Angle Inequalities in Tetrahedra

Mark Chen

Top performers in the International Mathematical Olympiad are undoubtedly fa-

miliar with the standard inequalities regarding angles in triangles. Several of these
standard inequalities regarding the most basic of 2-D figures also have analogues
in the most basic of 3-D figures, namely tetrahedra. While proving one of these
analogues, we will also discover some interesting properties of tetrahedra.
We start with a basic theorem:
Theorem 1. If A, B, and C are the angles in an acute triangle ABC, then
3
cos A + cos B + cos C ≤ .
2
Proof. Because the cosine function is concave on the interval 0, π2 , we have
 

A+B+C π 3
cos A + cos B + cos C ≤ 3 · cos = 3 · cos =
3 3 2
by Jensen’s inequality with f (x) = cos x.

This leads to a main theorem regarding angles in tetrahedra:

Theorem 2. Let A1 A2 A3 A4 be a tetrahedron, and let fi be the face which opposes
the vertex Ai . Furthermore, let fij = fji be the angle between the faces fi and fj .
Then X
cos fij ≤ 2.
1≤i<j≤4

From this point onwards, we will use the notation from Theorem 2. But before we
can approach Theorem 2 itself, it is helpful to familiarize ourselves with some basic
properties of tetrahedra. One of the these properties is the observation that every
tetrahedron has a centroid.
Definition 3. Let A1 , A01 ; A2 , A02 ; A3 , A03 ; A4 , A04 be the four pairs of diagonally op-
posite vertices of a parallelepiped, drawn so that face A1 A02 A3 A04 ∼ = face A03 A4 A01 A2 .
Through one of its vertices, say A1 , draw the three diagonals A1 A2 , A1 A3 , and A1 A4 .
We say that parallelepiped A1 A02 A3 A04 A03 A4 A01 A2 is circumscribed about tetrahedron
A1 A2 A3 A4 .
Definition 4. A median of a tetrahderon is a line segment connecting a vertex with
the centroid of the opposite face.
Lemma 5. The medians of a tetrahedron lie along the diagonals of the tetraheron’s
circumscribed parallelepiped and are two-thrids the lengths of the respective diago-
nals.

Mathematical Reflections 6 (2008) 1

Proof. Let A1 A2 A3 A4 be our tetrahdedron, and let its circumscibed parallelepiped
be named as indicated in Definition 3. The line A3 B joining the vertex A3 to the
midpoint B of A01 A03 lies in the plane of the parallelogram A1 A03 A01 A3 . Hence, A3 B
meets A1 A01 in the centroid C of triangle A3 A03 A01 , and thus ACB 3C
= 2. Now B is
the common point to the diagonals A1 A3 and A2 A4 of face A1 A2 A03 A4 , so A3 B is
0 0 0

a median of the face A3 A2 A4 of the tetrahedron, and C is the centroid of this face.
Therefore, the median A1 C of the tetrahedron lies along the diagonal A1 CA01 of the
parallelepiped. Let O be the point of intersection of diagonals A1 A01 and A3 A03 . The
length of the segment A01 C is equal to two thirds of A01 O, so it is equal to one third
of A1 A01 . Hence A1 C is equal to two thirds of A1 A01 .

We will also need the following lemmas concerning tetrahedra:

Lemma 6. Let the sides of a tetrahedron be ai , 1 ≤ i ≤ 6. If the radius of the

circumsphere of the tetrahedron is R, then
6
X
a2i ≤ 16R2 .
i=1

Proof. Let the coordinates of Ai be (xi , yi , zi ), 1 ≤ i ≤ 4. It is a well-known fact

that the centroid G of the tetrahedron is the point
 
x1 + x2 + x3 + x4 y1 + y2 + y3 + y4 z1 + z2 + z3 + z4
, , .
4 4 4

Let W be a point (x, y, z). Then

 !2 
P4
i=1 xi
X X X
16W G2 + (Ai Aj )2 = 16  x− + (xi − xj )2  . (1)
cyc
4
1≤i≤j≤4 1≤i≤j≤4

Because X X X
(x − xi )2 = x2 − 2x · xi + x2i
and X X X
(xi − xj )2 = 3 · x2i − 2 · xi xj ,
equation (1) reduces to:

Mathematical Reflections 6 (2008) 2

  !2 
X 4
X 4
X X X
16x2 − 8x · xi + xi +3· x2i − 2 · xi xj 
cyc i=1 i=1 1≤i≤j≤4
4 4
" #
X X X
= 16x2 − 8x · xi + 4 · x2i
cyc i=1 i=1
X 4
X
=4· (x − xi )2 = 4 · (W Ai )2
cyc i=1

Letting point W be equidistant

P from A1 , A22and A3 ,2A4 , we have W Ai = R for all
i. Since W G ≥ 0, we have 1≤i<j≤4 (Ai Aj ) ≤ 16R , as desired. We immediately
see that equality holds when W G = 0.

