You are on page 1of 5

ON THE MAXIMUM AREA OF A TRIANGLE WITH THE FIXED

DISTANCES FROM ITS VERTICES TO A GIVEN POINT


IVAN BORSENCO, ROBERTO BOSCH CABRERA, DANIEL LASAOSA
Introduction. In problem solving, the search for ideas that lead to a solution or to a better
understanding of mathematical concepts is more important than the ability to conclude that a
problem is solved. The problem that we discuss in this article is a good example of a question
whose answer is not particularly of interest, but rather the dierent approaches that lead to it.
Problem 1. Let P be a point in the plane. Points A, B, C are located at distances R
1
= 1,
R
2
= 2, R
3
= 3 from P, respectively. Find the maximum area of triangle ABC.
Such a problem could appear in a mathematical contest, where participants might be asked to
nd, with proof, a particular lower or upper bound. There is also the general problem, for which
we would like to nd the answer.
Problem 2. Let P be a point in the plane. Points A, B, C are located at distances R
1
, R
2
, R
3
from P, respectively. Find the maximum area of triangle ABC.
In the rst part of this article we explore some sharp bounds that estimate the maximum of
K
ABC
for the general problem. We use the conditions from the rst problem to verify the eec-
tiveness of our approaches. In the second part, we solve the general problem and show how to nd
the exact value of the maximum area.
Lower and Upper Bounds. To nd a lower bound for K
ABC
it is sucient to provide
a construction of a triangle ABC and compute its area. The simplest method is to construct
segments AP, BP, CP at equal angles of 120

from each other. Summing up the areas of triangles


APB, BPC, CPA we obtain
K
ABC

3
4
(R
1
R
2
+R
2
R
3
+R
3
R
1
) .
We could also construct triangle ABC by placing P on the segment joining two of the points from
the triplet (A, B, C) and take the line joining P with the third point perpendicular to this segment.
There are three possible constructions from which we conclude that
K
ABC
max
_
1
2
R
1
(R
2
+R
3
),
1
2
R
2
(R
1
+R
3
),
1
2
R
3
(R
1
+R
2
)
_
.
Using the values R
1
= 1, R
2
= 2, R
3
= 3 from Problem 1, we nd K
ABC

3
4
11 4.7631 and
K
ABC
max(
5
2
,
8
2
,
9
2
) = 4.5. We observe that our construction with equal angles separating the
segments yields a better lower bound.
The upper bounds require totally dierent approaches. The rst one is unique in its nature. We
reformulate the problem and map it in the coordinate plane. Let A(x
1
, y
1
), B(x
2
, y
2
), C(x
3
, y
3
) be
the coordinates of points A, B, C, respectively with P the origin. We have
x
2
1
+y
2
1
= R
2
1
,
x
2
2
+y
2
2
= R
2
2
,
x
2
3
+y
2
3
= R
3
3
,
and
K
ABC
=
1
2

1 x
1
y
1
1 x
2
y
2
1 x
3
y
3

.
We will use Hadamards Inequality.
Hadamards Inequality. If A is a n n matrix with entries a
ij
and determinant |A|, then
|A|
2

i=1
_
_
n

j=1
|a
ij
|
2
_
_
,
where equality holds if and only if the rows are orthogonal.
From Hadamards inequality,

x
1
y
1
1
x
2
y
2
1
x
3
y
3
1

2
(R
2
1
+ 1)(R
2
2
+ 1)(R
2
3
+ 1),
yielding
K
ABC

1
2
_
(R
2
1
+ 1)(R
2
2
+ 1)(R
2
3
+ 1).
Using this result in Problem 1, we obtain K
ABC

1
2

2 5 10 = 5.
It is dicult to interpret this result geometrically, because when we add R
2
1
and 1 we lose the
metric context. The upper bound we found using Hadamards inequality is surprising. It may not
be that useful in the general case, but it seems eective if we restrict P, A, B, C to be lattice points.
Another approach to nd an upper bound would be to use properties of the Fermat-Torricelli
point.
Lemma 1. Let T be the point in the plane of triangle ABC that minimizes the sum of the
distances to the vertices of triangle: f(M) = AM +BM +CM. Then
4

