Calculus: Multivariable Functions

The world is not one-dimensional. To model the temperature across a room, the profit of a company based on multiple inputs, or the gravitational pull from multiple celestial bodies, you need mathematics that can handle more than one variable at a time. Multivariable calculus extends the powerful tools of single-variable calculus—limits, derivatives, and integrals—to functions with two, three, or more independent variables. Mastering this subject unlocks your ability to analyze and optimize complex systems in physics, engineering, economics, and data science, providing the mathematical language for describing multidimensional change and accumulation.

Functions of Several Variables and Their Visualization

A multivariable function is a rule that assigns a single output value to an ordered pair or triple of input values. We denote a function of two variables as $z=f(x,y)$, where $x$ and $y$ are the independent variables and $z$ is the dependent variable. For three variables, we write $w=f(x,y,z)$.

Visualizing these functions is crucial. The primary method is the graph, which for $z=f(x,y)$ is a surface in three-dimensional space. Each point $(x,y,z)$ on the surface satisfies $z=f(x,y)$. Another essential visualization tool is the level curve (or contour line). For a given constant $c$, the level curve is the set of points $(x,y)$ in the plane such that $f(x,y)=c$. Imagine slicing the surface with a horizontal plane at height $c$; the intersection projected down is the level curve. For functions of three variables, we have level surfaces, where $f(x,y,z)=c$ defines a surface in 3D space, like the concentric spheres representing levels of a gravitational potential field.

Partial Derivatives and The Gradient Vector

For a function $z=f(x,y)$, how does $f$ change if we nudge $x$ while holding $y$ constant? This is captured by the partial derivative with respect to $x$, denoted $\frac{\partial f}{\partial x}$ or $f_x$. You compute it by treating all other variables as constants and differentiating with respect to $x$ using single-variable rules. Similarly, $\frac{\partial f}{\partial y}$ measures the rate of change in the $y$-direction.

Example: For $f(x,y)=x^{2}y+\sin (y)$, we have: $$\frac{\partial f}{\partial x}=2xy\text{(treat y as a constant)},$$ $$\frac{\partial f}{\partial y}=x^2+\cos (y)\text{(treat x as a constant)}.$$

The fundamental geometric interpretation is that $f_x(a,b)$ gives the slope of the tangent line to the curve formed by intersecting the surface $z=f(x,y)$ with the vertical plane $y=b$, at the point $(a,b,f(a,b))$.

The crowning achievement of first-order derivatives is the gradient vector, denoted $\nabla f$ (read "nabla f"). For a function $f(x,y)$, it is the vector of all its first partial derivatives: $$\nabla f(x,y)=⟨\frac{\partial f}{\partial x},\frac{\partial f}{\partial y}⟩.$$ For $f(x,y,z)$, it is $\nabla f=⟨f_x,f_y,f_z⟩$. The gradient is not just a list of derivatives; it points in the direction of the steepest ascent of the function at a point, and its magnitude $|\nabla f|$ is the rate of increase in that direction. Conversely, $-\nabla f$ points in the direction of steepest descent.

Directional Derivatives and The Multivariable Chain Rule

The partial derivative tells you the rate of change along the coordinate axes. What if you want the rate of change in an arbitrary direction? This is the directional derivative. Given a unit vector $\mathbf{u}=⟨u_1,u_2⟩$, the directional derivative of $f$ at $(a,b)$ in the direction of $\mathbf{u}$ is denoted $D_{\mathbf{u}}f(a,b)$. It can be computed elegantly using the gradient: $$D_{\mathbf{u}}f(a,b)=\nabla f(a,b)\cdot \mathbf{u}.$$ This dot product projects the gradient onto the desired direction, giving the instantaneous rate of change. If $\mathbf{u}$ is not a unit vector, you must normalize it first.

When a multivariable function's inputs are themselves functions of another variable (e.g., $f(x(t),y(t))$), the multivariable chain rule governs how $f$ changes with $t$. For this case: $$\frac{df}{dt}=\frac{\partial f}{\partial x}\frac{dx}{dt}+\frac{\partial f}{\partial y}\frac{dy}{dt}.$$ Think of it as summing the contributions of all pathways from $t$ to $f$. For more complex dependencies, like $f(x(u,v),y(u,v))$, the rule extends naturally: $$\frac{\partial f}{\partial u}=\frac{\partial f}{\partial x}\frac{\partial x}{\partial u}+\frac{\partial f}{\partial y}\frac{\partial y}{\partial u},$$ and similarly for $\frac{\partial f}{\partial v}$.

