Lagrange Error Bound - AP Calculus BC

SECTION 1

Historical Context & Motivation

Long before electronic calculators, mathematicians needed reliable methods for computing transcendental quantities such as e, sin 1, and ln 2 to many decimal places. Taylor polynomials offered a powerful tool—replace a complicated function with a polynomial and evaluate that polynomial instead. The deeper question, however, was always the same: how many terms are enough to guarantee a desired accuracy? Without a rigorous answer, a polynomial approximation was merely a hopeful guess.

1715 • Brook Taylor's Series

Brook Taylor publishes Methodus Incrementorum Directa et Inversa, introducing the general formula for expanding a function as an infinite power series about a point.

1742 • Maclaurin's Treatise

Colin Maclaurin popularizes the special case of Taylor series centered at x = 0, later called Maclaurin series, and begins discussing convergence behavior.

1797 • Lagrange's Remainder Theorem

Joseph-Louis Lagrange provides a precise formula for the remainder term Rₙ(x) in his Théorie des fonctions analytiques, giving mathematicians a concrete bound on truncation error.

1821 • Cauchy's Rigor

Augustin-Louis Cauchy reformulates the remainder theorem using the Mean Value Theorem, solidifying the modern proof that appears in calculus textbooks today.

Lagrange's contribution answered the critical question: if you truncate a Taylor series after n terms, how large can the error possibly be? The bound he established depends on the behavior of the (n + 1)-th derivative of the function—an elegant result that bridges local derivative information with global approximation quality. On the AP Calculus BC exam, this result appears as the Lagrange Error Bound and is one of the most tested topics in the series unit.

SECTION 2

Core Principles & Definitions

Before diving into the bound itself, you need to be fluent in several interconnected ideas. The Taylor polynomial of degree n centered at c is the unique polynomial that matches the function and its first n derivatives at x = c. The difference between the actual function value and this polynomial at any point x is called the remainder or truncation error. The Lagrange Error Bound places an upper limit on the absolute value of this remainder.

Taylor Polynomial Pₙ(x)

The degree-n polynomial Pₙ(x) = Σ f⁽ᵏ⁾(c)(x − c)ᵏ / k! for k = 0 to n. It 'matches' f at the center through order n.

Remainder Rₙ(x)

Defined as Rₙ(x) = f(x) − Pₙ(x). This is the exact error when the Taylor polynomial replaces f. The goal is to bound |Rₙ(x)| without knowing f(x) exactly.

The (n + 1)-th Derivative Bound M

Find M such that |f⁽ⁿ⁺¹⁾(z)| ≤ M for all z between c and x. This M is the engine of the bound—it captures how 'wild' the next derivative is on the interval.

The Lagrange Error Bound

|Rₙ(x)| ≤ M · |x − c|ⁿ⁺¹ / (n + 1)!. This is a guaranteed worst-case ceiling on the error; the actual error is usually smaller.

✦

KEY TAKEAWAY

Think of a Taylor polynomial as a GPS route and the remainder as how far off-course you might drift. The Lagrange Error Bound is the maximum radius of that drift zone. The smoother the road ahead (the smaller the next derivative), the tighter the bound—just as a well-paved highway keeps you closer to the planned route than a winding mountain pass.

SECTION 3

Visual Explanation

The following diagram shows a function f(x) = eˣ (in cyan) alongside its first-, second-, and third-degree Taylor polynomials centered at c = 0. The shaded error band around each polynomial represents the region allowed by the Lagrange Error Bound. Observe how the band narrows dramatically as the degree increases and how it widens as x moves farther from the center.

The cyan curve is f(x) = eˣ. The dashed violet line is the first-degree Taylor polynomial P₁(x) = 1 + x, and the dashed pink curve is the third-degree polynomial P₃(x). The amber bracket at x = 1 shows the error R₁(1) for P₁, while the green bracket shows R₃(1) for P₃—notice how the higher-degree polynomial dramatically reduces the gap.

Two key visual takeaways emerge from this diagram. First, the error grows as |x − c| increases: near the center the polynomial practically sits on top of f, but farther away it peels off. Second, increasing the polynomial degree shrinks the error much faster than you might expect, because the factorial (n + 1)! in the denominator of the bound grows very rapidly. These geometric observations are exactly what the Lagrange Error Bound formula captures algebraically.

