Skip to main content

Section 2.4 Applications of Taylor Series and Remainder Theorem

Physicists often use Taylor series to approximate functions in order to simplify calculations. The first example is the kinetic energy of an object in relativistic mechanics. The second example is the approximation of the period of a pendulum. We will discuss the error in using these approximations and when it is valid to use them. Additionally, we will discuss how to use the taylor series to estimate value of \(\pi\) in the last example.

Subsection 2.4.1 Approximating Relativistic Kinetic Energy

In relativistic mechanics, the mass of an object moving with velocity \(v\) is given by:
\begin{equation*} m = \frac{m_0}{\sqrt{1 - \frac{v^2}{c^2}}} \end{equation*}
where \(m_0\) is the rest mass of the object and \(c\) is the speed of light. Then the kinetic energy of an object of mass \(m\) moving with velocity \(v\) is:
\begin{equation*} K = mc^2-m_0c^2 \end{equation*}
In the case when \(v \ll c\text{,}\) we can use the Taylor series to approximate the kinetic energy. In example Example 2.13 we will show that the kinetic energy can be approximated by the formula \(K \approx \frac{1}{2} m_0v^2\) when \(v \ll c\text{.}\) See Figure 2.12 for a comparison of the relativistic kinetic energy and its Newtonian approximation.
Diagram Exploration Keyboard Controls
Key Action
Enter, A Activate keyboard driven exploration
B Activate menu driven exploration
Escape Leave exploration mode
Cursor down Explore next lower level
Cursor up Explore next upper level
Cursor right Explore next element on level
Cursor left Explore previous element on level
X Toggle expert mode
W Extra details if available
Space Repeat speech
M Activate step magnification
Comma Activate direct magnification
N Deactivate magnification
Z Toggle subtitles
C Cycle contrast settings
T Monochrome colours
L Toggle language (if available)
K Kill current sound
Y Stop sound output
O Start and stop sonification
P Repeat sonification output
Figure 2.11. Relativistic versus Newtonian kinetic energy. The relativistic energy races toward a wall at the speed of light \(c\text{,}\) while the Newtonian energy follows a gentle parabola; the two are approximately in agreement only when \(v \ll c\text{.}\)
Diagram Exploration Keyboard Controls
Key Action
Enter, A Activate keyboard driven exploration
B Activate menu driven exploration
Escape Leave exploration mode
Cursor down Explore next lower level
Cursor up Explore next upper level
Cursor right Explore next element on level
Cursor left Explore previous element on level
X Toggle expert mode
W Extra details if available
Space Repeat speech
M Activate step magnification
Comma Activate direct magnification
N Deactivate magnification
Z Toggle subtitles
C Cycle contrast settings
T Monochrome colours
L Toggle language (if available)
K Kill current sound
Y Stop sound output
O Start and stop sonification
P Repeat sonification output
Figure 2.12. Relativistic kinetic energy compared with its Newtonian approximation. The two nearly coincide when \(v \ll c\text{,}\) but the relativistic energy grows without bound as \(v \to c\text{.}\)

Example 2.13. Newtonian kinetic energy as an approximation to relativistic kinetic energy when \(v \ll c\).

