\ie\cf
\newcommand{\eqdef}{\mathbin{\stackrel{\rm def}{=}}}  \newcommand{\R} % real numbers  \newcommand{\N}} % natural numbers  \newcommand{\Z} % integers  \newcommand{\F} % a field  \newcommand{\Q} % the rationals  \newcommand{\C}{\mathbb{C}} % the complexes  \newcommand{\poly}}  \newcommand{\polylog}}  \newcommand{\loglog}}}  \newcommand{\zo}{\{0,1\}}  \newcommand{\suchthat}  \newcommand{\pr}[1]{\Pr\left[#1\right]}  \newcommand{\deffont}{\em}  \newcommand{\getsr}{\mathbin{\stackrel{\mbox{\tiny R}}{\gets}}}  \newcommand{\Exp}{\mathop{\mathrm E}\displaylimits} % expectation  \newcommand{\Var}{\mathop{\mathrm Var}\displaylimits} % variance  \newcommand{\xor}{\oplus}  \newcommand{\GF}{\mathrm{GF}}  \newcommand{\eps}{\varepsilon}  \notag
\no
Back to MB3

Week 13 Notes

100%

This week, we'll again be focusing on geometry.

Coordinate Geometry

In coordinate geometry, we add a coordinate grid to our geometric objects. For two-dimensional objects like circles and squares, the coordinate grid is the familiar \(x-y\) plane.

In a coordinate grid, there are two axes: the horizontal \(x\)-axis and the vertical \(y\)-axis. Both axes are real number lines, and any point in the grid is denoted by the corresponding points on the \(x\) and \(y\) axes. For example, the point \(A\) in the diagram above, if moved downwards towards the \(x\) axis, will meet the axis at value \(20\), and if moved leftwards toward the \(y\)-axis, will meet the axis at the value \(15\). So, the point is written as \((20,15)\).

The process of associating each point in the grid with \((x,y)\) coordinates is called the cartesian coordinate system. We've already seen alternate coordinate systems, like polar coordinates, in previous weeks. For now, we'll exclusively use the Cartesian system.

The origin is the point \((0,0)\). The beauty of coordinate grids is that given a geometric object, we can place the origin of the coordinate grid on the object wherever we like. We'll see examples of how that is useful soon.

Parameterizing Lines

Once we pick a coordinate system, we want to express our geometric objects as equations. Let's start by parameterizing line segments.

Given points \((a_1,b_1)\) and \((a_2,b_2)\), how do we write an equation to represent the line segment between the points?

We can think about it like a travelling problem: Suppose you are at your friend's house at point \((a_1,b_1)\), and you walk in a straight line starting at time \(t=0\) to your house at point \((a_2,b_2)\). Suppose you reach your house at time \(t=1\). At what point are you located at time \(0\leq t\leq 1\)?

The answer is the following: $$(a_2t+a_1(1-t),b_2t+b_1(1-t))$$As an example, if \((a_1,b_1)=(5,2)\) and \((a_2,b_2)=(2,5)\), the following Desmos plot will show your walking path between the two houses.

In other words, our formula for the line segment is the following: $$\{(a_2t+a_1(1-t),b_2t+b_1(1-t))\mid 0\leq t\leq 1\}$$

What about the line through the two points? The formula is quite similar:$$\{(a_2t+a_1(1-t),b_2t+b_1(1-t))\mid -\infty\lt t\lt\infty\}$$This is called the parameterization of a line.

Equations of conics

What about other shapes? How do we come up with equations for them? For polygons, we can treat the shapes as a collection of line segments. But curved surfaces are a bit more complicated.

Luckily for us, there's an important class of curved shapes that have quadratic formulas.

A conic is the set of points \((x,y)\) satisfying the following equation $$Ax^2+Bxy+Cy^2+Dx+Ey+F=0$$for some coefficients \(A,B,C,D,E\).

In other words, conics are the graphs of arbitrary quadratic polynomials. But what do they look like?

The Double Cone

The double cone is a geometric shape, composed of two infinitely tall cones that share the same vertex. It looks something like this:

Although not pictured here, both cones go off to infinity. The double cone is quite important, since the set of shapes formed by intersecting a plane with the double cone are the conics. Here are some examples:

The pictured conics are

  1. Circle
  2. Ellipse
  3. Parabola
  4. Hyperbola

The circle is a type of ellipse, so really, there are only \(3\) types of conics in the picture above.

Types of Conics

Given two points \(F_1,F_2\) in a plane, the ellipse of diameter \(D\) focused on the two points is the set of points \(X\) such that \(\overline{XF_1}+\overline{XF_2}=D\).

Some examples of ellipses with the same diameter are as follows:

Note that when \(F_1=F_2\), we get a circle centered at the focus points. It goes without saying that the diameter must be bigger than \(\overline{F_1F_2}\) for there to be an ellipse (what happens when \(D=\overline{F_1F_2}\)?)

So what is the formula of an ellipse? If \(F_1\) and \(F_2\) are both on the \(x\)-axis (or in other words, the ellipse is horizontal), and the midpoint of \(\overline{F_1F_2}\) is the origin (or in other words, the ellipse is centered at the origin), then the formula for the ellipse is $$\frac{4x^2}{D^2}+\frac{4y^2}{D^2-\overline{F_1F_2}^2}=1$$Replacing \(\frac 4{D^2}\) with \(A\) and \(\frac4{D^2-\overline{F_1F_2}^2}\) with \(C\) gives us $$Ax^2+Cy^2=1$$where \(A,C\geq0\). This is the general form of the horizontal ellipse. The general form of the equation of an ellipse can be found here.

Given a focus point \((x_f,y_f)\) and a line not containing the point \(ax+by+c=0\), a parabola focused on \((x_f,y_f)\) around the directrix \(ax+by+c=0\) is defined as the set of points \((x,y)\) such that the distance from the point \((x,y)\) to the directrix is identical to the distance from the point to the focus.

You are likely unfamiliar with this definition of a parabola, but you can easily see why it is identical to the one you are familiar with.

The pink lines are equal in length, showing that the distance to the focus and directrix are the same. So how do we find the formula of the parabola? We simply use the distance formula to find the distance to the focus, and then realize that the shortest distance between a line and a point is the line segment from the point perpendicular to the line. With this knowledge, we get the following formula for the parabola:

$$\frac{(ax+by+c)^2}{a^2+b^2}=(x-x_f)^2+(y-y_f)^2$$This formula is probably not that familiar, but let's set \(a=0\). Then, we get $$y^2+\frac cby+\frac{c^2}{b^2}=x^2-2xx_f+x_f^2+y^2-2yy_f+y_f^2$$$$y=\frac{b^2x^2-2b^2xx_f+b^2x_f^2+b^2y_f^2-c^2}{2b^2y_f+bc}$$Letting \(A=\frac{b^2}{2b^2y_f+bc}\), \(B=-2x_fA\), \(C=\frac{b^2x_f^2+b^2y_f^2-c^2}{2b^2y_f+bc}\), we get $$y=Ax^2+Bx+C$$Whenever the directrix is parallel to the \(x\)-axis, the formula for the parabola is the one we know well.