Optimization: Critical Points and Lagrange Multipliers

Finding maximum and minimum values is a central application. A point $(a,b)$ is a critical point of $f$ if $\nabla f(a,b)=0$ (i.e., $f_x(a,b)=0$ and $f_y(a,b)=0$) or if a partial derivative does not exist. Critical points include local maxima, local minima, and saddle points (which look like a mountain pass). To classify them, you use the Second Derivative Test. Compute the discriminant $D$ at the critical point: $$D=f_{xx}(a,b)f_{yy}(a,b)-{\left[f_{xy}(a,b)\right]}^2.$$

• If $D>0$ and $f_{xx}>0$, then $(a,b)$ is a local minimum.
• If $D>0$ and $f_{xx}<0$, then $(a,b)$ is a local maximum.
• If $D<0$, then $(a,b)$ is a saddle point.
• If $D=0$, the test is inconclusive.

Often, we need to optimize a function subject to a constraint, like maximizing volume given a fixed surface area. The method of Lagrange multipliers solves these problems. To find the extreme values of $f(x,y)$ subject to $g(x,y)=k$, you solve the system of equations: $$\nabla f(x,y)=\lambda \nabla g(x,y),g(x,y)=k.$$ Here, $\lambda$ is the Lagrange multiplier. Geometrically, this states that at an optimum, the level curve of $f$ is tangent to the constraint curve $g=k$, meaning their gradient vectors are parallel.

Multiple Integrals: Double and Triple Integrals

Just as single integrals compute area under a curve, double integrals compute volume under a surface. The double integral of $f(x,y)$ over a region $R$ in the $xy$-plane is denoted: $${\iint }_{R}f(x,y)dA.$$ You evaluate it as an iterated integral. For a rectangular region $R=[a,b]\times [c,d]$: $${\iint }_{R}f(x,y)dA={\int }_{x=a}^b{\int }_{y=c}^{d}f(x,y)dydx.$$ The order of integration ($dydx$ vs. $dxdy$) can be swapped, but you must adjust the limits of integration to describe the region $R$ correctly. For non-rectangular regions, the limits are functions of the outer variable.

Double integrals also calculate area (${\iint }_{R}1dA$), mass, center of mass, and other physical quantities over two-dimensional regions.

Triple integrals extend this concept to three dimensions. The triple integral of $f(x,y,z)$ over a solid region $E$ is: $${\iiint }_{E}f(x,y,z)dV.$$ It is evaluated as a three-fold iterated integral. Applications include computing the volume of a solid (${\iiint }_{E}1dV$), total mass given a density function, and average value of a function over a 3D region. Switching between rectangular, cylindrical ($r$, $\theta$, $z$), and spherical ($\rho$, $\varphi$, $\theta$) coordinates is often essential to simplify the integration, especially for regions with circular or spherical symmetry.

Common Pitfalls

1. Treating Partial Derivatives as Ordinary Derivatives: When computing $\frac{\partial f}{\partial x}$, you must treat all other variables as constants. A common error is to implicitly differentiate them as if they were functions of $x$. For $f(x,y)=x^{2}y$, $\frac{\partial f}{\partial x}=2xy$, not $2xy+x^2\frac{dy}{dx}$.
2. Misapplying the Chain Rule: For $z=f(x(t),y(t))$, the derivative $\frac{dz}{dt}$ is a sum of terms, not a product. Remember the structure: (partial of $f$ wrt $x$) * (derivative of $x$ wrt $t$) + (partial of $f$ wrt $y$) * (derivative of $y$ wrt $t$).
3. Confusing Gradient and Directional Derivative: The gradient is a vector pointing in the direction of steepest ascent. The directional derivative is a scalar representing the rate of change in a given direction. They are related by $D_{\mathbf{u}}f=\nabla f\cdot \mathbf{u}$.
4. Incorrect Limits in Iterated Integrals: The limits of integration for the inner integral(s) can depend on the outer variable(s). Always sketch the region of integration $R$ or the solid $E$ to correctly determine these functional limits. Switching integration order without adjusting limits is a frequent source of error.

Summary

• Multivariable functions like $f(x,y)$ model phenomena that depend on more than one factor. They are visualized through graphs, level curves, and level surfaces.
• Partial derivatives ($f_x$, $f_y$) measure the rate of change along coordinate axes. The gradient vector $\nabla f$ combines these into a single vector pointing in the direction of steepest increase.
• The directional derivative $D_{\mathbf{u}}f$ generalizes the rate of change to any direction and is computed via a dot product with the gradient. The multivariable chain rule systematically handles differentiation when variables are themselves functions of other variables.
• To find local extrema, locate critical points where the gradient is zero and use the Second Derivative Test to classify them. For optimization under constraints, the method of Lagrange multipliers is indispensable.
• Double and triple integrals calculate volume, mass, and other quantities over 2D and 3D regions. Mastery involves setting up correct iterated integrals, often utilizing coordinate transformations like polar, cylindrical, or spherical coordinates for simplification.