac12 \left(f\left(\frac{z}{\sqrt2}\right) - f(0) \right) \\ &= \frac12f\left(\frac{z}{\sqrt2}\right) \qed \end{align*}

ff is called the “error function”.

We made use of

x1x2f(ax+b)dx=1aax1+bax2+bf(x)dx\int_{x_1}^{x_2} f(ax+b)\d x = \frac 1a\int_{ax_1+b}^{ax_2+b} f(x)\d x

and

00f(x)dx=0\int_0^0 f(x)\d x = 0

With x=μ+zσx = \mu + z\sigma, we obtain

z=xμσz = \frac{x - \mu}{\sigma}

In other words, probability from mean to zz standard deviations more than the mean, depends only on zz.

Notes on trig

Calculus of trigonometric functions are only available in radians. Ensure that your calculator is in radians mode.