Home » Pi

Tag: Pi

Compute π using the Taylor polynomial of arctan x

For this exercise define

    \[ \alpha = \arctan \frac{1}{5}, \qquad \beta = 4 \alpha - \frac{1}{4} \pi. \]

  1. Using the trig identity

        \[ \tan (A+B) = \frac{\tan A + \tan B}{1 - \tan A \tan B} \]

    twice, once with A = B = \alpha, and then the second time with A = B = 2 \alpha, show that

        \[ \tan (2 \alpha) = \frac{5}{12}, \qquad \tan (4 \alpha) = \frac{120}{119}. \]

    Then use the same identity again with A = 4 \alpha and B = -\frac{1}{4} \pi to show

        \[ \tan \beta = \frac{1}{239}. \]

    This establishes the identity

        \[ \pi = 16 \arctan \frac{1}{5} - 4 \arctan \frac{1}{239}. \]

  2. Using the Taylor polynomial approximation T_{11} (\arctan x) at x =\frac{1}{2} prove that

        \[ 3.158328934 < 16 \arctan \frac{1}{5} < 3.158328972. \]

  3. Using the Taylor polynomial approximation T_3 (\arctan x) at x = \frac{1}{239} prove that

        \[ -0.016736309 < -4\arctan \frac{1}{239} < -0.016736300. \]

  4. Using the above parts show that the value of \pi to seven decimal places is 3.1415926.

  1. Proof. Letting A = B = \alpha = \arctan \frac{1}{5} we have

        \[ \tan (A+B) = \tan (2 \alpha) = \frac{\tan A + \tan B}{1 - \tan A \tan B} = \frac{(2/5)}{1 - (1/25)} = \frac{5}{12}. \]

    Letting A = B = 2 \alpha = 2 \arctan \frac{1}{5} we have

        \[ \tan (A+B) = \tan (4 \alpha) = \frac{\tan A + \tan B}{1 - \tan A \tan B} = \frac{(10/12)}{1 - (25/144)} = \frac{120}{119}. \]

    Letting A = 4 \alpha and B = -\frac{\pi}{4} we have (recalling that \beta = 4 \alpha - \frac{\pi}{4})

        \[ \tan (\beta) = \frac{\tan (4\alpha) + \tan\left(-\frac{\pi}{4}\right)}{1 - \tan (4\alpha) \tan \left( -\frac{\pi}{4} \right)} = \frac{(120/119) - 1}{1 - (120/119)(-1)} = \frac{1}{239}. \]

    But then

        \begin{align*}  && \tan \left( 4 \alpha - \frac{\pi}{4} \right) &= \frac{1}{239}\\[9pt] \implies && 4 \alpha - \frac{\pi}{4} &= \arctan \left( \frac{1}{239} \right) \\[9pt] \implies && \frac{\pi}{4} &= 4 \arctan \frac{1}{5} - \arctan \frac{1}{239} \\[9pt] \implies && \pi &= 16 \arctan \frac{1}{5} - 4 \arctan \frac{1}{239}. \qquad \blacksquare \end{align*}

  2. Proof. We know the Taylor polynomial approximation for \arctan x from this exercise (Section 7.8, Exercise #3):

        \[ T_{11} (\arctan x) = x - \frac{x^3}{3} + \frac{x^5}{5} - \frac{x^7}{7} + \frac{x^9}{9} - \frac{x^{11}}{11}. \]

    Therefore, we can compute an approximation to \arctan x,

        \begin{align*}  &&\arctan x &= \left( x - \frac{x^3}{3} + \frac{x^5}{5} - \frac{x^7}{7} + \frac{x^9}{9} - \frac{x^{11}}{11} \right) + E_{10}(x) \\[9pt] \implies && 16 \arctan \left(\frac{1}{5} \right) &= 16 \left( \frac{1}{5} - \frac{1}{375} + \frac{1}{15625} - \frac{1}{546875} + \frac{1}{17578125} - \frac{1}{537109375} \right) + E_{10}\left( \frac{1}{5} \right)\\[9pt] \implies && 16 \arctan \left( \frac{1}{5} \right) &= 3.158328957 + E_{10} \left( \frac{1}{5} \right), \end{align*}

    where

        \[ \left|E_{10}\left(\frac{1}{5} \right)\right| \leq \frac{(1/5)^{11}}{11} = .000000018. \]

    Therefore,

        \[ 3.158328934 < 16 \arctan \left( \frac{1}{5} \right) < 3.158328972. \qquad \blacksquare\]

