An Introduction to Ptolemy's Theorem

Qi Zhu

[email protected]

Contents

1 Introductory questions concerning the handout 2

2 Ptolemy's Theorem - The key of this Handout 2

3 Example Problems 5
3.1 Carnot's Theorem - Noticing Congurations . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 IMO 1995, #5 - Symmetry in Hexagon . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4 Problems - Practice makes perfect 8

1
Qi Zhu Ptolemy's Theorem 2016, September 05

1 Introductory questions concerning the handout

Ptolemy's Theorem can be powerful in easy problems, as well as in tough Olympiad problems. Often,
it is hard to spot the ingenious use of Ptolemy. As there are not many introductions to Ptolemy's
Theorem, I dedicated my time to write a fullling, but rather easy introduction.
You should read this article, if
• You don't know Ptolemy's Theorem.
• You don't know Ptolemy's Theorem very well.
• You know Ptolemy's Theorem, but you are rusty.

• You are an expert, but still want to learn more. (Or you just want to criticize my failures.)
• You do not know at least 6 proofs of the theorem.
• You want to help me improve my writing skills.
• You just want to make me happy. ♥

And so on! In general, we will nd several proofs to Ptolemy's Theorem, discuss a few examples and
at last, there will be a bunch of problems to practice for yourself.
The handout itself is not too dicult. It starts from easier problems and goes up to early Olympiad
level. Basic geometry understanding suce to understand this handout. In specic, you should be
condent with angle chasing, similar triangles and low tech theorems. Having some knowledge in basic
trigonometry wouldn't hurt, though.

2 Ptolemy's Theorem - The key of this Handout

Ptolemy's Theorem
If ABCD is a (possibly degenerate) cyclic quadrilateral, then
| AB| · |CD | + | AD | · | BC | = | AC | · | BD |.

D C

Figure 1: Cyclic quadrilateral ABCD

Proof. Spoilers ahead! Try proving it by yourself rst, then come back. There are many possibilities
to do so.

2
Qi Zhu Ptolemy's Theorem 2016, September 05

Pick point E on line BD, such that ∠DEA = ∠CBA. For convenience, call
a = | AB|, b = | BC |, c = |CD |, d = | DA|, e = | AC |, f = | BD |.

Ptolemy's Theorem would then be rewritten into

ac + bd = e f .

Now by construction, observe 4 ABE ∼ 4 ACD and 4 AED ∼ 4 ABC. (Why? Hint: Basic angle
chasing theorems in the circle. Look at Figure 2.) Thus, we have |BE
a
|
= ce and |ED
d
|
= de , so
ac = e · | BE| and bd = e · | ED |.
Recall | BE| + | ED | = f . Hence, adding them gives
ac + bd = e · | BE| + e · | ED | = e(| BE| + | ED |) = e f ,

which ends the proof. As for the degenerate case, note that quadrilateral ABCD then lies on a circle
with radius ∞. Doing Exercise 2.1 will make this clearer for you. If you're not too condent with
that, you may also prove it by bashing out some lengths.

D C

Figure 2: Angle Chasing for similar triangles

In fact, Ptolemy's Theorem is just a tiny part of what we will be looking at. You have probably heard
of the triangle inequality. But have you ever wondered about whether there is a similar inequality for
quadrilaterals? Then Ptolemy's Inequality is what you were looking for.
Ptolemy's Inequality
If A, B, C, D are four points in the plane a , then
| AB| · |CD | + | AD | · | BC | ≥ | AC | · | BD |.

Equality is achieved, if and only if ABCD is a (possibly degenerate) cyclic quadrilateral.

a Interestingly enough, that can actually be generalised for all dimensions.

Proof. Again, take a look by yourself rst.

We want to kind of mimic the proof of the equality case. That is, our proof now will be motivated
by the proof you've already seen. Let X and Y be points on the diagonals BD and AC, such that

3
Qi Zhu Ptolemy's Theorem 2016, September 05

∠ BAX = ∠CAD and ∠YBA = ∠ DCA. Let E be the intersection of AX and BY. Then, it is easy
to see that ∠BAC = ∠EAD, so 4 ABE ∼ 4 ACD. That gives
| AB| | BE|
= ⇐⇒ | AB| · |CD | = | AC | · | BE|. (1)
| AC | |CD |

The similarity also gives us || AC

AB|
|
= ||AD
AE|
|
. Since ∠BAE = ∠EAD, by S : A : S similarity, we get
4 AED ∼ 4 ABC. This similarity then gives
| AD | | ED |
= ⇐⇒ | AD | · | BC | = | AC | · | ED |. (2)
| AC | | BC |

Adding the equations (1) and (2) yields

| AB| · |CD | + | DA| · | BC | = | AC | · (| BE| + | ED |) ≥ | AC | · | BD |

by the triangle inequality. Equality holds, if and only if 4 BED is degenerate, that is E lies on BD. That
happens if and only if ∠DCA = ∠YBA = ∠DBA or equivalently if ABCD is a (possibly degenerate)
cyclic quadrilateral.

