Radii of Circles in Apollonius' Problem
Radii of Circles in Apollonius' Problem
Forum Geometricorum
Volume 17 (2017) 359–372. b b
FORUM GEOM
ISSN 1534-1178
Abstract. The paper presents the relation for radii of the eight circles in Apollo-
nius’ problem for circles which are tangent to three given circles. Analogously,
we derived the relations for radii of the 16 spheres which are tangent to four
given spheres, with coordinates of their centers and with their radii.
1. Introduction
It is well known that for three given circles generically there are eight different
circles that are tangent to them. The problem of ruler and compass constructability
of these eight circles is well-known. There are famous Apollonius’ and Gergonne’s
solutions to this problem [3]. Special cases of the three given circles are considered
and a number of other problems is known [2]. The first case is when we consider
three sides of the original triangle as three circles with infinite radii. The incircle
and three excircles of the original triangle are four solutions to Apollonius’ prob-
lem. The second case is when we have three excircles as a starting point. Three
sides of the original triangle are three solutions to Apollonius’ problem with infi-
nite radii [1]. The nine-point circle is tangent externally to the three excircles, by
Feuerbach theorem, and a relatively new object - the Apollonius circle is tangent
internally to three exircles (for some results about this circle see [4]-[7]). To these
five circles we can add three Jenkins circles which are tangent to three excircles,
by adding two of them externally and the third one internally.
2. Main result
Let us assume that the three given circles are K1 (O1 , r1 ), K2 (O2 , r2 ), K3 (O3 , r3 ),
Figure 1, with distances between centers O2 O3 = a, O3 O1 = b, O1 O2 = c and
with the area (O1 O2 O3 ) = ∆, which is different from 0. The following theorem
holds true:
Theorem 1. Let us assume that the radii of the eight circles with centers Si given
in Figure 1 ((a), (b), (c) and (d)) are pj , (j = 1, 2, ..., 8). Then
1 1 1 1 1 1 1 1
− + − + − + − = 0. (1)
p1 p2 p3 p4 p5 p6 p7 p8
Proof. Let us introduce the angles ϕ1 = ∠O2 SO3 , ϕ2 = ∠O3 SO1 , ϕ3 = ∠O1 SO2 ,
where S = S1 for Figure 1 (a), so as to obtain
cos2 ϕ1 + cos2 ϕ2 + cos2 ϕ3 − 2 cos ϕ1 cos ϕ2 cos ϕ3 = 1. (2)
O3
O3
b a
S1
b a S4
S2
S3
O1 c O2
O1 c O2
1A 1B
O3
S8
b a
O3
O1 c O2
S6 b a
S7
S5
O1 c O2
1C 1D
Figure 1. The eight different circles that are tangent to the three given circles.
If we substitute
ϕ1 ϕ2 ϕ3
t1 = sin2
, t2 = sin2 , t3 = sin2 ,
2 2 2
from relation (2) we have
t21 + t22 + t23 − 2(t1 t2 + t2 t3 + t3 t1 ) + 4t1 t2 t3 = 0. (3)
If we generally denote the center of the circle and radius by S and p, then for
the first unknown circle L1 (see Figure 1 (a)) is SO1 = p + r1 , SO2 = p + r2 ,
Radii of circles in Apollonius’ problem 361
SO3 = p + r3 and
(p + r2 )2 + (p + r3 )2 − a2 a2 − (r2 − r3 )2
cos ϕ1 = =⇒ t1 = ,
2(p + r2 )(p + r3 ) 4(p + r2 )(p + r3 )
and analogously
b2 − (r3 − r1 )2 c2 − (r1 − r2 )2
t2 = , t3 = .