Given two points \(F_1,F_2\) in a plane, the hyperbola of diameter \(D\) focused on the two points is the set of points \(X\) such that \(|\overline{XF_1}-\overline{XF_2}|=D\).

Hyperbolas looks like two swooshes on the same graph. The following graph plots the following hyperbolas: \(xy=1,\;xy=4,\;xy=9\).

The horizontal hyperbola (the hyperbola with the \(x,y\)-axes as the lines of symmetry) has the following standard equation: $$Ax^2-Cy^2=1$$where \(A,C\geq0\).

The \(3\) types of conics mentioned above are called non-degenerate conics, but there are other types of conics. What are the other possible intersections between a plane and the double cone?


Degenerate conics are called degenerate since their shapes don't have any curves. So given a quadratic equation \(Ax^2+Bxy+Cy^2+Dx+Ey+F=0\), how can we tell if it is an ellipse, a hyperbola, or parabola, assuming it isn't degenerate? We use the discriminant. In the quadratic equation, we use the discriminant \(b^2-4ac\) to determine whether there are \(2\) real roots, \(1\) real root, or no real roots. For conics, we have the following result:

Dealing with Points

In Coordinate Geometry, we often have a collection of points, and want to either find the area of the polygon with the points as vertices, or find a polynomial that goes through the points. Luckily for us, there are ways to do both.

Suppose that our points are \((a_1,b_1),\;(a_2,b_2)\), ..., \((a_n,b_n)\). The area of the polygon with these points as consecutive vertices is $$A=\frac12|a_1b_2+a_2b_3+\ldots+a_{n-1}b_n+a_nb_1-b_1a_2-b_2a_3+\ldots-b_na_1|$$

The shoelace theorem is named as such because the pattern of the products look like shoelaces.

Given points \((a_1,b_1),\;(a_2,b_2)\), ..., \((a_n,b_n)\) such that all of the \(a_i\) are distinct, there exists a unique polynomial \(p\) with degree \(\lt n\) such that \(p(a_i)=b_i\) for each \(1\leq i\leq n\).


Existence: We first want to find \(p\). For each \(1\leq i\leq n\), let \(g_i\) be the polynomial$$g_i(x)=(x-a_1)(x-a_2)\ldots(x-a_{i-1})(x-a_{i+1})\ldots(x-a_n)$$In other words, \(g_i\) is the product of all terms \(x-a_j\) except \(x-a_i\). Because of this \(g_i(a_j)=0\) for all \(1\leq j\leq n\) except \(i\). Let \(S_i=\frac{b_i}{g_i(a_i)}\cdot g_i\). Since \(S_i\) is a constant times \(g_i\), and \(g_i\) has degree \(n-1\), \(S_i\) has degree \(n-1\) and
  • \(S_i(a_j)=0\) whenever \(i\neq j\)
  • \(S_i(a_i)=b_i\)
Finally, let \(p(x)=\sum\limits_{i=1}^nS_i(x)\). Thus, $$p(a_j)=\sum\limits_{i=0}^nS_i(a_j)=0+0+\ldots+0+S_i(a_i)+0+\ldots+0=b_i$$And, since each \(S_i\) has degree \(n-1\), \(p\) has degree less than \(n\).

Uniqueness: Now, we want to show that there is only one \(p\) that satisfies the desired property. Let \(f,g\) both have degree \(\lt n\) and satisfy \(f(a_i)=g(a_i)=b_i\) for all \(i\). Thus, \(a_i\) is a root of \(f-g\) for each \(i\). Thus, \((x-a_1)(x-a_2)\ldots(x-a_n)\) must be a factor of \(f-g\). Since this product has degree \(n\), which is higher than the degree of \(f-g\), \(f-g\) must be \(0\), and thus, \(f=g\).

Conclusion

Coordinate Geometry is an alternate approach to geometry that focuses on establishing a coordinate system, and exploiting that coordinate system to find quantities that would be hard to find otherwise. The study of conics isn't all that important for competition math, but is hugely important for computation coordinate geometry, since computers use conics all the time to compute the trajectories of geometric objects. Here are an exercise that shows the difference in approach between coordinate geometry and standard geometry. Feel free to use whichever is more convenient.

Rectange \(ABCD\) has \(\overline{AB}=5\) and \(\overline{BC}=4\). \(E\) lies on \(\overline{AB}\) such that \(\overline{EB}=1\), \(G\) lies on \(\overline{BC}\) such that \(\overline{CG}=1\), and \(F\) lies on \(\overline{CD}\) such that \(\overline{DF}=1\). In the diagram below, what is \(\frac{PQ}{EF}\)?


Solution 1 (Coordinate Geometry): Let \(D\) be the origin. We want to find the equations of the lines \(AG,AC,EF\). By the given information, their slopes are \(-\frac35,\;-\frac45,\;2\) respectively. Using the point-slope formula, we can find their equations to be
  • \(AG:\;-\frac35x+4\)
  • \(AC:\;-\frac45x+4\)
  • \(EF:\;2x-4\)
Since \(P\) is the intersection of lines \(AC\) and \(EF\), and \(Q\) the intersection of lines \(AG\) and \(EF\), we can find the points to be \(P=\left(\frac{20}7,\frac{12}7\right)\) and \(Q=\left(\frac{40}{13},\frac{28}{13}\right)\). We already know that \(E=(4,4)\) and \(F=(2,0)\). So, we find the answer using the distance formula to be \(\frac{10}{91}\).


Solution 2 (Standard Geometry): Let \(H\) be the intersection between lines \(AG\) and \(CD\)

Let \(x=\overline{HC}\). Thus, \(\triangle HGC\sim\triangle HDA\), so $$\frac{HC}{CG}=\frac{HD}{AD}\to\frac x1=\frac{x+5}4\to x=\frac53$$Next, observe that \(\triangle AEQ\sim\triangle HFQ\) and \(\triangle AEP\sim\triangle CFP\). Thus, $$\frac{AE}{HF}=\frac{EQ}{QF}\to \frac{4}{3+\frac53}=\frac67=\frac{EQ}{QF}$$So, \(\frac{EQ}{EF}=\frac6{13}\) and $$\frac{AE}{CF}=\frac{EP}{FP}\to\frac43=\frac{EP}{FP}$$Thus \(\frac{FP}{EF}=\frac37\). Finally, this implies \(\frac{PQ}{EF}=1-\frac6{13}-\frac37=\frac{10}{91}\).