SECTION 4

Mathematical Framework

The Lagrange form of the remainder originates from Taylor's theorem with the remainder expressed in a form analogous to the Mean Value Theorem. If f has continuous derivatives through order n + 1 on an interval containing both c and x, then there exists some value z strictly between c and x such that the remainder equals exactly f⁽ⁿ⁺¹⁾(z) · (x − c)ⁿ⁺¹ / (n + 1)!. Because we typically cannot identify z precisely, we replace |f⁽ⁿ⁺¹⁾(z)| with a convenient upper bound M.

TAYLOR POLYNOMIAL

Pₙ(x) = Σₖ₌₀ⁿ f⁽ᵏ⁾(c) · (x − c)ᵏ / k!

Pₙ(x) is the degree-n Taylor polynomial centered at c. Each term uses the k-th derivative of f evaluated at the center.

EXACT REMAINDER (LAGRANGE FORM)

Rₙ(x) = f⁽ⁿ⁺¹⁾(z) · (x − c)ⁿ⁺¹ / (n + 1)! for some z between c and x

Rₙ(x) = f(x) − Pₙ(x). The unknown z is guaranteed to exist by a generalization of the Mean Value Theorem, but its exact location is usually unknown.

LAGRANGE ERROR BOUND

|Rₙ(x)| ≤ M · |x − c|ⁿ⁺¹ / (n + 1)!

where M = max |f⁽ⁿ⁺¹⁾(z)| for all z between c and x. This replaces the unknown (n+1)-th derivative at z with the worst-case value on the interval, yielding a guaranteed upper bound on the absolute error.

⚠

AP EXAM TIP

On the AP Calculus BC exam, the most common error is choosing M incorrectly. Remember: M must be an upper bound for |f⁽ⁿ⁺¹⁾(z)| on the entire interval between c and x, not just at the endpoints. For functions like sin x and cos x, whose derivatives are always bounded by 1 in absolute value, M = 1 is almost always the correct choice.

Notice how two competing forces govern the size of the bound. The numerator |x − c|ⁿ⁺¹ grows when x is far from c, but the denominator (n + 1)! grows factorially in n. For a fixed x and a well-behaved function, increasing n causes the factorial to dominate, driving the bound—and therefore the error—toward zero. This is precisely the mechanism by which Taylor series converge.

SECTION 5

Finding the Derivative Bound M

The most nuanced step in applying the Lagrange Error Bound is determining a valid value for M, the maximum of |f⁽ⁿ⁺¹⁾(z)| on the interval between c and x. The strategy depends entirely on the function in question. For some functions the derivatives cycle (like sine and cosine), making M obvious; for others, you must use calculus techniques—such as checking whether the (n + 1)-th derivative is increasing or decreasing—to identify the maximum.

This flowchart guides you through the decision process for selecting M. For trig functions, M = 1 always works. For exponentials and reciprocals, check whether the (n + 1)-th derivative is monotone—if so, evaluate at the endpoint that gives the larger absolute value.

Function	f⁽ⁿ⁺¹⁾(x) Pattern	Typical M Choice
sin x	±sin x or ±cos x	M = 1
cos x	±sin x or ±cos x	M = 1
eˣ	eˣ (always increasing)	M = eˣ evaluated at endpoint farther from 0
ln(1 + x)	(−1)ⁿ · n! / (1+x)ⁿ⁺¹	M = n! / (1+a)ⁿ⁺¹ where a is closer to −1
1/(1 − x)	(n+1)! / (1−x)ⁿ⁺²	Evaluate at x closest to 1

Common functions and their typical M values for the Lagrange Error Bound

💡

SAFE OVERESTIMATE

You can always choose a larger M than the true maximum of |f⁽ⁿ⁺¹⁾(z)|. The bound will still be valid—it just won't be as tight. On the AP exam, a generous but correct M earns full credit; a too-small M that violates the bound does not.

SECTION 6

Worked Example

Let us compute an upper bound on the error when the fourth-degree Maclaurin polynomial for sin x is used to approximate sin(0.5). This example mirrors the style and complexity of AP Calculus BC free-response questions.

Bounding the Error of a Maclaurin Polynomial for sin x

1
Step 1 — Write the Maclaurin polynomial

The Maclaurin series for sin x is x − x³/3! + x⁵/5! − x⁷/7! + ⋯ . The fourth-degree Maclaurin polynomial P₄(x) actually stops at the x³ term because the x⁴ coefficient is 0 (since sin x has only odd powers). More precisely, P₄(x) = x − x³/6. Note that P₄(x) = P₃(x) for this function.

