AP Calculus BC: Lagrange Error Bound

In calculus, we often use Taylor polynomials to approximate complex functions with simpler polynomials. But how do we know if our approximation is any good? The Lagrange error bound (or Taylor's Remainder Theorem) provides the crucial, quantitative answer. It tells you the maximum possible error between the true function value and your polynomial's estimate, allowing you to build reliable models for everything from engineering systems to financial predictions and know exactly how accurate they are.

Taylor Polynomials and the Need for Error Bounds

A Taylor polynomial $P_n(x)$ centered at $x=a$ approximates a function $f(x)$ using the function's derivatives at that center point. The $n$th-degree polynomial is given by: $$P_n(x)=f(a)+f^{\prime }(a)(x-a)+\frac{f^{\prime \prime }(a)}{2!}(x-a{)}^2+...+\frac{f^{(n)}(a)}{n!}(x-a{)}^n$$

While $P_n(x)$ is built from local information at $x=a$, we want to use it to approximate $f(x)$ at some other point $x$. The remainder or error, $R_n(x)$, is defined as the difference: $$f(x)=P_n(x)+R_n(x)\text{or}{R}_n(x)=f(x)-P_n(x).$$ The Lagrange error bound gives us a way to put a guaranteed upper limit on the absolute value of this remainder, $|R_n(x)|$, without knowing the exact value of $f(x)$. This is essential for applications where approximations must meet a specific tolerance for error.

Statement of the Lagrange Error Bound Theorem

The theorem provides a formula for the maximum possible error. If a function $f$ is differentiable $n+1$ times on an interval $I$ containing $x$ and $a$, then for each $x$ in $I$, there exists a real number $z$ between $x$ and $a$ such that the remainder is: $$R_n(x)=\frac{f^{(n+1)}(z)}{(n+1)!}(x-a{)}^{n+1}.$$ This exact form is powerful, but we don't know the value of $z$. The Lagrange error bound is the worst-case scenario. It states: $$|R_n(x)|\le \frac{M}{(n+1)!}|x-a{|}^{n+1}.$$

Here, $M$ is the maximum absolute value of the $(n+1)$th derivative of $f$ on the interval $I$ between $a$ and $x$. In other words, $M$ is a number such that $|f^{(n+1)}(z)|\le M$ for all $z$ in that interval. You find $M$ by analyzing the $(n+1)$th derivative over the relevant interval.

Calculating the Error Bound: A Step-by-Step Process

Applying the bound requires a systematic approach. Let's walk through an example: Find the Lagrange error bound for the fourth-degree Taylor polynomial of $f(x)=e^x$ centered at $a=0$ when approximating $e^{0.5}$.

1. Identify the components.

• Function: $f(x)=e^x$
• Center: $a=0$
• Degree of polynomial, $n=4$
• $x$-value to approximate: $x=0.5$

1. Determine the next derivative. We need the $(n+1)=5$th derivative: $f^{(5)}(x)=e^x$.

1. Find the maximum value $M$ of $|f^{(5)}(z)|$ on the interval between $a=0$ and $x=0.5$. Since $e^z$ is strictly increasing, its maximum absolute value on $[0,0.5]$ occurs at the right endpoint: $M=e^{0.5}$.

1. Plug into the error bound formula.

$$|R_4(0.5)|\le \frac{M}{(5)!}|0.5-0{|}^5=\frac{e^{0.5}}{120}(0.5{)}^5$$

1. Calculate the numerical bound.

$$\frac{e^{0.5}}{120}\cdot (0.03125)\approx \frac{1.64872}{120}\cdot 0.03125\approx 0.0004295$$ So, $|R_4(0.5)|\le 0.0004295$. This means our fourth-degree Maclaurin polynomial's estimate for $e^{0.5}$ is guaranteed to be within about 0.00043 of the true value.

Determining the Number of Terms for a Given Accuracy

A more powerful application is working backwards: finding the smallest degree $n$ of a Taylor polynomial required to achieve a desired precision. This is a hallmark of engineering and computational mathematics.

Scenario: Approximate $\sin (1)$ (where 1 is in radians) using a Maclaurin series ($a=0$) with an error less than $10^{-5}$.

1. Set up the inequality. We want $|R_n(1)|<0.00001$.

The general bound is $|R_n(1)|\le \frac{M}{(n+1)!}|1{|}^{n+1}$.

1. Find a general expression for $M$. For $f(x)=\sin x$, all derivatives are $\pm \sin x$ or $\pm \cos x$. Therefore, for any $z$, $|f^{(n+1)}(z)|\le 1$. We can always use $M=1$.

1. Solve for $n$. We need:

$$\frac{1}{(n+1)!}\cdot (1{)}^{n+1}<0.00001\text{or}(n+1)!>100,000.$$

1. Test integer values.

• For $n=6$: $(7)!=5040$. 5040 < 100,000. Not enough.
• For $n=7$: $(8)!=40,320$. 40,320 < 100,000. Not enough.
• For $n=8$: $(9)!=362,880$. 362,880 > 100,000. Success.

Therefore, the smallest degree is $n=8$. A 9th-term (since we use the $(n+1)$th derivative) would be the first unused term, confirming we need the 8th-degree polynomial.

Common Pitfalls

Misidentifying the value of $n$: The most frequent error is confusing the degree of the polynomial with the derivative order in the formula. Remember: If you are using an nth-degree polynomial $P_n(x)$, the error bound uses the next, unused derivative, $f^{(n+1)}(z)$. The factorial and exponent in the denominator are also $(n+1)$.

Selecting the wrong interval for finding $M$: The number $M$ must be a maximum for the entire interval between the center $a$ and the approximation point $x$, not just at the endpoints. For example, if the derivative has a critical point (like a local maximum) inside $(a,x)$, that point could yield the true maximum $M$. Always check the behavior of $f^{(n+1)}(x)$ over the whole closed interval.

Forgetting absolute values in $M$: The theorem requires $M$ to be an upper bound for the absolute value of the derivative, $|f^{(n+1)}(z)|$. If you simply find the maximum of the derivative itself (which could be negative), you may underestimate the maximum absolute error. You are bounding $|R_n(x)|$, so $M$ must be positive and cover the worst-case deviation.

Misapplying to alternating series: For alternating series where the terms decrease monotonically to zero, the Alternating Series Error Bound is simpler and often tighter: $|R_n|\le |a_{n+1}|$. If a Taylor series is alternating, you can use this simpler bound. The Lagrange bound always works but may overestimate. Recognize which tool is most efficient.

Summary

• The Lagrange error bound $|R_n(x)|\le \frac{M}{(n+1)!}|x-a{|}^{n+1}$ provides a guaranteed upper limit for the error in a Taylor polynomial approximation, where $M$ is the maximum value of $|f^{(n+1)}(z)|$ on the interval between $a$ and $x$.
• Its primary uses are to quantify the accuracy of a given polynomial approximation and to determine the minimum polynomial degree required to achieve a specific error tolerance, a key step in designing reliable approximations.
• Critical steps in calculation involve correctly identifying $n$ (the degree of the polynomial used), finding the $(n+1)$th derivative, and carefully analyzing that derivative's absolute maximum over the relevant interval.
• Avoid common mistakes by remembering the formula uses the next unused derivative term ($n+1$), ensuring $M$ bounds the absolute value of that derivative, and selecting the correct interval for analysis.
• Mastering this bound transforms Taylor polynomials from mere approximations into powerful, quantifiable tools for solving real-world problems in science and engineering where precision is non-negotiable.