4(p + r3 )(p + r1 ) 4(p + r1 )(p + r2 )
From relation (3) we now have relation (4):
(a2 − (r2 − r3 )2 )2 (p + r1 )2 + (b2 − (r3 − r1 )2 )2 (p + r2 )2
+ (c2 − (r1 − r2 )2 )2 (p + r3 )2
− 2(a2 − (r2 − r3 )2 )(b2 − (r3 − r1 )2 )(p + r1 )(p + r2 )
− 2(a2 − (r2 − r3 )2 )(c2 − (r1 − r2 )2 )(p + r1 )(p + r3 )
− 2(b2 − (r3 − r1 )2 )(c2 − (r1 − r2 )2 )(p + r2 )(p + r3 )
+ (a2 − (r2 − r3 )2 )(b2 − (r3 − r1 )2 )(c2 − (r1 − r2 )2 )
= 0. (4)
or in another form
F1 (p, r1 , r2 , r3 ) = 0. (5)
Equation (4) is of the second degree and is of the form
A1 p2 + B1 p + C1 = 0, (6)
where
A1 = 4(a2 (r1 − r2 )(r1 − r3 ) + b2 (r2 − r3 )(r2 − r1 ) + c2 (r3 − r1 )(r3 − r2 ))
− 16∆2
= f (r1 , r2 , r3 ), (7)
2 2 2 2 2 2 2 2 2
B1 = 2{r1 (a − (r2 − r3 ) ) + r2 (b − (r3 − r1 ) ) + r3 (c − (r1 − r2 ) )
− (a2 − (r2 − r3 )2 )(b2 − (r3 − r1 )2 )(r1 + r2 )
− (b2 − (r3 − r1 )2 )(c2 − (r1 − r2 )2 )(r2 + r3 )
− (c2 − (r1 − r2 )2 )(a2 − (r2 − r3 )2 )(r3 + r1 )}
= g(r1 , r2 , r3 ), (8)
C1 = r12 a4 + r22 b4 + r32 c4 + a2 b2 c2
− a2 b2 (r12 + r22 ) − b2 c2 (r22 + r32 ) − c2 a2 (r32 + r12 )
+ a2 (r12 − r22 )(r12 − r32 ) + b2 (r22 − r32 )(r22 − r12 )
+ c2 (r32 − r12 )(r32 − r22 )
= h(r1 , r2 , r3 ). (9)
362 M. R. Stevanović, P. B. Petrović, and M. M. Stevanović
For the second unknown circle L2 (see Figure 1 (a)) we have SO1 = p − r1 ,
SO2 = p − r2 , SO3 = p − r3 and a corresponding equation in the form of equation
(4), and
F1 (p, −r1 , −r2 , −r3 ) = 0, (10)
2
A2 p + B2 p + C2 = 0, (11)
with
A2 = f1 (−r1 , −r2 , −r3 ) = A1 ,
B2 = g1 (−r1 , −r2 , −r3 ) = B1 ,
C2 = h1 (−r1 , −r2 , −r3 ) = C1 ,
which implies that
A1 p21 + B1 p1 + C1 = 0, A1 p22 − B1 p2 + C1 = 0,
and
1 1 B1
− =− . (12)
p1 p2 C1
For the third circle L3 (see Figure 1 (b)) we have SO1 = p − r1 , SO2 = p + r2 ,
SO3 = p + r3 and we get F1 (p, −r1 , r2 , r3 ) = 0 with A3 p2 + B3 p + C3 = 0,
A3 = f1 (−r1 , r2 , r3 ), B3 = g1 (−r1 , r2 , r3 ), C3 = h1 (−r1 r2 , r3 ) = C1 .
For the fourth circle L4 (see Figure 1 (b)) we have SO1 = p+r1 , SO2 = p−r2 ,
SO3 = p − r3 and we get F1 (p, r1 , −r2 , −r3 ) = 0 with A4 p2 + B4 p + C4 =
0, A4 = f1 (r1 , −r2 , −r3 ) = A3 , B4 = g1 (r1 , −r2 , −r3 ) = −B3 , C4 =
h1 (r1 , −r2 , −r3 ) = C1 and again we get
1 1 B3
− =− . (13)
p3 p4 C1
Analogously, we have
1 1 B5
− = − , (14)
p5 p6 C1
1 1 B7
− = − , (15)
p7 p8 C1
where
B5 = g1 (r1 , −r2 , r3 ), B7 = g1 (r1 , r2 , −r3 ).
Formula (1) follows from (12), (13), (14), (15) because
g1 (r1 , r2 , r3 ) + g1 (−r1 , r2 , r3 ) + g1 (r1 , −r2 , r3 ) + g1 (r1 , r2 , −r3 ) = 0.
Remark 1. If the index j of the circle with radius pj is an even (odd) number, then
1/pj (in formula (1)) has the sign + (−).
Radii of circles in Apollonius’ problem 363
Remark 2. In the first case of the three given circles mentioned in the introduction,
we get the formula
1 1 1 1
= + +
r r1 r2 r3
where r, r1 , r2 , r3 are the inradius and exradii of triangle ABC.