In general, the strategy for coordinate geometry is to come up with equations for all of the important shapes (lines and curves), and then use algebra to find their points of intersection. For the remainder of these notes, we'll be focusing on the most special of conics: circles.

Circle Geometry

Let's begin our discussion of circles with some definitions.

We often talk about the angle subtended by a chord. By this, we mean the angle formed by the endpoints of the chord and the center of the circle, as in the following diagram:

In the diagram above, the chord \(AB\) is subtended by \(\angle AOB\). We also say that the measure of arc \(AB\), the part of the circumference of the circle between \(A\) and \(B\), is \(\angle AOB\).

If we are using radians, there is a nice formula relating the length of an arc and its measure: \(s=r\theta\). \(s\) is the length of the arc, \(r\) is the radius, and \(\theta\) is the measure of the arc in radians.

Chords of the same length have the same measure. To see why, consider the following diagram:

By SSS, both triangles are congruent, and thus the angles \(\angle AOB\) and \(\angle POQ\) are congruent.

An interior angle \(\angle APB\) is an angle with \(A,P,B\) all points on the circle. For any interior angle \(\angle APB\), the measure of the arc \(AB\) is twice \(\angle APB\), as in the following diagrams:

In each of the three cases above, \(\angle AOB=2\angle APB\). In particular, when \(\angle AOB\) is \(180^\circ\), i.e., it is a straight line, \(\angle APB\) is a right angle. So, for any right triangle, the hypotenuse will be the diameter of the circumscribing circle. The proof of this fact relies on the following diagram.

Since \(\triangle AVO\) is isosceles, \(\angle AVO=\angle VAO\), and thus \(\angle AOB=\angle AVO+\angle VAO\), so \(\angle AOB=2\angle AVO\).

Secant Theorems

There are many special relationships between secants and angles. The simplest one is that tangents are always perpendicular to the radius at the intersection point.

There are three interesting types of secant diagrams we'll look at, as shown below. \(O\) stands for the center of the circle.

In diagram \(1\), \(\angle AEB=\frac{\angle AOB+\angle DOC}2\). In diagram \(2\), \(\angle BAC=\angle DAC\). In diagram \(3\), \(\angle ACE=\frac{\angle AOE-\angle BOD}2\).


Diagram 1: Since \(E\) is on \(AC\), \(\angle ACB=\angle ECB\). By the interior angle theorem, \(\angle ECB=\angle ACB=\frac12\angle AOB\). Likewise, \(\angle EBC=\angle DBC=\frac12\angle DOC\). And, \(\angle AEB=\angle ECB+\angle EBC\).


Diagram 2: Note that triangles \(\triangle ACO\), \(\triangle ADO\), and \(\triangle CDO\) are all isosceles. So, let \(\angle ACO=\angle CAO=\alpha\), \(\angle DCO=\angle CDO=\beta\), \(\angle ADO=\angle DAO=\gamma\). Since triangles have \(180^\circ\), we have $$2\alpha+2\beta+2\gamma=180^\circ\to\alpha=90^\circ-\beta-\gamma$$Since \(\angle ADC=\angle ADO+\angle CDO\), \(\angle ADC=\beta+\gamma\). So, \(\angle CAO+\angle ADC=90^\circ\). Since tangents are perpendicular to radii, \(\angle CAO+\angle BAC=90^\circ\). Thus, \(\angle BAC=\angle ADC\).


Diagram 3: Since \(\angle ACE +\angle BED=\angle ABE\), \(\angle ABE=\frac12\angle AOE\), and \(\angle BED=\frac12\angle BOD\), the desired claim is true.

Hopefully you should be able to see that these three statements are really all the same statement, just in different forms. With this theorem in hand, we can prove one of the most important theorems in circle geometry.

Given a circle and a point, there exists a constant \(C\) such that for any line through the point that intersects the circle, the distance between the first intersection and the point, times the distance between the second intersection and the point equals \(C\). We call \(C\) the power of the point. In particular, in the secant diagrams above:
  1. There are two chords that intersect inside the circle. Thus, \(AE\cdot EC=DE\cdot EB\).
  2. There is a secant and tangent eminating from a point outside the circle. Thus, \(AB\cdot AB=BC\cdot BD\).
  3. There are two secants that intersect at a point outside the circle. Thus, \(AC\cdot BC=EC\cdot DC\).


Diagram 1: As we proved in the Secant Angles Theorem via looking at the subtended arc, \(\angle ECB=\angle EDA\) and \(\angle EAD=\angle EBC\). Thus, \(\triangle AED\sim\triangle BEC\). So, \(\frac{EC}{ED}=\frac{AE}{BE}\). Multiplying by \(ED\cdot BE\) gives us the desired result.


Diagram 2: Since \(\angle ABC=\angle ABD\) and \(\angle BAC=\angle ADC\) via the Secant Angles Theorem, \(\triangle ABD\sim\triangle ABC\). So, \(\frac{AB}{AD}=\frac{BC}{AB}\). Multiplying by \(AD\cdot AB\) gives us the desired relation.


Diagram 3: Draw a tangent from point \(C\) to the circle, and apply the result from Diagram \(2\) to both secants. This completes the proof.
Show that the power of point \(P\) is \(\left\vert\overline{OP}^2-R^2\right\vert\), where \(R\) is the radius and \(O\) is the center of the circle.


If \(P\) is inside or on the circle, then consider the points of intersection \(A,B\) of the diameter that goes through \(P\). Thus, \(\overline{AP}=R-\overline{OP}\) and \(\overline{BP}=R+\overline{OP}\), or vice versa, so $$\overline{AP}\cdot\overline{BP}=(R+\overline{OP})(R-\overline{OP})=R^2-\overline{OP}^2$$If \(P\) is outside the circle, we know that the power of \(P\) is the square of the length of the tangent from \(P\) to the circle. Let \(T\) be the point of tangency. Then, \(\triangle OPT\) is a right triangle, so $$\overline{TP}^2=\overline{OP}^2-R^2$$

We often will define the power of the point as \(\overline{OP}^2-R^2\), which is identical to the definition above, except that powers of points inside the circle are negated. This gives us the following important result.

Given two circles \(\omega_1,\omega_2\) with distinct centers \(O_1,O_2\) and radii \(R_1,R_2\), we define the radical axis of the two circles to be the set of points such that their powers with respect to both circles are the same. The following statements hold:
  1. The radical axis of \(\omega_1\) and \(\omega_2\) is a line perpendicular to \(\overline{O_1O_2}\)
  2. The radical axis goes through any points of intersection of the circles \(\omega_1,\omega_2\)
  3. There exists a unique point \(P\) on the line \(O_1O_2\) such that \(P\) is on the radical axis.