D
C

X
A E

Figure 3: Mimicing the equality case with two new points

We'll proceed by discovering more proofs of the theorems. You may skip this, if you want to, but it is
good practice and you should denitely do it somewhen. There are elegant proofs to Ptolemy, as well
as instructive ones that explain the method being used very well.
Exercise 2.1. Prove Ptolemy's Inequality with Inversion . Try to cover the equality case with that
proof as well.
Exercise 2.2. Prove Ptolemy's Inequality with Complex Numbers . Again, try to cover the equality
case.
Exercise 2.3. Prove Ptolemy's Inequality with Simson's Line. Cover the equality case.
Exercise 2.4. Prove Ptolemy's Theorem with trigonometry. Think of Addition Theorems or the Law
of Cosine .
Exercise 2.5. Can you nd dierent proofs that weren't mentioned in this handout? 1

1 If you do, then feel free to send me your proof ! Contact me at [email protected].

4
Qi Zhu Ptolemy's Theorem 2016, September 05

3 Example Problems

In this section, I will be presenting 2 problems to give a general idea of how Ptolemy's Theorem may
be used. In specic, I will try to explain motivational steps and include a write-up as I'd do it in a
contest. You should try them yourself before reading the solution.

3.1 Carnot's Theorem - Noticing Congurations

Problem 3.1. Let d a , d b , d c be the distances from the circumcenter of an acute triangle to its sides
and let R and r be its circumradius and inradius respectively. Prove that da + db + dc = R + r.
Solving 3.1. It's good to start a problem labeling key objects. Let's do that. Let the feet of the
perpendiculars from the circumcenter O to triangle ABC be X, Y, Z and let a, b, c be the sidelengths.
Then, I see that XBYO, YCZO, ZAXO are all cyclic as they have two opposite right angles. I also
note that X, Y, Z are actually midpoints of the sides. That also tells me that 4 XYZ is actually the
Medial Triangle. A cool conguration I have back in my mind! So | XY | =
2 b, |YZ | = 2 c, | ZX | = 2 a.
1 1 1

For convenience, call 21 a = a0 , 12 b = b0 , 12 c = c0 . Also, we dened a lot of heights with da , db , dc onto

those sides. It tells me that I should look out for areas. That technique is often used when the lengths
of the perpendiculars from a point to triangle sides is given.

Z Y
O

A X B

Figure 4: The Medial Triangle and all its might

Anyway, I proceed by using Ptolemy three times on the three cyclic quadrilaterals XBYO, YCZO, ZAXO,
since I know the side lengths of those quadrilateral well and it is connected to the relation I am trying
to prove. It gives me
a0 dc + c0 d a = Rb0
a0 db + b0 d a = Rc0
b0 dc + c0 db = Ra0

It seems like, somehow combining them could lead to some canceling. So I try some combinations.
One of them being adding all three equations. That yields
a 0 ( d b + d c ) + b 0 ( d c + d a ) + c 0 ( d a + d b ) = R ( a 0 + b 0 + c 0 ).

I'd like to divide by a0 + b0 + c0 . I'd already have the summand R on the right hand side. And the left
hand side almost has factor a0 + b0 + c0 . So perhaps I'd add something to the equation. That is why I
add a0 da + b0 db + c0 dc to both sides. This gives
( a0 + b0 + c0 )(d a + db + dc ) = R( a0 + b0 + c0 ) + a0 d a + b0 db + c0 dc .

5
Qi Zhu Ptolemy's Theorem 2016, September 05

So now it suces to prove

a0 d a + b0 db + c0 dc
= r.
a0 + b0 + c0
I understand the denominator very well, it is just the semiperimeter, which we call s. So it is equivalent
to
a0 d a + b0 db + c0 dc = rs = [ ABC ]
where [ ABC] is the area of 4 ABC. Note that [ ABC] = rs is well known. But wait. The left hand
side can be easily expressed as [ ABC]. I see
a0 d a + b0 db + c0 dc = [ BCO] + [CAO] + [ ABO] = [ ABC ].

So that's where the area came into play!

Proof. Let 4 ABC be one such triangle and denote its side lengths with a, b, c. Call the feet of the
perpendiculars from circumcenter O to the sides X, Y, Z. Then XBYO is cyclic, since opposite angles
add up to 180◦ due to ∠BXO = ∠OYB = 90◦ . Similarly, YCZO, ZAXO are also cyclic. Note that
X, Y, Z are the midpoints of the sides and | XY | = 12 b, |YZ | = 12 c, | ZX | = 12 a as 4 XYZ is the medial
triangle of 4 ABC. Let 12 a = a0 , 12 b = b0 , 21 c = c0 . Then, by Ptolemy on XBYO, YCZO, ZAXO, we
get
a0 dc + c0 d a = Rb0
a0 db + b0 d a = Rc0
b0 dc + c0 db = Ra0

Adding them and a0 da + b0 db + c0 dc yields

a0 d a + b0 db + c0 dc
d a + db + dc = R + .
a0 + b0 + c0
But
a0 d a + b0 db + c0 dc [ BCO] + [CAO] + [ ABO] [ ABC ]
0 0 0
= 0 0 0
= 0 = r.
a +b +c a +b +c a + b0 + c0
So we are done.

3.2 IMO 1995, #5 - Symmetry in Hexagon

Problem 3.2. Let ABCDEF be a convex hexagon with | AB| = | BC| = |CD| and | DE| = | EF| =
| FA|, such that ∠ BCD = ∠EFA = 60◦ . Suppose G and H are points in the interior of the hexagon
such that ∠ AGB = ∠DHE = 120◦ . Prove that | AG| + |GB| + |GH | + | DH | + | HE| ≥ |CF |.
Solving 3.2. There are certain conditions that seem like they could be dicult to handle. Hexagons
can be dicult and the inequality uses a good number of dierent lengths, that could be disgusting.
But on the other hand, there are also several conditions that seem easy to handle. I like 60◦ or 120◦
angles and a lot of lengths are equal.
Now, the rst diculty I have had with this problem is that I did not know how to draw a diagram.
And things get dicult in geometry, if you cannot visualize it, so a good diagram is crucial. Therefore,
the rst task should be trying to draw such a hexagon. For that, I made a quick scratch without the
ruler to see, if there is any special property that I could use to draw the hexagon. And with that, I
noticed that 4 BCD and 4 EFA are actually equilateral, as they are both isosceles with a 60◦ angle.
That lets me label the diagram a bit more by letting
| AB| = | BC | = |CD | = | BD | = a and | DE| = | EF | = | FA| = | EA| = b.

Now, I'd start by drawing segment AB. I could follow up with a segment BD with equal length. I knew
that I could just construct point C on the far right, so I didn't care about that for now. I knew that

6
Qi Zhu Ptolemy's Theorem 2016, September 05

4 ADE was isosceles with | DE| = | EA| = b as I was looking at the diagram, so point E must lie on
the perpendicular bisector of segment AD. I drew the perpendicular bisector, picked a suitable point E
and constructed equilateral triangles with segments BD and EA outside of quadrilateral ABDE. I've
succeeded in drawing the hexagon ABCDEF!
Time to look at the problem itself. I notice that the given angles are 60◦ and 120◦ and they conveniently
add up to 180◦ . That smells like cyclic quadrilaterals. Thus, it makes me think about a point C0 , such
that AC0 BG is a cyclic quadrilateral. A good point to choose would probably be a point that makes
4 AC 0 B equilateral. So I chose that. Now AC 0 BG is cyclic and | AC 0 | = |C 0 B| = | BA| = a. The
diagonals and side lengths of that quadrilateral are all interesting to me, so that's when I use Ptolemy
on the newly constructed AC0 BG. It gives me
| AG | · a + | GB| · a = |C 0 G | · a ⇐⇒ | AG | + | GB| = |C 0 G |.

By symmetry, I know that we also get | DH | + | HE| = | HF0 |.

F0

H D

F
C
G

B
A

C0

Figure 5: So much symmetry!

Now the given inequality seems easier to handle. It is equivalent to

|C 0 G | + | GH | + | HF 0 | ≥ | FC |

I then note that points G and H are arbitrary points on those circles I've discovered with the cyclic
quadrilaterals. Hence, it could be possible that C0 , G, H, F0 are collinear. It seems good to use the
triangle inequality now to say
|C 0 G | + | GH | + | HF 0 | ≥ | F 0 C 0 |.
Now, it suces to prove | F0 C0 | ≥ | FC |. But look at the construction. Looking at the gure, this
seems obvious. That construction is so symmetric, by symmetry, we'd get | F0 C0 | = | FC |, it should not