  3. Proof. Again using the Taylor polynomial approximation to \arctan x we have

        \begin{align*}  &&T_3 (\arctan x) &= x - \frac{x^3}{3} \\[9pt] \implies && \arctan x &= x - \frac{x^3}{3} + E_2 (x), \qquad |E_2 (x)| \leq \frac{x^3}{3} \\[9pt] \implies && \arctan \left( \frac{1}{239} \right) &= \left( \frac{1}{239} \right) - \frac{(1/239)^3}{3} + E_2 \left( \frac{1}{239} \right) \\[9pt] \implies && -4 \arctan \left( \frac{1}{239} \right) &= -0.01673630401 + E_2 \left( \frac{1}{239} \right) \\[9pt] \implies && -0.016736309 &< -4 \arctan \left( \frac{1}{239} \right) < -0.016736300. \qquad \blacksquare \end{align*}

  4. Finally,

        \begin{align*}   \pi &= 16 \arctan \frac{1}{5} - 4 \arctan \frac{1}{239} \\  &\approx 3.158328957 - 0.01673630401 \\  &\approx 3.1415926.  \end{align*}

Use integrals to establish the formula for the area of an ellipse

Denoting the unit circle, x^2 + y^2 = 1, by C we define an ellipse E to be the set of points

    \[ E = \{ (ax, by) \mid (x,y) \in C,  a > 0,  b > 0 \}. \]

  1. Show that the points on E satisfy the equation:

        \[ \left( \frac{x}{a} \right)^2 + \left( \frac{y}{b} \right)^2 = 1. \]

  2. Prove that the area of E is measurable and that

        \[ a(E) = \pi a b. \]


  1. Proof. If (x,y) is a point on E then \left( \frac{x}{a}, \frac{y}{b} \right) is a point on C (since all points of E are obtained by taking a point of C and multiplying the x-coordinate by a and the y-coordinate by b). By definition of C, we must then have,

        \[ \left( \frac{x}{a} \right)^2 + \left(\frac{y}{b} \right)^2 = 1. \qquad \blacksquare \]

  2. Proof. From part (a) we know E is the set of points (x,y) such that \left( \frac{x}{a} \right)^2 + \left( \frac{y}{b} \right)^2 = 1. But, this implies,

        \[ y = b \cdot \sqrt{ 1 - \left( \frac{x}{a} \right)^2}, \qquad \text{or} \qquad y = -b \cdot \sqrt{1 - \left( \frac{x}{a} \right)^2}. \]

    Hence, the area of E is the area enclosed from -a to a by the graphs of

        \[ g(x) = b \cdot \sqrt{1 - \left( \frac{x}{a} \right)^2} \qquad \text{and} \qquad f(x) = -b \cdot \sqrt{1 - \left( \frac{x}{a} \right)^2}. \]

    To show this region is measurable and has area \pi a b we begin with this identity (from Apostol, 2.4 Exercise 17),

        \begin{align*}  \pi = 2 \int_{-1}^1 \sqrt{1-x^2} && \implies && \pi b &= 2 \int_{-1}^1 b \sqrt{1-x^2} \, dx \\ && \implies && \pi a b &= 2a \int_{-1}^1 b \sqrt{1-x^2} \, dx \\ && \implies && \pi a b &= 2 \int_{-a}^a b \sqrt{1 - \left( \frac{x}{a} \right)^2} \, dx \\ && \implies && \pi ab &= \int_{-a}^a \left( b \sqrt{1 - \left( \frac{x}{a} \right)} - \left(-b \sqrt{1 - \left( \frac{x}{a} \right)^2} \right)\right) \, dx. \end{align*}

    In the second to last line we have used the expansion/contraction of the interval of integration. Hence, we know the integral from -a to a of g(x) - f(x) exists and has value \pi a b. Thus, E is measurable and a(E) = \pi a b. \qquad \blacksquare

Use dodecagons to deduce an inequality about π

By considering dodecagons inscribed and circumscribed about a unit disk, establish the inequalities

    \[ 3 < \pi < 12(2-\sqrt{3}). \]


First, we draw some pictures of the situation for reference.

