The Definition of Extrema and the Necessary Conditions

The Definition of an Extremum

Extremum of a multivariable function

Suppose the function z=f(x,y)z = f(x, y) is defined in some neighbourhood of the point P0(x0,y0)P_0(x_0, y_0). If for every point P(x,y)P(x, y) in that neighbourhood other than P0P_0 we have

f(x,y)<f(x0,y0)(or f(x,y)>f(x0,y0))f(x, y) < f(x_0, y_0) \quad \text{(or } f(x, y) > f(x_0, y_0) \text{)}

then the function is said to attain a local maximum (or a local minimum) f(x0,y0)f(x_0, y_0) at P0P_0, and the point P0P_0 is called a local maximum point (or a local minimum point).

Local maxima and local minima are collectively called extrema, and the corresponding points are collectively called extremum points.

Extrema of Functions of Three or More Variables

Similarly, a function of nn variables u=f(x1,x2,…,xn)u = f(x_1, x_2, \ldots, x_n) attains an extremum at the point P0P_0 if, in some neighbourhood of P0P_0, f(P0)f(P_0) is a maximum (local maximum) or a minimum (local minimum).

The Necessary Condition for an Extremum

Theorem1

Fermat’s theorem (the necessary condition for an extremum): suppose the function z=f(x,y)z = f(x, y) has partial derivatives at the point (x0,y0)(x_0, y_0) and attains an extremum there. Then

fx(x0,y0)=0,fy(x0,y0)=0f_x(x_0, y_0) = 0, \quad f_y(x_0, y_0) = 0

For a function of nn variables, all the first-order partial derivatives vanish at an extremum point:

fxi(P0)=0(i=1,2,…,n)f_{x_i}(P_0) = 0 \quad (i = 1, 2, \ldots, n)

Critical Points and Stationary Points

Critical point (stationary point)

A point (x,y)(x, y) at which the first-order partial derivatives of the function f(x,y)f(x, y) vanish simultaneously, that is, which satisfies

fx(x,y)=0,fy(x,y)=0f_x(x, y) = 0, \quad f_y(x, y) = 0

is called a stationary point (or critical point) of the function.

More generally, a point at which the gradient ∇f=0⃗\nabla f = \vec{0} is called a stationary point.

Saddle Points

Saddle point

A stationary point that is not an extremum point is called a saddle point. At a saddle point the function increases in some directions and decreases in others, and the shape is reminiscent of a saddle.

A typical example: f(x,y)=x2−y2f(x,y) = x^2 - y^2 at (0,0)(0,0). Here fx=2x=0f_x = 2x = 0 and fy=−2y=0f_y = -2y = 0, so it is a stationary point. But along the xx-axis the function increases (f(x,0)=x2f(x,0)=x^2), while along the yy-axis it decreases (f(0,y)=−y2f(0,y)=-y^2), so it is not an extremum point but a saddle point.

Worked Examples

Example 1: Finding the Stationary Points

Find the stationary points of the function f(x,y)=x3+y3−3xyf(x, y) = x^3 + y^3 - 3xy.

stationary pointscritical pointsthe partial derivatives vanish

Approach:

A stationary point is where the two partial derivatives vanish simultaneously, so you simply solve the system of equations.

Steps:

  1. The partial derivatives: fx=3x2−3yf_x = 3x^2 - 3y, fy=3y2−3xf_y = 3y^2 - 3x
  2. Setting them to zero: {3x2−3y=03y2−3x=0\begin{cases} 3x^2 - 3y = 0 \\ 3y^2 - 3x = 0 \end{cases}, that is, {y=x2x=y2\begin{cases} y = x^2 \\ x = y^2 \end{cases}
  3. Substituting: x=(x2)2=x4x = (x^2)^2 = x^4, that is, x4−x=0x^4 - x = 0 and x(x3−1)=0x(x^3-1) = 0
  4. Solving gives x=0x = 0 or x=1x = 1
  5. When x=0x=0 we have y=0y=0; when x=1x=1 we have y=1y=1
  6. The stationary points are (0,0)(0, 0) and (1,1)(1, 1)

Answer: the stationary points are (0,0)(0,0) and (1,1)(1,1)

Example 2: An Extremum Point Where the Partial Derivatives Do Not Exist

Discuss whether f(x,y)=x2+y2f(x, y) = \sqrt{x^2 + y^2} attains an extremum at (0,0)(0,0).

an extremumthe partial derivatives do not exista local minimum

Approach:

An extremum point need not be a stationary point — the points where the partial derivatives do not exist have to be examined separately, and that is exactly the point of this problem.

