Interesting math problem that I saw online

  • Level: Undergrad 
  • Thread starter Charles Link
  • Start date
  • Featured
84 replies · 11K views · last reply
I understand there is no general formula for higher order figures inscribed in a circle in terms of the radius as a function of side lengths. At least not without high powered math and special functions. But still we can use the equation above to numerically solve it for higher order with the caveat that for an ##n##-sided figure, the longest side is less than the sum of the other sides. ##c_{n}<\sum _{i=1}^{n-1}c_{i}##

We can start with the general formula under the assumption the given sides make a closed figure inscribed by a circle.

##\sum _{i=1}^{n}\arcsin \left(\frac{c_{i}}{2R}\right)=\pi ##


Depending on the side lengths, there are two possible cases for the center of the circumscribed circle:
  1. Polygon includes the center of the circle (Center Inside)
  2. Center of the circle lies outside the polygon (Center Outside)
We can check this by calculating the Critical Radius ##R_{\text{crit}}=\frac{L_{n}}{2}## (Where ##L_{n}## is the longest side). Then we calculate the angle_sum defined as

##A=\sum _{i=1}^{n-1}\arcsin \left(\frac{c_{i}}{2R_{\text{crit}}}\right)##

Then,
  • If ##A >\frac{\pi}{2}##: The center is inside the polygon.
  • If ##A < \frac{\pi}{2}##: The center is outside the polygon.
  • If ##A = \frac{\pi}{2}##: The longest side ##L## is a literal diameter of the circle.

We can then use a solver (I used Desmos) to compute the intersection of two functions as follows;
  • If Case 1 (Inside):
    ##y=\sum _{i=1}^{n}\arcsin \left(\frac{c_{i}}{2R}\right)\quad \text{and}\quad y=\pi ##
  • If Case 2 (Outside):
    ##y=\sum _{i=1}^{n-1}\arcsin \left(\frac{c_{i}}{2R}\right)\quad \text{and}\quad y=\arcsin \left(\frac{c_{n}}{2R}\right)##
    (Where ##c_{n}##is the longest side)


 
  • Like
Likes   Reactions: anuttarasammyak
Here is an example for the case (2,3,4,5,6). In this case the sum angle is 2.578 rads which is greater than ##\frac{\pi}{2}## so the circle center is inside the polygon.

Plot of a cyclic pentagon with vertices A (3.52, 0.00), B (-1.59, 3.14), C (-3.13, -1.62), D (0.40, -3.50) and E (2.95, -1.92) inscribed in a dashed circle centred at the origin, with the centre marked inside the pentagon.


We use the equations from case 1 above to calculate the intersection.
IMG_6700.webp



And here is a case for (2,3,4,5,11). In this case the sum angle is 1.063 rads which is less than ##\frac{\pi}{2}## so the circle center is outside the polygon.

Plot titled Inscribed Polygon with Chord Lengths 2, 3, 4, 5, 11: five vertices V1 to V5 on a dashed circle centred near the origin, joined by chords labelled 2, 3, 4, 5 and 11, with the 11 chord running from V5 to V1 and passing below the origin.


We use the equations from case 2 above to calculate the intersection.
Line graph titled Central Angle Curves Intersection for Chord Lengths, plotting central angle in degrees against radius R from 5.5 to 9.0. A blue curve for the large angle and an orange curve for the sum of small angles cross at R about 5.86, theta about 139.8 degrees.
 
Last edited:
  • Like
Likes   Reactions: anuttarasammyak
anuttarasammyak said:
There are m rods length of each does not surpass sum of the others. With these rods we can make m polygons which has a circumcircle. It seems to hold.
Sure. Shown below is a quadrilateral in which one side (green) is equal to the sum of the other three.

Horizontal bar divided into four colored segments in sequence: red, blue, magenta, and a green strip running along the full length beneath them.


There is no circle that passes through all 4 points where the color changes.

If you shorten the green segment by ##\epsilon## keeping the other three at the same length, there will be a closed quadrilateral and a circle passing through all 4 points.

If you lengthen the green segment by ##\epsilon## keeping the other three at the same length, you will no longer have a closed polygon.