3K
ABC
f
2
(T).
Proof of Lemma 1. It is known that, if all angles of triangle ABC are less than 120

, then T
is the Fermat-Torricelli point. Otherwise, if there is an angle of triangle ABC that is greater than
or equal to 120

, say A 120

, then T A. In the rst case, the area of triangle ABC can be


expressed as
K
ABC
=

3
4
(AT BT +BT CT +CT AT).
Using the basic inequality AT BT + BT CT + CT AT
1
3
(AT + BT + CT)
2
, we obtain the
desired result. In the second case, A 120

, hence
K
ABC

1
2
AB AC sin(A)

3
4
AB AC.
Mathematical Reflections 2 (2014) 2
But f(T) = f(A) = AB +AC, yielding
K
ABC

3
4
AB AC

3
4

(AB +AC)
2
4
=

3
16
f
2
(T),
an even tighter bound than we desired. Thus, the lemma is proved.
Returning back to the problem, assume that we found triangle ABC with maximum area. Let
T be the point that minimizes the sum of the distances to the vertices of the triangle. Then
AP +BP +CP f(T), and using Lemma 1, we obtain
K
ABC

1
4

3
(R
1
+R
2
+R
3
)
2
.
Substituting R
1
= 1, R
2
= 2, R
3
= 3, we nd K
ABC
3

3 5.1961. This upper bound is


close, but nonethless weaker than the one obtained by using Hadamards inequality.
Finding the exact value. Denote u = BPC, v = CPA, w = APB. Observe that our
problem is equivalent to the following maximization problem:
Given the distances R
1
, R
2
, R
3
, and angles u, v, w, nd the maximum value of the function
f(u, v, w) =
1
2
(R
1
R
2
sin w +R
2
R
3
sin u +R
3
R
1
sin v) ,
on the compact set K =
_
(u, v, w) [0, ]
3
: u +v +w = 2
_
.
Dene the function
L(u, v, w, ) =
1
2
(R
1
R
2
sin w +R
2
R
3
sin u +R
3
R
1
sin v) +(u +v +w 2).
By Weierstrasss Theorem, the maximum of K
ABC
exists; it is attained either on the boundary
of the set K or when all partial derivatives are equal to 0. The boundary points of K describe
the situation as when one of the angles (u, v, w) is equal to 0, then K
ABC
= 0 or when one of the
angles (u, v, w) = . In the latter case, the maximum of K
ABC
occurs when the other two angles
are both equal to

2
. This brings us to the case already discussed, where we looked for lower bounds
of K
ABC
, and found that
K
ABC

1
2
max (R
1
(R
2
+R
3
), R
2
(R
1
+R
3
), R
3
(R
1
+R
2
)) .
After inspecting the maximum of K
ABC
on the boundary of K, we look for the extreme points
inside this set. We have
L
u
=
1
2
R
2
R
3
cos u = 0,
L
v
=
1
2
R
3
R
1
cos v = 0,
L
w
=
1
2
R
1
R
2
cos w = 0,
L

= u +v +w 2 = 0.
Solving the system of equations, we nd that there is a number such that R
1
= 2 cos u, R
2
=
2 cos v, R
3
= 2 cos w, and = 2
2
cos ucos v cos w. Note that u +v +w = 2 if and only if
1 + 2 cos ucos v cos w = cos
2
u + cos
2
v + cos
2
w.
Mathematical Reflections 2 (2014) 3
Because cos w = cos(u +v), we have
1 + 2 cos ucos v cos w = 1 cos w(cos(u +v) + cos(u v))
= 1 cos w(cos w + cos(u v))
= 1 + cos
2
w cos wcos(u v)
= 1 + cos
2
w + cos(u +v) cos(u v)
= 1 + cos
2
w +
1
2
(cos 2u + cos 2v)
= cos
2
u + cos
2
v + cos
2
w.
Substituting cos u =
R
1
2
, cos v =
R
2
2
, cos w =
R
3
2
, we nd that should satisfy the following
equation
1 +
2R
1
R
2
R
3
8
3