Steps:

  1. At (0,0)(0,0): f(0,0)=0f(0,0) = 0
  2. For every (x,y)≠(0,0)(x,y) \neq (0,0): f(x,y)=x2+y2>0=f(0,0)f(x,y) = \sqrt{x^2+y^2} > 0 = f(0,0)
  3. Hence (0,0)(0,0) is a local minimum point, with local minimum value 00.
  4. But the partial derivatives: fx(0,0)=lim⁡Δx→0∣Δx∣Δxf_x(0,0) = \lim_{\Delta x\to0} \frac{|\Delta x|}{\Delta x} does not exist, and similarly fy(0,0)f_y(0,0) does not exist.
  5. So (0,0)(0,0) is not a stationary point but an extremum point where the partial derivatives do not exist.

Answer: (0,0)(0,0) is a local minimum point (with local minimum value 0), but the partial derivatives do not exist there, so it is not a stationary point.

Example 3: Recognising a Saddle Point

Prove that f(x,y)=xyf(x, y) = xy has a saddle point at (0,0)(0,0).

a saddle pointa stationary pointnot an extremum point

Approach:

At a stationary point, if the value of the function increases in some directions and decreases in others, the point is a saddle point.

Steps:

  1. fx=yf_x = y and fy=xf_y = x, and at (0,0)(0,0) we have fx=fy=0f_x = f_y = 0, so (0,0)(0,0) is a stationary point.
  2. Along the direction y=xy = x: f(x,x)=x2>0f(x,x) = x^2 > 0 (for x≠0x \neq 0), so the function increases.
  3. Along the direction y=−xy = -x: f(x,−x)=−x2<0f(x,-x) = -x^2 < 0 (for x≠0x \neq 0), so the function decreases.
  4. Hence every neighbourhood of (0,0)(0,0) contains points where the function value is greater than 0 and points where it is less than 0, while f(0,0)=0f(0,0)=0.
  5. So (0,0)(0,0) is not an extremum point but a saddle point.

Answer: (0,0)(0,0) is a saddle point.


Exercises

Exercise 1

Find the stationary points of the function f(x,y)=x2−xy+y2+3x−2y+1f(x, y) = x^2 - xy + y^2 + 3x - 2y + 1.

stationary pointscritical points

Approach:

The usual business of solving a system of equations for the stationary points.

Steps:

  1. fx=2x−y+3f_x = 2x - y + 3 and fy=−x+2y−2f_y = -x + 2y - 2
  2. Setting them to zero: {2x−y+3=0−x+2y−2=0\begin{cases} 2x - y + 3 = 0 \\ -x + 2y - 2 = 0 \end{cases}
  3. Solving gives x=−43x = -\frac{4}{3} and y=13y = \frac{1}{3}

Answer: the stationary point is (−43,13)(-\frac{4}{3}, \frac{1}{3})

Exercise 2

Find the stationary points of the function f(x,y)=e−(x2+y2)f(x, y) = e^{-(x^2+y^2)} and decide whether they can be extremum points.

stationary pointsan extremumthe exponential function

Approach:

The outer exponential function is positive and monotone, so the stationary point corresponds directly to a local maximum.

Steps:

  1. fx=−2xe−(x2+y2)f_x = -2x e^{-(x^2+y^2)} and fy=−2ye−(x2+y2)f_y = -2y e^{-(x^2+y^2)}
  2. Setting fx=fy=0f_x = f_y = 0, and since e−(x2+y2)>0e^{-(x^2+y^2)} > 0, we get x=0,y=0x = 0, y = 0
  3. The stationary point is (0,0)(0,0) and f(0,0)=1f(0,0) = 1
  4. For every (x,y)≠(0,0)(x,y) \neq (0,0): f(x,y)=e−(x2+y2)<1=f(0,0)f(x,y) = e^{-(x^2+y^2)} < 1 = f(0,0)
  5. Hence (0,0)(0,0) is a local maximum point, with local maximum value 11.

Answer: the stationary point is (0,0)(0,0), and it is a local maximum point (with local maximum value 1).


Summary

Symbols Used in This Article

SymbolTypePronunciation/ExplanationMeaning in This Article
P0P_0Pointpoint zeroAn extremum point
∇f\nabla fGradientgradientThe gradient of the function
0⃗\vec{0}Zero vectorzero vectorThe zero vector
fx,fyf_x, f_yPartial derivativespartial derivativesThe first-order partial derivatives

Found an issue with this content?

Tell us what could be improved.