MULTIVARIABLE CALCULUS • PARTIAL DERIVATIVES

Tangent Planes & Linear Approximation — Tangent planes and linear approximation

Learn how flat planes can approximate curved surfaces, extending the tangent-line idea into three dimensions.

Historical Context & Motivation

Long before GPS navigation or computer graphics, mathematicians grappled with a fundamental question: how do you describe the shape of a curved surface at a single point? In single-variable calculus, you already know how to draw a tangent line to a curve — a straight line that just touches the curve and matches its slope at one specific location. But what happens when your function depends on two variables and traces out a surface in three-dimensional space? A single line is no longer enough; you need an entire flat plane.

The idea of approximating complicated shapes with simpler ones stretches back centuries. Mathematicians realized that zooming in closely enough on any smooth surface makes it look almost flat, just as zooming in on Earth's surface makes the ground beneath your feet seem perfectly level. Formalizing this intuition required the development of partial derivatives and the careful study of functions of multiple variables.

1684
Leibniz Publishes Calculus
Gottfried Wilhelm Leibniz publishes his foundational calculus work, introducing the notation for derivatives and integrals that would later extend to multiple variables.
1734
Euler's Partial Derivatives
Leonhard Euler develops the concept of partial differentiation, enabling mathematicians to measure how a function changes when only one of its input variables is varied while the others are held constant.
1795
Monge & Differential Geometry
Gaspard Monge formalizes the study of surfaces and tangent planes in his work on differential geometry, giving engineers precise tools for describing curved objects.
1823
Cauchy Rigorizes Limits
Augustin-Louis Cauchy provides rigorous definitions of limits and continuity, solidifying the logical foundations for tangent planes and linear approximation.

The central question this lesson addresses is: Given a surface z = f(x, y), how can we build the best flat approximation near a specific point? Answering this question will unlock powerful estimation techniques and deepen your understanding of how derivatives work in higher dimensions.

Core Principles & Definitions

Before we dive into formulas, let's establish the key ideas that make tangent planes and linear approximation work. Each of these principles builds on your knowledge of single-variable calculus and extends it into the world of surfaces.

1

Partial Derivatives

A partial derivative measures the rate of change of f(x, y) with respect to one variable while holding the other constant. fₓ gives the slope in the x-direction; fy gives the slope in the y-direction.
2

Tangent Plane

The tangent plane at a point on a surface is the unique flat plane that touches the surface at that point and has the same slopes (partial derivatives) in every direction.
3

Linearization

The linearization L(x, y) of a function at a point is the equation of the tangent plane, used as a simple approximation of f(x, y) near that point.
4

Differentiability

A function is differentiable at a point if the tangent plane exists and provides a good local approximation. Having both partial derivatives is necessary but not always sufficient.
KEY TAKEAWAY
Think of the tangent plane like a small piece of paper pressed onto the surface of a basketball at one point. Right at the point of contact, the paper lies perfectly flat against the curved surface. If you stay close to that contact point, the paper is a great stand-in for the ball's surface — but wander too far and the curvature makes the paper a poor match. Linear approximation works the same way: it's excellent near the point of tangency and gets worse as you move away.

Visual Explanation

The diagram below shows a curved surface z = f(x, y) in three-dimensional space with a tangent plane touching it at a specific point. Notice how the plane and the surface share the same point and the same slopes in both the x- and y-directions. The two dashed lines on the tangent plane represent the tangent lines in the x-direction and y-direction, whose slopes are given by the partial derivatives fₓ and fy.

A surface z = f(x, y) with its tangent plane at point P. The amber dashed line shows the tangent in the x-direction (slope = fₓ), and the pink dashed line shows the tangent in the y-direction (slope = fy).

At the point P = (a, b, f(a, b)), the tangent plane is the best possible flat approximation to the surface. It captures both the height of the surface and its tilt in every direction. If the surface were a mountain, the tangent plane would be like laying a perfectly flat board on the mountainside at one spot — it tells you exactly how steep the ground is and which way it's sloping.

Mathematical Framework

Now let's translate our geometric intuition into precise formulas. Suppose you have a differentiable function z = f(x, y) and you want to find the tangent plane at the point (a, b). You need three pieces of information: the value of the function at that point, f(a, b), and the two partial derivatives, fₓ(a, b) and fy(a, b).

TANGENT PLANE EQUATION
z = f(a, b) + fₓ(a, b) · (x − a) + f_y(a, b) · (y − b)
Here (a, b) is the point of tangency, f(a, b) is the surface height at that point, fₓ(a, b) is the partial derivative with respect to x evaluated at (a, b), and fy(a, b) is the partial derivative with respect to y evaluated at (a, b).