Example
The segments shown above are
Red = 1.0 u
Blue =1.6 u
Magenta = 2.2 u
Green = 4.8 u

Using the formula for the radius of the circumscribed circle, I found that if I shorten the green segment by ε = 1×10-6 u,
##R\approx2,000~##u and that the angle between green and red is 0.04° and between blue and magenta is 179.96°.
 
anuttarasammyak said:
In generalc case of m polygon
∑i=1msin−1⁡Lid=π
seems to hold with choice of parameter d for any L1,L2,..,Lm. So there seems to exist a circumcircle.
I find the above is for polygon which includes centern of circumcircle in it. As shown in #61 including polygon which does not include the center, the condition is
$$(\sum_{i=1}^m \sin^{-1}\frac{f(L_i)L_i}{d})(\sum_{i=1}^m \sin^{-1}\frac{L_i}{d}-\pi)=0$$
defined for d > the longest ##L_i## and ##f(L_i) = -1## for the longest rod, 1 otherwise. When we have plural longest rods, any only one of them should be chosen.

As far as {Li} can form m-polygons, there exist some of them which have circumcircle whose diameter d would satisfy above equation,i.e., have circumicircle.

Proof
Say ##L_1 = max\{L_i\}##
case
$$\sum_{i=1}^m\sin^{-1}\frac{L_i}{L_1}-\pi>0$$
$$f(d):=\sum_{i=1}^m\sin^{-1}\frac{L_i}{d}-\pi$$ is continuous monotonous decreasing function to ##-\pi## as d increase from ##L_1##. Thus there exists unike value d which satisfies f(d)=0,i.e.,
$$\sum_{i=1}^m\sin^{-1}\frac{L_i}{d}-\pi=0$$

case
$$\sum_{i=1}^m\sin^{-1}\frac{L_i}{L_1}-\pi<0$$
$$\sum_{i=2}^m\sin^{-1}\frac{L_i}{L_1}-\sin^{-1}\frac{L_1}{L_1}<0$$
$$g(d):=\sum_{i=2}^m\sin^{-1}\frac{L_i}{d}-sin^{-1}\frac{L_1}{d}<0$$
for ##d = L_1+0##. By expanding g(d) by 1/d
$$g(d)=\frac{\sum_{i=2}^m L_i -L_1}{d}+o(1/d^2) > 0 $$
Thus there exists value d which satisfies g(d)=0,i.e..
$$\sum_{i=2}^m\sin^{-1}\frac{L_i}{d}-\sin^{-1}\frac{L_1}{d}=0$$

QED
 
Last edited:
kuruman said:
Well, since the line joining P2 and P4 is a diameter, the triangles on either side of it are right triangles no matter where you put P1 and P3. Referring to the diagram in post #51, the Pythagorean theorem says, $$b^2+c^2=a^2+d^2=4R^2\implies R=\frac{1}{2}\sqrt{b^2+c^2}= \frac{1}{2}\sqrt{a^2+d^2}. $$ I agree, the problem isn't hard. What is the surprise?
Using the formula of post 22 for ## r ##, I found using ## a^2+d^2=b^2+c^2 ## that the algebra is non-trivial to show that ## 4r^2=a^2+d^2=b^2+c^2 ##, but I got it to work. Perhaps there is a shortcut to the algebra. Of course there is also Thales' theorem that it had to be the case.
 
I have a follow-on to the 2,3,4,5 that I came up with yesterday. The 2,3,4,5 turns out to have a radius of ## r=\sqrt{986.7}/12 ##. The 5 turns out to show up on the far half of the circle. For a new case, take sides ## x,2,4 ## and ## 5 ##,( instead of ## 3,2,4 ## and ## 5 ##), with the same radius, with the ## 5 ## this time on the same half of the circle as the other 3 sides, and find ## x ##. I believe I solved it correctly. The solution that I have is not tremendously difficult, (about 5 or 6 steps), and I think it makes for an interesting problem.
 
@Charles Link the general solution is
$$r=\sqrt{\frac{(ab+cd)(ac+bd)(ad+bc)}{(b+c+d-a)(c+d+a-b)(d+a+b-c)(a+b+c-d)}}$$
For the case of a,b,c,d=2,3,4,5
$$r=\frac{1}{4}\sqrt{\frac{11*13*23}{30}}$$
It seems to be same as your solution,
 
Last edited:
@anuttarasammyak I got it correct. What I'm asking for now is to find ## x ## of post 66. I computed it and got ## x \approx .043 ##. For a hint or two on solving it, draw the two 5's meeting at the right side of the "stick" of length 4. Then draw in the ## x ## and the 3, having those meet at the left side of the "stick" of length 2. I wish I cold draw the picture, but I have zero computer graphics skills. The 5's and the ## x ## and 3 subtend identical angles. The angle between the 5's is readily found. It remains to use the law of cosines to solve for ## x ## with the ## x ## and the 3 as the two sides in the law of cosines formula.
 