Assume that the line \(O_1,O_2\) is the \(x\)-axis (rotate the plane if this isn't the case).
  1. Let \(P\) be any point in the plane, and \(P'\) the intersection of the line \(O_1O_2\) and the line perpendicular to it through \(P\). By the Pythagorean Theorem, for \(i=1,2\)$$O_iP^2=O_iP'^2+PP'^2$$Thus, \(P\) is on the radical axis if and only if \(P'\) is. When combined with the result of part \(3\), this completes the proof.
  2. Any point on the intersection of \(\omega_1\) and \(\omega_2\) has power \(0\) with respect to both circles, and thus is on the radical axis.
  3. For some point \(P\) on line \(O_1O_2\), let \(d\) be the length \(PO_1\), and \(w\) the length \(O_1O_2\). If point \(P\) is to the left of \(O_1\), then the powers of \(P\) with respect to both circles are equal if and only if $$\begin{aligned}(d+w)^2-d^2&=R_2^2-R_1^2\\d&=\frac{R_2^2-R_1^2-w^2}{2w}\end{aligned}$$If \(P\) is to the right of \(O_1O_2\), the powers of \(P\) with respect to both circles are equal if and only if $$\begin{aligned}(d-w)^2-d^2&=R_2^2-R_1^2\\-d&=\frac{R_2^2-R_1^2-w^2}{2w}\end{aligned}$$So, if \(\frac{R_2^2-R_1^2-w^2}{2w}\) is \(0\), the only point that satisfies \(d=0\) is \(O_1\). If \(\frac{R_2^2-R_1^2-w^2}{2w}\gt0\), the second equation is impossible to satisfy as \(d\geq0\), and there is exactly one point on line to the left of \(O_1\) with distance \(d\) from \(O_1\). If \(\frac{R_2^2-R_1^2-w^2}{2w}\lt0\), the first equation is impossible to satisfy as \(d\geq0\), and there is exactly one point on the line to the right of \(O_1\) with distance \(d\) from \(O_1\).
Given \(3\) circles \(\omega_1,\omega_2,\omega_3\) with centers not all on the same line, their pairwise radical axes intersect at a single point.


Let \(P\) be the point of intersection between the radical axis of circles \(\omega_1\) and \(\omega_2\), and the radical axis of circles \(\omega_2\) and \(\omega_3\). If \(P_1,P_2,P_3\) are the powers of \(P\) with respect to circles \(\omega_1,\omega_2,\omega_3\), then since \(P\) is on the radical axis of \(\omega_1\) and \(\omega_2\), \(P_1=P_2\), and because \(P\) is on the radical axis of \(\omega_2\) and \(\omega_3\), \(P_2=P_3\). Thus, \(P_1=P_3\), so \(P\) is on the radical axis of \(\omega_1\) and \(\omega_3\).

Cyclic Quadrilaterals

Any quadrilateral whose vertices are on a circle is called a cyclic quadrilateral. Cyclic quadrilaterals are defined by the following theorem.

A quadrilateral \(ABCD\) is a cyclic quadrilateral if and only if \(\angle ABC+\angle CDA=180^\circ\).


If \(ABCD\) is a cyclic quadrilateral, then if the circle has center \(O\), \(\angle AOC+\angle COA=360^\circ\), so \(\angle ABC+\angle CDA=180^\circ\) by the interior angle theorem.


If \(\angle ABC+\angle CDA=180^\circ\), then consider the circumcircle of \(\triangle ABC\). Call the center \(O\) and radius \(r\). There are \(3\) cases
  1. \(D\) is inside the circle: Let \(E\) be the other intersection of \(AD\) with the circle, and let \(F\) be the other intersection of \(CD\) with the circle. Then, by the Secant Angles Theorem, $$\angle CDA=\frac{\angle COA+\angle EOF}2=\angle BFE+\angle AFC$$So, $$\angle CDA+\angle ABC=\angle BFE+\angle AFC+\angle ABC=\angle BFE+180^\circ$$This is a contradiction.
  2. If \(D\) is outside the circle, then define \(E,F\) identically to the first part. Thus, $$\angle CDA=\frac{\angle AOC-\angle EOF}2=\angle AFC-\angle BFE$$So, $$\angle CDA+\angle ABC=\angle AFC-\angle BFE+\angle ABC=180^\circ-\angle BFE$$a contradiction.
  3. \(D\) is on the circle: This is the only possibility, finishing the theorem.

Now that we have a way to check if a quadrilateral is cyclic, we can introduce two useful theorem involving cyclic quadrilaterals.

Given cyclic quadrilateral \(ABCD\), the following equation is true: $$\overline{AB}\cdot\overline{CD}+\overline{BC}\cdot\overline{DA}=\overline{AC}\cdot\overline{BD}$$


Pick a point \(P\) on \(\overline{BD}\) such that \(\angle DAP=\angle CAB\)


Since \(\angle ADP=\angle ADB=\angle ACB\), \(\triangle DAP\sim\triangle CAB\), so $$\frac{DP}{BC}=\frac{AD}{AC}$$Also, \(\angle APB=\angle180^\circ-\angle APD=180^\circ-\angle ABC=\angle ADC\) and \(\angle ABP=\angle ACD\), so \(\triangle APB\sim\triangle ADC\). Hence,$$\frac{BP}{CD}=\frac{AB}{AC}$$Combining these equations gives us$$\frac{BC\cdot AD+AB\cdot CD}{AC}=BP+DP=BD$$Multiplying by \(AC\) gives us Ptolemy's Theorem.
Given a cyclic quadrilateral with side lengths \(a,b,c,d\), if we let \(s=\frac{a+b+c+d}2\), the area of the quadrilateral is $$A=\sqrt{(s-a)(s-b)(s-c)(s-d)}$$

Challenge: Projective Geometry

Projective Geometry is a very important concept, but likely one you haven't seen in mathematics. In fact, the first place most people see it is in art class, not in math class. But what is it?

History of Projective Thinking

One of the first historical records involving projective thinking was that of Italian artist Filippo Brunelleschi. Brunelleschi demonstrated the geometrical method of perspective by painting the outlines of various Florentine buildings onto a mirror. Soon after, nearly every artist in Florence and in Italy used geometrical perspective in their paintings, notably Masolino da Panicale and Donatello. Not only was perspective a way of showing depth, it was also a new method of composing a painting. Paintings began to show a single, unified scene, rather than a combination of several.

