Approximating Values of a Function Using Local Linearity and Linearization - AP Calculus AB

SECTION 1

Historical Context & Motivation

Long before electronic calculators and computer algebra systems, mathematicians and scientists needed efficient methods to evaluate complicated functions — square roots, trigonometric ratios, logarithms — by hand. The idea that a smooth curve looks increasingly like a straight line the closer you zoom in is both geometrically intuitive and analytically powerful. This principle, known as local linearity, underpins the technique of linearization: replacing a function with its tangent line near a known point to approximate values that would otherwise be difficult to compute. The development of this idea spans centuries and connects Newton's fluxions to modern numerical analysis.

~250 BCE • Archimedes' Method of Exhaustion

Archimedes approximated areas under curves by inscribing polygons — straight-sided figures — that approached the curve. This foreshadowed the idea of replacing curves with linear segments for practical computation.

1665–1687 • Newton's Calculus of Fluxions

Isaac Newton developed the concept of the derivative and recognized that a differentiable function could be approximated near a point by the first term of its power-series expansion, effectively using the tangent line as a local proxy.

1715 • Taylor's Theorem Published

Brook Taylor generalized Newton's insight into a full polynomial approximation framework. Linearization is the first-order (degree-one) case of a Taylor polynomial centered at a point a.

1797 • Lagrange's Analytic Formulation

Joseph-Louis Lagrange provided rigorous error bounds for polynomial approximations, enabling mathematicians to quantify precisely how accurate a tangent-line estimate could be.

The central question that linearization answers is deceptively simple: if you know the exact value of a function and its derivative at one point, how well can you estimate the function's value at a nearby point? This question matters not only for pencil-and-paper approximation but also for understanding error propagation in physics measurements, for building numerical algorithms, and for justifying the use of differentials throughout applied mathematics.

SECTION 2

Core Principles & Definitions

Linearization rests on a single geometric observation: at any point where a function is differentiable, the graph of the function and the graph of its tangent line are virtually indistinguishable over a sufficiently small interval. This observation converts a potentially complex calculation into simple arithmetic with a linear function. Below are the foundational ideas that make the technique work.

Differentiability Implies Local Linearity

If f is differentiable at x = a, then f can be well-approximated by a linear function near a. The closer x is to a, the better the approximation. This is the geometric meaning of the derivative.

The Linearization Formula

The linearization of f at a is L(x) = f(a) + f ′(a)(x − a). This is simply the equation of the tangent line to f at the point (a, f(a)), written in point-slope form.

Choosing the Base Point a

Select a value of a that is close to the target x-value and at which f(a) and f ′(a) are easy to compute exactly. The quality of the approximation depends on this choice.

Over- and Under-Estimates via Concavity

If f is concave up near a, the tangent line lies below the curve, so L(x) is an underestimate. If f is concave down, L(x) is an overestimate. The second derivative determines which case applies.

✦

KEY TAKEAWAY

Think of linearization like using a ruler against a curved surface. If you press a ruler tangent to a basketball, the ruler perfectly touches the surface at one point and stays extremely close to the surface for a small region around that point. The ruler is the tangent line, and the basketball's surface is the function. The farther you move from the point of tangency, the more the ruler diverges from the curved surface — just as the linearization loses accuracy as x moves away from a.

SECTION 3

Visual Explanation — The Tangent-Line Approximation

The violet-to-cyan curve represents f(x), while the dashed amber line is the tangent line L(x) at the point (a, f(a)), shown in pink. At the nearby value x₀, the tangent-line value L(x₀) closely approximates the true value f(x₀). The vertical gap between the two points represents the approximation error, which shrinks as x₀ approaches a.

The diagram above captures the essence of linearization. At the point of tangency (a, f(a)), the curve and the tangent line share the same value and the same slope. As you move horizontally to a nearby input x₀, the tangent line's output L(x₀) stays close to the curve's true output f(x₀). The key insight is that the error between f(x₀) and L(x₀) is roughly proportional to (x₀ − a)², meaning a small horizontal displacement produces an even smaller vertical error. This is why linearization works so well for points near a but deteriorates rapidly when x₀ is far from a.

SECTION 4

Mathematical Framework

The linearization of f at a is built directly from the definition of the derivative. Since f ′(a) = lim (f(x) − f(a))/(x − a) as x → a, for x near a we can drop the limit and write the approximation f(x) − f(a) ≈ f ′(a)(x − a). Rearranging gives the linearization formula below.

