Stationary points

We are often interested in the location of the highest and lowest points on the graph of a function ff (see for example the section about optimisation), or other special points.

Definition 1

Consider the graph below. We call point AA a local minimum (or a high point), and point CC a local maximum (or a low point). We also call AA or BB a local extremum (or turning point). The plural forms are local maxima, local minima, and local extrema. Note that AA and BB are the highest and lowest points only in a local environment, that is, AA forms the top of a hill, and BB is at the bottom of a valley. Or to be more formal, CC is a local maximum because if we move away from this point (to the left or right) we go downwards. And AA is a local minimum because moving away from this point we go upwards.

Points on the graph where the tangent is horizontal are called stationary points. Local maxima and local minima are clearly stationary points. Their is also a third type of stationary point, called a saddle point, in the figure it is point BB. Moving to the left from such a point, we go downwards on the graph, moving to the right we go upwards (or vice versa).

Let's summarise:

Theorem 1

Consider a function ff, and a point P(xy)P(x|y) on the graph of ff. A point PP on the graph with xx-coordinate xx is a stationary point, if the tangent at PP is horizotnal, that is, we have

Equation 1
stationary point at xf(x)=0\text{stationary point at $x$}\leftrightarrow f'(x)=0

Stationary point at point PP with xx-coordinate xx

A stationary point is either a local extremum (local maximum or minimum), or a saddle point.

Example 1

Find the stationary points of the function f(x)=x32x2f(x)=x^3-2x^2 by calculation. Then draw the graph in order to classify the stationary points as local maximum, local minimum or saddle point.

Solution

To find the xx-coordinate of the stationary points, determine all values of xx with

f(x)=0f'(x)=0

Because f(x)=3x24xf'(x)=3x^2-4x, we have to find all values of xx with

3x24x=03x^2-4x=0

Use the midnight formula to solve the equation or alternatively, observe that I can write

x(3x4)=0x(3x-4)=0

from which follows that x1=0x_1=0 and x2=43x_2=\frac{4}{3}. With f(0)=0f(0)=0 and f(43)=3227f(\frac{4}{3})=-\frac{32}{27}, we get the stationary points P1(00)\underline{P_1(0|0)} and P2(433227)\underline{P_2(\frac{4}{3} | -\frac{32}{27})}. From the drawing (not shown) follows that P1P_1 is a local maximum and P2P_2 is a local minimum.

Can we somehow determine the type of stationary point (maximum, minimum, saddle point)? This might be convenient if we cannot draw the graph, or if there are so many stationary points that it is hard to identify and classify all of them (e.g. plot the graph of the function f(x)=sin(1/x)f(x)=\sin(1/x)).

Fortunately, there are some simple criteria to classify stationary points.

Theorem 2

Assume that P(xy)P(x|y) is a point on the graph of the function ff. Then we have the following is true:

Equation 2
f(x)=0 andf(x)<0local max at xf(x)>0local min at xf(x)=0 and f(x)0saddle pt at xf(x)=0 and f(x)=0draw f to decide\begin{array}{lll} f^{\prime}(x)=0 \text{ and}&&\\ \quad\dots\, f^{\prime\prime}(x)<0 & \rightarrow & \text{local max at $x$}\\ \quad\dots\, f^{\prime\prime}(x)>0 & \rightarrow & \text{local min at $x$}\\ \quad\dots\, f^{\prime\prime}(x)=0 \text{ and } f^{\prime\prime\prime}(x)\neq 0 & \rightarrow & \text{saddle pt at $x$} \\ \quad\dots\, f^{\prime\prime}(x)=0 \text{ and } f^{\prime\prime\prime}(x)= 0 & \rightarrow & \text{draw $f$ to decide}\end{array}

Classifying stationary points.

Proof

In order to see this, let's first have a closer look at the "hills" and "valleys" of a graph ff, and how their derivatives look like (see below). You can see that in the case of a valley, the graph of ff' has a positive slope at the deepest point xx, that is, f(x)>0f^{\prime\prime}(x)>0, and in the case of a hill, the graph of ff' has a negative slope at the highest point xx, that is, f(x)<0f^{\prime\prime}(x)<0.