As shown by the quick proliferation of accurate perspective paintings in Florence, Brunelleschi likely understood (with help from his friend the mathematician Toscanelli), but did not publish, the mathematics behind perspective. Decades later, his friend Leon Battista Alberti wrote De pictura in 1435, a treatise on proper methods of showing distance in painting based on Euclidean (standard) geometry. However, these ideas remained the work of artists for around 100 years, when the architect Gerard Desargues rediscovered the notion of "points at infinity". Desargues advanced Brunelleschi's work by establishing the mathematics of not just lines and distance, but also that of conics, in paintings with vanishing points. Blaise Pascal, finding the writings of Desargues, was so inspired that he came up with the now famous Pascal's Theorem.

Unfortunately, just a few years later, Rene Descartes published a breakthrough work introducing cartesian coordinates, and work on projective geometry was mostly abandoned.

It was only in 1822, when mathematician Jean-Victor Poncelet published a foundational treatise on projective geometry, that interest in projective geometry reemerged.

Since then, it has revolutionized the foundations of algebraic geometry, influencing the works of Riemann, Dirac, Einstein, Bohr, etc.

Motivation

Draw a picture of a large, flat plain with a pair of railroad tracks running through it. It looks something like the picture below. When we draw a perspective picture like this, the parallel train tracks appear to actually meet at the horizon. Let’s change the rules of geometry a little so that this actually happens.

So, we define the projective plane initially as just the ordinary Real plane. But, like in our railroad, we want our parallel lines to eventually meet. So to our real plane, we add points, one for each set of parallel lines in the plane. We call these points the points at infinity. Thus, any two parallel lines intersect at the point at infinity corresponding to their slope.

We also add a line at infinity. It just passes through all of the points at infinity.

So, a normal line and the line at infinity intersect at the point at infinity corresponding to the slope of the normal line. So, every pair of distinct lines intersect at a single point.

And, every pair of distinct points has a single line passing through them. There's a notion of duality here: every two points determines a line, and every two lines determines a point. This is a key property of projective geometry.

If you can't visualize what this space looks like, don't worry! For now, as long as you can appreciate the duality of points and lines in projective space, you are on your way to understanding it.

Homothety

Before dealing with the projective plane, let's try to understand the duality between lines and points via projections in the standard plane. The simplest example of projection is the introduction of a vanishing point into the plane. A vanishing point is simply a point around which we project objects, like in the diagram below

Point \(O\), the vanishing point, projects the square \(ABCD\) to the new square \(A'B'C'D'\). We call this projection homothety. The scale factor is the ratio \(\frac{OD'}{OD}\). We describe this homothety as \(H(O,D,D')\), or \(H\left(O,\frac{OD'}{OD}\right)\).

Given a vanishing point \((a,b)\) and a scale factor \(k\), \(H((a,b),k)\) sends point \((x,y)\) to \((k(x-a)+a,k(y-b)+b)\). Some examples of homothety are presented below.

Prove the following basic properties of homothety:
  • The image of any object under homothety is similar to the object
  • A homothety is uniquely defined by its action on two points
  • The slope of a line under homothety is preserved.
  • Angles are preserved under homothety
  • For \(k\neq1\), the lines preserved by the homothety \(H(P,k)\) are the lines that go through \(P\)
  • A shape with area \(A\) after homothety \(H(P,k)\) has area \(k^2A\), and a line segment with length \(l\) after the same homothety has length \(|k|l\).

As in art, the fun really begins when we introduce multiple vanishing points. How do multiple homotheties interact?

The composition of homotheties \(H((a,b),k)\) and \(H((c,d),l)\) is a homothety \(H((e,f),kl)\) where \((a,b)\), \((c,d)\), \((e,f)\) are collinear, assuming \(kl\neq1\).


Let $$\begin{aligned}e&=\frac{kla+(c-a)l-c}{kl-1}\\f&=\frac{klb+(d-b)l-d}{kl-1}\end{aligned}$$Check that \(H((a,b),k)\) followed by \(H((c,d),l)\) is the same as \(H((e,f),kl)\) via the formula. And, the line through points \((a,b)\) and \((c,d)\) is given by \(y-b=\frac{d-b}{c-a}\cdot(x-a)\). Thus, we want to show that \((e,f)\) is a solution to this equation.$$\begin{aligned}klb+(d-b)l-d-klb+b&=\frac{d-b}{c-a}\cdot(kla+(c-a)l-c-kla+a)\\(d-b)(l-1)&=\frac{d-b}{c-a}\cdot(c-a)(l-1)\\l-1&=l-1\end{aligned}$$
If \(kl=1\), the composition of homotheties \(H((a,b),k)\) and \(H((c,d),l)\) is a translation


By the formula, applying both homotheties to point \((x,y)\) results in \((x+(a-c)(l-1),y+(b-d)(l-1))\).

As a corollary of the above theorem, a translation composed with a homothety is a homothety.

The projective plane

Now that we've developed an appreciation for projection in the standard plane, let's extend the standard plane to build the projective plane. First, we'll develop a formal notion of a projective point (more commonly known as a homogenous coordinate).

Let \((x,y)\) be a point in the real plane. If there are real numbers \(a,b,c\) such that \(x=\frac ac\) and \(y=\frac bc\), we call \([a;b;c]\) homogenous coordinates for point \((x,y)\).

Note that if we are given homogenous coordinates \([a;b;c]\), we can find the corresponding cartesian point by division as \(\left(\frac ac,\frac bc\right)\). But, the other way around isn't possible. Given a cartesian point \((x,y)\), there are multiple homogenous points corresponding to that cartesian point. For example, corresponding to \((1,1)\) are homogenous coordinates \([1;1;1]\), \([2;2;2]\), \([-5;-5;-5]\), etc.

What if \(c=0\) in the above example? For now, we'll define \(\frac x0\) for any \(x\) to be \(\infty\). In the world of projective geometry, \(\infty=-\infty\), so we don't need to worry about sign.

Describe all of the homogenous coordinates corresponding to the cartesian coordinates \((1,-1)\). Is there a pattern here?

That isn't great. Ideally, we'd want there to be exactly one homogenous coordinate for each cartesian coordinate. So how can we modify our space of homogenous coordinates to have this property?

The answer is simple: mods! We already used mods to a similar effect when constructing polynomial rings (see the Challenge Section on Week 11). Here, we observe the following phenomenon. If \((x,y)\) has homogenous representations \([a;b;c]\) and \([d;e;f]\), then \(\frac ac=\frac df\) and \(\frac bc=\frac ef\). If \(\lambda = \frac cf\), then $$[a;b;c]=[\lambda d;\lambda e;\lambda f]$$In other words, any two homogenous coordinates with the same cartesian coordinate representation are multiples of each other!