Show that the kinetic energy of an object moving with velocity \(v\) can be approximated by the formula \(K \approx \frac{1}{2} m_0v^2\) when \(v \ll c\text{.}\)
Hint.
We can use the Taylor series for \(f(x) = \frac{1}{\sqrt{1 - x}}\) at \(x=0\) to approximate \(K\text{.}\) Use \(x = \frac{v^2}{c^2}\text{.}\)
Solution.
We can rewrite the kinetic energy as follows:
\begin{equation*} K = mc^2-m_0c^2 = m_0c^2 \left( \frac{1}{\sqrt{1 - \frac{v^2}{c^2}}} - 1 \right) \end{equation*}
Let \(x = \frac{v^2}{c^2}\text{.}\) Then we have:
\begin{equation*} K = m_0c^2 \left( \frac{1}{\sqrt{1 - x}} - 1 \right) \end{equation*}
Now we can use the Taylor series for \(f(x) = \frac{1}{\sqrt{1 - x}}\) at \(x=0\) to approximate \(K\text{.}\) The Taylor series for \(f(x)\) at \(x=0\) is given by:
\begin{equation*} f(x) = 1 + \frac{1}{2}x + \frac{3}{8}x^2 + \frac{5}{16}x^3 + \cdots \end{equation*}
Therefore, we have:
\begin{equation*} K = m_0c^2 \left( \frac{1}{\sqrt{1 - x}} - 1 \right) = m_0c^2 \left( \frac{1}{2}x + \frac{3}{8}x^2 + \frac{5}{16}x^3 + \cdots \right) \end{equation*}
Substituting \(x = \frac{v^2}{c^2}\text{,}\) we get:
\begin{align*} K \amp = m_0c^2 \left( \frac{1}{2}\frac{v^2}{c^2} + \frac{3}{8}\frac{v^4}{c^4} + \frac{5}{16}\frac{v^6}{c^6} + \cdots \right) \\ \amp = \frac{1}{2} m_0v^2 + \frac{3}{8} m_0\frac{v^4}{c^2} + \frac{5}{16} m_0\frac{v^6}{c^4} + \cdots \end{align*}
When \(v \ll c\text{,}\) the higher order terms in the series become negligible, and we can approximate the kinetic energy as:
\begin{equation*} K \approx \frac{1}{2} m_0v^2. \end{equation*}

Example 2.14. Estimating the error in using the Newtonian kinetic energy formula to approximate the relativistic kinetic energy.

Assume that a car a moving with a velocity of \(v = 30 m/s \) (\(\approx 67\) miles per hour). Use the remainder’s theorem to estimate the error in using the Newtonian kinetic energy formula \(K \approx \frac{1}{2} m_0v^2\) to approximate the relativistic kinetic energy of the car. The speed of light is \(c = 3 \times 10^8 m/s\text{.}\)
Solution.
Note that when using the Newtonian kinetic energy formula, we are using the Taylor polynomial of degree 1 at \(x=0\) to approximate \(K\text{.}\) Therefore, the error term is given by:
\begin{equation*} R_1(x) = m_0 c^2 \frac{f^{(2)}(c)}{2!} x^2 \end{equation*}
Since \(f^{(2)}(x) = \frac{3}{4} (1-x)^{-5/2}\text{,}\) we have:
\begin{equation*} |R_1(x)| \leq m_0 c^2 \frac{3}{8} (1-x)^{-5/2} x^2 \end{equation*}
Since \(x = \frac{v^2}{c^2}\text{,}\) we have:
\begin{equation*} |R_1(x)| \leq m_0 c^2 \frac{3}{8} \left( 1 - \frac{v^2}{c^2} \right)^{-5/2} \left( \frac{v^2}{c^2} \right)^2 \end{equation*}
Substituting \(v/c = 10^{-7}\) and \(c = 3 \times 10^8 m/s\text{,}\) we get:
\begin{equation*} |R_1(x)| \leq m_0 (3 \times 10^8)^2 \frac{3}{8} \left( 1 - 10^{-14} \right)^{-5/2} \left(10^{-14} \right)^2 \lt 4 \times 10^{-11} m_0 \end{equation*}
This shows that the error is extremely small, and the Newtonian kinetic energy formula is a very good approximation to the relativistic kinetic energy for a car moving at \(30 m/s\text{.}\)

