Maximum Circumcircle

Geometry Level 4

Given a triangle of area 10 10 and perimeter 16 16 , find the largest possible circumradius, R max R_\text{max} and submit your answer as 1 0 6 R max \lfloor 10^6 \cdot R_\text{max}\rfloor .


The answer is 3551815.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

2 solutions

Chris Lewis
Apr 12, 2021

Call the triangle's area T T and its semiperimeter s s . Let its sidelengths be a , b , c a,b,c (as usual opposite vertices A , B , C A,B,C respectively).

We know T T and s s ; let's make a a a parameter. By definition, b + c = 2 s a b+c=2s-a

Now, from Heron's formula, s ( s a ) ( s b ) ( s c ) = T 2 s ( s a ) ( s 2 ( b + c ) s + b c ) = T 2 s ( s a ) ( s 2 ( 2 s a ) s + b c ) = T 2 s ( s a ) ( b c s ( s a ) ) = T 2 b c = s ( s a ) + T 2 s ( s a ) \begin{aligned} s(s-a)(s-b)(s-c) &=T^2 \\ s(s-a)\left( s^2-(b+c)s+bc\right) &= T^2 \\ s(s-a)\left( s^2-(2s-a)s+bc\right) &= T^2 \\ s(s-a)( bc - s(s-a)) &= T^2 \\ bc &= s(s-a)+\frac{T^2}{s(s-a)} \end{aligned}

Knowing b + c b+c and b c bc means we can find b b and c c as the roots of a quadratic (see below).

The circumradius is given by R = a b c 4 T = 1 4 T ( a s ( s a ) + a T 2 s ( s a ) ) R=\frac{abc}{4T}=\frac{1}{4T} \left(as(s-a)+\frac{aT^2}{s(s-a)} \right)

In this case, s = 8 s=8 and T = 10 T=10 ; so R = 1 40 ( 8 a ( 8 a ) + 25 a 2 ( 8 a ) ) R=\frac{1}{40} \left(8a(8-a)+\frac{25a}{2(8-a)} \right)

Setting the derivative with respect to a a equal to zero (and tidying up) we get 4 a 3 80 a 2 + 512 a 1049 = 0 4 a^3 - 80 a^2 + 512 a - 1049 =0

This has three positive roots; the maximum value of R R is found at the root a 6.3792 a\approx 6.3792 which gives R = 3.551815 R=3.551815\ldots , and the answer 3551815 \boxed{3551815} .


Additional consideration: OK, this is a "maximum"; but when a = 7.5 a=7.5 , we get R = 5.4375 R=5.4375 ; so what's going on?

Recall we have a quadratic to find the sides b b and c c . Explicitly, b b and c c are the roots of x 2 ( 2 s a ) x + s ( s a ) + T 2 s ( s a ) = 0 x^2-(2s-a)x+s(s-a)+\frac{T^2}{s(s-a)}=0

We need both roots to be real; so the discriminant of this quadratic needs to be non-negative, ie ( 2 s a ) 2 4 ( s ( s a ) + T 2 s ( s a ) ) 0 (2s-a)^2-4\left(s(s-a)+\frac{T^2}{s(s-a)}\right)\ge 0

This simplifies to a 2 4 T 2 s ( s a ) a^2\ge \frac{4T^2}{s(s-a)}

Now, if a s a\ge s , we get b + c = 2 s a a b+c=2s-a \le a , which breaks the triangle inequality (we can't form a triangle with these sidelengths). So s a s-a is positive and a 2 s ( s a ) 4 T 2 a^2 s(s-a) \ge 4T^2

Solving this inequality in this case we find (approximately) 3.24 < a < 6.97 3.24<a<6.97

If we let a a be the larger root here we find the same value of R R as above . So it is indeed the maximum (but I'm a bit surprised by this particular result!)

Thanks for the thoughtful and thorough analysis. I find problems of this type (where triangle area and perimeter are specified) fascinating.

Fletcher Mattox - 2 months ago

Log in to reply

Yes - more generally I'm quite enjoying the theme of families of triangles given two parameters.

For this particular case did you use a similar approach? It feels a little clunky to introduce sidelengths but it's a quick way to make things more familiar.

Chris Lewis - 2 months ago

Log in to reply

Yes. I used a similar approach. I started with a different form of Heron's formula:

T = 1 4 ( a 2 + b 2 + c 2 ) 2 2 ( a 4 + b 4 + c 4 ) T = \dfrac{1}{4}\sqrt{(a^2+b^2+c^2)^2 - 2(a^4+b^4+c^4)}

and, if P P = perimeter, I eventually arrived at this quartic polynomial in a a to describe the family of triangles:

4 P 2 a 4 4 P 3 a 3 + P 4 a 2 + 32 T 2 P a 16 T 2 P 2 > 0 4P^2a^4 - 4P^3a^3 + P^4a^2 + 32T^2Pa - 16T^2P^2 > 0

I suspect this is exactly equivalent to the relationship you found.

