I was hanging around with circles when I did the following "workout" :

(Sorry for not including any diagram of my construction. I could not find any suitable software for drawing. I have tried my best to describe each and every step so that you can follow along.)

Construction : I took a circle with $O$ as center and $\overline{AB}$ as diameter. The perpendicular bisector of $AB$ meets the "upper" arc $\overline{AB}$ at $C$. Hence, $\overline{OC}$ divides the upper hemisphere into two quadrants. Next I constructed $\angle OCD = 60^o$, $D$ lying in the quadrant $BOCB$ and on the circumference. This gives me an equilateral triangle $\triangle OCD$. I joined $\overline{AD}$ and $\overline{BD}$.

Method :

Let $r$ be the radius of the circle.

As $\overline {OD}=\overline {OB}=r$, so $\angle DBO=\angle ODB$. But $\angle DOB = 90^o-60^o = 30^o$. So $\angle DBO = \angle ODB=75^o$.

Using Sine Rule in $\triangle BDO$ :

$$\frac{\sin 30^0}{\overline {BD}} = \frac {\sin 75^o}{r}$$

$$\implies \overline{BD}=\frac {r}{2} \csc 75^o$$


$$\overline {AC}+\overline{CD}+\overline{BD} < \frac {2\pi r}{2}$$

By Pythagorean Theorem, $\overline{AC}=r\sqrt{2}$.

$$\therefore r\sqrt{2}+r+\frac{r}{2}\csc 75^o < \pi r$$

Cancelling out the common factor $r$ :

$$\sqrt{2}+1+\frac{\csc 75^o}{2} < \pi$$


$$\pi > 2.9318516525781364$$

My thoughts : This "workout" gives a very "broad" approximation for $\pi$. I was overjoyed when I derived this result myself, but I am looking for a better approximation. So here is/are my question(s) :

Is there a way to refine/fine-tune/improve my approximation by using the same geometric setup ? Can $\pi$ be approximated ("in the same diagram") without using trigonometry? Is my "method" something which already has been "done" or tested ?

Any help will be gratefully appreciated.

Thanks in advance ! :)

  • 1
    $\begingroup$ You could simply use $\pi>6\overline{BD}$. $\endgroup$ Feb 9 '17 at 20:38
  • 1
    $\begingroup$ As $\angle DOB=30°$, $BD$ is the side of a regular dodecagon inscribed in the circle, so $12 BD<2\pi r$. This gives a better approximation. $\endgroup$ Feb 10 '17 at 7:52
  • 1
    $\begingroup$ Now I realise the above is exactly what wrote @markfischler in his answer. $\endgroup$ Feb 10 '17 at 7:55
  • 1
    $\begingroup$ For the upper bound, just think of the circumscribed dodecagon, for instance. $\endgroup$ Feb 10 '17 at 7:58
  • 1
    $\begingroup$ And of course, you can go up to a $24$-gon, and more. Using trig bisection formulas to have an exact result. $\endgroup$ Feb 10 '17 at 8:00

A first step in refinement is to construct point $E$ on the circumference in quadrant $AOCA$ such that angle $COE$ is $60^\circ$. And draw line segments $\overline{OE}$, $\overline{AE}$ and $\overline{CE}$.

Then, by the same trig reasoning you applied above, $$ \overline{AE} = \frac r2 \csc(75^\circ) = \frac{\sqrt{2}(\sqrt{3}-1)}{2}\\ \overline{EC} = r\\ \overline{CD} = r\\ \overline{BD} = \frac r2 \csc(75^\circ) = \frac{\sqrt{2}(\sqrt{3}-1)}{2} $$ And again the same form of reasoning you used above gives $$ \pi > \frac12 \csc(75^\circ) = \frac{\sqrt{2}(\sqrt{3}-1)}{2} + 1 + 1 + \frac12 \csc(75^\circ) + \frac{\sqrt{2}(\sqrt{3}-1)}{2} $$ or $$\pi > 2 + \sqrt{2}(\sqrt{3}-1) \approx 3.03528 $$ Next, you can place points $F$ and $G$ on the midpoints of arcs $DC$ and $EC$ respectively, and draw line segments $\overline{EG}$, $\overline{GC}$, $\overline{CF}$, and $\overline{FB}$. The same trig reasoning then says $$ \pi > 6 \frac{\sqrt{2}(\sqrt{3}-1)}{2}\approx 3.10583 $$ This is the method Archimedes used; it inscribes a regular $12$-sided figure in the circle, and observes that the circumference of the circle exceeds the perimeter of the dodecagon.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy