Triangle Identities via Elimination Theory

Dam Van Nhi and Luu Ba Thang

Abstract
The aim of this paper is to present some new identities via resultants and elimination theory,
by presenting the sidelengths, the lengths of the altitudes and the exradii of a triangle as roots of
cubic polynomials.

Introduction

Finding identities between elements of a triangle represents a classical topic in elementary geometry.
In Recent Advances in Geometric Inequalities [3], Mitrinovic et al presents the sidelengths, the lengths
of the altitudes and the various radii as roots of cubic polynomials, and collects many interesting
equalities and inequalities between these quantities. As far as we know, almost of them are built using
only elementary algebra and geometry. In this paper, we use transformations of rational functions
and resultants to build similar nice equalities and inequalities might be difficult to prove with different
methods.

Main results

To fix notations, suppose we are given a triangle ABC with sidelengh a, b, c. Denote the radius of
the circumcircle by R, the radius of incircle by r, the area of ABC by S, the semiperimeter as p,
the radii of the excircles as r1 , r2 , r3 , and the altitudes from sides a, b and c, respectively, as ha , hb and
hc . We have the results from [3, Chapter 1].
Theorem 1. [3] Using the above notations,
(i) a, b, c are the roots of the cubic polynomial
x3 2px2 + (p2 + r2 + 4Rr)x 4Rrp.

(1)

(ii) ha , hb , hc are the roots of the cubic polynomial

x3

S 2 + 4Rr3 + r4 2 2S 2
2S 2
x
+
x

.
2Rr2
Rr
R

(2)

(iii) r1 , r2 , r3 are the roots of the cubic polynomial

x3 (4R + r)x2 + p2 x p2 r.

(3)

Next, we present how to use transformations of rational functions to build new geometric equalities
ax + b
and inequalities. The transformation that we use here is given by y =
, a, b, c, d R.
cx + d
Proposition 2. Let
T1 =
T2 =
T3 =

3r1 2r 3r2 2r 3r3 2r

+
+
2r1 + r
2r2 + r
2r3 + r
3r1 2r 3r2 2r 3r2 2r 3r3 2r 3r3 2r 3r1 2r
+
+
2r1 + r 2r2 + r
2r2 + r 2r3 + r
2r3 + r 2r1 + r
3r1 2r 3r2 2r 3r3 2r
.
2r1 + r 2r2 + r 2r3 + r
1

We have
T3 =

3r1 2r 3r2 2r 3r3 2r

100r2 + 9p2
.
>
2r1 + r 2r2 + r 2r3 + r
19r2 + 12p2

3ri 2r
3
r(yi + 2)
, i = 1, 2, 3 we have yi 6= and ri =
. Since r1 , r2 , r3 are the roots of
2x + r
2
3 2yi
the polynomial (3), if we substitute ri into the polynomial (3), we obtain

Proof. Let yi =

(3r2 + 8Rr + 12p2 )yi3 + (11r2 + 20Rr 40p2 )yi2 + (8r2 16Rr + 39p2 )yi (4r2 + 48Rr + 9p2 ) = 0;
therefore y1 , y2 , y3 are the roots of
(3r2 + 8Rr + 12p2 )y 3 + (11r2 + 20Rr 40p2 )y 2 + (8r2 16Rr + 39p2 )y (4r2 + 48Rr + 9p2 ).
By Vietes formula, we have
40p2 20Rr 11r2
3r2 + 8Rr + 12p2
8r2 16Rr + 39p2
= y1 y2 + y2 y3 + y3 y1 =
3r2 + 8Rr + 12p2
4r2 + 48Rr + 9p2
= y1 y2 y3 = 2
.
3r + 8Rr + 12p2

T1 = y1 + y2 + y3 =
T2
T3

4r2 + 48rt + 9p2

then f is an increasing function over [2r, +]. Therefore, from R 2r
3r2 + 8rt + 12p2
we have T3 f (2r) i.e
3r1 2r 3r2 2r 3r3 2r
100r2 + 9p2
>
.
2r1 + r 2r2 + r 2r3 + r
19r2 + 12p2

Let f (t) =

Proposition 3. Using the above notations, let

T1 =
T2 =
T3 =

2r1 r 2r2 r 2r3 r

+
+
r1 + r
r2 + r
r3 + r
2r1 r 2r2 r 2r2 r 2r3 r 2r3 r 2r1 r
+
+
r1 + r r2 + r
r2 + r r3 + r
r3 + r r1 + r
2r1 r 2r2 r 2r3 r
.
r1 + r r2 + r r3 + r

then we have an equality 4T3 + T2 2T1 = 5.

