Art of Problem Solving

WOOT 2011–12
Inversion

1 Introduction

Geometric inversion is a technique for transforming diagrams, like using homothety or reflection. (In fact,
inversion can be thought of as reflecting through a circle, as opposed to reflecting through a line.) The
power of inversion lies in the fact that an inverted diagram has many of the same properties as the original
diagram, but it can be much simpler to work with. In this handout, we describe the properties of inversion,
and how it can be applied to solve geometry problems.

2 Definitions
−−→
Given a circle ω centered at O with radius r, and a point P other than O, let P 0 be the point on ray OP
such that OP · OP 0 = r2 . Then P 0 is the inverse of P with respect to the circle ω. (The inverse of O is the
point at infinity, denoted by P∞ .)

P0
ω
P

The following properties are evident:

• If P 0 is the inverse of P , then P is the inverse of P 0 . Thus, we can refer to pairs of inverse points.

• A point is its own inverse if and only if it lies on the circle.

• A point lies inside the circle if and only if its inverse lies outside the circle.

1
 Art of Problem Solving
WOOT 2011–12
Inversion

3 Inverses of Two Points

Let A and B be two points, and let A0 and B 0 be their inverses with respect to circle ω, respectively.

A0
ω
A
O
B
B0

Then
OA0 r2 /OA OB
= = ,
OB 0 r2 /OB OA
and ∠A0 OB 0 = ∠BOA, so triangles A0 OB 0 and BOA are similar. Hence, the distance A0 B 0 is given by

OA0 r2 /OA r2 · AB
A0 B 0 = · AB = · AB = .
OB OB OA · OB
We will refer to this formula as the inversion distance formula.
Furthermore, ∠AA0 B 0 = ∠OA0 B 0 = ∠OBA = 180◦ − ∠ABB 0 , so quadrilateral ABB 0 A0 is cyclic. (This also
follows from OA · OA0 = OB · OB 0 = r2 .)

A0
ω
A
O
B
B0

2
 Art of Problem Solving
WOOT 2011–12
Inversion

4 Inverses of Lines and Circle

4.1 The Inverse of a Line Passing Through the Center

If ` is a line passing through the center O of circle ω, then the inverse of ` with respect to ω is itself.

`
ω

4.2 The Inverse of a Line Not Passing Through the Center

Let ` be a line that does not pass through the center O of circle ω. Let P be an arbitrary point on `, and let
Q be the projection of O onto `. Let P 0 and Q0 be the inverses of P and Q with respect to ω, respectively.
Then quadrilateral P QQ0 P 0 is cyclic.

ω P
0
P

O Q
Q0

Since ∠P QQ0 = 90◦ , it follows that ∠P P 0 Q0 = 90◦ . Then ∠OP 0 Q0 = 90◦ , which means that P 0 lies on the
circle with diameter OQ0 . Conversely, if P 0 is a point on the circle with diameter OQ0 other than O, then let
P be the inverse of P 0 . Then by the same argument, ∠P QQ0 = 90◦ , so P lies on `. Therefore, the inverse
of ` with respect to ω is the circle with diameter OQ0 . (In particular, the circle passes through O.)

3
 Art of Problem Solving
WOOT 2011–12
Inversion

4.3 The Inverse of a Circle Passing Through the Center

Let γ be a circle that passes through the center O of circle ω. Then by the same argument as in the previous
subsection, the inverse of γ is a line (not passing through O).

ω
γ

4.4 The Inverse of a Circle Not Passing Through the Center

Let γ be a circle that does not pass through the center O of circle ω. Let O1 be the center of circle γ, and
let OO1 intersect γ at Q and R. Let P be an arbitrary point on γ, and let P 0 , Q0 , and R0 be the inverses of
P , Q, and R with respect to ω, respectively.

ω
P γ
P0
O
R0 Q0 Q O1 R

4
 Art of Problem Solving
WOOT 2011–12
Inversion

Since P lies on γ, ∠QP R = 90◦ . Quadrilateral P QQ0 P 0 is cyclic, so ∠P 0 Q0 R0 = 180◦ − ∠P 0 Q0 Q = ∠P 0 P Q.