Fletcher Mattox - 2 months ago

Log in to reply

@Fletcher Mattox Solving a different problem I just came across a very nice relationship: if we're given the inradius r r and circumradius R R of a triangle, then

b c = 4 r R + A I 2 bc=4rR+AI^2

where A A is a vertex of the triangle, I I is its incentre and b b and c c are the lengths of the sides that meet at A A (ie standard notation).

I haven't seen this elsewhere and my derivation is horrible (gigantic trig expressions that suddenly collapse). Is this something you've come across? If not, any ideas for a nice way to prove it?

Chris Lewis - 1 month, 4 weeks ago
Mark Hennings
Apr 13, 2021

If the sides of the triangle are a = v + w a = v+w , b = u + w b = u+w and c = u + v c = u+v for u , v , w > 0 u,v,w > 0 , then we have that u + v + w = 8 u+v+w = 8 . Since 10 = 1 2 b c sin A 10 = \tfrac12bc\sin A we deduce that sin A = 20 ( u + v ) ( u + w ) \sin A = \tfrac{20}{(u+v)(u+w)} , and since 2 R = a sin A 2R = \tfrac{a}{\sin A} we deduce that R = 1 40 ( u + v ) ( u + w ) ( v + w ) R \; = \; \tfrac{1}{40}(u+v)(u+w)(v+w) Also note that 10 = s ( s a ) ( s b ) ( s c ) = u v w ( u + v + w ) = 8 u v w 10 = \sqrt{s(s-a)(s-b)(s-c)} = \sqrt{uvw(u+v+w)} = \sqrt{8uvw} , so we have that u v w = 25 2 uvw = \tfrac{25}{2} . But then R = 1 40 [ 8 3 8 2 ( u + v + w ) + 8 ( u v + u w + v w ) u v w ] = 1 5 ( u v + u w + v w ) 5 16 R \; = \; \tfrac{1}{40}\big[8^3 -8^2(u+v+w) + 8(uv + uw + vw) - uvw\big] \; = \; \tfrac15(uv +uw + vw) - \tfrac{5}{16} and hence we want to minimize R = 1 5 ( u v + u w + v w ) 5 16 R = \tfrac15(uv + uw + vw) - \tfrac{5}{16} subject to the constraints u + v + w = 8 u+v+w=8 and u v w = 25 2 uvw = \tfrac{25}{2} . Using the method of Lagrange multipliers we are looking to solve the equations 1 5 ( v + w ) λ μ v w = 0 1 5 ( u + w ) λ μ u w = 0 1 5 ( u + v ) λ μ u v = 0 \begin{aligned} \tfrac15(v+w) - \lambda - \mu vw & = \; 0 \\ \tfrac15(u+w) - \lambda - \mu uw & = \; 0 \\ \tfrac15(u+v) - \lambda - \mu uv & = \; 0 \end{aligned} and these equations imply that ( u v ) ( 1 5 μ w ) = ( u w ) ( 1 5 μ v ) = ( v w ) ( 1 5 μ u ) = 0 (u-v)(\tfrac15 - \mu w) = (u-w)(\tfrac15 - \mu v) = (v-w)(\tfrac15 - \mu u) = 0 . If u , v , w u,v,w were distinct, we would deduce that u = v = w = 1 5 μ u=v=w=\tfrac{1}{5\mu} , which is absurd. Since ( 8 3 ) 3 25 2 (\tfrac83)^3 \neq \tfrac{25}{2} , we cannot have u = v = w u=v=w . Thus, without loss of generality, we have v = u v=u , w = 8 2 u w = 8 - 2u , λ = 1 5 u \lambda =\tfrac15u , μ = 1 5 u 1 \mu = \tfrac15u^{-1} , where u 2 ( 8 2 u ) = 25 2 u^2(8-2u) = \tfrac{25}{2} .

There are two positive real roots to this cubic equation, 1.62077230 1.62077230 and 3.48555973 3.48555973 . The first corresponds to the minimum possible value of R m i n = 3.29782965 R_{\mathrm{min}} = 3.29782965 , while the second corresponds to the maximum possible value of R m a x = 3.55181516 R_{\mathrm{max}} = 3.55181516 , and hence 1 0 6 R m a x = 3551815 \lfloor 10^6 R_{\mathrm{max}}\rfloor = \boxed{3551815} .

Can't I do it like this:

Circumradius= abc/4∆ and then use Am-Gm inequality. For maximum radius, a=b=c=16/3. Then radius=16³/3³.40 and I got myself the answer 3.792592.

Baibhab Chakraborty - 1 month, 4 weeks ago

Log in to reply

If a = b = c = 16 / 3 a=b=c=16/3 then the area is not 10 10 .

Mark Hennings - 1 month, 4 weeks ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...