Remark 3. In the second case we get
1 1 1 1 1
+ + + = ,
p1 p2 p3 q m
where p1 , p2 , p3 are the radii of Jenkins circles, q is the radius of Apollonius’ circle
and m = R/2 is the radius of Euler circle or nine-point circle.
This formula can be proved independently since
a b c
p1 = q, p2 = q, p3 = q.
b+c c+a a+b
Remark 4. In the same way, the same result can be proved for the three given
circles, provided that two of them are inside of the third one.
3. Positions of 8 circles
The radical circle of the three given circles K1 (O1 , r1 ), K2 (O2 , r2 ), K3 (O3 , r3 ),
is the circle orthogonal to all of them, and pairs of circles (L1 , L2 ), (L3 , L4 ),
(L5 , L6 ), (L7 , L8 ) – Figure 1, are inversive with respect to the radical circle. For
this radical circle K0 (S0 , r0 ), Figure 2, the following holds true.
O3
b a
S1
S0
S2
O1 c O2
Figure 2. The radical circle of the three given circles with the circles L1 and L2 ,
inversive to the radical circle, based on Figure 1 (a)
364 M. R. Stevanović, P. B. Petrović, and M. M. Stevanović
where
u5 (r1 , r2 , r3 ) = u1 (r1 , r2 , −r3 ),
v5 (r1 , r2 , r3 ) = v1 (r1 , r2 , −r3 ),
w5 (r1 , r2 , r3 ) = w1 (r1 , r2 , −r3 ).
x7 = 2p7 u7 + x0 , y7 = 2p7 v7 + y0 , z7 = 2p7 w7 + z0 ,
x8 = −2p8 u7 + x0 , y8 = −2p8 v7 + y0 , z8 = −2p8 w7 + z0 ,
where
u7 (r1 , r2 , r3 ) = u1 (r1 , −r2 , r3 ),
v7 (r1 , r2 , r3 ) = v1 (r1 , −r2 , r3 ),
w7 (r1 , r2 , r3 ) = w1 (r1 , −r2 , r3 ).
(2) There are collinear triplets of points (S0 , S1 , S2 ), (S0 , S3 , S4 ), (S0 , S5 , S6 )
and (S0 , S7 , S8 ), and
S0 S1 ⊥ q1 : r1 x + r2 y + r3 z = 0,
S0 S3 ⊥ q3 : − r1 x + r2 y + r3 z = 0,
S0 S5 ⊥ q5 : r1 x − r2 y + r3 z = 0,
S0 S7 ⊥ q7 : r1 x + r2 y − r3 z = 0,
where q1 , q3 , q5 , q7 are the lines M12 M22 M32 , M12 M21 M31 , M11 M22 M31 , and
M11 M21 M32 respectively.
(3)
1 1 2
− = , (23)
S0 S1 S2 S0 S0 V1
where U1 = S0 S1 ∩ q1 and V1 and U1 are inversive to each other with respect to
the radical circle. Analogously, this is assumed for the other centers Sj .
4. Three-dimensional case
In this case we have four spheres and a maximum of 16 spheres, each of which
being tangent to all of the four given spheres. Analogous relations for radii of these
16 spheres will be found subsequently. Let us assume that the four spheres are
Φ1 (O1 , r1 ), Φ2 (O2 , r2 ), Φ3 (O3 , r3 ), Φ4 (O4 , r4 ). We can take the basic tetrahedron
ABCD to be tetrahedron O1 O2 O3 O4 with O1 = A(1 : 0 : 0 : 0), O2 = B(0 : 1 :
0 : 0), O3 = C(0 : 0 : 1 : 0), O4 = D(0 : 0 : 0 : 1) given in the barycentric
coordinates with mutual distances AB = c, AC = b, AD = d, BC = a, BD = e,
CD = f . An important role in further investigations is played by Cayley-Menger
determinant ∆0 of tetrahedron ABCD given as follows:
0 1 1 1 1
1 0 c2 b2 d2
∆0 = 1 c2 0 a2 e2 . (24)
1 b2 a2 0 f 2
1 d 2 e2 f 2 0
The known result is that the volume V of tetrahedron ABCD is given by for-
mula
∆0 = 288V 2 .
366 M. R. Stevanović, P. B. Petrović, and M. M. Stevanović
If we apply ∆ij to denote the algebraic cofactor of element with row-column posi-
tion (i, j) in corresponding Cayley-Menger matrix, we obtain the following result.