It is a well-known fact that every tetrahedron has a circumsphere (a shphere contain-

ing all four vertices of the tetrahedron) and an insphere (a sphere tangent to all four
faces of the tetrahedron). We call the center of the circumsphere the circumcenter
and the center of the insphere the incenter.
Now we find a necessary and sufficient condition for the coincidence of the circum-
center, incenter, and centroid of a tetrahedron.

Lemma 7. A tetrahedron is isosceles (each pair of its opposite sides is congruent)

if and only if its centroid, circumcenter, and incenter are coincident.

Proof. We will first prove that a tetrahedron is isosceles if and only if its incenter
coincides with its centroid. The “only if” direction is trivial. Now suppose that the
incenter and the centroid of a tetrahedron are coincident at point G. Because G
is a centroid, the volumes of GA1 A2 A3 , GA1 A3 A4 , GA1 A4 A2 , and GA2 A3 A4 are
equal. Also, because G is an incenter, G is the same distance from all faces fi for
1 ≤ i ≤ 4. Hence all four faces have equal area. Again, let us denote fij = fji as
the angle between the faces that oppose vertices Ai and Aj . Projecting the areas of
A1 A2 A3 , A1 A3 A4 , and A1 A4 A2 onto A2 A3 A4 , we have the equation:

[A1 A2 A3 ] cos f14 + [A1 A3 A4 ] cos f12 + [A1 A4 A2 ] cos f13 = [A2 A3 A4 ] (1)
Because all faces are of equal area, we have the system of equations:

cos f12 + cos f13 + cos f14 = 1,

cos f21 + cos f23 + cos f24 = 1,

cos f31 + cos f32 + cos f34 = 1,

Mathematical Reflections 6 (2008) 3

 cos f41 + cos f42 + cos f43 = 1.
After performing multiple substitutions, we find that

cos f12 = cos f34 , cos f13 = cos f24 and cos f14 = cos f23 .

Also, because 0 < fij < π for 1 ≤ i < j ≤ 4, we have f12 = f34 , f13 = f24 ,
and f14 = f23 . Now drop perpendicular A1 B onto face A2 A3 A4 and perpendicular
A2 C onto face A1 A3 A4 . Furthermore, drop perpendicular A1 B 0 onto side A2 A3 and
perpendicular A2 C 0 onto side A1 A4 . Calculating the volume of the tetrahdron in
two ways, we have
1 1
[A2 A3 A4 ]A1 B = [A1 A3 A4 ]A2 C, . (2)
3 3
from which we find that A1 B = A2 C. Noting that BB 0 ⊥ A2 A3 , and CC 0 ⊥ A1 A4 ,
we have A1 BB 0 ∼= A2 CC 0 by the AAS Postulate (f23 = f14 ). Hence A1 B 0 = A2 C 0 .
Because all faces are of equal area,
1 1
A1 B 0 × A2 A3 = A2 C 0 × A1 A4 .
2 2
Therefore, A1 A4 = A2 A3 , and similarly, other pairs of opposite sides are congruent.
Hence our tetrahdron is isosceles. Now, it remains to prove that a tetrahedron is
isosceles if and only if its centroid and circumcenter are coincident. Suppose we have
an isosceles tetrahedron. It is easy to see that the circumscribed parallelepiped is
rectangular. Hence the parallelepiped’s diagonals are equal, which implies that the
tetrahedron’s medians are equal (median lengths equal 23 corresponding diagonal
lengths by Lemma 5). Therefore the centroid is equidistant from the vertices of the
tetrahedron, and is also a circumcenter. Now suppose that the centroid is equidistant
from the vertices. Then all medians of the the tetrahedron are equal, implying that
all diagonals of the circumscribed parallelopiped are equal. Hence the circumscribed
parallelepiped is rectangular and the tetrahedron is isosceles.

Now, our final lemma:

Lemma 8. Let the incenter of tetrahedron A1 A2 A3 A4 meet face fi at Pi for 1 ≤

i ≤ 4. Then A1 A2 A3 A4 is isosceles if and only if P1 P2 P3 P4 is isosceles.