This equation is a direct extension of the tangent line formula from single-variable calculus, y = f(a) + f'(a)(x − a). Instead of one slope term, there are now two — one for each independent variable. The function linearization is simply the right-hand side of the tangent plane equation, given a special name because of how often it's used for approximation.

LINEARIZATION (LINEAR APPROXIMATION)
L(x, y) = f(a, b) + fₓ(a, b) · (x − a) + f_y(a, b) · (y − b)
For (x, y) near (a, b), f(x, y) ≈ L(x, y). The closer (x, y) is to (a, b), the better the approximation.

There is also a useful related concept: the total differential. While the linearization gives you the approximate value of the function, the total differential tells you the approximate change in the function when x and y each shift by small amounts dx and dy.

TOTAL DIFFERENTIAL
dz = fₓ(a, b) · dx + f_y(a, b) · dy
dx = x − a represents a small change in x, and dy = y − b represents a small change in y. The total differential dz estimates the resulting change in z.
💡 Connection to Single-Variable Calculus
In Calc 1, the tangent line at x = a is y = f(a) + f'(a)(x − a), and the differential is dy = f'(a) · dx. The multivariable versions simply add a second slope term for the second variable. If you set fy = 0 (no dependence on y), you recover the single-variable formulas exactly.

Geometric Meaning & Error Analysis

Understanding the geometry behind linear approximation helps you see why the formula works and when it might fail. The diagram below compares the actual surface value f(x, y) with the tangent-plane approximation L(x, y) at several points near the point of tangency. Notice that the approximation error — the vertical gap between the surface and the plane — grows as you move farther from the base point (a, b).

A cross-section of the surface (purple curve) and the tangent plane (green line) along y = b. The red segments show the approximation error, which increases with distance from the point of tangency.

The approximation error can be quantified. If f has continuous second partial derivatives near (a, b), then the error satisfies |f(x, y) − L(x, y)| ≤ M · (|x − a| + |y − b|)², where M depends on the magnitudes of the second partial derivatives. This tells you two important things: first, the error shrinks quadratically as you approach (a, b), meaning halving your distance roughly quarters the error. Second, large curvature (big second derivatives) leads to worse approximations because the surface bends away from the plane more quickly.

⚠️ When Does the Approximation Fail?
The tangent plane may not exist at points where the function is not differentiable — for example, at sharp corners or cusps on the surface. If the partial derivatives are not continuous near a point, proceed with caution: the tangent plane might technically exist but still provide a poor approximation.

Worked Example

Let's put the formula to work with a concrete example. We'll find the tangent plane and then use linear approximation to estimate a function value.

Find the tangent plane to f(x, y) = x² + 2xy + y² at (1, 2), then approximate f(1.05, 1.98).
1
Step 1 — Evaluate the function at the pointCompute f(1, 2) = (1)² + 2(1)(2) + (2)² = 1 + 4 + 4 = 9.
f(1, 2) = 9
2
Step 2 — Find the partial derivative with respect to xDifferentiate f with respect to x while treating y as a constant: fₓ(x, y) = 2x + 2y. Evaluate at (1, 2): fₓ(1, 2) = 2(1) + 2(2) = 6.
fₓ(1, 2) = 6
3
Step 3 — Find the partial derivative with respect to yDifferentiate f with respect to y while treating x as a constant: fy(x, y) = 2x + 2y. Evaluate at (1, 2): fy(1, 2) = 2(1) + 2(2) = 6.
fy(1, 2) = 6
4
Step 4 — Write the tangent plane equationSubstitute into the formula z = f(a, b) + fₓ(a, b)(x − a) + fy(a, b)(y − b): z = 9 + 6(x − 1) + 6(y − 2). Simplify: z = 9 + 6x − 6 + 6y − 12 = 6x + 6y − 9.
z = 6x + 6y − 9
5
Step 5 — Use the linearization to approximate f(1.05, 1.98)L(1.05, 1.98) = 9 + 6(1.05 − 1) + 6(1.98 − 2) = 9 + 6(0.05) + 6(−0.02) = 9 + 0.30 − 0.12 = 9.18. The exact value is f(1.05, 1.98) = (1.05)² + 2(1.05)(1.98) + (1.98)² = 1.1025 + 4.158 + 3.9204 = 9.1809. The error is only |9.18 − 9.1809| = 0.0009.
L(1.05, 1.98) ≈ 9.18 (error ≈ 0.0009)

This example shows how remarkably accurate the linear approximation can be when (x, y) is close to (a, b). The tangent plane yielded an estimate within about 0.01% of the true value, which is impressive for such a simple formula. In practical applications — engineering estimates, physics calculations, or computer graphics — this kind of quick approximation saves enormous computational effort.