2ri r
r(yi + 1)
, i = 1, 2, 3 we have yi 6= 2 and x =
. Since r1 , r2 , r3 are the roots of
ri + r
2 yi
the polynomial (3), if we substitute ri into the polynomial (3), we obtain

Proof. Let yi =

2(r2 + 2Rr + p2 )yi3 3(r2 + 3p2 )yi2 12(Rr p2 )yi (r2 + 8Rr + 4p2 ) = 0
Therefore, y1 , y2 , y3 are the roots of the polynomial
2(r2 + 2Rr + p2 )y 3 3(r2 + 3p2 )y 2 12(Rr p2 )y (r2 + 8Rr + 4p2 ).
Applying the Vietes formulas for this polynomial, we obtain 4T3 + T2 2T1 = 5.
Elimination theory [1] is one of the most effective method to solve polynomial equations, so if we
combine this method and Theorem 1, we can obtain some nice equalities. Consequently, we give a brief
overview of the resultant of two polynomials. Given a field K C and two polynomials f, g K[x] of
positive degree
f

= a0 xm + a1 xm1 + + am , a0 6= 0, m > 0

g = b0 xn + b1 xn1 + + bn , b0 6= 0, n > 0.
2

We build the square matrix of size (m + n) (m + n) associated to f, g

follows:

a0 a1 am

a0 a1 am

.. ..
..

.
.
.

a0 a1 am
S(f, g) :=
b
b

bn

0
1

b0 b1 bn

.. ..
..

.
.
.

b0 b1 bn

(called Sylvester matrix) as

(4)

where n first rows only depend on the coefficients of f , m last rows depend on the coefficients of g and
the blank spaces are filled with 0. Then the resultant of f and g, denoted Res(f, g), is the determinant
of Sylvester matrix S(f, g).
If we want to emphasize the dependence on x, we write Res(f, g, x) instead of Res(f, g). Two
important properties of resultants [2, Chapter 3] are
(Common factor) Res(f, g) = 0 if and only if f and g have a non trivial common factor in K[x].
(Elimination) There are polynomials U, V K[x] such that U f +V g = Res(f, g). The coefficients
of U, V are integer polynomial in the coefficients of f and g.
Proposition 4. Using the above notations,
(i) (a2 + 2Rr)(b2 + 2Rr)(c2 + 2Rr) = 2Rr(ab + bc + ca 2Rr)2 .
(ii) (a2 2Rr)(b2 2Rr)(c2 2Rr) = 4a2 b2 c2 2Rr(ab + bc + ca + 2Rr)2 .
a2 + p2 + 4Rr + r2 b2 + p2 + 4Rr + r2 c2 + p2 + 4Rr + r2
(r1 + r2 + r3 + r)2
=
.
S
S
S
S
p
(a2 + p2 + 4Rr + r2 )(b2 + p2 + 4Rr + r2 )(c2 + p2 + 4Rr + r2 )
(iv) S =
.
r1 + r2 + r3 + r

2

r1 r2 r3 2
(v) (r12 + 2Rr)(r22 + 2Rr)(r32 + 2Rr) = r1 r2 r3 2Rr(r1 + r2 + r3 ) +2Rr 2Rr
.
r
 r2
 r2
 r2

h (2R + r)2
i
1
2
3
(vi)

1
=
4

1
p2
p2
p2
p2

(iii)

Proof. (i) Let f := x3 2px2 + (p2 + 4Rr + r2 )x 4Rrp = 0 and g := x2 + 2Rr y. By elimination
property, two polynomials f (x), g(x) have a common root if and only if resultant Res(f, g; x) = 0. So,
y 3 (4p2 + 2Rr 2T )y 2 + (T 2 2Rr)y 2RrT 2 = 0 where T = p2 + 2Rr + r2 .

