Kvant Math Problem 976

Place the square in coordinates:

Verified: no
Verdicts: SKIP + SKIP
Solve time: 1m22s
Source on kvant.digital

Problem

From vertex $A$ of the square $ABCD$, two rays are drawn forming an angle of $45^\circ$ between them. One ray intersects side $BC$ at point $E$ and diagonal $BD$ at point $P$; the other intersects side $CD$ at point $F$ and diagonal $BD$ at point $O$. Prove that the area of triangle $AEF$ is twice the area of triangle $APO$.

E. G. Gotman

Exploration

Place the square in coordinates:

$$A=(0,0),\quad B=(1,0),\quad C=(1,1),\quad D=(0,1).$$

The diagonal $BD$ has equation

$$x+y=1.$$

Let the ray through $E$ and $P$ have slope $m>0$. Its equation is

$$y=mx.$$

Since it meets $BC$, we must have $m\le 1$.

The second ray forms an angle $45^\circ$ with the first. If its slope is $n$, then

$$\tan 45^\circ=\frac{n-m}{1+mn},$$

hence

$$n=\frac{m+1}{1-m}.$$

Since $m\le 1$, this ray indeed meets $CD$.

Compute the intersection points:

$$E=(1,m),\qquad P=\left(\frac1{1+m},\frac{m}{1+m}\right),$$

$$F=\left(\frac1n,1\right),\qquad O=\left(\frac1{1+n},\frac{n}{1+n}\right).$$

The area of a triangle with one vertex at the origin equals half the absolute value of the determinant of the other two vertices. Thus

$$[AEF] =\frac12\left| \begin{vmatrix} 1&m\[2mm] 1/n&1 \end{vmatrix} \right| =\frac12\left(1-\frac mn\right).$$

Likewise

$$[APO] =\frac12 \left| \begin{vmatrix} \dfrac1{1+m}&\dfrac{m}{1+m}\[3mm] \dfrac1{1+n}&\dfrac{n}{1+n} \end{vmatrix} \right| = \frac12\cdot \frac{n-m}{(1+m)(1+n)}.$$

The relation $n=(m+1)/(1-m)$ should convert both expressions to the same parameter. Carrying this out gives

$$1-\frac mn=\frac{1+m^2}{1+m},$$

while

$$\frac{n-m}{(1+m)(1+n)} =\frac{1+m^2}{2(1+m)}.$$

Hence

$$[AEF]=2[APO].$$

The step most likely to hide an error is the algebraic simplification after substituting

$$n=\frac{m+1}{1-m}.$$

That computation must be checked carefully.

Problem Understanding

A square $ABCD$ is given. From $A$ two rays are drawn making an angle of $45^\circ$. One ray meets side $BC$ at $E$ and diagonal $BD$ at $P$. The other meets side $CD$ at $F$ and diagonal $BD$ at $O$. The goal is to prove that

$$[AEF]=2[APO],$$

where brackets denote area.

This is a Type B problem, a pure proof.

The core difficulty is relating the geometry of the two rays to the positions of their intersections with the diagonal $BD$. A coordinate model turns the $45^\circ$ condition into an explicit relation between the slopes of the rays, after which both areas can be computed directly.

Proof Architecture

Let the square have coordinates $A=(0,0)$, $B=(1,0)$, $C=(1,1)$, $D=(0,1)$.

The first lemma states that if the first ray has slope $m$, then the second ray has slope

$$n=\frac{m+1}{1-m}.$$

This follows from the tangent formula for the angle between two lines.

The second lemma computes the coordinates of $E,F,P,O$:

$$E=(1,m),\quad F=\left(\frac1n,1\right),\quad P=\left(\frac1{1+m},\frac{m}{1+m}\right),\quad O=\left(\frac1{1+n},\frac{n}{1+n}\right).$$

These come from intersecting the rays with the sides of the square and with the diagonal $x+y=1$.

The third lemma computes the two areas:

$$[AEF]=\frac12\left(1-\frac mn\right),$$

$$[APO] = \frac12\cdot \frac{n-m}{(1+m)(1+n)}.$$

This uses the determinant formula for the area of a triangle with one vertex at the origin.

The final step substitutes the expression for $n$ and verifies

$$[AEF]=2[APO].$$

The most delicate point is the algebraic reduction after substituting

$$n=\frac{m+1}{1-m}.$$

Solution

Let

$$A=(0,0),\quad B=(1,0),\quad C=(1,1),\quad D=(0,1).$$

Then the diagonal $BD$ has equation

$$x+y=1.$$

Let the ray passing through $E$ and $P$ have slope $m$. Since it intersects side $BC$, its equation is

$$y=mx,$$

with $0<m\le1$.

Let the second ray, passing through $F$ and $O$, have slope $n$. Since the angle between the rays equals $45^\circ$,

$$\frac{n-m}{1+mn}=\tan 45^\circ=1.$$

Hence

$$n-m=1+mn,$$

and therefore

$$n=\frac{m+1}{1-m}.$$

The point $E$ lies on $BC$, whose equation is $x=1$. Thus

$$E=(1,m).$$

The point $F$ lies on $CD$, whose equation is $y=1$. Since the second ray is $y=nx$,

$$F=\left(\frac1n,1\right).$$

To find $P$, solve

$$y=mx,\qquad x+y=1.$$

Substituting gives

$$x+mx=1,$$

so

$$x=\frac1{1+m},\qquad y=\frac{m}{1+m}.$$

Hence

$$P=\left(\frac1{1+m},\frac{m}{1+m}\right).$$

Similarly, $O$ is the intersection of

$$y=nx$$

with

$$x+y=1.$$

Therefore

$$O=\left(\frac1{1+n},\frac{n}{1+n}\right).$$

Since $A$ is the origin, the area of a triangle with vertices $A,X,Y$ equals

$$\frac12\left| \begin{vmatrix} x_X&y_X\ x_Y&y_Y \end{vmatrix} \right|.$$

Applying this to $AEF$,

$$[AEF] = \frac12 \left| \begin{vmatrix} 1&m\[2mm] 1/n&1 \end{vmatrix} \right| = \frac12\left(1-\frac mn\right).$$

Applying the same formula to $APO$,

$$[APO] = \frac12 \left| \begin{vmatrix} \dfrac1{1+m}&\dfrac{m}{1+m}\[3mm] \dfrac1{1+n}&\dfrac{n}{1+n} \end{vmatrix} \right|.$$

The determinant equals

$$\frac{n}{(1+m)(1+n)} - \frac{m}{(1+m)(1+n)} = \frac{n-m}{(1+m)(1+n)}.$$

Hence

$$[APO] = \frac12\cdot \frac{n-m}{(1+m)(1+n)}.$$

Now substitute

$$n=\frac{m+1}{1-m}.$$

First,

$$1-\frac mn = 1-\frac{m(1-m)}{1+m} = \frac{1+m-m+m^2}{1+m} = \frac{1+m^2}{1+m}.$$

Next,

$$n-m = \frac{m+1}{1-m}-m = \frac{1+m^2}{1-m},$$

and

$$1+n = 1+\frac{m+1}{1-m} = \frac2{1-m}.$$

Therefore

$$\frac{n-m}{(1+m)(1+n)} = \frac{\frac{1+m^2}{1-m}} {(1+m)\frac2{1-m}} = \frac{1+m^2}{2(1+m)}.$$

Thus

$$[APO] = \frac12\cdot \frac{1+m^2}{2(1+m)} = \frac{1+m^2}{4(1+m)},$$

while

$$[AEF] = \frac12\cdot \frac{1+m^2}{1+m} = \frac{1+m^2}{2(1+m)}.$$

Comparing the two expressions,

$$[AEF]=2[APO].$$

This completes the proof.

Verification of Key Steps

The first delicate step is the slope relation. For lines of slopes $m$ and $n$,

$$\tan\theta=\frac{n-m}{1+mn}.$$

Since $\theta=45^\circ$, the right-hand side equals $1$. Solving

$$n-m=1+mn$$

gives

$$n(1-m)=1+m,$$

hence

$$n=\frac{1+m}{1-m}.$$

No sign ambiguity occurs because both rays lie inside the first quadrant.

The second delicate step is the determinant for $[APO]$. Using the coordinates directly,

$$\det(P,O) = \frac1{1+m}\cdot\frac n{1+n} - \frac m{1+m}\cdot\frac1{1+n}.$$

Factoring out $(1+m)^{-1}(1+n)^{-1}$ leaves $n-m$, producing

$$\frac{n-m}{(1+m)(1+n)}.$$

A common mistake is to forget one denominator factor.

The third delicate step is the substitution of $n$. Computing separately,

$$n-m=\frac{1+m^2}{1-m},$$

and

$$1+n=\frac2{1-m}.$$

Substituting these values yields

$$\frac{n-m}{(1+m)(1+n)} = \frac{1+m^2}{2(1+m)},$$

which gives exactly the factor $2$ between the two areas.

Alternative Approaches

A synthetic proof can be obtained by projecting points on the diagonal $BD$. The diagonal divides the square into two congruent right isosceles triangles, and the condition that the rays differ by $45^\circ$ allows the use of angle-chasing inside triangle $ABD$. Expressing the areas through distances from points on $BD$ to the sides of the square leads to proportional segments on the diagonal and eventually to the same factor $2$.

The coordinate method is preferable because the geometry is encoded by a single parameter $m$. The $45^\circ$ condition becomes one algebraic identity, every relevant point has an explicit coordinate, and both areas reduce to straightforward determinant computations.