7
Qi Zhu Ptolemy's Theorem 2016, September 05

be hard to prove it rigorously. Some congruency or rotating should do it. Cool, I've solved an IMO
problem. Yay.
Proof. Let C0 and F0 be points outside of hexagon ABCDEF such that 4 AC0 B and 4 DF0 E are
equilateral. As ∠ AGB + ∠BC0 A = 180◦ , quadrilateral AC0 BD is cyclic. Therefore, by Ptolemy's
Theorem we have
| AG | · | BC 0 | + | GB| · |C 0 A| = |C 0 G | · | AB| ⇐⇒ | AG | + | GB| = |C 0 G |,

since | BC0 | = |C0 A| = | AB| by construction. Similarly, | DH | + | HE| = | F0 H |. Thus,

| AG | + | GB| + | GH | + | DH | + | HE| = |C 0 G | + | GH | + | HF 0 | ≥ | F 0 C 0 |,

with two uses of the triangle inequality |C0 G | + |GH | ≥ |C0 H | and |C0 H | + | HF0 | ≥ | F0 C0 |. But we
also have | FC | = | F0 C0 |, as hexagons AXBDYE and ABCDEF are congruent because all correspond-
ing sides and angles are equal.

4 Problems - Practice makes perfect

In the concluding section 2 , I will include several problems for you to practice on your own. They are
roughly arranged in diculty, even though I cannot assure that you will feel the same way. Take your
time for the problems and don't give up too early.
Most problems will be solvable with Ptolemy, you might nd solutions without using Ptolemy, though.
I even threw in a few problems that can not necesarilly by solved with Ptolemy to let you think about
all possible approaches, instead of being too focused on Ptolemy. (Or maybe I didn't? Find it out by
yourself!)
Also, if you know cool problems related to Ptolemy or any of the problems in this article, please contact
me as well! Have fun and good luck!
Problem 4.1 (NIMO 14.1). Let A, B, C, D be four points on a line in this order. Suppose that
| AC | = 25, | BD | = 40, and | AD | = 57. Compute | AB| · |CD | + | AD | · | BC |.
Problem 4.2. Point P is chosen on the arc CD of the circumcircle of a square ABCD. Prove that
√
| PA| + | PC | = 2 · | PB|.

Problem 4.3. Let triangle ABC be isosceles with | AC | = | BC| and let P be a point of arc AB of its
circumcircle. Then prove that |PA||+|
PC |
PB|
is constant for all possible points P.
Problem 4.4. The angle bisector of ∠BAC of triangle ABC meets its circumcircle at D. Prove that
| AB| + | AC | ≤ 2 · | AD |.

Problem 4.5. Let ABCD be a cyclic quadrilateral with ∠CBA = ∠ ADC = 90◦ . Prove that
| BD | = | AC | · sin ∠ BAD.

Problem 4.6 (Baltic Way 2001, #7). A parallelogram ABCD is given. A circle passing through A
meets the line segments AB, AC and AD at inner points M, K, N , respectively. Prove that
| AB| · | AM| + | AD | · | AN | = | AK | · | AC |.
2I might make documents with hints and solutions in the near future. But until that happens, feel free to ask me via
[email protected], if any questions come up. I'd also take a look at your solutions, if you want to.

8
Qi Zhu Ptolemy's Theorem 2016, September 05

Problem 4.7 (Geometry Revisited, 1.9.1). Let Q be a point of segment BC of an equilateral triangle
ABC. Let line AQ meet the circumcenter again at P. Then prove
1 1 1
+ = .
| PA| | PC | | PQ|

Problem 4.8 (Germany 2014). Let ABCDEFG be a regular heptagon with side length 1. Prove that
1 1
+ = 1.
| AC | | AD |

Problem 4.9. Let α = π7 . Prove that

1 1 1
= + .
sin α sin 2α sin 3α
Problem 4.10 (HMMT 2013 Team - #6). Let triangle ABC satisfy 2| BC | = | AB | + | AC | and have
incenter I and circumcircle ω. Let D be the intersection of AI and ω (with A, D distinct). Prove
that I is the midpoint of AD.
Problem 4.11 (IMO 2001, #6). Let a > b > c > d be positive integers and suppose that
ac + bd = (b + d + a − c)(b + d − a + c).

Prove that ab + cd is not prime.

References

[1] Oleg Golberg, Ptolemy's Theorem. URL: http://www.ideamath.org/samplegeo6.pdf

[2] Evan Chen, Euclidean Geometry in Mathematical Olympiads, MAA Press, 2016.
[3] Eckard Specht and Robert Strich, Geometria - Scientiae Atlantis , Otto-von-Guericke-Universität
Magdeburg, 2009.
[4] Nguyen Ngoc Giang, M.Sc., Ptolemy's Inequality, Mathematical Excalibur, 2013.
[5] Kin-Yin Li, Ptolemy's Theorem, Mathematical Excalibur, 1996.
[6] H.S.M Coxeter and S.L.Greitzer, Geometry Revisited, Mathematical Association of America