So, we set the following equivalence. \([a;b;c]\equiv[d;e;f]\) if there exists a number \(\lambda\) such that \(a=\lambda d\), \(b=\lambda e\), \(c=\lambda f\).

There is one issue with this construction: every homogenous coordinate has infinitely many equivalent homogenous coordinates, except for \([0;0;0]\). Thus, we disallow this as a homogenous coordinate. Why we do this will become obvious in a moment.

Now, we are ready to define the projective plane.

The projective plane \(\mathbb RP^2\) is defined as the set of homogenous coordinates \([a;b;c]\) with at least one of \(a,b,c\) nonzero under the equivalence \([a;b;c]\equiv[\lambda a;\lambda b;\lambda c]\) for all nonzero \(\lambda\).

This is probably still very confusing. What is the difference between this and the normal cartesian coordinate system. How does this relate to the vanishing points that we talked about earlier? What does this space even look like? We'll answer all of these questions next.

Visualizing Projective Space

If we were to draw projective space, we might draw something like this:

Inside the squiggly dotted line lies the real plane \(\mathbb R^2\). On the border of the real plane, we have the points at infinity, which we call \(d_\infty\). Each one corresponds to a possible direction of line, since any two parallel lines head in the same direction (however, each direction and the direction in the exact opposite direction have the same point at infinity: thus, \(d_\infty=d_{-\infty}\)). So the circle labeled \(r\) is really the "line at infinity" which goes through all of the points at infinity.

This visual should bother you. What does it mean for there to be a boundary on the plane. Doesn't the plane go off to infinity? There isn't any border! However, we are just building our intuition before we see the actual picture.

Let's go back to our construction of the projective plane via homogenous coordinates. What does the set of homogenous points that are equivalent to a points \([a;b;c]\) look like if we plotted them on a graph?

If you haven't seen linear algebra before, try drawing a 3D diagram, or perhaps using WolframAlpha to draw one for you. Once you do so, you'll see that the set is just the points on the line that goes through the origin \((0,0,0)\) and \((a,b,c)\)! In other words, the projective plane is the set of lines in 3D space that go through the origin.

But how can we rectify this perspective with the ones we've discussed earlier? That's where projection (and hence the name projective plane) comes into play. In our 3D space, draw the plane \(w=1\) as in the diagram below. Then, each point \(P\) in the plane \(w=1\) corresponds to the line \(PO\) in projective space (where \(O\) is the origin). And, the points at infinity are represented by the lines parallel to the plane \(w=1\).

If this image still isn't clear, here's another way to see it. Add to the picture above the unit sphere in 3D space. Every line through the origin intersects this sphere at exactly two points. In fact, since the lines goes through the center of the sphere, projective space can be viewed as the space of diameters of the unit sphere.

Diameters are often identified by their endpoints. So what if we associate points in projective space with points on the sphere? Would that work? Almost. For any point \([x;y;z]\) on the unit sphere, the homogenous point \([-x;-y;-z]\) is also on the unit sphere and is equivalent to \([x;y;z]\). So we can define projective space in a new way.

The projective plane \(\mathbb RP^2\) is defined as the set of points \((x,y,z)\) on the unit sphere under the equivalence \((x,y,z)\equiv(-x,-y,-z)\) for all points.

In other words, the projective plane can be viewed as half of the sphere, since we are removing half of the points.

What are the differences between a hemisphere (half a sphere) and the projective plane?

Our new diagram is as follows: point \(P\) in the cartesian plane is represented by the line \(PO\), which is also represented by points \(P_1,P_2\) in the unit sphere.

We now have \(5\) ways of thinking about the projective plane.

  1. The cartesian plane with points at infinity and a line at infinity added in.
  2. The set of homogenous coordinates under an equivalence relation.
  3. The set of lines through the origin in 3D space.
  4. The projections of the origin onto a plane \(w=1\).
  5. The set of points on the unit sphere under an equivalence relation.

The more of these ways of thinking you master, the better your understanding of the projective plane will be.

Where are the points at infinity and the line at infinity in each of the \(5\) approaches above?


  1. On the border of the cartesian plane
  2. The points at infinity are the points \([a;b;0]\), and the line at infinity is the set of points at infinity
  3. The points at infinity are the lines parallel to the \(x-y\) plane, and the line at infinity is just the \(x-y\) plane.
  4. The points at infinity are the projections that don't meet the plane \(w=1\), and the line at infinity is the set of these points.
  5. The points at infinity are the points on the unit sphere in the \(x-y\) plane, and the line at infinity is the unit circle in the \(x-y\) plane.

Solving the above exercise should test your understanding of the projective plane. Now, let's do some things with it!

Homogenous Polynomials

The projective plane is very useful because any two points determine a line, and any two lines determine a point. This symmetry leads to a cleaner way to handle line and conics than in Coordinate Geometry.

Suppose we take the standard formula for a line in the cartesian plane: \(ax+by+c=0\). If we replace \((x,y)\) with \([X;Y;Z]\), we get the formula $$a\frac XZ+b\frac YZ+c=0$$Multiplying by \(Z\), we get $$aX+bY+cZ=0$$This is called a homogenous polynomial, since all of the terms have the same degree (\(1\)).

We can similarly homogenize a quadratic polynomial. \(ax^2+bxy+cy^2+dx+ey+f=0\) becomes $$aX^2+bXY+cY^2+dXZ+eYZ+fZ^2=0$$ As we can see, homogenizing a line gives us a plane in 3D space, as in the following diagram:

Now why would we want to homogenize polynomials? One of the great benefits is the following theorem:

Given any non-degenerate conic \(ax^2+bxy+cy^2+dx+ey+f=0\), the homogenization of the conic is isomorphic to a circle.

Let's understand what this theorem means. It say, if we move any conic from the real plane to the projective plane, it becomes isomorphic to a circle. By isomorphic, we roughly mean, we can stretch or compress one shape to form the other (for example, a circle is isomorphic to an ellipse, since we can stretch the circle out. The circle is not isomorphic to a parabola, since the parabola has an open end.)

This is not an obvious theorem! In the real plane, ellipses, parabolas, and hyperbolas are all non-isomorphic (you can't transform any one of them into any other of them.) However, in the projective plane, they are all really just the same shape.

To see why this is the case, let's imagine the image of these conics on the sphere (our 5th way of imagining the projective plane.)

As the diagram above shows, a line through the origin in the cartesian plane is a great circle on the sphere. What about our conics?

In the diagram above, \(l_\infty\) represents the line at infinity. If a circle on our sphere doesn't touch the line at infinity, as in the first example, the projection onto the real plane of the circle will be an ellipse, as in the following picture.