Quadrilateral P RR0 P 0 is cyclic, so ∠P 0 R0 Q0 = 180◦ − ∠P 0 P R. Then

∠Q0 P 0 R0 = 180◦ − ∠P 0 Q0 R0 − ∠P 0 R0 Q0
= 180◦ − ∠P 0 P Q − (180◦ − ∠P 0 P R)
= ∠P 0 P R − ∠P 0 P Q
= ∠QP R
= 90◦ ,

which means that P 0 lies on the circle with diameter Q0 R0 . Conversely, if P 0 is a point on the circle with
diameter Q0 R0 , then let P be the inverse of P 0 . Then by the same argument, ∠QP R = 90◦ , so P lies on γ.
Therefore, the inverse of γ with respect to ω is the circle with diameter Q0 R0 . (In particular, the circle does
not pass through O.)
Note: If O2 is the center of the circle with diameter Q0 R0 , then in general, O2 is not the inverse of O1 with
respect to ω. In other words, when inversion maps circles to circles, it does not necessarily map the centers
of those circles to each other.
We can summarize our results as follows:

• The inverse of a line passing through O is itself.

• The inverse of a line not passing through O is a circle (passing through O).
• The inverse of a circle passing through O is a line (not passing through O).
• The inverse of a circle not passing through O is another circle (not passing through O).

5
 Art of Problem Solving
WOOT 2011–12
Inversion

5 Inversion and Angles

Let α and β be two curves in the plane that intersect at P , and assume there exist tangents t and u to
these curves at P , respectively. Then the angle between curves α and β at P is defined as the angle between
tangents t and u.

P
t β
ω
t0
P0

O u0

Let P 0 , t0 , and u0 be the inverses of P , t, and u with respect to circle ω, respectively. (We assume that the
center O of circle ω does not lie on t or u; these cases can be dealt with easily, using the same argument.)
Then t0 and u0 are circles that intersect at O and P 0 . Note that the angle between t0 and u0 at P 0 is equal
to the angle between t0 and u0 at O.
The tangent to t0 at O is parallel to t, and the tangent to u0 at O is parallel to u. (These observations follow
from our proof in Subsection 4.2.) Then the angle between t0 and u0 at O is equal to the angle between t and
u. Therefore, the angle between t0 and u0 at P 0 is also equal to the angle between t and u. Hence, inversion
preserves angles.
Note: We are not saying that if A0 , B 0 , and C 0 are the inverses of A, B, and C, respectively, then ∠B 0 A0 C 0 =
∠BAC. We are saying that inversion preserves angles at intersections of curves.

6
 Art of Problem Solving
WOOT 2011–12
Inversion

We say that two curves are orthogonal if they intersect at right angles. In particular, two circles α and β
with centers O1 and O2 , respectively, are orthogonal if and only if ∠O1 P O2 = 90◦ , where P is a point where
the two circles intersect. Equivalently, the tangent from O1 to β is equal to the radius of α (and the tangent
from O2 to α is equal to the radius of β).

α
P β

O1 O2

Let γ be a circle that is orthogonal to ω. Let γ and ω intersect at A and B, and let γ 0 be the inverse of
circle γ with respect to ω. Then γ 0 is also a circle that passes through A and B.

P0
γ
A
ω
P
B
O

But since inversion preserves angles as intersections of curves, and ω is its own inverse, γ 0 is also orthogonal
to ω. Since there is only one circle passing through A and B that is orthogonal to ω, circle γ is its own
inverse. (In other words, if P is a point on γ, then the inverse P 0 of P with respect to ω also lies on γ.)
Conversely, by the same argument, any circle (other than ω) that is its own inverse with respect to ω is
orthogonal to ω.
We can prove another converse of this result: Let P and P 0 be two distinct points that are inverses with
respect to ω, and let γ be a circle that passes through P and P 0 . We will show that γ is its own inverse with
respect to ω.

7
 Art of Problem Solving
WOOT 2011–12
Inversion

Since P and P 0 are distinct, and inverses with respect to ω, one of them lies inside ω, and the other lies
outside ω. Hence, γ must intersect ω at two points, say A and B.
Let γ 0 be the inverse of γ with respect to ω. Then γ 0 passes through P , P 0 , A, and B. But γ also passes
through these points, and four points uniquely determine a circle, so again, circles γ and γ 0 coincide, which
means that γ is its own inverse.
For a circle γ other than ω, we have established that the following statements are equivalent:

• The circle γ is orthogonal to ω.