Let a,b,c,d=x,2,4,5 and r=##\frac{1}{4}\sqrt{\frac{11*13*23}{2*3*5}}##, x=3 is unike solution. I may be misunderstanding the problem which @Charles Link posts.
 
Last edited:
anuttarasammyak said:
Let a,b,c,d=x,2,4,5 and r=##\frac{1}{4}\sqrt{\frac{11*13*23}{2*3*5}}##, x=3 is unike solution. I may be misunderstanding the problem which @Charles Link posts.
## x=3 ## works, but find ## x ## for the case with the same ## r ## where the 5 is no longer on the far side of the circle. I wish I could draw a picture for you. @bob012345 Hopefully you understood what I was asking.
 
  • Like
Likes   Reactions: anuttarasammyak
I wanted to post this last night but was looking for a more exact solution but here it is. Using case2 from Post #61

##y = \sum_{i=1}^{n-1} \arcsin\left(\frac{c_i}{2R}\right)## and ##y = \arcsin\left(\frac{c_n}{2R}\right)##

where ##c_n## is the longest side which is ##5##.

##\therefore y = \arcsin\left(\frac{2}{2R}\right) + \arcsin\left(\frac{4}{2R}\right) + \arcsin\left(\frac{x}{2R}\right) = \arcsin\left(\frac{5}{2R}\right)##

So ##\arcsin\left(\frac{x}{2R}\right) = \arcsin\left(\frac{5}{2R}\right) - \arcsin\left(\frac{4}{2R}\right) - \arcsin\left(\frac{2}{2R}\right)##

Taking the sine of both sides:

##x = 2R \sin\left[\arcsin\left(\frac{5}{2R}\right) - \arcsin\left(\frac{4}{2R}\right) - \arcsin\left(\frac{2}{2R}\right)\right]##

##x = 2R \sin\left[1.2698475 - 0.869559 - 0.3919836\right]##

##x = 0.043478 ##

I’ll try and draw this a bit later.
 
  • Like
Likes   Reactions: Charles Link and anuttarasammyak
Charles Link said:
## x=3 ## works, but find ## x ## for the case with the same ## r ## where the 5 is no longer on the far side of the circle. I wish I could draw a picture for you. @bob012345 Hopefully you understood what I was asking.
Yes, this is a case where the center of the circle is outside the figure.
 
  • Like
Likes   Reactions: Charles Link
Charles Link said:
## x=3 ## works, but find ## x ## for the case with the same ## r ## where the 5 is no longer on the far side of the circle. I wish I could draw a picture for you. @bob012345 Hopefully you understood what I was asking.
There are two values of ##x## that fulfill the equation. Consider the following construction with straight edge and compass to find them.
  1. PolygonGeo.webp
    Draw a circle of radius ##R=\frac{\sqrt{986.7}}{12}.##
  2. Mark point A anywhere on the circumference. Open the compass to radius ##5## and draw an arc centered at A that intersects the circle at point B. Connect A and B with a straight line to get the first side AB = ##5##.
  3. Open the compass to radius ##4##. A circle of radius 4 centered at point A has two intersection points with the circle of radius ##R##. Choose one of the two intersection points and call it C. One choice puts the center of the circle inside the quadrilateral, the other puts it outside. Connect points A and C with a straight line to get the second side AC = ##4.##
  4. Open the compass to radius ##2##. A circle of radius 3 centered at point B has two intersection points with the circle of radius ##R##. Select the one such that a line connecting it with point C does not intersect side AB and call it D . Connect points B and D with a straight line to get the third side BD = ##2.##
  5. Connect points C and D to get the first value of the fourth side ##x##.
  6. Repeat steps 4 and 5 with the second choice of intersection point in step 3.
 
kuruman said:
There are two values of ##x## that fulfill the equation. Consider the following construction with straight edge and compass to find them.
  1. View attachment 375397
I literally was just about to post almost the exact same diagram. The only thing I can add then is there are more cases. Hint: they involve self interactions.
 