Subsection 2.4.2 Small-Angle Approximation for a Pendulum

We begin this section by briefly reviewing the forces acting on a simple pendulum and how the small-angle approximation allows us to treat its motion as simple harmonic motion.
The bob moves along the arc, so only the component of gravity tangent to that arc drives the motion. Resolving the weight \(m\vec{\mathbf{g}}\) into a component along the string (\(mg\cos\theta\text{,}\) balanced by the string tension \(\vec{\mathbf{F}}_T\)) and a component tangent to the arc, as shown in Figure 2.15, gives the restoring force
\begin{equation*} F = -mg\sin\theta . \end{equation*}
The minus sign indicates that the force always points back toward the equilibrium (straight-down) position. This is not Hooke’s law: the force is proportional to \(\sin\theta\text{,}\) not to the displacement itself, so the motion is not exactly simple harmonic. Writing the arc displacement as \(x = \ell\theta\text{,}\) we would need \(F\) to be proportional to \(x\)—that is, to \(\theta\)—for the motion to be simple harmonic.
The small-angle approximation bridges this gap. From the Taylor series
\begin{equation*} \sin\theta = \theta - \frac{\theta^3}{3!} + \frac{\theta^5}{5!} - \cdots, \end{equation*}
when \(\theta\) is small (in radians) the higher-order terms are negligible and \(\sin\theta \approx \theta\text{.}\) The restoring force then becomes
\begin{equation*} F \approx -mg\theta = -\frac{mg}{\ell}\,x, \end{equation*}
which is Hooke’s law \(F = -kx\) with effective spring constant \(k = \dfrac{mg}{\ell}\text{.}\) Thus, for small swings the pendulum behaves as a simple harmonic oscillator, with period
\begin{equation*} T = 2\pi\sqrt{\frac{\ell}{g}}, \end{equation*}
independent of both the amplitude and the mass. Taylor’s Remainder Theorem (see Example 2.17) enables us to quantify how small \(\theta\) must be for this approximation—.
Diagram Exploration Keyboard Controls
Key Action
Enter, A Activate keyboard driven exploration
B Activate menu driven exploration
Escape Leave exploration mode
Cursor down Explore next lower level
Cursor up Explore next upper level
Cursor right Explore next element on level
Cursor left Explore previous element on level
X Toggle expert mode
W Extra details if available
Space Repeat speech
M Activate step magnification
Comma Activate direct magnification
N Deactivate magnification
Z Toggle subtitles
C Cycle contrast settings
T Monochrome colours
L Toggle language (if available)
K Kill current sound
Y Stop sound output
O Start and stop sonification
P Repeat sonification output
Figure 2.15. Forces on a simple pendulum. The weight \(m\vec{\mathbf{g}}\) resolves into a component \(mg\cos\theta\) along the string (balanced by the tension \(\vec{\mathbf{F}}_T\)) and a component \(mg\sin\theta\) tangent to the arc, which acts as the restoring force.
Figure 2.16. Forces on a simple pendulum. The weight \(m\vec{g}\) resolves into a radial component \(mg\cos\theta\) along the string, balanced by the tension \(\vec{F}_T\text{,}\) and a tangential component \(mg\sin\theta\) directed toward equilibrium, which acts as the restoring force. For small angles, \(\sin\theta \approx \theta\) gives \(F \approx -\dfrac{mg}{\ell}x\text{.}\)

Example 2.17. Simple Harmonic Motion of a Pendulum as an Approximation.

Use the remainder theorem to analyze the claim made in Giancoli’s textbook that for small angles, \(\sin\theta \approx \theta\text{.}\) Specifically, show that the error is less than \(1\%\) for angles below \(15^\circ\text{.}\)
Here is the exact quote from Giancoli’s textbook:
“For angles less than \(15^\circ\text{,}\) the difference between \(\theta\) (in radians) and \(\sin\theta\) is less than \(1\%\text{.}\)”
Solution.
In deriving simple harmonic motion for a pendulum, the restoring force is \(F = -mg\sin\theta\text{,}\) which is proportional to \(\sin\theta\) rather than to the angular displacement \(\theta\) itself. The motion is therefore not exactly simple harmonic. Giancoli resolves this by appealing to the smallness of the angle: for angles below \(15^\circ\text{,}\) he states that “the difference between \(\theta\) (in radians) and \(\sin\theta\) is less than \(1\%\)” He notes that this can be seen “by looking at the series expansion” of \(\sin\theta\text{,}\) namely
\begin{equation*} \sin\theta = \theta - \frac{\theta^3}{3!} + \frac{\theta^5}{5!} - \frac{\theta^7}{7!} + \cdots . \end{equation*}
We can turn this observation into a quantitative statement using Taylor’s Remainder Theorem. Let \(f(\theta) = \sin\theta\text{,}\) expanded about \(\theta = 0\text{.}\) Replacing \(\sin\theta\) by \(\theta\) amounts to using the first-degree Taylor polynomial \(p_1(\theta) = \theta\text{.}\) Since the coefficient of \(\theta^2\) in the series above is zero, we have \(p_1(\theta) = p_2(\theta) = \theta\text{,}\) so the error is controlled by the third derivative: for some \(c\) with \(0 \lt c \lt \theta\text{,}\)
\begin{equation*} \sin\theta - \theta = R_2(\theta) = \frac{f'''(c)}{3!}\,\theta^3 = -\frac{\cos c}{6}\,\theta^3 . \end{equation*}
Because \(|\cos c| \le 1\text{,}\) the absolute error satisfies
\begin{equation*} \left| \sin\theta - \theta \right| = \frac{\cos c}{6}\,\theta^3 \le \frac{\theta^3}{6}, \end{equation*}
and dividing by \(\theta\) bounds the relative error:
\begin{equation*} \frac{\left| \theta - \sin\theta \right|}{\theta} \le \frac{\theta^2}{6}. \end{equation*}
This confirms Giancoli’s claim and even sharpens it. The relative error stays below \(1\%\) precisely when \(\dfrac{\theta^2}{6} \lt 0.01\text{,}\) that is, when \(\theta \lt \sqrt{0.06} \approx 0.245\) radians, or about \(14^\circ\text{.}\) At Giancoli’s stated cutoff of \(15^\circ = \pi/12\) radians, the bound gives
\begin{equation*} \frac{\left| \theta - \sin\theta \right|}{\theta} \le \frac{(\pi/12)^2}{6} \approx 0.0114, \end{equation*}
i.e. about \(1.1\%\text{.}\) So \(15^\circ\) is exactly the round-number boundary of the “less than \(1\%\)” regime, and Taylor’s Remainder Theorem locates the true \(1\%\) threshold at \(\theta \approx 14^\circ\text{.}\) This is precisely the range in which the pendulum behaves as a simple harmonic oscillator with period \(T = 2\pi\sqrt{\ell/g}\text{.}\)
Diagram Exploration Keyboard Controls
Key Action
Enter, A Activate keyboard driven exploration
B Activate menu driven exploration
Escape Leave exploration mode
Cursor down Explore next lower level
Cursor up Explore next upper level
Cursor right Explore next element on level
Cursor left Explore previous element on level
X Toggle expert mode
W Extra details if available
Space Repeat speech
M Activate step magnification
Comma Activate direct magnification
N Deactivate magnification
Z Toggle subtitles
C Cycle contrast settings
T Monochrome colours
L Toggle language (if available)
K Kill current sound
Y Stop sound output
O Start and stop sonification
P Repeat sonification output
Figure 2.18. The relative error \(\dfrac{|\theta - \sin\theta|}{\sin\theta}\) rises above the \(1\%\) threshold \(y = 0.01\) at \(\theta^* \approx 0.244\) radians (about \(14^\circ\)), the root of \(\theta = 1.01\sin\theta\text{.}\)
Figure 2.19. A pendulum swinging through decreasing amplitudes while the relative error of the approximation \(\sin\theta \approx \theta\) is computed at each angle, followed by the Taylor remainder bound \(\dfrac{|\theta - \sin\theta|}{\theta} \le \dfrac{\theta^2}{6}\) and the graph locating the \(1\%\) threshold at \(\theta^* \approx 0.244 \approx 14^\circ\text{.}\)

Subsection 2.4.3 Approximating the value of \(\pi\) using Taylor series

In this subsection, we will discuss how to approximate the value of \(\pi\) using Taylor series. We will use the Taylor series for arctan(x) to approximate \(\pi\text{.}\)

Example 2.20. Approximating \(\pi\) using the Taylor series for \(\arctan(x)\).