Strengths, Limitations & Comparisons

Linear approximation is a powerful tool, but like any tool it has ideal use cases and situations where it falls short. The table below summarizes the key advantages and disadvantages.

Strengths and limitations of tangent-plane linear approximation
AspectStrengthsLimitations
ComputationOnly requires evaluating f, fₓ, and f_y at one point — fast and simple.Accuracy degrades quickly for points far from (a, b).
ApplicabilityWorks for any differentiable function of two (or more) variables.Cannot be used at points where the function is not differentiable.
Error behaviorError shrinks quadratically as you approach the base point.High curvature (large second derivatives) leads to larger errors.
InsightProvides geometric intuition about local surface behavior.Captures only first-order effects; misses curvature entirely.
KEY TAKEAWAY
Linear approximation is like using a weather forecast that only looks at the current temperature and its rate of change to predict the next hour. For a short window, it's remarkably reliable. But if a cold front is barreling in (high curvature), the forecast breaks down fast. For better accuracy over a larger range, you would need second-order (quadratic) approximation, which accounts for curvature — but that's a topic for later.

Connection to Advanced Theory

The tangent plane and linear approximation are actually just the first step in a broader framework. In more advanced courses, you'll learn about higher-order approximations using Taylor polynomials in two variables. Just as a quadratic Taylor polynomial improves on a tangent line in single-variable calculus, a second-degree Taylor polynomial improves on the tangent plane by adding curvature correction terms involving the second partial derivatives.

Linear vs. quadratic approximation comparison
FeatureTangent Plane (1st Order)Quadratic Approx. (2nd Order)
Derivatives usedf, fₓ, f_y (first partials only)f, fₓ, f_y, fₓₓ, fₓ_y, f_yy (first and second partials)
ShapeFlat planeCurved paraboloid
Error orderO(distance²) — quadratic decayO(distance³) — cubic decay, much faster
ComplexitySimple — just three evaluationsModerate — requires six derivative evaluations

The tangent plane also connects to the concept of the gradient vector, ∇f = ⟨fₓ, fy⟩. The gradient points in the direction of steepest ascent on the surface and is perpendicular to the level curves. In fact, the normal vector to the tangent plane is ⟨fₓ(a, b), fy(a, b), −1⟩, which encodes the same information as the gradient. These ideas show up everywhere in optimization, physics, and machine learning, making the tangent plane one of the most important foundational concepts in multivariable calculus.

Practice Problems

PROBLEM 1CONCEPTUAL
In your own words, explain why a tangent plane requires two partial derivatives, while a tangent line in single-variable calculus requires only one derivative. What geometric role does each partial derivative play?
PROBLEM 2BASIC CALCULATION
Find the equation of the tangent plane to f(x, y) = 3x² − y² + 4 at the point (2, 1).
PROBLEM 3INTERMEDIATE
Let f(x, y) = √(x² + y²). Find the linearization L(x, y) at the point (3, 4) and use it to approximate f(3.1, 3.9).
PROBLEM 4APPLIED
The volume of a cylinder is V = πr²h. A metal can is manufactured with radius r = 5 cm and height h = 12 cm, but there are small measurement errors: dr = 0.03 cm and dh = −0.05 cm. Use the total differential to estimate how much the actual volume differs from the intended volume.
PROBLEM 5CRITICAL THINKING
Consider f(x, y) = sin(xy) at the point (π, 1). Find the tangent plane, then explain why linear approximation will be less accurate here than it was for f(x, y) = x² + 2xy + y² at (1, 2) (from the worked example), even for the same size displacement. Hint: think about the second partial derivatives.

Lesson Summary

The tangent plane to a surface z = f(x, y) at the point (a, b) is a flat plane that touches the surface at that point and matches its slopes in both the x- and y-directions. Its equation is z = f(a, b) + fₓ(a, b)(x − a) + fy(a, b)(y − b), which requires knowing the function value and both partial derivatives at the point. This formula extends the familiar tangent-line equation from single-variable calculus by adding a second slope term for the second independent variable.

The linearization L(x, y) uses the tangent plane equation to approximate f(x, y) near (a, b). The total differential dz = fₓ dx + fy dy estimates the change in the function for small shifts in the inputs. The approximation error shrinks quadratically with distance from the base point, making the method extremely accurate for nearby evaluations. The function must be differentiable at the point for the tangent plane to exist and provide a valid approximation. Looking ahead, second-order Taylor polynomials extend this idea by incorporating curvature, yielding even more accurate approximations over wider regions.

Varsity Tutors • Multivariable Calculus • Tangent Planes & Linear Approximation