Critical Points in Multivariable Calculus

students, imagine trying to find the highest hill or lowest valley on a landscape 🏔️. In single-variable calculus, you look for points where the slope is zero or undefined. In multivariable calculus, the idea is similar, but now a function depends on more than one input, so the graph can rise, fall, or flatten in different directions at once. Critical points are the places where this search begins.

What a Critical Point Means

For a function of two variables, such as $f(x,y)$, a critical point is a point in the domain where the gradient is zero or where the gradient does not exist, as long as the function is defined there. The gradient is the vector of partial derivatives:

$$\nabla f(x,y)=\left\langle \frac{\partial f}{\partial x},\frac{\partial f}{\partial y} \right\rangle$$

So a point $(a,b)$ is critical if either

$$\nabla f(a,b)=\mathbf{0}$$

or the partial derivatives do not exist at $(a,b)$ while $f(a,b)$ still exists.

Why does this matter? Because critical points are the first places to check when looking for local maxima, local minima, or saddle points. They are the multivariable version of the points where $f'(x)=0$ or $f'(x)$ does not exist in single-variable calculus.

A simple example is

$$f(x,y)=x^2+y^2$$

Its gradient is

$$\nabla f(x,y)=\langle 2x,2y\rangle$$

Setting both components equal to zero gives $x=0$ and $y=0$, so $(0,0)$ is a critical point. In fact, it is a local minimum because the function measures squared distance from the origin and is never negative.

How to Find Critical Points

Finding critical points usually follows a clear process:

Compute the partial derivatives $f_x$ and $f_y$.
Set them equal to zero and solve the system

$$f_x(x,y)=0, \quad f_y(x,y)=0$$

Check any points where $f_x$ or $f_y$ fail to exist, but $f$ still exists.
List all candidate critical points.

This process is important because not every critical point is a maximum or minimum. Some are neither. They may be saddle points, which behave like a mountain pass ⛰️—rising in one direction and falling in another.

Example: let

$$f(x,y)=x^2-y^2$$

Then

$$f_x(x,y)=2x, \quad f_y(x,y)=-2y$$

Setting these equal to zero gives $x=0$ and $y=0$, so $(0,0)$ is a critical point. But this point is not a max or min. If $y=0$, then $f(x,0)=x^2\ge 0$, but if $x=0$, then $f(0,y)=-y^2\le 0$. The function goes up in one direction and down in another, which is the signature of a saddle point.

This is a key multivariable idea: unlike in one-variable calculus, $f_x=0$ and $f_y=0$ only tell you that the surface is flat in the coordinate directions. The point still might not be an extremum.

Why Critical Points Are the Starting Point for Optimization

Optimization means finding the best value of a function, such as the largest profit, smallest cost, greatest area, or lowest energy. In many real problems, the best point occurs at a critical point because there is no easy way to increase or decrease the function locally.

For example, suppose a company models profit by

$$P(x,y)$$

where $x$ and $y$ represent two decision variables, like amounts of two products. A local maximum of $P$ may occur where

$$\nabla P(x,y)=\mathbf{0}$$

because at that point, small changes in either variable do not immediately improve profit.

Critical points are also linked to the geometry of a surface. If the surface is perfectly flat in the tangent plane at a point, then the gradient is zero there. The gradient points in the direction of steepest increase, so if it is zero, the surface has no preferred uphill direction at that instant.

However, not every optimization problem is solved by critical points alone. Sometimes the domain has boundaries. Then the absolute maximum or minimum may occur on an edge or corner instead of in the interior. That is why optimization problems often require checking both critical points and boundaries.

Interpreting Different Types of Critical Points

A critical point can be one of three main types:

Local maximum: the function is larger nearby than at the point.
Local minimum: the function is smaller nearby than at the point.
Saddle point: the function is neither a local max nor a local min.

The function

$$f(x,y)=x^2+y^2$$

has a local minimum at $(0,0)$.

