Note 1
[G] Euclidean Geometry
Euclidean geometry in my opinion is the hardest section in olympiads despite being so widely used beyond olympiads and competitions. Many problems are solved by angle chasing, finding similar triangles, and only with elementary concepts. However, it is very difficult to utilize such concepts unless you have really good eyes, draws perfectly, or in the best condition. In this note, we will discuss more in-depth about the topics that many of us are familiar with geometry in competitions and gaining better eyes for euclidean geometry.
1.1 Power of a Point
You might already be very familiar with this topic if you have math competition background. However, I wanted to revisit this topic as it is surprisingly common topic in math olympiad. However, the difference could be finding patterns and more advanced topics from power of a point. With that in mind, let’s review some of the topics.
1.1.1 Definitions and Theorems
As always, let’s get started with the definition.
Definition 1.1.1
The power of a point of \(P\) with respect to a circle \(\omega \) represented as \(P_\omega (P) = PO^2 - r^2\) where \(O\) and \(r\) the center and radius of \(\omega \) respectively.
The power of a point is a numeric value that describes the relationship between the circle and the point. Before reviewing the relationship, let’s first derive the formula in the definition. The theorem below is the heart of all concepts in this section.
Theorem 1.1.2: Power of a Point Theorem
Let \(P\) be a point and \(\omega \) be a circle on the same space. If lines \(l_1\) and \(l_2\) that passes through \(P\) meets \(\omega \) at \(A\), \(B\) and \(C\), \(D\) (not necessarily distinct) respectively, then \(PA \cdot PB = PC \cdot PD\).
If you read the other notes, you may have noticed that I try my best to include proofs for all theorems included. However, let’s skip the proof for now for the purpose of this note.
From the theorem above, we can deduce that \(PX \cdot PY\), where \(X\) and \(Y\) are points on \(\omega \) such that \(P, X, Y\) are collinear, is always consistent. With that in mind, consider the following diagram.
By the theorem, \(P_\omega (P) = PX \cdot PY = PB \cdot PB\). Moreover by the definition of a line tangent to a circle, \(PB \perp BO\) and \(PO^2 = PB^2 + r^2\) by Pythagorean Theorem. Therefore, \(P_\omega (P) = PO^2 - r^2\).
With the definition, we can derive trivial relationship between the power of a point and the location of the point. If \(P_\omega (P) < 0\), then \(P\) is inside \(\omega \). If \(P_\omega (P) = 0\), then \(P\) is on \(\omega \) and \(P\) is outside \(\omega \) if and only if \(P_\omega (P) > 0\).
Continuing from the definition, we can derive another priceless definition that is one of the foundational concept in olympiad geometry.
Definition 1.1.3
The radical axis of two circles \(\omega _1\) and \(\omega _2\) with different center is the locus of \(P\) such that \(P_{\omega _1}(P) = P_{\omega _2}(P)\).
Another way of thinking about this is that a point \(P\) is on the radical axis if and only if \(O_1P^2 - r_1^2 = O_2P^2 - r_2^2\) by definition.
Continuing, there is another very properties of radical axis.
Theorem 1.1.4: Properties of Radical Axis
- 1.
- The radical axis is always a straight line.
- 2.
- The radical axis of \(\omega _1\) and \(\omega _2\) is always perpendicular to the line that pass through two origins of \(\omega _1\) and \(\omega _2\).
Let’s first prove the first property.
Proof.
Let \(\omega _1\) be a circle defined as \((x - a_1)^2 + (y - b_1)^2 = r_1^2\). Similarly, let \(\omega _2\) be a circle \((x - a_2)^2 + (y - b_2)^2 = r_1^2\). Let \(P(x, y)\) be any points such that \(P_{\omega _1}(P) = P_{\omega _2}(P)\). Substituting, the following equation holds by definition. \[ (x - a_1)^2 + (y - b_1)^2 - r_1^2 = (x - a_2)^2 + (y - b_2)^2 - r_2^2 \] Expanding and rearranging the equation, \begin{align*} &\quad \ x^2 - 2a_1 x + a_1^2 + y^2 - 2b_1 y + b_1^2 - r_1^2 \\ &= x^2 - 2a_2 x + a_2^2 + y^2 - 2b_2 y + b_2^2 - r_2^2 \\ &(2a_2 - 2a_1) x + (2b_2 - 2b_1) y + C = 0 \end{align*}
holds where \(C\) is the constant values. Notice that the locus of all \((x, y)\) that satisfies the equation above is a linear function. Thus, the radical axis is always linear.
Continuing, we can prove the second property. There are few ways of proving this, but let’s take an algebraic approach with our equation above.
Proof.
Continuing with the equation \[ (2a_2 - 2a_1) x + (2b_2 - 2b_1) y + C = 0 \] from the previous proof, it is evident that the slope of the line is \(-\frac {2a_2 - 2a_1}{2b_2 - 2b_1}\). Continuing, the slope of the line that passes through the two centers of \(\omega _1\) and \(\omega _2\) is \(\frac {b_2 - b_1}{a_2 - a_1}\). Because \(-\frac {2a_2 - 2a_1}{2b_2 - 2b_1} \cdot \frac {b_2 - b_1}{a_2 - a_1} = -1\), the two lines are perpendicular to each other.