If we are in the second case, in which the circle touches the line at infinity at one point (as a tangent), the projection will be a parabola. And in the third case, where the line at infinity is a secant to the circle, the projection of the circle will be a hyperbola.

In the quadratic equation, if \(b^2-4ac>0\), there are two roots. If \(b^2-4ac=0\), there is one root, and if \(b^2-4ac\lt0\), there are zero roots. Similarly, if \(B^2-4AC>0\), our projection is a hyperbola, and the circle projecting it intersects the line at infinity at two points. If \(B^2-4AC=0\), our projection is a parabola, and the circle projecting it intersects the line at infinity at one point. If \(B^2-4AC\lt0\), our projection is an ellipse, and the circle projecting it intersects the line at infinity at zero points. The quantity \(b^2-4ac\) is called the discriminant, and shows up all over mathematics.

Another way of thinking about the differences between these \(3\) conics is as follows. In the projective space, the two "U"s in the hyperbola are glued together on each of their strands at the two points at infinity which they intersect at. The parabola's two strands intersect at one point at infinity. Thus, you can begin to imagine both of these shapes as very stretched out ellipses.

For the following conics, find the points at infinity that they intersect
  • \(x^2=1\)
  • \(x^2-y^2=1\)


  • \([0;1;0]\)
  • \([1;1;0],\;[1;-1;0]\)

Duality

Now that we've developed the projective plane, we can formalize our understanding of the duality between lines and points.

Let \(\omega\) be the unit circle \(\{[\cos\theta;\sin\theta;1]\mid\theta\in[0,2\pi)\}\), and let \(O\) be the origin \([0;0;1]\). Let \(A\) be any point in the projective plane. If \(A=O\), the polar of \(A\) is the line at infinity. If \(A\) is on the line at infinity, the polar of \(A\) is the line through \(A\) perpendicular to \(OA\). Otherwise, let \(A'\) be the point on the ray \(\vec{OA}\) that is \(\frac1d\) away from \(O\), where \(d\) is the length of the line segment \(OA\). The polar of \(A\) is the line \(a\) through \(A'\) perpendicular to \(OA'\).

Given a line \(l\), if it intersects \(O\), the pole of \(l\) is the point on the intersection of the line perpendicular to \(l\) through \(O\) with the line at infinity. If \(l\) is the line at infinity, the pole of \(l\) is \(O\). Otherwise, let \(A'\) be the point on \(l\) closest to \(O\) as in the diagram above, and let \(A\) be the point on ray \(\vec{OA'}\) such that the distance \(OA=\frac1{d'}\), with \(d'\) being the length of the line segment \(OA'\).

With these definitions, we've associated every point in projective space with a line, and vice versa. This association has several nice properties that will formalize our understanding of duality.

The polar of a pole of a line is itself, and the pole of a polar of a point is itself.


Let's do some casework
  • If line \(l\) is the line at infinity, its pole is \(O\), whose polar is the line at infinity
  • If \(O\) is on line \(l\), then the pole is the intersection of the line perpendicular to \(l\) through \(O\) with the line at infinity. And, the polar of this point is just \(l\)
  • If neither of the above cases hold, then \(OA'\) has length \(0\lt d\lt\infty\), where \(A'\) is the closest point to \(O\) on line \(l\). Thus, \(A\) is the pole of \(l\) as defined above. And, the polar of \(A\) is the line that goes through \(A'\) that is perpendicular to \(OA'\). By standard geometry (prove it if it isn't obvious), the point closest to \(O\) on a line \(l\) satisfies \(OA'\perp l\).
The other part of the proof is essentially identical, and is left to you as an exercise.
If point \(p\) is on line \(l\), the pole of \(l\) is on the polar of \(p\).


  • If \(l\) is the line at infinity, then \(p\) is a point at infinity, meaning that the polar of \(p\) is a line through the origin and the pole of \(l\) is the origin.
  • If \(l\) goes through the origin, then the pole of \(l\) is the point at infinity perpendicular to \(l\). Thus, we just need to show that the polar of \(p\) is perpendicular to \(l\). If \(p\) is a point at infinity, the polar of \(p\) is the line perpendicular to \(l\) through the origin. If \(p\) is the origin, the polar is the line of infinity, containing the pole of \(l\). If \(p\) is any other point, the polar of \(p\) is a line perpendicular to \(l\).
  • If \(l\) is any other line, then the situation is akin to the diagram above with \(l=a\) and \(p=A'\). Thus, the pole of \(a\) is \(A\) and the polar of \(A'\) is the line through \(A\) perpendicular to \(OA\).

Note that the incidence theorem implies that a point is on a line if and only if its polar contains the pole of the line (can you see why?). With the isomorphism and incidence theorems, we've established that the projective plane is dual, meaning that there is a one to one mapping between points and lines that preserves points being on lines. With self-duality, we gain a lot of power.

Self-duality implies that for any theorem in the projective plane that involves points, lines, concurrency (lines intersecting at the same point), and collinearity (points on the same line), the dual theorem, which is the theorem with lines and points switched, and concurrency and collinearity switched, is also true. Let's see a few examples of dual theorems.

If \(\triangle ABC\) and \(\triangle DEF\) have the property that lines \(AD\), \(BE\), and \(CF\) are concurrent, then \(X=AB\cap DE\), \(Y=BC\cap EF\), and \(Z=CA\cap FD\) are collinear.


This proof requires linear algebra, which we haven't learned, so instead, I'm going to do a bit of handwaving. We'll use the perspective of the projective plane as the set of lines through the origin in 3D space. In this perspective, lines in the projective plane are planes in 3D space. Thus, \(X\) and \(Y\) are lines in 3D space, and there exists a unique plane that contains \(X,Y\). Since this plane goes through the origin, there exists some rotation of 3D space that makes this plane into the plane \(z=0\). This makes \(X,Y\) points at infinity. Thus, all we want to show is that \(Z\) is a point at infinity.

\(X,Y\) being points at infinity imply \(AB\parallel DE\) and \(BC\parallel EF\). Consider the homothety \(H(P,B,E)\) (the homothety at \(P\) that sends \(B\) to \(E\)). Since homothety preserves slope, sends \(B\) to \(E\), and sends lines \(PA\) and \(PC\) to themselves, it must send \(A\) to \(D\) and \(C\) to \(F\). This implies that \(CA\parallel FD\) as homothety preserves slope, proving that \(Z\) is on the line of infinity.

We call the point of concurrency the center of perspective and the line of collinearity the axis of perspective.

So what is the dual of Desargue's Theorem? It's just the converse!