Use the Taylor series for \(\arctan(x)\) to approximate the value of \(\pi\text{.}\) Use Taylor’s Remainder Theorem to find an upper bound for the error in this approximation, and show that the error decreases as the order \(n\) of the Taylor polynomial increases.
Solution.
The Taylor series for \(f(x) = \arctan(x)\) centered at \(a = 0\) is
\begin{align*} \arctan(x) \amp = \sum_{k=0}^{\infty} \frac{(-1)^k}{2k+1}\, x^{2k+1}\\ \amp = x - \frac{x^3}{3} + \frac{x^5}{5} - \frac{x^7}{7} + \cdots, \qquad |x| \le 1. \end{align*}
Since \(\arctan(1) = \dfrac{\pi}{4}\text{,}\) evaluating the series at \(x = 1\) gives
\begin{equation*} \pi = 4\arctan(1) = 4\left(1 - \frac{1}{3} + \frac{1}{5} - \frac{1}{7} + \cdots\right). \end{equation*}
Therefore, if \(T_n(x)\) denotes the Taylor polynomial of order \(n\) for \(\arctan(x)\) centered at \(0\text{,}\) we obtain the approximation
\begin{equation*} \pi \approx 4\,T_n(1). \end{equation*}
To bound the error, we apply Taylor’s Remainder Theorem: if \(\left|f^{(n+1)}(t)\right| \le M\) for all \(t\) between \(0\) and \(x\text{,}\) then
\begin{equation*} |R_n(x)| = |f(x) - T_n(x)| \le \frac{M}{(n+1)!}\,|x|^{n+1}. \end{equation*}
For \(f(x) = \arctan(x)\text{,}\) the derivatives satisfy the closed form
\begin{equation*} f^{(n)}(t) = (n-1)!\,\cos^n(\theta)\,\sin\!\left(n\left(\theta + \tfrac{\pi}{2}\right)\right), \qquad \theta = \arctan(t), \end{equation*}
which can be verified by induction on \(n\text{.}\) Since \(|\cos^{n}(\theta)| \le 1\) and \(|\sin\!\left(n\left(\theta + \tfrac{\pi}{2}\right)\right)| \le 1\text{,}\) it follows that
\begin{equation*} \left|f^{(n+1)}(t)\right| \le n! \quad \text{for all } t \in \mathbb{R}, \end{equation*}
so we may take \(M = n!\text{.}\) With \(x = 1\text{,}\) Taylor’s Remainder Theorem gives
\begin{equation*} |R_n(1)| = \left|\arctan(1) - T_n(1)\right| \le \frac{n!}{(n+1)!}\,(1)^{n+1} = \frac{1}{n+1}. \end{equation*}
Multiplying by \(4\text{,}\) the error in the approximation of \(\pi\) satisfies
\begin{equation*} \left|\pi - 4\,T_n(1)\right| \le \frac{4}{n+1}. \end{equation*}
Since \(\dfrac{4}{n+1} \to 0\) as \(n \to \infty\text{,}\) the error decreases to zero as the order \(n\) of the Taylor polynomial increases. However, the convergence is very slow. For instance, with \(n = 9\) we get \(4\,T_9(1) \approx 3.339683\) with guaranteed error at most \(\tfrac{4}{10} = 0.4\text{,}\) and to guarantee an error of at most \(0.01\) one needs to take \(n \ge 399\text{.}\)
Figure 2.21. The approximation \(4\,T_n(1)\) converging to \(\pi\text{,}\) and the error \(\left|\pi - 4\,T_n(1)\right|\) decreasing below the upper bound \(\frac{4}{n+1}\) from Taylor’s Remainder Theorem as \(n\) increases.
As we saw in the previous example, Taylor series approached \(\pi\) very slowly. In the project below, we will see how to use the so-called Euler’s formula to approximate \(\pi\) much faster. As you may know, there are many other methods to approximate \(\pi\text{,}\) which we will not cover here.

Project 2.4.1. Computing \(\pi\) with Euler’s identity.