bob012345 said:
I literally was just about to post almost the exact same diagram. The only thing I can add then is there are more cases. Hint: they involve self interactions.
"Self interactions" doesn't say much to me. I have constructed a quadrilateral inside a given circle with a straight edge and compass and argued in step 3 that the construction branches out to two separate solutions. If you look at my drawing, once you draw side AB, you must draw AC and BD either above the AB to include the center of the circle or below AB to exclude it. Connecting the free ends C and D completes the construction. It should be possible to construct any additional solution you may have with a straight edge and compass. Please show me how.
 
kuruman said:
"Self interactions" doesn't say much to me. I have constructed a quadrilateral inside a given circle with a straight edge and compass and argued in step 3 that the construction branches out to two separate solutions. If you look at my drawing, once you draw side AB, you must draw AC and BD either above the AB to include the center of the circle or below AB to exclude it. Connecting the free ends C and D completes the construction. It should be possible to construct any additional solution you may have with a straight edge and compass. Please show me how.


I meant like this. Here ##x=\large\frac{41}{11}=3.72727..##


IMG_6860.webp
 
@bob012345 You got the 5 in the right place,but not the 2. Suggestion is to put the other 5 in there as well, and draw two cases: one with an ## x ##, (very small), and the other with the 3 going to the other 5.

For @kuruman It helps to put the 2 next to the 4. Then you only need one 2, one 4, two 5's, an ## x ##, and one 3.
 
@bob012345 Looks like you got the same answer I did in post 72, thank you, but there is an easier way to do it. If one of you can draw up the cases I described above, with the two 5's (with a point in common, each meeting at the right end of the 4), one 2, one 4, an ## x ##, and a 3, it would make my explanation easier to follow.
 
Last edited:
Charles Link said:
@bob012345 Looks like you got the same answer I did in post 72, thank you, but there is an easier way to do it. If one of you can draw up the cases I described above, with the two 5's (with a point in common, each meeting at the right end of the 4), one 2, one 4, an ## x ##, and a 3, it would make my explanation easier to follow.
This is what you want but it’s hard to see the detail so I include a blowup of the region of small ##x##
IMG_6864.webp
IMG_6865.webp
 
Charles Link said:
@bob012345 You got the 5 in the right place,but not the 2. Suggestion is to put the other 5 in there as well, and draw two cases: one with an ## x ##, (very small), and the other with the 3 going to the other 5.

In post #77, I was giving a new solution.
 
@bob012345 Post 80 is great, but I almost need ## x ## to be about one unit across, so that you can see the two angles that are identical. The two 5's form an angle that cover the same section of the circle as the angle going from the 3 to the second 5, (originating at the end of the 2), and the line going from the ## x ## to the first 5, (also originating at the end of the 2. These two angles are identical, and once you solve the ## \cos{\theta} ## for the 5's, it is a simple matter to solve for ## x ## setting the two expressions for the law of cosines the same, with the second expression involving the 3 and the ## x ##. The ## x ## if drawn to scale is too small to see this with any kind of diagram.
 
Charles Link said:
x=3 works, but find x for the case with the same r where the 5 is no longer on the far side of the circle. I wish I could draw a picture for you. @bob012345 Hopefully you understood what I was asking.
Thanks. I think I have got it.
From the general solution we get the eqation
$$11*13*23(x+1)(x+3)(11-x)(x+7)-64*30(x+10)(5x+8)(2x+5)=0$$
$$(x-3)(11x+41)(13x+67)(23x-1)=0$$
$$x=3,\frac{1}{23},-\frac{11}{41},-\frac{13}{67}$$
Positive answers are
$$x=3,\frac{1}{23}$$
 
  • Informative
Likes   Reactions: Charles Link
@anuttarasammyak I believe you began with the general solution from post 68, and if I understood it, your solution is great. I thought I had a fairly good solution, but yours appears to be at least one better. We all seem to be getting the same answer, about .043, but yours is exact. Very good. Excellent. :)

and it took me about fifteen minutes to do the algebra and math to compare your simple exact solution to my complicated exact solution, and yes, the two are identical. :)
 
Last edited:
  • Like
Likes   Reactions: anuttarasammyak
bob012345 said:
I meant like this. Here ##x=\large\frac{41}{11}=3.72727..##


View attachment 375398
Sure. I thought that self-intersecting polygons were to be excluded, that is why I excluded them in my construction.