Before we move on to further theorems, lemmas, definitions, and problems, here are five special forms that makes our lives easier when graphing to solve problems.
Also for the third diagram, one tip when drawing is that you can find the radical axis by drawing two lines, or segments, that are tangent to both circles. Then, the radical axis is the line that connects the centers of the two segments. Before moving on to common patterns and theorems, here is one more definition to keep in mind.
Definition 1.1.5
Coaxial circles are circles that share the common pairwise radical axis.
Below is an example of three coaxial circles.
The three circles \(\omega _1, \omega _2, \omega _3\) are coaxial because the radical axis for \(\omega _1, \omega _2\), for \(\omega _2, \omega _3\), and for \(\omega _3, \omega _1\) are all equal. It is also worth noting that three circles in this case all passes through two common intersections. Coaxial circles are not limited to three circles, but most problems would be on three circles for olympiads.
With the definition in mind, we can observe some key properties of coaxial circles.
Proof.
Let the centers of \(\omega _1, \omega _2, \omega _3\) be \(O_1, O_2, O_3\) and the radical axis be \(l\). By definition, the radical axis of \(\omega _1\) and \(\omega _2\) is \(l\) and \(O_1O_2 \perp l\). Similarly, \(O_1O_3 \perp l\) since \(l\) is the radical axis of \(\omega _1\) and \(\omega _3\). Because there exists exactly one line perpendicular to \(l\) that passes through \(O_1\), \(O_1O_2\) and \(O_1O_3\) lines are the same line and \(O_1,O_2,O_3\) are collinear.
This theorem can look somewhat trivial. However, finding such conditions and looking for coaxial circles can come very handy.
Never underestimate! Facts like \(OA = OB\) for a circle with center \(O\) and \(P_\omega (A) = P_\omega (B)\) can be the key to the solution. Whenever you see three circles, circles defined by reflection, collinearity of their centers, and/or the intersection of three circles, we should be reminded of coaxial circles and suspect it as the key at least once.
Continuing, we can see the relationship between the three radical axes with three distinct circles. One thing to keep in mind if you are taking Korean Mathematical Olympiad, then the following theorem is known as Monge’s Theorem in addition to one that is related to homothety. However let’s call this Radical Axis Concurrence Theorem since I think that is more universal.
Theorem 1.1.7: Radical Axis Concurrence Theorem
Given three distinct circles, the pairwise radical axes of the circles are concurrent.
Proof.
Let \(\omega _1, \omega _2, \omega _3\) be three distinct circles with the centers \(O_1, O_2, O_3\). Let \(l_1, l_2, l_3\) be the radical axes of the pairs \(O_1, O_2\) and \(O_2, O_3\) and \(O_3, O_1\) respectively. Let \(K\) be the intersection of \(l_1\) and \(l_2\). By definition, \(P_{\omega _1}(K) = P_{\omega _2}(K)\) and \(P_{\omega _2}(K) = P_{\omega _3}(K)\). Therefore, \(P_{\omega _1}(K) = P_{\omega _3}(K)\) and \(K \in l_3\). Therefore, \(l_1, l_2, l_3\) are concurrent.
This is it! Continuing from the theorem, we can define the following term.
Definition 1.1.8
The radical center of three distinct circles is the concurrent point of pairwise radical axes.
Last, but not least, here is a very powerful theorem on coaxial circles. Honestly, this theorem made me suspect power of a point every time I see more than two circles on a problem.
Theorem 1.1.9
Let \(\omega _1\) and \(\omega _2\) be two distinct circles and \(A, B, C\) be three distinct points. If \[ \frac {P_{\omega _1}(A)}{P_{\omega _2}(A)} = \frac {P_{\omega _1}(B)}{P_{\omega _2}(B)} = \frac {P_{\omega _1}(C)}{P_{\omega _2}(C)} \] holds, then either the circumcircle of \(\triangle {ABC}\), \(\omega _1\), and \(\omega _2\) are coaxial circles or \(A, B, C\) is the radical axis of \(\omega _1\) and \(\omega _2\).
Proof.
First, notice that for any circles \(x^2 + y^2 + ax + by + c = 0\) and a point \(A(x_1, y_1)\), \(P_\omega (A) = (x_1 - p)^2 + (y_2 - q)^2 - r^2 = x_1^2 + y_1^2 + ax_1 + by_1 + c\). Moreover, if \(\frac {P_{\omega _1}(A)}{P_{\omega _2}(A)} = \frac {P_{\omega _1}(B)}{P_{\omega _2}(B)} = \frac {P_{\omega _1}(C)}{P_{\omega _2}(C)} = k\), then \(P_{\omega _1}(X) - k P_{\omega _2}(X) = 0\) for \(X = A, B, C\).