If \(\triangle ABC\) and \(\triangle DEF\) have the property that \(AB\cap DE\), \(BC\cap EF\), and \(CA\cap FD\) are collinear, then lines \(AD\), \(BE\), and \(CF\) are concurrent.

The next example is a bit more relevant in contest mathematics.

Given a \(\triangle ABC\) and some line \(l\), let \(D=BC\cap l\), \(E=CA\cap l\), \(F=AB\cap l\). Then, \(D,E,F\) are collinear if and only if $$\frac{AF}{FB}\cdot\frac{BD}{DC}\cdot\frac{CE}{EA}=1$$


Consider the homotheties \(H(D,B,C)\), \(H(E,C,A)\), \(H(F,A,B)\). Clearly, composing these homotheties leads to \(B\) staying in the same place. Moreover, the composition of the first two homotheties keeps line \(DE\) in the same place since the centers are \(D\) and \(E\). So, there are two cases
  • If \(F\) is on line \(DE\), then the final homothety also keeps line \(DE\) fixed. Since points \(B,D,E\), which aren't collinear, are all fixed by the composition, the composition must be the identity. That means that the product of the scale factors must be \(1\). In other words, $$\frac{CD}{BC}\cdot\frac{AE}{EC}\cdot\frac{BF}{FA}=1$$
  • If \(F\) is not on \(DE\), then \(F\) doesn't fix \(DE\), so the composition doesn't either. Since \(B\) is fixed by the homothety, the composition of the three homotheties must be a homothety centered at \(B\) with scale factor not equal to \(1\). Thus, $$\frac{CD}{BC}\cdot\frac{AE}{EC}\cdot\frac{BF}{FA}\neq1$$
Prove Composing Homotheties 1 with Menelaus' Theorem

The dual of Menelaus's Theorem is Ceva's Theorem, which says the following:

Given a \(\triangle ABC\) with point \(D\) on line \(BC\), point \(E\) on line \(AC\), and point \(F\) on line \(AB\), lines \(AD\), \(BE\), and \(CF\) are concurrent if and only if $$\frac{AF}{FB}\cdot\frac{BD}{DC}\cdot\frac{CE}{EA}=1$$


Let \(P\) be the intersection of lines \(AD\) and \(CF\). Applying Menelaus' Theorem on \(\triangle BDA\) with collinear points \(C,P,F\) gives $$\frac{BC}{CD}\cdot\frac{DP}{PA}\cdot\frac{AF}{FB}=1$$If we apply Menelaus' Theorem to \(\triangle DCA\) with points \(B,E,P\), we get $$\frac{DB}{BC}\cdot\frac{CE}{EA}\cdot\frac{AP}{PD}=1\text{ iff }P\text{ on }BE$$Multiplying the left hand sides of both equations results in$$\frac{BD}{DC}\cdot\frac{CE}{EA}\cdot\frac{AF}{FB}=1\text{ iff }P\text{ on }BE$$This completes the proof.
Prove that medians, altitudes, and angle bisectors of triangles are concurrent with Ceva's Theorem.

These theorems and homothety/projection as general tools are immensely valuable in geometry problems. Here's an example.

Let \(ABC\) be any triangle, with orthocenter \(H\), points \(D,E,F\) midpoints of sides \(BC\), \(CA\), \(AB\), points \(K,L,M\) the feet of altitudes to sides \(BC\), \(CA\), \(AB\), and points \(P,Q,S\) midpoints of segments \(AH\), \(BH\), and \(CH\). Then, points \(D,E,F,K,L,M,P,Q,S\) all lie on a circle.


Let's first show some primary results. If \(Q\) was the orthocenter of \(\triangle ABC\), and \(Q'\) the reflection of point \(Q\) over side \(BC\), we'd have the following diagram:



Note by right triangle properties that \(\angle Q'AC+\angle A=90^\circ\), and \(\angle QCA+\angle C=90^\circ\). Thus, $$\angle CQ'A=\angle Q'QC=\angle Q'AC+\angle QCA=180^\circ-\angle A-\angle C=\angle B$$Thus, by the Secant Angles Theorem, \(Q'\) lies on the circumcircle. So, if \(H_1,H_2,H_3\) are the reflections of point \(H\) over the sides \(BC\), \(CA\), \(AB\), they'd all lie on the circumcircle.

Next, let \(A',B',C'\) be the reflection of \(H\) over \(D,E,F\). Then, by right triangle properties$$\begin{aligned}\angle BCH+\angle ABC&=90^\circ\\\angle CBH+\angle ACB&=90^\circ\end{aligned}$$So, $$\angle BHC=180^\circ-\angle BCH-\angle CBH=\angle ABC+\angle ACB=180^\circ-\angle BAC$$Since \(HD=DA'\) and \(BD=DC\), \(BHCA'\) is a parallelogram, and thus \(\angle BHC=\angle BA'C\). So, \(\angle BA'C+\angle BAC=180^\circ\), and by the cyclic quadrilateral theorem, \(A'\) is on the circumcircle. Thus, \(A',B',C'\) are all on the circle.

Consider \(H(H,0.5)\). By definition, it maps \(A,B,C\) to \(P,Q,S\). By our work above, it maps \(H_1,H_2,H_3\) to \(K,L,M\) and \(A',B',C'\) to \(D,E,F\). Since all of these points are on the circumcircle, and the homothety of a circle is a circle, the theorem is proved.

For any triangle \(ABC\), show that the orthocenter \(H\), circumcenter \(X\), centroid \(G\), and the center of the nine point circle \(N\) are collinear.


In the previous exercise, we showed that the homothety \(H(H,0.5)\) sends \(O\) to \(N\), so \(H,O,N\) are collinear. Now, consider the homothety \(H(G,-0.5)\). We've shown in Week \(4\) that this homothety sends \(A,B,C\) to \(D,E,F\), the midpoints of sides \(BC,AC,AB\) respectively. Thus, this homothety sends \(AB\) to \(DE\), implying that \(BC\parallel EF\). And, it sends line \(AH\) to \(DH'\), where \(H'\) is the point that \(H\) is sent to under the homothety. So, $$DH'\parallel AH\bot BC\parallel EF$$Thus, \(H'\) is on the altitude from \(D\) to \(EF\), and by parallel logic, \(H'\) is on all three altitudes. Thus, \(H'\) is the orthocenter of \(\triangle DEF\).

Moreover, since altitudes of \(DEF\) are perpendicular bisectors of \(ABC\), \(H'\) is the point of intersection of the perpendicular bisectors of \(ABC\), which implies that \(H'=O\). This proof also implies the following relations:$$HG=2GO,\;ON=NH,\;OG=2GN,\;NH=3GN$$

The Euler line also happens to contain a lot of other special points in the triangle. Look at this link for more details.

Top