LINEARIZATION FORMULA

L(x) = f(a) + f ′(a) · (x − a)

L(x) is the linearization of f at a. Here, f(a) is the known function value, f ′(a) is the derivative at a, and (x − a) is the displacement from the base point. The approximation f(x) ≈ L(x) is valid when x is close to a.

DIFFERENTIAL FORM

Δy ≈ dy = f ′(a) · dx

The change in y, denoted Δy = f(x) − f(a), is approximated by the differential dy = f ′(a) · dx, where dx = x − a. This is the same linearization expressed in the language of differentials.

ERROR CHARACTERIZATION

f(x) − L(x) ≈ ½ f ″(c) · (x − a)²

By Taylor's theorem, the error equals ½ f ″(c)(x − a)² for some c between a and x. If f ″ > 0 on the interval (concave up), then f(x) > L(x) and the linearization underestimates. If f ″ < 0 (concave down), the linearization overestimates.

💡

AP Exam Note

On the AP Calculus AB exam, you will be expected to use linearization to estimate function values and to determine whether the estimate is an overestimate or underestimate using the sign of the second derivative. Both multiple-choice and free-response questions may require you to justify your reasoning about concavity.

SECTION 5

Concavity, Error, and Choosing the Base Point

Selecting an appropriate base point a and understanding how concavity affects the direction of the error are two skills tested frequently on the AP exam. The base point must satisfy two requirements: f(a) and f ′(a) must be exactly computable (no approximation needed for the base itself), and a must be as close to the target x-value as practical. The diagram below illustrates how the sign of f ″ determines whether the tangent line sits above or below the curve.

Left panel: when f ″ > 0 (concave up), the tangent line sits below the curve, producing an underestimate. Right panel: when f ″ < 0 (concave down), the tangent line sits above the curve, producing an overestimate. This relationship is a direct consequence of the error term ½ f ″(c)(x − a)².

Concavity of f near a	Sign of f ″	Tangent Line Position	L(x) is a …
Concave up	f ″(a) > 0	Below the curve	Underestimate
Concave down	f ″(a) < 0	Above the curve	Overestimate

Concavity determines the direction of approximation error

SECTION 6

Worked Example — Approximating √4.1

Suppose you need to approximate √4.1 without a calculator. The function f(x) = √x is differentiable for x > 0, and the nearest value at which f is easy to evaluate exactly is a = 4, where f(4) = 2. This worked example demonstrates the full linearization procedure, including an over/under determination.

Approximate √4.1 using linearization

1
Step 1 — Identify the function, base point, and target

Let f(x) = √x = $x^{1/2}$ . Choose a = 4 because 4 is the perfect square closest to 4.1. The target value is x = 4.1, so dx = x − a = 0.1.
2
Step 2 — Compute f(a) and f ′(a)

f(4) = √4 = 2. The derivative is f ′(x) = 1/(2√x), so f ′(4) = 1/(2·2) = 1/4 = 0.25.

f(4) = 2, f ′(4) = 0.25
3
Step 3 — Write the linearization L(x)

L(x) = f(a) + f ′(a)(x − a) = 2 + 0.25(x − 4).
4
Step 4 — Evaluate L at the target x = 4.1

L(4.1) = 2 + 0.25(4.1 − 4) = 2 + 0.25(0.1) = 2 + 0.025 = 2.025.

√4.1 ≈ 2.025
5
Step 5 — Determine over- or underestimate

f ″(x) = −1/( $4x^{3/2}$ ). At x = 4, f ″(4) = −1/(4·8) = −1/32 < 0. Since f is concave down near a = 4, the tangent line lies above the curve, and L(4.1) = 2.025 is an overestimate. (The true value is √4.1 ≈ 2.02485…, confirming the tangent-line value is slightly too large.)

Overestimate (f ″ < 0)

SECTION 7

Strengths and Limitations of Linearization

Linearization is among the most versatile approximation techniques in calculus, but like every tool, it has boundaries. Understanding when the method excels and when it breaks down is essential both for the AP exam and for applications in physics and engineering.

Strengths	Limitations
Requires only f(a) and f ′(a) — no higher derivatives needed	Accuracy degrades rapidly as \|x − a\| increases
Provides quick mental or pencil-and-paper estimates	Cannot capture curvature — the method is blind to concavity in its output
Over/under nature can be determined via f ″	Useless at points where f is not differentiable (corners, cusps)
Foundation for differentials, error propagation, and Newton's method	For better accuracy, higher-order Taylor polynomials (Calc BC) are needed

Balancing the utility and constraints of tangent-line approximation

⚡

WHEN TO USE LINEARIZATION

Linearization is analogous to a GPS that gives you an accurate bearing for the next hundred meters but becomes unreliable over longer distances. Just as recalibrating your GPS at a new reference point refreshes the prediction, choosing a base point a close to your target ensures the tangent-line approximation remains trustworthy. When you need accuracy over a wider interval, upgrading to higher-order Taylor polynomials is like switching to a more detailed map.