• The circle γ is its own inverse with respect to ω.

• The circle γ passes through two distinct points P and P 0 that are inverses with respect to ω.

6 Inverting Through the Incircle

Let ω, I, and r denote the incircle, incenter, and inradius of triangle ABC, respectively. Let the incircle ω
be tangent to sides BC, CA, and AB at D, E, and F , respectively.

ω
E
F I

B D C

We invert the diagram through the incircle ω.

8
 Art of Problem Solving
WOOT 2011–12
Inversion

First, we consider the inverse of line BC. The inverse of line BC is a circle passing through I, say ωA . Since
BC is tangent to ω at D, the circle ωA is also tangent to ω at D. Therefore, ωA must be the circle with
diameter ID.

ω
E
F
ωA I

B D C

Similarly, the inverse of AC is the circle with diameter IE, say ωB , and the inverse of AB is the circle with
diameter IF , say ωC .
Now, let A0 , B 0 , and C 0 be the inverses of A, B, and C, respectively. Since A is the intersection of AB and
AC, A0 is the intersection of ωB and ωC other than I. Similarly, B 0 is the intersection of ωA and ωC other
than I, and C 0 is the intersection of ωA and ωB other than I.

ω
ωB
E
ωC

A0
F

I
C0

B0 ωA

Since A0 lies on ωB , the circle with diameter IE, ∠IA0 E = 90◦ . But A0 also lies on ωC , the circle with
diameter IF , so ∠IA0 F = 90◦ . Hence, E, A0 , and F are collinear. In fact, A0 is the midpoint of EF , since

9
 Art of Problem Solving
WOOT 2011–12
Inversion

IA0 is perpendicular to chord EF . By the same argument, B 0 is the midpoint of DF , and C 0 is the midpoint
of DE. We conclude that the circumcircle of triangle A0 B 0 C 0 (which is the inverse of the circumcircle of
triangle ABC) is also the nine-point circle of triangle DEF .

ω
ωB
E
ωC

A0
F

I
C0

B0 ωA

Hence, the circumradius of triangle A0 B 0 C 0 is r/2, half the circumradius of triangle DEF . If O and R are
the circumcenter and circumradius of triangle ABC, respectively, then clearly IO < R, so by Exercise 1 (at
the end of the handout),
r r2 · R
= 2 ,
2 R − IO2
which implies that
IO2 = R2 − 2Rr.

7 Worked-out Examples

Problem. (Ptolemy’s Theorem) Let ABCD be a cyclic quadrilateral. Prove that

AB · CD + BC · DA = AC · BD.

10
 Art of Problem Solving
WOOT 2011–12
Inversion

Solution. Let ω be a circle centered at A with radius r, and let B 0 , C 0 , and D0 be the inverses of B, C, and
D with respect to ω, respectively.

A B B0

D
C0
C

D0

Since B, C, and D lie on a circle passing through A, the points B 0 , C 0 , and D0 lie on a line, in that order,
so B 0 C 0 + C 0 D0 = B 0 D0 . Then from the inversion distance formula,
r2 · BC r2 · CD r2 · BD
+ = .
AB · AC AC · AD AB · AD
Multiplying both sides by AB · AC · AD and dividing by r2 , we get
AB · CD + BC · DA = AC · BD.

Problem. (Apollonius Circle) Let A and B be two points in the plane, and let k 6= 1 be a positive real
number. Show that the locus of points P such that AP/P B = k is a circle.

Solution. Let P be a point such that AP/P B = k. Let ω be a circle centered at A with radius r, and let
B 0 and P 0 be the inverses of B and P with respect to ω, respectively.

ω
P
P0

A
B0 B

11
 Art of Problem Solving
WOOT 2011–12
Inversion

Then by inversion distance formula,

r2 · BP r2
B0P 0 = = ,
AB · AP k · AB
which is a fixed quantity. Hence, the locus of P 0 is a circle centered at B 0 . Furthermore,

r2
AB 0 = ,
AB
which is not equal to B 0 P 0 since k 6= 1. Therefore, the locus of P 0 is a circle that does not pass through A,
which means that locus of P is a circle. 
Problem. Let α and β be two circles that do not intersect. Prove that there exists a circle ω such that the
inverses α0 and β 0 of α and β with respect to ω, respectively, are concentric.