Looking at a saddle and inspecting the derivative, we see that ff' has a horizontal tangent at the saddle point xx, and thus f(x)=0f^{\prime\prime}(x)=0, but clearly f(x)0f^{\prime\prime\prime}(x)\neq 0.

Finally, it f(x)=f(x)=0f^{\prime\prime}(x)=f^{\prime\prime\prime}(x)=0 if could either be a flat hill, flat valley, or flat saddle point (see figure below). In this case we have to draw the graph to decide if PP is a local maximum, local minimum, or saddle point.

Exercise 1
  1. Determine by calculation the stationary points, and classify them using the criteria given above. Draw the graph with the calculator to verify your results.
    1. f(x)=x2f(x)=x^2
    2. g(x)=x3g(x)=x^3
    3. h(x)=5x4h(x)=5-x^4
    4. k(x)=x412x2k(x)=x^4-\frac{1}{2}x^2
  2. Consider the function f(x)=x3ux2f(x)=x^3-u\cdot x^2, where uu is a fixed, but unknown number (a so called parameter).
    1. Show that ff has a stationary point at x=0x=0 for all values of uu.

    2. For what values of uu does ff have a local maximum, a local minimum, or a saddle point at x=0x=0?

Solution
  1. It is

    1. Find stationary point PP: f(x)=2x=0x=0f^\prime(x)=2x=0 \rightarrow x=0, thus P(00)\underline{P(0|0)}. Type? Because f(x)=2f(0)=2>0f^{\prime\prime}(x)=2 \rightarrow f^{\prime\prime}(0)=2>0 \rightarrow local minimum.
    2. Find stationary point PP: g(x)=3x2=0x=0g^\prime(x)=3x^2=0 \rightarrow x=0, thus P(00)\underline{P(0|0)}.Because g(x)=6xg(0)=0g^{\prime\prime}(x)=6x \rightarrow g^{\prime\prime}(0)=0 and g(x)=60g^{\prime\prime\prime}(x)=6\neq 0 it is a saddle point.
    3. Find stationary point PP: h(x)=4x3=0x=0h^\prime(x)=-4x^3=0 \rightarrow x=\underline{0}, thus P(05)\underline{P(0|5)}.Type? Because h(x)=12x2h(0)=0h^{\prime\prime}(x)=-12x^2 \rightarrow h^{\prime\prime}(0)=0 and h(x)=24xh(0)=0h^{\prime\prime\prime}(x)=-24x \rightarrow h^{\prime\prime\prime}(0)=0. Draw graph \rightarrow "flat" local maximum.
    4. Find stationary point PP: k(x)=4x3x=0x(4x21)=0x1=0,x2=0.5,x3=0.5k^\prime(x)=4x^3-x=0 \rightarrow x(4x^2-1)=0 \rightarrow x_1=\underline{0}, x_2=-0.5, x_3=0.5, thus P1(00),P2(0.50.0625),P3(0.50.0625)\underline{P_1(0|0)}, \underline{P_2(-0.5|-0.0625)}, \underline{P_3(0.5|-0.0625)}.Type? Because k(x)=12x21k(0)=1<0k^{\prime\prime}(x)=12x^2-1 \rightarrow k^{\prime\prime}(0)=-1<0 \rightarrow local maximum, k(0.5)=2>0k^{\prime\prime}(-0.5)=2>0 \rightarrow local minimum, k(0.5)=2>0k^{\prime\prime}(0.5)=2>0 \rightarrow local minimum.
  2. f(x)=3x22uxf^\prime(x)=3x^2-2ux, f(x)=6x2uf^{\prime\prime}(x)=6x-2u, f(x)=6f^{\prime\prime\prime}(x)=6

    1. For every fixed value of uu we have f(0)=3022u0=0x=0f^\prime(0)=3\cdot 0^2-2u\cdot 0=0 \rightarrow x=0 \rightarrow is stationary point for all values uu.
    2. f(0)=602u=2u<0f^{\prime\prime}(0)=6\cdot 0-2u=-2u<0 \rightarrow at x=0x=0 there is a local maximum for u>0\underline{u>0}, and a local minimum for u<0\underline{u<0}. For u=0\underline{u=0} it is f(0)=0f^{\prime\prime}(0)=0 and f(0)=60f^{\prime\prime\prime}(0)=6\neq 0, so at x=0x=0 there is a saddle point.