Let \(\omega _1\) be defined as \(x^2 + y^2 + a_1x + b_1y + c_1 = 0\) and \(\omega _2\) as \(x^2 + y^2 + a_2x + b_2y + c_2 = 0\). Rewriting the expression \(P_{\omega _1}(X) - k P_{\omega _2}(X)\), the following equations are obtained. \begin{align*} P_{\omega _1}(X) - k P_{\omega _2}(X) &= x^2 + y^2 + a_1x + b_1y + c_1 \\ &\qquad \qquad \qquad \,- k(x^2 + y^2 + a_2x + b_2y + c_2) \\ &= (1 - k) x^2 + (1 - k) y^2 + (a_1 - ka_2) x \\ &\qquad \qquad \qquad \,+ (b_1 - kb_2) y + (c_1 - kc_2) \end{align*}
Define \(\omega _3\) as \((1 - k) x^2 + (1 - k) y^2 + (a_1 - ka_2) x + (b_1 - kb_2) y + (c_1 - kc_2) = 0\) for \(k \neq 1\). Then, \(A, B, C \in \omega _3\) and \(\omega _3\) is the circumcircle of \(\triangle {ABC}\). Moreover by construction, the power of any point \(X\) satisfy the following equation. \[ P_{\omega _3}(X) = \frac {P_{\omega _1}(X) - k P_{\omega _2}(X)}{1 - k} \] Therefore, if \(X\) lies on the radical axis of \(\omega _3\) and \(\omega _1\), i.e. \(P_{\omega _3}(X) = P_{\omega _1}(X)\), then \((1 - k) P_{\omega _1}(X) = P_{\omega _1}(X) - k P_{\omega _2}(X)\) and \(P_{\omega _1}(X) = P_{\omega _2}(X)\). In other words, \(P_{\omega _3}(X) = P_{\omega _1}(X)\) if and only if \(P_{\omega _1}(X) = P_{\omega _2}(X)\) and \(\omega _1, \omega _2, \omega _3\) are coaxial circles.
If \(k = 1\), then \(P_{\omega _1}(A) = P_{\omega _2}(A)\), \(P_{\omega _1}(B) = P_{\omega _2}(B)\), and \(P_{\omega _1}(C) = P_{\omega _2}(C)\) holds and \(A, B, C\) lies on the radical axis of \(\omega _1\) and \(\omega _2\). Because radical axis is a straight line, \(A, B, C\) are collinear and the points form the radical axis.
With the definition and theorems in mind, let’s solve some practice problems to see how they are hidden.
1.1.2 Brain Teasers
The first problem is quite nostalgic and elegant problem that I wished to share.
Exercise 1.1.10: 2025 Korean MO Problem 5 (\(\bullet \circ \circ \))
Let \(D\) be a point on the segment \(BC\) of triangle \(ABC\) such that \(\overline {AD} = \overline {BD} = \frac {1}{2} \overline {CD}\). Let \(P (\neq A, D)\) be a point on segment \(AD\), \(M\) be the midpoint of \(PD\), and \(J\) be the incenter of triangle \(MDC\). If the circumcircles of \(\triangle {MJC}\) and \(\triangle {ABP}\) intersect at two distinct points \(Q\) and \(R\), show that \(D, Q, R\) are collinear.
Solution.
First, let’s draw the diagram.
Notice that because \(Q\) and \(R\) both lies of the circles \(\omega _1\) (left) and \(\omega _2\) (right), it suffices to show that \(D\) lies on the radical axis of \(\omega _1\) and \(\omega _2\). In other words, we must show that \(P_{\omega _1} (D) = P_{\omega _2} (D)\). Consider the following equation. \[ P_{\omega _1} (D) = \overline {AD} \cdot \overline {PD} = \overline {BD} \cdot (2 \cdot \overline {MD}) = 2 \overline {BD} \cdot \overline {MD} = \overline {DC} \cdot \overline {MD} \] Therefore, it suffices to show that \(P_{\omega _2} (D) = \overline {DC} \cdot \overline {MD}\).
To find \(P_{\omega _2}\), we can use another classic pattern that we can see from the problem. We see the point \(D\), the incenter \(J\) and the circumcircle of \(\triangle {MJC}\). Naturally trying to implement incenter/excenter lemma here, we can draw and find excenter of \(\triangle {MDC}\) at \(D\).
Notice that \(\angle {MDJ} = \angle {JDC}\) by construction. Similarly, \(\angle {MI_DJ} = \angle {MCJ} = \angle {JCD}\). Thus \(\triangle {MDI_D}\) and \(\triangle {JDC}\) are similar. Because \(MD : JD = DI_D : DC\), \(P_{\omega _2}(D) = DJ \cdot DI_D = MD \cdot DC\).
Never overestimate! I remember trying my best not to use power of a point during the exam because the morning problems were hard that day, and ended up wasting lots of time in the afternoon exam. I panicked, and learned a lesson.