Kvant Math Problem 853
A square $ABCD$ rotates about its fixed center $O$, while a fixed line $l$ is taken as the $x$-axis.
Verified: yes
Verdicts: PASS + PASS
Solve time: 29m33s
Source on kvant.digital
Problem
The square $ABCD$ rotates about its fixed center. Find the locus traced by the midpoint of the segment $PQ$, where $P$ is the foot of the perpendicular dropped from the point $D$ to the fixed line $l$, and $Q$ is the midpoint of the side $AB$.
Y. Tabov (Bulgaria)
Exploration
A square $ABCD$ rotates about its fixed center $O$, while a fixed line $l$ is taken as the $x$-axis. The midpoint $Q$ of $AB$ and the foot $P$ of the perpendicular from $D$ to $l$ both depend on the rotation angle $\theta$. The midpoint $M$ of $PQ$ is expressed in terms of $\theta$, and the locus is obtained by eliminating the parameter.
The configuration is handled by placing the square in a coordinate system centered at $O$, then applying a rotation matrix to all relevant points.
Problem Understanding
The square is initially centered at the origin with vertices
$A\left(-\frac a2,\frac a2\right),\quad B\left(\frac a2,\frac a2\right),\quad C\left(\frac a2,-\frac a2\right),\quad D\left(-\frac a2,-\frac a2\right).$
A rotation by angle $\theta$ about the origin is applied using
$R_\theta=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix}.$
The midpoint $Q$ of $AB$ is rotated, the vertex $D$ is rotated, the point $P$ is obtained by orthogonal projection of $D$ onto the $x$-axis, and $M$ is the midpoint of $PQ$.
Proof Architecture
The midpoint of $AB$ before rotation is
$Q_0=\left(0,\frac a2\right).$
After rotation,
$Q=\left(-\frac a2\sin\theta,\frac a2\cos\theta\right).$
The vertex $D_0=\left(-\frac a2,-\frac a2\right)$ transforms as
$D=\left(\frac a2(\sin\theta-\cos\theta),-\frac a2(\sin\theta+\cos\theta)\right).$
The projection of $D$ onto the $x$-axis keeps the $x$-coordinate and sets the $y$-coordinate to zero, hence
$P=\left(\frac a2(\sin\theta-\cos\theta),0\right).$
The midpoint $M$ of $PQ$ is computed coordinatewise. For the $x$-coordinate,
$x_M=\frac12\left(\frac a2(\sin\theta-\cos\theta)-\frac a2\sin\theta\right)=-\frac a4\cos\theta.$
For the $y$-coordinate,
$y_M=\frac12\left(0+\frac a2\cos\theta\right)=\frac a4\cos\theta.$
Thus
$M=\left(-\frac a4\cos\theta,\frac a4\cos\theta\right).$
Solution
The parametric equations of the locus are
$x=-\frac a4\cos\theta,\qquad y=\frac a4\cos\theta.$
These relations imply the linear constraint
$x+y=0.$
The parameter $\cos\theta$ varies between $-1$ and $1$, so $y$ ranges from $-\frac a4$ to $\frac a4$. The same restriction applies to $x$ through $x=-y$.
The locus is therefore a segment of the line $y=-x$ bounded by the extreme values of $\cos\theta$.
When $\cos\theta=1$, the point is
$M\left(-\frac a4,\frac a4\right),$
and when $\cos\theta=-1$, the point is
$M\left(\frac a4,-\frac a4\right).$
Verification of Key Steps
The rotation of $Q$ follows directly from applying the rotation matrix to the midpoint of $AB$. The rotation of $D$ is obtained by applying the same matrix to $\left(-\frac a2,-\frac a2\right)$, which produces a linear combination of $\sin\theta$ and $\cos\theta$ with correct sign structure.
The projection step preserves the horizontal coordinate of $D$ because orthogonal projection onto the $x$-axis removes only the vertical component. The midpoint computation is an affine average of two correctly rotated points, which reduces both coordinates to expressions proportional to $\cos\theta$ alone after cancellation of $\sin\theta$ terms.
The resulting dependence shows that all motion collapses onto a single scalar parameter $\cos\theta$, which forces the locus to be a line segment rather than a general conic.
Alternative Approaches
The point $Q$ depends linearly on the rotated vector corresponding to the midpoint of a side of the square, while $P$ depends linearly on the rotated vertex $D$ followed by projection onto a fixed axis. Both operations preserve linear dependence on $\sin\theta$ and $\cos\theta$, but the midpoint operation eliminates the $\sin\theta$ component entirely, leaving both coordinates proportional to $\cos\theta$ with opposite signs. This reduction forces collinearity of all positions of $M$, so the image of the motion is a one-dimensional segment rather than a quadratic curve.