Solution. If α and β are already concentric, then we can take ω to be any circle that is concentric with α
and β, so assume that α and β are not concentric.
We start by looking at the completed diagram. Let m be the line joining the centers of α and β. By
symmetry, it makes sense to look for a circle ω whose center O lies on m.

β ω

α β0
α0 O
m

12
 Art of Problem Solving
WOOT 2011–12
Inversion

To help us locate O, let γ 0 be the line passing through the common centers of α0 and β 0 that is perpendicular
to m. Then line γ 0 is orthogonal to both α0 and β 0 , so γ is a circle that is orthogonal to both α and β. (Also,
γ passes through O.) Thus, we must find a circle γ that is orthogonal to both α and β.

β ω
γ
α β0
0
α O m

γ0

Let R be the center of circle γ. If γ is orthogonal to α, then the tangent from R to α is equal to the radius of
γ. Similarly, if γ is orthogonal to β, then the tangent from R to β is also equal to the radius of γ. Therefore,
R must lie on the radical axis of α and β.

β ω
γ
α β0
R α0 O m

γ0

13
 Art of Problem Solving
WOOT 2011–12
Inversion

We can now construct the circle ω as follows: Let m be the line joining the centers of α and β, let ` be the
radical axis of α and β, and let R be the intersection of ` and m.

β ω
γ
α β0
R α0 O m

` γ0

Since R lies on the radical axis `, the tangents from R to α and β are equal. (Since α and β do not intersect,
` lies outside both α and β, so R lies outside both α and β.) Let γ be the circle centered at R whose radius
is equal to the common tangent. Then γ is orthogonal to both α and β.
Let O be an intersection of γ and m, and let ω be any circle centered at O. Let α0 and β 0 be the inverses of
α and β with respect to ω, respectively. Since the centers of both α and β lie on m, the centers of both α0
and β 0 also lie on m.
Since γ is a circle that passes through O and whose center R lies on m, the inverse of γ with respect to ω is
a line γ 0 that is perpendicular to m. And since α and β are orthogonal to γ, α0 and β 0 are orthogonal to γ 0 ,
which means that the centers of both α0 and β 0 lie on γ 0 . Therefore, circles α0 and β 0 are concentric. (Both
circles are centered at the intersection of m and γ 0 .) 

14
 Art of Problem Solving
WOOT 2011–12
Inversion

An elegant corollary of this problem is a result known as Steiner’s Porism: Let α and β be two circles, so
that α lies inside β. Given a circle that is externally tangent to α and internally tangent to β, we construct
a chain of circles starting with that circle, so that each circle in the chain is also externally tangent to α and
internally tangent to β, and tangent to the previous circle in the chain. If the chain closes for some initial
circle, then the chain closes for any initial circle.

α β0

α0
β

We can invert circles α and β to obtain two concentric circles α0 and β 0 , and the result follows.
Problem. Let A, B, and C be three collinear points. We construct the semicircles with diameters AB, AC,
and BC. Let ω0 be the circle with diameter BC, and for all n ≥ 1, let ωn be the circle that is externally
tangent to circle ωn−1 and semicircle BC, and internally tangent to semicircle AC, as shown below.
Let On and rn denote the center and radius of circle ωn , respectively. Show that the distance from On to
AC is 2nrn for all n.

ω3 ω2
ω4
ω1

A B C

15
 Art of Problem Solving
WOOT 2011–12
Inversion

Solution. We invert the diagram through a circle centered at A, so that circle ωn is its own inverse. (For
example, we can take the radius of inversion to be the length of the tangent from A to circle ωn .) The case
n = 3 is illustrated below.

O40

O3 O2
O4
O20 O1

O10

O00 O0
A C0 B0 B C

Let B 0 and C 0 be the inverses of B and C, respectively. The inverse of semicircle AB is a ray starting
at B 0 that is perpendicular to AC. Similarly, the inverse of semicircle AC is a ray starting at C 0 that is
perpendicular to AC.
Let ωi0 be the inverse of circle ωi . Each circle ωi is tangent to semicircles AB and AC, so each circle ωi0 is
tangent to both rays. Hence, all circles ωi0 are congruent. But circle ωn is its own inverse, so each circle ωi0
has radius rn . Furthermore, consecutive circles are externally tangent, and the center of circle ω00 lies on
AC, so the distance from On to AC is 2nrn . 