P₄(x) = x − x³/6
2
Step 2 — Identify n and compute f⁽ⁿ⁺¹⁾(x)

We are using the degree-4 polynomial, so n = 4. The (n + 1)-th derivative is f⁽⁵⁾(x). Since f(x) = sin x, the successive derivatives cycle: cos x, −sin x, −cos x, sin x, cos x, … . Thus f⁽⁵⁾(x) = cos x.

f⁽⁵⁾(x) = cos x
3
Step 3 — Determine M

We need |f⁽⁵⁾(z)| = |cos z| ≤ M for all z between c = 0 and x = 0.5. Since |cos z| ≤ 1 for all real z, we can safely take M = 1.

M = 1
4
Step 4 — Apply the Lagrange Error Bound

|R₄(0.5)| ≤ M · |x − c|ⁿ⁺¹ / (n + 1)! = 1 · |0.5 − 0|⁵ / 5! = (0.5)⁵ / 120 = 0.03125 / 120.

|R₄(0.5)| ≤ 0.03125 / 120 ≈ 0.000260
5
Step 5 — Interpret the result

The Lagrange Error Bound guarantees that the approximation P₄(0.5) = 0.5 − (0.5)³/6 = 0.5 − 0.020833… ≈ 0.479167 differs from the true value sin(0.5) ≈ 0.479426 by no more than about 2.60 × 10⁻⁴. The actual error is |0.479426 − 0.479167| ≈ 2.59 × 10⁻⁴, confirming the bound is valid (and in this case, quite tight).

Error < 2.60 × 10⁻⁴ ✓

SECTION 7

Strengths & Limitations

The Lagrange Error Bound is the most frequently used error bound in introductory calculus courses, but it is not the only tool available. Understanding its advantages and shortcomings helps you decide when to use it and when another method—such as the Alternating Series Estimation Theorem—might be more efficient.

Aspect	Lagrange Error Bound	Alternating Series Estimation
Applicability	Works for any Taylor polynomial of a sufficiently differentiable function	Only applies when the series is alternating and satisfies the alternating series conditions
What you need	An upper bound M on \|f⁽ⁿ⁺¹⁾\| on the relevant interval	The first omitted term of the series (must be decreasing in absolute value)
Tightness	Can overestimate error significantly if M is loose	Often gives a tighter bound because it uses the actual next term
AP exam usage	Used when FRQ specifically asks for 'Lagrange error bound' or for non-alternating series	Preferred when the series alternates and the problem asks for a general error estimate
Difficulty	Requires computing the (n+1)-th derivative and bounding it	Simpler—just evaluate the next term in the series

Comparison of the two main error-bounding tools in AP Calculus BC

⚖️

WHEN TO USE WHICH

Think of the Lagrange Error Bound as a universal wrench—it fits every bolt but may not give the tightest grip. The Alternating Series Estimation Theorem is a precision socket wrench—when it fits (alternating series), it's often faster and tighter. On the AP exam, read the prompt carefully: if it says 'use the Lagrange error bound,' you must use the formula with M and (n + 1)!.

SECTION 8

Connections to Advanced Theory

The Lagrange Error Bound is, at its heart, a consequence of the Mean Value Theorem applied iteratively. In more advanced analysis courses, you encounter other forms of the Taylor remainder—the Cauchy form and the integral form—each useful in different contexts. The integral form, Rₙ(x) = ∫ from c to x of f⁽ⁿ⁺¹⁾(t)(x − t)ⁿ / n! dt, provides a direct connection to numerical integration and is often used in proofs of convergence for specific series.

Remainder Form	Formula	Best Used For
Lagrange	f⁽ⁿ⁺¹⁾(z)(x − c)ⁿ⁺¹/(n+1)!	Bounding error without integration; AP Calculus BC
Cauchy	f⁽ⁿ⁺¹⁾(z)(x − z)ⁿ(x − c)/n!	L'Hôpital-style arguments; proving convergence rates
Integral	∫ from c to x f⁽ⁿ⁺¹⁾(t)(x−t)ⁿ/n! dt	Exact computation; real analysis proofs; multivariable extensions

Three forms of the Taylor remainder

In multivariable calculus, the Lagrange remainder generalizes to functions of several variables, where the bound involves partial derivatives of order n + 1 and norms of displacement vectors. In numerical analysis, the same structural idea underlies error bounds for interpolation polynomials, quadrature rules, and finite difference methods. Mastering the single-variable version now builds the conceptual foundation you will rely on in linear algebra, differential equations, and beyond.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