_
R
2
1
4
2
+
R
2
2
4
2
+
R
2
3
4
2
_
= 0.
It follows that is solution of a cubic equation
x
3

1
4
(R
2
1
+R
2
2
+R
2
3
)x +
1
4
R
1
R
2
R
3
= 0.
Let p =
1
4
_
R
2
1
+R
2
2
+R
2
3
_
and q =
1
4
R
1
R
2
R
3
. Then our equation becomes x
3
+px +q = 0. The
discriminant is equal to
=
q
2
4
+
p
3
27
=
1
64
_
R
2
1
R
2
2
R
2
3

1
27
(R
2
1
+R
2
2
+R
2
3
)
3
_
.
By the AM-GM inequality, 0, therefore the cubic has three real roots. If x
1
, x
2
, x
3
are the
roots of the cubic, then by Vi`etas Formulas x
1
x
2
x
3
= q < 0 and x
1
+ x
2
+ x
3
= 0. From these
equations it follows that two of the roots are positive and, the other one is negative.
Assume > 0. Then equations cos u =
R
1
2
, cos v =
R
2
2
, and cos w =
R
3
2
, imply u, v, w <

2
so
u +v +w = 2 does not hold. Therefore the only root that we should consider is the negative one.
Suppose < 0. Then we can nd angles u, v, w, where

2
< u, v, w < . Construct a circle with
radius || and a triangle ABC with sides 2|| sin(180

u), 2|| sin(180

v), and 2|| sin(180

w).
The angles of this triangle are equal to 180

u, 180

v, and 180

w. Let H be the orthocenter


of triangle ABC. Then BHC = u, CHA = v, AHB = w, and AH = 2 cos u, BH = 2 cos v,
CH = 2 cos w. Thus in triangle ABC, the orthocenter satises all the necessary conditions. It
follows that ABC is a candidate that produces the maximum area, with P being its orthocenter
and being its circumradius.
Using this method we determine the fourth value that is a candidate for the maximum of K
ABC
.
It remains to nd a negative root of the cubic x
3

1
4
(R
2
1
+R
2
2
+R
2
3
)x +
1
4
R
1
R
2
R
3
= 0. Since
sin u =
_
4
2
R
2
1
2||
, sin v =
_
4
2
R
2
2
2||
, sin w =
_
4
2
R
2
3
2||
,
we get that the fourth value, which we denote by K

, is equal to
K

=
1
4||
_
R
1
R
2
_
4
2
R
2
3
+R
2
R
3
_
4
2
R
2
1
+R
3
R
1
_
4
2
R
2
2
_
.
Mathematical Reflections 2 (2014) 4
Finally,
max(K
ABC
) = max
_
1
2
R
1
(R
2
+R
3
),
1
2
R
2
(R
1
+R
3
),
1
2
R
3
(R
1
+R
2
), K

_
.
If we want the maximum area of the triangle in Problem 1, we need the negative root of the cubic
equation x
3

7
2
x +
3
2
= 0. Using Maple, we obtain = 2.0565, yielding max(K
ABC
) 4.9044.
We also determine angles u, v, w as
u = arccos
_
R
1
2
_
104

,
v = arccos
_
R
2
2
_
119

,
w = arccos
_
R
3
2
_
137

.
Setting A(1, 0), the position of points B and C is shown in the gure:
We observe that point P is the orthocenter of triangle ABC.
The main result that we proved is that, except for three extreme boundary cases, triangle ABC
that has the maximum area must have P as its orthocenter. The conditions of the problem imply
that we know the distances AH, BH, CH. In order to nd the maximum area, we need the
circumradius R, found by solving a cubic equation. In conclusion, by knowing R, we can construct
triangle ABC and nd its maximum area.
Ivan Borsenco
Massachusetts Institute of Technology, USA.
ivan.borsenco@gmail.com
Roberto Bosch Cabrera
University of Havana, Cuba.
bobbydrg@gmail.com
Daniel Lasaosa
University of Navarra, Spain.
daniel.lasaosa@unavarra.es
Mathematical Reflections 2 (2014) 5

You might also like