Proposition 4. (1) The center O of the circumscribed sphere of tetrahedron ABCD
(or circumcenter) has barycentric coordinates
O(∆12 : ∆13 : ∆14 : ∆15 ). (25)
(2) The circumradius R of the upper circumsphere is given by
∆11
R2 = − . (26)
2∆0
(3) If for point P (x : y : z : t) we introduce two relevant expressions
τ = τ (P ) = x + y + z + t, (27)
T = T (P ) = a2 yz + b2 zx + c2 xy + d2 xt + e2 yt + f 2 zt, (28)
then we have
1
τ (O) = ∆0 , T (O) = ∆0 · ∆11 . (29)
2
Let us now introduce the radical sphere Φ0 (S0 , r0 ), i.e., the sphere with property
S0 A2 − r12 = S0 B 2 − r22 = S0 C 2 − r32 = S0 D2 − r42 = r02 , (30)
or the sphere which is orthogonal to the four given spheres Φ1 , Φ2 , Φ3 , Φ4 . Then
we have
Proposition 5. (1) This radical sphere corresponds to the equation
T = τ (r12 x + r22 y + r32 z + r42 t). (31)
(2) The center S0 (x0 : y0 : z0 : t0 ) has coordinates
x0 = ∆12 + r12 ∆22 + r22 ∆32 + r32 ∆42 + r42 ∆52 , (32)
y0 = ∆13 + r12 ∆23 + r22 ∆33 + r32 ∆43 + r42 ∆53 , (33)
z0 = ∆14 + r12 ∆24 + r22 ∆34 + r32 ∆44 + r42 ∆54 , (34)
t0 = ∆15 + r12 ∆25 + r22 ∆35 + r32 ∆45 + r42 ∆55 . (35)
(3) For the radius r0 , the following formula holds.
r02 = R2 − r12 x(M ) + r22 y(M ) + r32 z(M ) + r42 t(M ) ,
(36)
where M is the midpoint of the segment S0 O.
In Figure 3 we introduce corresponding ordered quadruplets. An appropriate
number j in illustration (from (a) to (p)) denotes that sphere Lj is tangent to the
four given spheres. The plus sign in the second position (given in brackets in each
figure) means that sphere with center B is outside-externally tangent to sphere Lj ,
while the minus sign at the fourth position means that sphere with center D is
inside sphere Lj – internally tangent, and similarly in other cases. For each of the
16 possible layouts, corresponding signs are given immediately under the figure,
Radii of circles in Apollonius’ problem 367
C D C D
C D C D
A A
B B
B B
A A
C D D C
C D C D
A A
B B
B B
A A
C D C D
D C C D
A A
B B
B B
A A
D C D C
C D C D
A A
B B
B B
A A
1 2
C1 = r (∆12 + r12 ∆22 + r22 ∆32 + r32 ∆42 + r42 ∆52 )
2 1
+ r22 (∆13 + r12 ∆23 + r22 ∆33 + r32 ∆43 + r42 ∆53 )
+ r32 (∆14 + r12 ∆24 + r22 ∆34 + r32 ∆44 + r42 ∆54 )
+ r42 (∆15 + r12 ∆25 + r22 ∆35 + r32 ∆45 + r42 ∆55 )
+ (∆11 + r12 ∆12 + r22 ∆13 + r32 ∆14 + r42 ∆15 )
= h(r1 , r2 , r3 , r4 ), (48)
For the second unknown sphere L2 (see Figure 3) we have SO1 = p − r1 ,
SO2 = p − r2 , SO3 = p − r3 , SO4 = p − r4 as illustrated by the equation
F1 (p, −r1 , −r2 , −r3 , −r4 ) ≡ A2 p2 +B2 p+C2 ≡ A1 p2 −B1 p+C1 = 0. (49)
Now we get
1 1 B1
− =− . (50)
p1 p2 C1
This means that to the difference 1/p1 − 1/p2 we can relate the ordered quadru-
ple (+, +, +, +) related to (r1 , r2 , r3 , r4 ) since B1 is linear with respect to all
rj . Since C1 is the same for all 16 combinations (ε1 r1 , ε2 r2 , ε3 r3 , ε4 r4 ) for all
ε ∈ {−1, 1}. When combining the signs of the ordered quadruplets, we obtain the
following results given in the next theorem.