(5)

Let y1 = a2 + 2Rr, y2 = b2 + 2Rr and y3 = c2 + 2Rr. Because a, b and c are the roots of f (x), y1 , y2 , y3
are the roots of the equation (5). Using Vietes formulas, we have
y1 y2 y3 = 2RrT 2 = 2Rr(ab + bc + ca 2Rr)2 .
(ii) Let f := x3 2px2 + (p2 + 4Rr + r2 )x 4Rrp = 0 and g := x2 2Rr y then
y 3 +(2Rr+2T 4p2 )y 2 +(T 2 +4RrT 32Rrp2 )y+2RrT 2 64R2 r2 p2 = 0 where T = p2 +6Rr+r2 . (6)
Let y1 = a2 2Rr, y2 = b2 2Rr and y3 = c2 2Rr, then y1 , y2 , y3 are the roots of the equation (6).
Using Vietes formulas, we have
y1 y2 y3 = 4a2 b2 c2 2Rr(ab + bc + ca + 2Rr)2 .

(iii) Let f (x) = x3 2px2 + (p2 + 4Rr + r2 )x 4Rrp and g(x) = x2 + p2 + 4Rr + r2 y, we have
y 3 (p2 + 4Rr + r2 )y 2 (2py 4Rrp 2r2 p)2 = 0.

(7)

Let y1 = a2 + p2 + 4Rr + r2 , y2 = b2 + p2 + 4Rr + r2 and y3 = c2 + p2 + 4Rr + r2 , then y1 , y2 , y3 are

the roots of the equation (7). Using Vietes formulas, we have y1 y2 y3 = (4R + 2r)2 r2 p2 . Hence,
(a2 + p2 + 4Rr + r2 )(b2 + p2 + 4Rr + r2 )(c2 + p2 + 4Rr + r2 ) = (4Rr + 2r2 )2 p2 .
Otherwise, r1 + r2 + r3 = 4R + r, we have
(r1 + r2 + r3 + r)2
a2 + p2 + 4Rr + r2 b2 + p2 + 4Rr + r2 c2 + p2 + 4Rr + r2
=
.
S
S
S
S
(iv) This equality is deduced from (iii).
(v) Let f := x3 (4R + r)x2 + p2 x p2 r and g := x2 + 2Rr y. Similar to (i), we have

2
(r12 + 2Rr)(r22 + 2Rr)(r32 + 2Rr) = r1 r2 r3 2Rr(r1 + r2 + r3 )

r1 r2 r3 2
+2Rr 2Rr
.
r
(vi) Similarly to (i), let f := x3 (4R + r)x2 + p2 x p2 r and g := x2 p2 y; the conclusion follows.

If we combine them with the well-known inequality R 2r [3], we can yield some interesting
inequalities. In particular,
Corollary 5. Using the above notations,
(i) (a2 + 2Rr)(b2 + 2Rr)(c2 + 2Rr) 6 R2 (ab + bc + ca 2Rr)2 .
(ii) (a2 + 2Rr)(b2 + 2Rr)(c2 + 2Rr) 2R6 .
(iii) (a2 2Rr)(b2 2Rr)(c2 2Rr) > 4a2 b2 c2 R2 (ab + bc + ca + R2 )2 .
(iv) (a2 2Rr)(b2 2Rr)(c2 2Rr) 6 4a2 b2 c2 4r2 (ab + bc + ca + 4r2 )2 .
Proof. (i) By Proposition 4(i) and R 2r, we are done.
(ii) Let T = (a2 +2Rr)(b2 +2Rr)(c2 +2Rr), by Proposition 4(i) we have (a2 +2Rr)(b2 +2Rr)(c2 +2Rr) =
4p2
2Rr(ab + bc + ca 2Rr)2 . Combining with ab + bc + ca
and 4p2 27R2 , we obtain
3
T 2Rr(9R2 2Rr)2 T 6 R3 .2r(9R 2r)2 .
Consider the function f (x) = x(9R x)2 where 0 < x 6 R. We have f 0 (x) > 0 with each x (0, R],
deduce that the function f is the increasing function in (0, R]. Therefore f (x) f (R) = 64R3 with
each x (0, R]. Notice that R 2r, so T 64R6 and the equality holds if and only if 4ABC is
equilateral.
By Proposition 4(ii) and R 2r, we have the proof of (iii) and (iv).
Corollary 6. Suppose given a convex quadrilateral ABCD of AB = a, BC = b, CD = c, DA = d
inscribed the circle of center O, radius R. Denoted r1 , r2 , r3 , r4 as the radius of incircle the triangles
26 R6
4ABC, 4BCD, 4CDA and 4DAB, respectively. Let T =
then