Proof. (⇒) Let I denote the incenter of tetrahedron A1 A2 A3 A4 . Because IP1 ⊥

4A2 A3 A4 and IP4 ⊥ 4A1 A2 A3 , it is easy to see that ∠P1 IP4 = π − f14 , the
supplement of the angle between faces f1 and f4 . Similarly, ∠P2 IP3 = π − f23 .
In Lemma 4, we proved that f14 = f23 for all isosceles tetrahedra, so we have the
equivalence ∠P1 IP4 = ∠P2 IP3 . Because IP1 = IP2 = IP3 = IP4 , 4P1 IP4 ∼ =
4P2 IP3 by the SAS Postulate. Hence P1 P4 = P2 P3 , and working similarly with
other opposite edges, we find that P1 P2 P3 P4 is isosceles.

Mathematical Reflections 6 (2008) 4

(⇐) As usual, we let I denote the incenter of A1 A2 A3 A4 . Draw a plane through I,
P1 , and P4 , intersecting A2 A3 at point B. Also, draw a plane through I, P2 , and
P3 , intersecting A1 A4 at point C. Note that IB ⊥ A2 A3 and that IB is the perpen-
dicular bisector of P1 P4 . Because IP1 = IP2 = IP3 = IP4 , I is the circumcenter of
tetrahedron P1 P2 P3 P4 . But because tetrahedron P1 P2 P3 P4 is isosceles, I is also its
incenter and centroid by Lemma 7. Let the intersection of IB and P1 P4 be D and
the intersection of IC and P2 P3 be E. Then D and E are the midpoints of P1 P4
and P2 P3 respectively. Because I is the centroid of P1 P2 P3 P4 ,
" #
~1 + P~2 + P~3 + P~4
P 1 ~1 + P~4 P~2 + P~3
P 1 ~
I~ = = + = [D ~
+ E].
4 2 2 2 2

Hence D, I, and E are collinear (in particular, B, D, I, E, and C are all collinear).
Since P1 P4 = P2 P3 , we have

4IP1 P4 ∼
= 4IP2 P3 ⇒ 4IBP1 ∼
= 4ICP2 ⇒ I

so I is the midpoint of segment BC. Now let the plane through P3 , P4 , and I
intersect A1 A2 at M and let the plane through P1 , P2 and I intersect A3 A4 at N.
Similarly, we find that I is the midpoint of segment M N . Likewise, let the plane
through P1 , P3 , and I intersect A2 A4 at P and let the plane through P2 , P4 , and I
intersect A1 A3 at Q. As before, we find that I is the midpoint of segment P Q. It
is not difficult to see that BC, M N , and P Q perpendicularly bisect one another at
point I. In particular, P Q is perpendicular to the plane through B, C, M, N , and
the lines A1 A3 and A2 A4 . As I is the midpoint of P Q, B, C, M, N are the midpoints
of A2 A3 , A1 A4 , A1 A2 and A3 A4 respectively. Hence I is the centroid of A1 A2 A3 A4 ,
and by Lemma 7, A1 A2 A3 A4 is isosceles.

Now we can finally prove Theorem 2!

Proof. As usual, let I be the incenter of our tetrahedron. Let plane S through
points I, P1 , P4 intersect A2 A3 at point X. Now connect AP1 and AP4 , and get
P1 XP4 = (f14 ) ⇒ P1 IP4 = π − f14 . If the inradius of the tetrahderon is r, then the
law of cosines yields:

(P1 P4 )2 = 2r2 − 2r2 cos ∠P1 IP4 = 2r2 (1 + cos f14 ).

Summing over all lengths Pi Pj , 1 ≤ i < j ≤ 4, we obtain:
 
X X
(Pi Pj )2 = 2r2 6 + cos fij  .
1≤i<j≤4 1≤i<j≤4

By Lemma 6, the above equation becomes

Mathematical Reflections 6 (2008) 5

 X X
(Pi Pj )2 ≤ 16r2 ⇔ cos fij ≤ 2.
1≤i<j≤4 1≤i<j≤4

By Lemmas 6, 7, and 8, we have equality if and only if A1 A2 A3 A4 is isosceles.

We now present several extensions of Theorem 2. The analogue of the inequality

√
A B C 3 3
cos cos cos ≤
2 2 2 8
in a triangle ABC is:
Y fij 8
cos ≤ . (3)
2 27
1≤i<j≤4
1
The analogue of the inequality cos A cos B cos C ≤ 8 in a triangle ABC is:
1
Π1≤i<j≤4 cos fij ≤ . (4)
36
A B C 9
Finally, the analoge of the inequality cos2 2 + cos2 2 + cos2 2 ≤ 4 in a triangle ABC
is:
X fij
cos2 ≤ 4. (5)
2
1≤i<j≤4

We leave the proofs of these simple extensions to the interested reader.

1 Bibliography
[1] Altshiller-Court, Nathan. Modern Pure Solid Geometry, Chelsea, 1964.

[2] Wang, Hsiang-Tung. Inequalities, He Nan Educational Press, 1994.

Mathematical Reflections 6 (2008) 6