Theorem 6. Let us assume that the radii of the sixteen spheres given in Figure 3
are pj (j = 1, 2, ..., 16) and the volume V is different from 0. Then
1 1 1 1 1 1 1 1
− − − − − − − = 0, (51)
p1 p2 p3 p4 p5 p6 p11 p12
1 1 1 1 1 1 1 1
− − − − − − − = 0, (52)
p1 p2 p3 p4 p7 p8 p13 p14
1 1 1 1 1 1 1 1
− − − − − − − = 0, (53)
p1 p2 p3 p4 p9 p10 p15 p16
1 1 1 1 1 1 1 1
− − − − − + − = 0, (54)
p1 p2 p5 p6 p7 p8 p15 p16
1 1 1 1 1 1 1 1
− − − − − + − = 0, (55)
p1 p2 p5 p6 p9 p10 p13 p14
1 1 1 1 1 1 1 1
− − − − − + − = 0, (56)
p1 p2 p7 p8 p9 p10 p11 p12
1 1 1 1 1 1 1 1
− + − − − − −
p1 p2 p3 p4 p5 p6 p7 p8
1 1 1 1 1 1 1 1
− − + − + − + − = 0,
p9 p10 p11 p12 p13 p14 p15 p16
(57)
370 M. R. Stevanović, P. B. Petrović, and M. M. Stevanović
1 1 1 1 1 1 1 1
− − − + − − −
p1 p2 p3 p4 p5 p6 p7 p8
1 1 1 1 1 1 1 1
− − + − − − − − = 0,
p9 p10 p11 p12 p13 p14 p15 p16
(58)
1 1 1 1 1 1 1 1
− − − − − + −
p1 p2 p3 p4 p5 p6 p7 p8
1 1 1 1 1 1 1 1
− − − − + − − − = 0,
p9 p10 p11 p12 p13 p14 p15 p16
(59)
1 1 1 1 1 1 1 1
− − − − − − −
p1 p2 p3 p4 p5 p6 p7 p8
1 1 1 1 1 1 1 1
+ − − − − − + − = 0.
p9 p10 p11 p12 p13 p14 p15 p16
(60)
Corollary 7. From the above formulae we can also obtain
1 1 1 1 1 1 1 1
− − − + − + − = 0, (61)
p3 p4 p5 p6 p13 p14 p15 p16
1 1 1 1 1 1 1 1
− − − + − + − = 0, (62)
p3 p4 p7 p8 p11 p12 p15 p16
1 1 1 1 1 1 1 1
− − − + − + − = 0, (63)
p3 p4 p9 p10 p11 p12 p13 p14
1 1 1 1 1 1 1 1
− − − + − − − = 0, (64)
p5 p6 p7 p8 p11 p12 p13 p14
1 1 1 1 1 1 1 1
− − − + − − − = 0, (65)
p5 p6 p9 p10 p11 p12 p15 p16
1 1 1 1 1 1 1 1
− − − + − − − = 0. (66)
p7 p8 p9 p10 p13 p14 p15 p16
These formulae and formulae listed in Theorem 6 are all possible formulae of
this type.
Corollary 8. For the radius r0 of the radical circle,
C1
r02 = − . (67)
∆0
Proof. From formula (37) we have
T1 = τ1 (c2 y1 + b2 z1 + d2 t1 ) = (p1 + r1 )2 τ1 .
Since
τ1 = τ (S1 ) = τ (S0 ) = ∆0 ,
Radii of circles in Apollonius’ problem 371
References
[1] H. S. M. Coxeter, The Problem of Apollonius, Amer. Math. Monthly, 75 (1968) 5–15.
[2] F. G.-M., Exercices de géométrie, Tours, France: Maison Mame, 18-20 and 663, 1912.
[3] M. Gergonne, Recherche du cercle qui en touche trois autres sur une sphere, Ann. math. pures
appl., 4 (1813–1814).
[4] D. Grinberg and P. Yiu, The Apollonius circle as a Tucker circle, Forum Geom., 2 (2002) 175–
182.
[5] J. C. Lagarias, C. L. Mallows, and A. R. Wilks, Beyond the Descartes Circle Theorem, Amer.
Math. Monthly 109 (2002) 338–361.
372 M. R. Stevanović, P. B. Petrović, and M. M. Stevanović
Predrag B. Petrović: University of Kragujevac, Faculty of Technical Sciences Čačak, Svetog Save
65, 32000 Čačak, Serbia
E-mail address: predrag.petrovic@ftn.kg.ac.rs
Marina M. Stevanović: University of Belgrade, Faculty of Mathematics, Studentski trg 16, 11000
Beograd, Serbia
E-mail address: marina.stevanovic42@gmail.com