ac + bd + 4 r1 r2 r3 r4
q
T (ab + 4r12 )(bc + 4r22 )(cd + 4r32 )(da + 4r42 ).

Proof. Let AC = x, BD = y, by Proposition 4 we have

(a2 + 2Rr1 )(b2 + 2Rr1 )(x2 + 2Rr1 ) 6 64R6 .
Otherwise, we have (a2 + 2Rr1 )(b2 + 2Rr1 ) (ab + 2Rr1 )2 , so we deduce that
(ab + 2Rr1 )2 (x2 + 2Rr1 ) 64R6 .
Similar, we obtain
(bc + 2Rr2 )2 (y 2 + 2Rr2 ) 64R6
(cd + 2Rr3 )2 (x2 + 2Rr3 ) 64R6
(da + 2Rr4 )2 (y 2 + 2Rr4 ) 64R6
Multiplying, we obtain

(ab+2Rr1 )(bc+2Rr2 )(cd+2Rr3 )(da+2Rr4 )]2 (x2 +2Rr1 )(x2 +2Rr3 )(y 2 +2Rr2 )(y 2 +2Rr4 ) 224 R24 .
(8)
Consequently,

(x2 + 2Rr1 )(x2 + 2Rr3 ) > (x2 + 2R r1 r3 )2

(y 2 + 2Rr2 )(y 2 + 2Rr4 ) > (y 2 + 2R r2 r4 )2

and

(x + 2R r1 r3 )(y + 2R r2 r4 ) > (xy + 2R 4 r1 r2 r3 r4 )2 .

2

(xy + 2R 4 r1 r2 r3 r4 )2 (x2 + 2Rr1 )(x2 + 2Rr3 )(y 2 + 2Rr2 )(y 2 + 2Rr4 )

(9)

From (8) and (9), it follows that

(ab + 2Rr1 )(bc + 2Rr2 )(cd + 2Rr3 )(da + 2Rr4 )(xy + 2R 4 r1 r2 r3 r4 ) 212 R12 .

(10)

From 2ri R, i = 1, 2, 3, 4 and by Ptolemys theorem, we furthermore have that

xy + 2R 4 r1 r2 r3 r4 = ac + bd + 2R 4 r1 r2 r3 r4 ac + bd + 4 r1 r2 r3 r4 .
Therefore, we deduce that
212 R12
212 R12
>

(ac + bd + 4 r1 r2 r3 r4 )2
(ac + bd + 2R 4 r1 r2 r3 r4 )2
> (ab + 2Rr1 )(bc + 2Rr2 )(cd + 2Rr3 )(da + 2Rr4 ).

T2 =

Using again the inequality 2ri R, i = 1, 2, 3, 4, we obtain

q
T > (ab + 4r12 )(bc + 4r22 )(cd + 4r32 )(da + 4r42 ).

References
[1] D.A.Cox, J. B. Little, and D. OShea. Using Algebraic Geometry. Volume 185 of Graduate Texts
in Mathematics. Springer-Verlag, NY, 499 pages, 1998.
[2] D. A. Cox, J. B. Little and D. OShea, Ideals, Varieties, and Algorithms 2nd edition, SpringerVerlag, New York, 1997.
[3] D.S. Mitrinovic, J.E. Pecaric and V. Volenec, Recent Advances in Geometric Inequalities, Acad.
Publ., Dordrecht, Boston, London 1989.
5

Luu Ba Thang
Department of Mathematics and Informatics, Hanoi National University of Education
136 Xuan Thuy, Cau Giay, Hanoi, Vietnam
thanglb@hnue.edu.vn
Dam Van Nhi
School for gifted student, Hanoi National University of Education
136 Xuan Thuy, Cau Giay, Hanoi, Vietnam