Rendered by QuickLaTeX.com

( Note: I don’t know a way to do this without using trig functions, which haven’t been introduced in the text yet. If you have an alternative approach without them, please leave a comment and let us know about it. )

Since these are dodecagons, the angle at the origin of the circle of each triangular sector is 2 \pi / 12 = \pi/6, and the angle of the right triangles formed by splitting each of these sectors in half (shown in the diagrams) is then \pi/12. Then we use the fact that

    \[ \tan \left( \frac{\pi}{12} \right) = 2 - \sqrt{3}, \]

    \[ \sin \left( \frac{\pi}{12} \right) = \frac{\sqrt{3} - 1}{2 \sqrt{2}}, \]

    \[ \cos \left( \frac{\pi}{12} \right) = \frac{\sqrt{3} + 1}{2 \sqrt{2}}. \]

Now, for the circumscribed dodecagon we have the area of the right triangle T with base 1 in the diagram on the left given by

    \[a(T) = \frac{1}{2} bh = \frac{1}{2} \cdot 1 \cdot (2 - \sqrt{3}) = 1 - \frac{\sqrt{3}}{2}. \]

Since there are 24 such triangles in the dodecahedron, we then have the area of the circumscribed dodecahedron D_c given by

    \[ a(D_c) = 24 \left(1 - \frac{\sqrt{3}}{2} \right) = 12 (2 - \sqrt{3}). \]

For the inscribed dodecagon we consider the right triangle T with hypotenuse 1 in the diagram. The length of one of the legs is then given by \sin \left( \frac{\pi}{12} \right) = \frac{\sqrt{3} - 1}{2} and the other is given by \cos \left( \frac{\pi}{12} \right). So the area of the triangle is

    \[ a(T) = \frac{1}{2} bh = \frac{1}{2} \cdot \frac{\sqrt{3}-1}{2 \sqrt{2}} \cdot \frac{\sqrt{3}+1}{2 \sqrt{2}} = \frac{2}{16} = \frac{1}{8}.\]

Since there are 24 such triangles in the inscribed dodecahedron, D_{i} we then have,

    \[ a(D_i) = 24 \cdot \frac{1}{8} = 3. \]

Since the area of the unit circle is, by definition, \pi, and it lies in between these two dodecahedrons, we have,

    \[ 3 < \pi < 12(2 - \sqrt{3}). \qquad \blacksquare \]

Compute some integrals in terms of π

Recall we have defined \pi as the area of the unit disk. We have proved (Apostol, Section 2.3, Example 3) that

    \[ \pi = 2 \int_{-1}^1 \sqrt{1-x^2} \, dx. \]

Use this and the theorems on the property of integrals to compute the following in terms of \pi:

  1. \displaystyle{\int_{-3}^3 \sqrt{9-x^2} \, dx}.
  2. \displaystyle{\int_0^2 \sqrt{1-\frac{1}{4}x^2} \, dx}.
  3. \displaystyle{\int_{-2}^2 (x-3) \sqrt{4-x^2} \, dx}.

  1. Using the expansion/contraction property of the integral (Apostol, Theorem 1.19) we compute:

        \begin{align*}   \int_{-3}^3 \sqrt{9-x^2} \, dx &= 3 \int_{-1}^1 \sqrt{9 - (3x)^2} \, dx & (\text{Exp/Cont property}) \\   &= 9 \int_{-1}^1 \sqrt{1-x^2} \, dx & (\text{Factoring out a constant}) \\   &= \frac{9}{2} \pi &(\text{Expression for } \pi \text{ given}).  \end{align*}

  2. Using the expansion/contraction property of the integeral (Apostol, Theorem 1.19) we compute:

        \begin{align*}   \int_0^2 \sqrt{1 - \frac{1}{4}x^2} \, dx &= 2 \int_0^1 \sqrt{1 - x^2} \, dx \\   &= \int_{-1}^1 \sqrt{1-x^2} \, dx \\   &= \frac{\pi}{2}  \end{align*}

  3. Using the linearity properties of the integral, as well as, expansion/contraction and the formula for \pi given above. We also use this exercise (Apostol, Section 1.25, Exercise #26 part (b)) which establishes that the integral from -b to b of an odd function is 0.

        \begin{align*}  \int_{-2}^2 (x-3) \sqrt{4 - x^2} \, dx &= \int_{-2}^2 x \sqrt{4-x^2} \, dx - 3 \int_{-2}^2 \sqrt{4-x^2} \, dx \\  &= - 6 \int_{-1}^1 \sqrt{4-4x^2} \, dx & (x \sqrt{4-x^2} \text{ is an odd function})\\   &= -12 \int_{-1}^1 \sqrt{1-x^2} \, dx \\  &= -6 \pi. \end{align*}