In Example 2.20 we approximated \(\pi\) by evaluating the Taylor series for \(\arctan(x)\) at \(x = 1\text{,}\) and Taylor’s Remainder Theorem gave the error bound \(\frac{4}{n+1}\text{,}\) which decreases very slowly. In this guided problem we compute \(\pi\) far more efficiently using Euler’s identity
\begin{equation*} \arctan\left(\tfrac{1}{2}\right) + \arctan\left(\tfrac{1}{3}\right) = \frac{\pi}{4}, \end{equation*}
which lets us evaluate the Taylor series at the small arguments \(x = \tfrac12\) and \(x = \tfrac13\text{,}\) where it converges much faster.

(a) Proving Euler’s identity.

Let \(\alpha = \arctan\left(\tfrac12\right)\) and \(\beta = \arctan\left(\tfrac13\right)\text{.}\) Use the addition formula
\begin{equation*} \tan(\alpha + \beta) = \frac{\tan\alpha + \tan\beta}{1 - \tan\alpha\,\tan\beta} \end{equation*}
to prove that \(\alpha + \beta = \dfrac{\pi}{4}\text{.}\)
Hint.
Compute \(\tan(\alpha+\beta)\) first. Then explain why \(\alpha + \beta\) must lie in the interval \(\left(0, \tfrac{\pi}{2}\right)\text{,}\) and why this pins down its value.
Solution.
Since \(\tan\alpha = \tfrac12\) and \(\tan\beta = \tfrac13\text{,}\)
\begin{equation*} \tan(\alpha+\beta) = \frac{\tfrac12 + \tfrac13}{1 - \tfrac12\cdot\tfrac13} = \frac{\tfrac56}{\tfrac56} = 1. \end{equation*}
Moreover \(0 \lt \tfrac13 \lt \tfrac12 \lt 1\) implies \(0 \lt \beta \le \alpha \lt \tfrac{\pi}{4}\text{,}\) so \(\alpha + \beta \in \left(0, \tfrac{\pi}{2}\right)\text{.}\) The only angle in this interval whose tangent equals \(1\) is \(\tfrac{\pi}{4}\text{,}\) so \(\alpha + \beta = \tfrac{\pi}{4}\text{.}\) (Without the interval check we could only conclude \(\alpha + \beta = \tfrac{\pi}{4} + k\pi\) for some integer \(k\text{.}\))

(b) The approximation.

Let \(T_n(x)\) be the Taylor polynomial of order \(n\) for \(\arctan(x)\) centered at \(0\text{.}\) Use Euler’s identity to explain why
\begin{equation*} \pi \approx 4\left[\,T_n\!\left(\tfrac12\right) + T_n\!\left(\tfrac13\right)\right]. \end{equation*}
Write out this approximation explicitly for \(n = 3\text{.}\)
Hint.
Multiply Euler’s identity by \(4\) and replace each arctangent by its Taylor polynomial.
Solution.
By Euler’s identity, \(\pi = 4\arctan\left(\tfrac12\right) + 4\arctan\left(\tfrac13\right)\text{.}\) Replacing \(\arctan(x)\) by \(T_n(x)\) at \(x=\tfrac12\) and \(x=\tfrac13\) gives the stated approximation. For \(n = 3\text{,}\) \(T_3(x) = x - \tfrac{x^3}{3}\text{,}\) so
\begin{align*} \pi \amp\approx 4\left[\left(\tfrac12 - \tfrac{1}{24}\right) + \left(\tfrac13 - \tfrac{1}{81}\right)\right]\\ \amp= 4\left[\tfrac{11}{24} + \tfrac{26}{81}\right] \approx 3.117284. \end{align*}

(c) Bounding the error with the Remainder Theorem.