In the Lagrange Error Bound formula |Rₙ(x)| ≤ M · |x − c|ⁿ⁺¹ / (n + 1)!, the value M represents:

A. The exact value of f⁽ⁿ⁺¹⁾(c)
B. The maximum value of |f⁽ⁿ⁾(z)| on the interval between c and x
C. An upper bound for |f⁽ⁿ⁺¹⁾(z)| for all z between c and x
D. The average value of f⁽ⁿ⁺¹⁾(z) on the interval [c, x]
E. The value of f⁽ⁿ⁺¹⁾(x)

Show answer (C)

The correct answer is C. The Lagrange remainder involves f⁽ⁿ⁺¹⁾(z) for an unknown z between c and x. Since z is unknown, we replace |f⁽ⁿ⁺¹⁾(z)| with an upper bound M that works for every z in that interval. Choice A is wrong because M need not equal the derivative at the center—it must bound the derivative everywhere between c and x. Choice B is wrong because M bounds the (n+1)-th derivative, not the n-th derivative. Choice D is wrong because the Mean Value Theorem guarantees existence of a specific z, not an average. Choice E is wrong because M must bound the derivative at all points between c and x, not just at x itself.

Question 2: BASIC CALCULATION

The third-degree Maclaurin polynomial for eˣ is P₃(x) = 1 + x + x²/2 + x³/6. Using the Lagrange Error Bound, what is the maximum error when P₃(x) is used to approximate e⁰·² (i.e., x = 0.2)? Use M = e⁰·² ≈ 1.2215.

A. 6.52 × 10⁻⁵
B. 8.14 × 10⁻⁵
C. 1.96 × 10⁻⁴
D. 3.26 × 10⁻⁴
E. 1.63 × 10⁻³

Show answer (B)

The correct answer is B. With n = 3, f(x) = eˣ, c = 0, and x = 0.2, the (n+1)-th derivative is f⁽⁴⁾(x) = eˣ. On [0, 0.2], the maximum of eˣ is e⁰·² ≈ 1.2215, so M = 1.2215. The bound gives |R₃(0.2)| ≤ 1.2215 × (0.2)⁴ / 4! = 1.2215 × 0.0016 / 24 = 0.0019544 / 24 ≈ 8.14 × 10⁻⁵. Choice A underestimates M (likely used M = 1). Choice C uses (0.2)³/3! instead of (0.2)⁴/4!. Choice D uses 3! in the denominator instead of 4!. Choice E uses (0.2)³ in the numerator.

Question 3: INTERMEDIATE

You want to approximate cos(1) using a Maclaurin polynomial with error less than 10⁻⁶. What is the smallest degree n of the Maclaurin polynomial that guarantees this accuracy using the Lagrange Error Bound?

A. n = 6
B. n = 8
C. n = 9
D. n = 10
E. n = 12

Show answer (C)

The correct answer is C. For f(x) = cos x, all derivatives satisfy |f⁽ⁿ⁺¹⁾(z)| ≤ 1, so M = 1. We need 1 · |1 − 0|ⁿ⁺¹ / (n + 1)! < 10⁻⁶, i.e., 1/(n + 1)! < 10⁻⁶. Testing factorials: 1/8! = 1/40320 ≈ 2.48 × 10⁻⁵ (too large); 1/9! = 1/362880 ≈ 2.76 × 10⁻⁶ (still too large); 1/10! = 1/3628800 ≈ 2.76 × 10⁻⁷ < 10⁻⁶ ✓. So we need (n + 1)! ≥ 10!, meaning n + 1 ≥ 10, so n ≥ 9. The smallest such n is 9. Choice A (n = 6) gives 1/7! ≈ 2.0 × 10⁻⁴. Choice B (n = 8) gives 1/9! ≈ 2.76 × 10⁻⁶, which exceeds 10⁻⁶. Choices D and E work but are not the smallest valid n.