SECTION 8

Connection to Taylor Polynomials and Advanced Theory

Linearization is the first-degree case of a broader approximation strategy that you will encounter in AP Calculus BC and in college-level analysis courses. A Taylor polynomial of degree n centered at a is a polynomial that matches the function and its first n derivatives at x = a. The linearization L(x) is simply the Taylor polynomial of degree 1, capturing only the function value and the first derivative. Higher-degree polynomials incorporate curvature (via f ″), rate of change of curvature (via f ‴), and so on, producing increasingly accurate approximations over wider intervals.

Feature	Linearization (Degree 1)	Taylor Polynomial (Degree n)
Formula	f(a) + f ′(a)(x − a)	Σ f⁽ᵏ⁾(a)/k! · (x − a)ᵏ for k = 0 to n
Derivatives used	f(a) and f ′(a) only	f(a), f ′(a), f ″(a), …, f⁽ⁿ⁾(a)
Error order	O((x − a)²)	O((x − a)ⁿ⁺¹)
AP Coverage	AP Calculus AB	AP Calculus BC

Linearization as the foundation for higher-order polynomial approximations

Even within the scope of AP Calculus AB, linearization connects to several other important ideas. The concept of the differential dy = f ′(x) dx is the infinitesimal version of linearization. Newton's method for finding roots of equations uses repeated linearizations to converge on a solution. L'Hôpital's Rule, in its proof, relies on the local linearity of the numerator and denominator near the point where both vanish. Mastering linearization therefore gives you an intellectual foothold on multiple advanced topics.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

If f is twice differentiable at x = a with f ″(a) > 0, what can be concluded about the linearization L(x) = f(a) + f ′(a)(x − a) for x near a?

A. L(x) overestimates f(x) for all x near a (x ≠ a)
B. L(x) underestimates f(x) for all x near a (x ≠ a)
C. L(x) overestimates f(x) for x > a and underestimates for x < a
D. L(x) = f(x) for all x near a
E. The relationship between L(x) and f(x) cannot be determined from f ″(a) alone

Show answer (B)

The correct answer is B. When f ″(a) > 0, the function is concave up near a, meaning the curve bends upward away from the tangent line. The tangent line therefore lies below the curve on both sides of a, so L(x) < f(x) for x ≠ a near a — an underestimate. Choice A is incorrect because it describes the concave-down case (f ″ < 0). Choice C is incorrect because the tangent line lies below the curve on both sides when f is concave up, not just one side. Choice D is incorrect because L(x) = f(x) only at x = a itself. Choice E is incorrect because the sign of f ″ is precisely the information needed to determine the over/under relationship.

Question 2: BASIC CALCULATION

Let f(x) = sin(x). Using the linearization of f at a = 0, what is the approximate value of sin(0.05)?

A. 0
B. 0.025
C. 0.05
D. 0.0998
E. 1.05

Show answer (C)

The correct answer is C. The linearization of f(x) = sin(x) at a = 0 is L(x) = f(0) + f ′(0)(x − 0) = 0 + 1·x = x. Therefore L(0.05) = 0.05. Choice A uses only f(0) = 0 and ignores the derivative term. Choice B incorrectly halves the value, possibly confusing the formula with ½·f ′(0)·x. Choice D appears to be sin(0.1) rounded, using the wrong input. Choice E adds 1 to x, perhaps confusing sin with the linearization of cos at 0, where L(x) = 1.

Question 3: INTERMEDIATE

Let g(x) = eˣ. Use the linearization of g at a = 1 to approximate

e^{1.1}

. Which of the following is the approximation, and is it an overestimate or underestimate?

A. e + 0.1e ≈ 2.990; underestimate
B. e + 0.1e ≈ 2.990; overestimate
C. 1 + 1.1 = 2.1; underestimate
D. e + 0.1 ≈ 2.818; overestimate
E. 1.1e ≈ 2.990; overestimate

Show answer (A)

The correct answer is A. L(x) = g(1) + g ′(1)(x − 1) = e + e(x − 1) = e(1 + (x − 1)) = e·x evaluated at structure: L(x) = e + e(0.1) = 1.1e ≈ 2.990. Since g ″(x) = eˣ > 0 for all x, eˣ is concave up everywhere, so the tangent line lies below the curve, making the approximation an underestimate. Choice B gives the correct numerical value but incorrectly labels it an overestimate. Choice C uses a = 0 instead of a = 1. Choice D forgets to multiply the displacement by f ′(1) = e and merely adds 0.1. Choice E has the right numerical value but incorrectly identifies the direction of error.