In Example 2.20 we showed that the derivatives of \(f(x) = \arctan(x)\) satisfy \(\left|f^{(n+1)}(t)\right| \le n!\) for all \(t\text{.}\) Use Taylor’s Remainder Theorem to show that for \(|x| \le 1\)
\begin{equation*} \left|\arctan(x) - T_n(x)\right| \le \frac{|x|^{n+1}}{n+1}, \end{equation*}
and conclude that
\begin{equation*} \left|\pi - 4\left[\,T_n\!\left(\tfrac12\right) + T_n\!\left(\tfrac13\right)\right]\right| \le \frac{4}{n+1}\left[\left(\tfrac12\right)^{n+1} + \left(\tfrac13\right)^{n+1}\right]. \end{equation*}
Hint.
Apply Taylor’s Remainder Theorem with \(M = n!\) exactly as in Example 2.20, but keep the factor \(|x|^{n+1}\text{.}\) Then use the triangle inequality on the two remainders.
Solution.
Taylor’s Remainder Theorem with \(M = n!\) gives
\begin{equation*} |R_n(x)| \le \frac{n!}{(n+1)!}\,|x|^{n+1} = \frac{|x|^{n+1}}{n+1}. \end{equation*}
Writing \(\pi = 4\arctan\left(\tfrac12\right) + 4\arctan\left(\tfrac13\right)\) and using the triangle inequality,
\begin{align*} \left|\pi - 4\left[T_n\!\left(\tfrac12\right) + T_n\!\left(\tfrac13\right)\right]\right| \amp\le 4\left|R_n\!\left(\tfrac12\right)\right| + 4\left|R_n\!\left(\tfrac13\right)\right|\\ \amp\le \frac{4}{n+1}\left[\left(\tfrac12\right)^{n+1} + \left(\tfrac13\right)^{n+1}\right]. \end{align*}
Unlike the bound \(\frac{4}{n+1}\) at \(x = 1\text{,}\) this bound decays geometrically: each increase of \(n\) by one cuts it by better than half.

(d) How much better is it?

Evaluate the error bound for \(n = 9\) and compare it with the bound \(\tfrac{4}{10} = 0.4\) obtained in Example 2.20 for the same order. Then find the smallest \(n\) for which the bound guarantees an error of at most \(10^{-6}\text{.}\)
Hint.
For the second part, the term \(\left(\tfrac12\right)^{n+1}\) dominates; try increasing odd values of \(n\text{.}\)
Solution.
For \(n = 9\text{,}\)
\begin{equation*} \frac{4}{10}\left[\left(\tfrac12\right)^{10} + \left(\tfrac13\right)^{10}\right] = \frac{4}{10}\left[\frac{1}{1024} + \frac{1}{59049}\right] \approx 3.97 \times 10^{-4}, \end{equation*}
roughly a thousand times smaller than the bound \(0.4\) at \(x = 1\text{;}\) the approximation itself is \(4\left[T_9\!\left(\tfrac12\right) + T_9\!\left(\tfrac13\right)\right] \approx 3.141741\text{.}\) For an error of at most \(10^{-6}\text{,}\) testing odd values gives, at \(n = 15\text{,}\) a bound of about \(3.8\times10^{-6}\text{,}\) while at \(n = 17\)
\begin{equation*} \frac{4}{18}\left[\left(\tfrac12\right)^{18} + \left(\tfrac13\right)^{18}\right] \approx 8.5 \times 10^{-7} \lt 10^{-6}, \end{equation*}
so \(n = 17\) suffices. By contrast, the bound \(\frac{4}{n+1}\) from \(x = 1\) would require \(n \ge 3{,}999{,}999\) for the same guarantee.
The video below shows the approximation \(4\left[T_n\!\left(\tfrac12\right) + T_n\!\left(\tfrac13\right)\right]\) converging to \(\pi\text{,}\) and compares its error, together with the Remainder-Theorem bound, against the much slower method at \(x = 1\text{.}\)
Figure 2.22. The error of Euler’s-identity approximation (with its bound \(\frac{4}{n+1}\left[\left(\frac12\right)^{n+1} + \left(\frac13\right)^{n+1}\right]\)) decreasing geometrically as \(n\) increases, compared with the series at \(x=1\text{.}\)
You have attempted of activities on this page.