8 Tips on Using Inversion

• Inversion works best on problems that involve circles that are either tangent to each other or tangent
to lines.
• Ideally, an inverted diagram should be simpler than the original diagram. Since inversion can take
circles to lines, and lines are easier to work with than circles, try to choose a point that produces as
many lines (and reduces as many circles) as possible. Points that generally achieve this are points of
tangency, or points that many circles pass through.
• In most cases, the radius of inversion is not important, because changing the radius of inversion only
changes the scale of the inverted diagram. However, in other cases, it is useful to superimpose the
inverted diagram on top of the original diagram. (See the problem on page 15 for an example.) If
there are elements in the original diagram and inverted diagram that can be aligned nicely, this is a
sign that the choice of radius may be important.

16
 Art of Problem Solving
WOOT 2011–12
Inversion

9 Exercises
1. Let ω be a circle centered at O with radius r. Let ω1 be a circle centered at O1 with radius r1 , and let
d1 = OO1 . Prove the following:
(a) If d1 > r1 , then the inverse of circle ω1 with respect to ω is a circle centered at O2 with radius
r2 , where
r2 d1
OO2 = 2 ,
d1 − r12
and
r2 r1
r2 = .
d21− r12
(b) If d1 = r1 , i.e. circle ω1 passes through O, then the inverse of circle ω1 with respect to ω is a line
whose distance from O is
r2
.
2r1
(c) If d1 < r1 , then the inverse of circle ω1 with respect to ω is a circle centered at O2 with radius
r2 , where
r2 d1
OO2 = 2 ,
r1 − d21
and
r2 r1
r2 = .
r12 − d21
2. Let D be a point in triangle ABC such that AB = ab, AC = ac, AD = ad, BC = bc, BD = bd, and
CD = cd. Prove that ∠ABD + ∠ACD = π/3. (Singapore, 2004)
3. Two circles Γ and Γ0 intersect at A and D, as shown below. A line is tangent to Γ and Γ0 at E and F ,
respectively. The line through D parallel to EF intersects Γ and Γ0 again at C and B, respectively.
Show that the circumcircles of triangles BDE and CDF intersect again on the line AD.

Γ0
B D C

A
F E

17
 Art of Problem Solving
WOOT 2011–12
Inversion

4. Let ω be a circle centered at O. Let A0 , B 0 , and C 0 be the inverses of A, B, and C with respect to ω,
respectively.
(a) Let I be the incenter of triangle ABC, and let I 0 be the inverse of I with respect to ω. Prove
that I 0 is the A-excenter of triangle AB 0 C 0 .
(b) Show that O is the orthocenter of triangle ABC if and only if O is the incenter or an excenter of
triangle A0 B 0 C 0 .
5. Let ` be a fixed line, and let O be a point not on `. Let A and B be variable points on ` such that
∠AOB is constant. Show that the circumcircle of triangle OAB is tangent to a fixed circle.

`
A B

6. A circle C0 of radius 1 km is tangent to a line L at Z. A circle C1 of radius 1 mm is drawn tangent

to both C0 and L, on the right-hand side of C0 . A family of circles Ci is constructed outwardly to the
right side so that each Ci is tangent to C0 , L, and to the previous circle Ci−1 . Eventually the members
become so large that it is impossible to enlarge the family any further. How many circles can be drawn
before this happens?

C0

L
Z

7. Let α and β be two circles, so that α lies inside β. Let ω1 , ω2 , and ω3 be three circles, such that each
circle ωi is externally tangent to α and internally tangent to β. For 1 ≤ i ≤ 3, let ωi be tangent to α
and β at Si and Ti , respectively. Prove that S1 T1 , S2 T2 , and S3 T3 are concurrent.

18
 Art of Problem Solving
WOOT 2011–12
Inversion

T1
ω1 β
α
S1

S3
S2

ω2 ω3
T2 T3

8. Watch the YouTube video “Moebius Transformations Revealed”:

[Link]

19