Question 4: APPLIED

4 pts

A particle's position along a straight track is given by s(t), where s is measured in meters and t in seconds. At time t = 3, the particle is at position s(3) = 10 meters and has velocity s ′(3) = −4 m/s. The acceleration satisfies s ″(t) > 0 for all t in the interval [2, 4]. (a) Write the linearization L(t) of s at t = 3. (b) Use L(t) to approximate the particle's position at t = 3.2. (c) Is your approximation in part (b) an overestimate or underestimate of the actual position s(3.2)? Justify your answer using s ″. (d) Explain the physical meaning of the linearization in this context. What assumption does it make about the particle's motion?

Show answer

Rubric (4 points): (a) [1 point] L(t) = s(3) + s ′(3)(t − 3) = 10 + (−4)(t − 3) = 10 − 4(t − 3). Award 1 point for the correct linearization with both f(a) and f ′(a) terms correctly substituted. (b) [1 point] L(3.2) = 10 − 4(3.2 − 3) = 10 − 4(0.2) = 10 − 0.8 = 9.2 meters. Award 1 point for correct evaluation. (c) [1 point] Since s ″(t) > 0 on [2, 4], the function s(t) is concave up near t = 3. When a function is concave up, the tangent line lies below the curve, so L(3.2) < s(3.2). Therefore 9.2 meters is an underestimate. Award 1 point for correct conclusion with valid justification referencing concavity and the sign of s ″. (d) [1 point] The linearization assumes that the particle continues moving at a constant velocity of −4 m/s from t = 3 onward — that is, it models the particle as having zero acceleration. In reality, s ″(t) > 0 means the particle is decelerating (since velocity is negative and acceleration is positive), so it does not move as far in the negative direction as the linearization predicts, which is consistent with the underestimate conclusion. Award 1 point for explaining that linearization assumes constant velocity (no acceleration) and connecting this to the physical scenario.

Question 5: CRITICAL THINKING

3 pts

Let h(x) = x³ − 6x. A student uses linearization at a = 2 to approximate h(1.98). (a) Compute the linearization L(x) and evaluate L(1.98). (b) Without computing h(1.98) exactly, determine whether L(1.98) is an overestimate or underestimate. Justify your answer. (c) Explain why choosing a = 2 is a better base point than a = 1 for approximating h(1.98), even though both yield exact values of h.

Show answer

Rubric (3 points): (a) [1 point] h(2) = 8 − 12 = −4. h ′(x) = 3x² − 6, so h ′(2) = 12 − 6 = 6. L(x) = −4 + 6(x − 2). L(1.98) = −4 + 6(−0.02) = −4 − 0.12 = −4.12. Award 1 point for correct linearization and evaluation. (b) [1 point] h ″(x) = 6x, so h ″(2) = 12 > 0. Since h is concave up at a = 2, the tangent line lies below the curve near x = 2, so L(1.98) < h(1.98). The approximation −4.12 is an underestimate. Award 1 point for correct conclusion with justification using the sign of h ″. (c) [1 point] The error in linearization is approximately ½|h ″(c)|(x − a)². With a = 2, |x − a| = 0.02, and the error is proportional to (0.02)² = 0.0004. With a = 1, |x − a| = 0.98, and the error is proportional to (0.98)² = 0.9604 — roughly 2400 times larger. The linearization is only accurate when x − a is small, so a = 2 produces a far more accurate approximation. Award 1 point for explaining that accuracy depends on |x − a| being small and quantitatively or qualitatively comparing the two displacements.

SUMMARY

Lesson Summary

Summary

Local linearity is the principle that a differentiable function closely resembles a straight line when viewed over a sufficiently small interval. The linearization of f at a base point a is the tangent-line function L(x) = f(a) + f ′(a)(x − a), which provides a rapid estimate of f(x) for values of x near a. The base point should be chosen so that f(a) and f ′(a) are exactly computable and a is as close to the target x-value as possible.

The direction of the approximation error is governed by concavity: when f ″ > 0 (concave up), the linearization underestimates; when f ″ < 0 (concave down), it overestimates. The error itself is approximately ½ f ″(c)(x − a)², growing with the square of the displacement. Linearization is the degree-one case of the Taylor polynomial and serves as the foundation for differentials, error propagation, and Newton's method.