The function

$$f(x,y)=-x^2-y^2$$

has a local maximum at $(0,0)$.

The function

$$f(x,y)=x^2-y^2$$

has a saddle point at $(0,0)$.

These examples show that the same condition $\nabla f=\mathbf{0}$ can lead to different behaviors. That is why later topics in optimization use the second derivative test to help classify critical points.

A helpful analogy is a terrain map 🗺️. A local minimum is like the bottom of a bowl, a local maximum is like the top of a hill, and a saddle point is like a mountain pass. At the pass, you can go up in one direction and down in another.

Example: Finding and Classifying Critical Points

Consider

$$f(x,y)=x^2+4y^2-4x-8y$$

First find the partial derivatives:

$$f_x(x,y)=2x-4$$

$$f_y(x,y)=8y-8$$

Set them equal to zero:

$$2x-4=0 \quad \Rightarrow \quad x=2$$

$$8y-8=0 \quad \Rightarrow \quad y=1$$

So $(2,1)$ is a critical point.

To understand what kind of point it is, complete the square:

$$f(x,y)=(x-2)^2+4(y-1)^2-8$$

Since $(x-2)^2\ge 0$ and $4(y-1)^2\ge 0$, the smallest value of $f$ happens when both squares are zero, at $(2,1)$. So $(2,1)$ is a local minimum, and the minimum value is $-8$.

This example shows a practical strategy: critical points give candidate locations, and then algebra or another test helps determine the type of point.

Critical Points in Real Life

Critical points appear in many applied settings. For example:

In business, they help identify a production level that maximizes profit or minimizes cost.
In physics, they can represent equilibrium positions where net force is zero.
In engineering, they help locate stable designs or stress points in a system.
In machine learning, they are related to points where a loss function stops decreasing locally.

Suppose a factory wants to minimize material cost for a box with dimensions controlled by two variables. The cost function may have a critical point where the cheapest design occurs. students, this is why critical points are not just abstract math—they are used to make real decisions 📦.

In science and engineering, a critical point may also suggest stability. If a small change in input causes the output to increase, the point may be unstable. If nearby changes increase the function in every direction, the point is more stable and likely a minimum.

Connection to the Bigger Optimization Topic

Critical points are the foundation of optimization in multivariable calculus. They are the first candidates you examine when searching for extrema inside a region. Later, the second derivative test helps decide whether a critical point is a max, min, or saddle. After that, constrained optimization with Lagrange multipliers handles problems where the answer must satisfy an equation such as

$$g(x,y)=c$$

So the full optimization process often looks like this:

Find critical points using the gradient.
Classify them if possible.
Check boundaries of the domain.
Use constraints when needed.

That means critical points are not the final answer by themselves. They are the gateway to solving the optimization problem.

Conclusion

Critical points are a central idea in multivariable calculus because they mark places where a function is flat in the sense of first partial derivatives. They are the natural starting point for finding local maxima, local minima, and saddle points. By setting the gradient equal to zero and checking where partial derivatives fail to exist, you can find candidate points for optimization. students, once you understand critical points, the rest of optimization becomes much easier to follow because every later method builds on this idea.

Study Notes

A critical point of $f(x,y)$ is a point where $\nabla f(x,y)=\mathbf{0}$ or where the partial derivatives do not exist while $f$ is defined.
The gradient is $\nabla f(x,y)=\left\langle \frac{\partial f}{\partial x},\frac{\partial f}{\partial y} \right\rangle$.
To find critical points, solve $f_x(x,y)=0$ and $f_y(x,y)=0$, then check where partial derivatives fail to exist.
Critical points are candidates for local maxima, local minima, or saddle points.
A saddle point is neither a max nor a min; the function rises in some directions and falls in others.
Critical points are the first step in optimization, but boundary points may also matter.
The second derivative test and constrained optimization methods, such as Lagrange multipliers, build on the idea of critical points.
Examples like $x^2+y^2$, $-x^2-y^2$, and $x^2-y^2$ show the main types of critical point behavior.