Question 4: APPLIED

4 pts

A function f has derivatives of all orders for all real numbers. The table below gives values of f and its first four derivatives at x = 2. | k | 0 | 1 | 2 | 3 | 4 | |---|---|---|---|---|---| | f⁽ᵏ⁾(2) | 3 | −1 | 4 | −6 | 10 | It is known that |f⁽⁵⁾(x)| ≤ 30 for all x in the interval [2, 2.5]. (a) Write the fourth-degree Taylor polynomial P₄(x) for f centered at x = 2. (b) Use P₄(2.5) to approximate f(2.5). (c) Use the Lagrange Error Bound to find an upper bound for |f(2.5) − P₄(2.5)|. (d) Could the actual value of f(2.5) be as large as 3.5? Justify using your results from parts (b) and (c).

Show answer

RUBRIC (4 points): Point 1 — Part (a): P₄(x) = 3 + (−1)(x−2) + (4/2!)(x−2)² + (−6/3!)(x−2)³ + (10/4!)(x−2)⁴ = 3 − (x−2) + 2(x−2)² − (x−2)³ + (5/12)(x−2)⁴. Point 2 — Part (b): P₄(2.5) = 3 − 0.5 + 2(0.25) − 0.125 + (5/12)(0.0625) = 3 − 0.5 + 0.5 − 0.125 + 0.02604… ≈ 2.901. Point 3 — Part (c): With n = 4, M = 30, c = 2, x = 2.5: |R₄(2.5)| ≤ 30 · |0.5|⁵ / 5! = 30 · (1/32) / 120 = 30/3840 = 1/128 ≈ 0.00781. Point 4 — Part (d): From parts (b) and (c), f(2.5) lies in the interval [P₄(2.5) − 0.00781, P₄(2.5) + 0.00781] = [2.893, 2.909] (approximately). Since 3.5 is well outside this interval, f(2.5) cannot be as large as 3.5.

Question 5: CRITICAL THINKING

3 pts

Let f(x) = sin x and let Pₙ(x) be the n-th degree Maclaurin polynomial for f. (a) Show that the Lagrange Error Bound guarantees |Rₙ(x)| → 0 as n → ∞ for every real number x, thereby proving that the Maclaurin series for sin x converges to sin x for all x. (b) Explain why this argument would not work if f⁽ⁿ⁺¹⁾ grew faster than (n + 1)!.

Show answer

RUBRIC (3 points): Point 1 — Part (a), bound setup: For f(x) = sin x, every derivative is ±sin x or ±cos x, so |f⁽ⁿ⁺¹⁾(z)| ≤ 1 for all z. Therefore |Rₙ(x)| ≤ 1 · |x|ⁿ⁺¹ / (n + 1)!. Point 2 — Part (a), limit argument: For any fixed x, the sequence |x|ⁿ⁺¹ / (n + 1)! → 0 as n → ∞ because factorial growth in the denominator dominates polynomial/exponential growth in the numerator. (This can be shown by the ratio test: the ratio of consecutive terms is |x|/(n + 2) → 0.) By the Squeeze Theorem, since 0 ≤ |Rₙ(x)| ≤ |x|ⁿ⁺¹/(n+1)! → 0, we conclude |Rₙ(x)| → 0, so sin x = Σ from k=0 to ∞ of (−1)ᵏx²ᵏ⁺¹/(2k+1)! for all x ∈ ℝ. Point 3 — Part (b): If |f⁽ⁿ⁺¹⁾(z)| grew as fast as or faster than (n+1)!, then M would depend on n and grow comparably with (n+1)!, so the bound M · |x|ⁿ⁺¹/(n+1)! would not necessarily tend to zero. The cancellation that drives the bound to zero relies on M being bounded independently of n (or at least growing slower than (n+1)!).

SUMMARY

Lesson Summary

Summary

The Lagrange Error Bound provides a guaranteed upper limit on the absolute error when a Taylor polynomial Pₙ(x) replaces a function f(x). The bound is |Rₙ(x)| ≤ M · |x − c|ⁿ⁺¹ / (n + 1)!, where M is an upper bound for |f⁽ⁿ⁺¹⁾(z)| on the interval between the center c and the evaluation point x. The key skill is selecting an appropriate M: for sin x and cos x, M = 1 always works; for eˣ, evaluate at the interval endpoint farther from zero.

This bound also serves as the theoretical engine for proving Taylor series convergence: if |Rₙ(x)| → 0 as n → ∞, the series converges to the function. On the AP exam, distinguish between the Lagrange Error Bound (which applies universally) and the Alternating Series Estimation Theorem (which applies only to alternating series but is often tighter). Practice identifying n, computing the (n + 1)-th derivative, choosing M, and evaluating the bound—these steps appear in nearly every FRQ on series.