IMO 1965
Read the problems at the official link. The 1965 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1965 problems (PDF) from the IMO's official site.
The topic labels, hints and solutions below are our own, written from scratch: our solutions, not official ones. If you spot a mistake, please use βReport a problemβ at the bottom of the page.
P1 Trigonometric double inequality with nested roots
- Hint 1. $1\pm\sin2x=(\sin x\pm\cos x)^2$.
- Hint 2. For real $u,v$, $\bigl||u+v|-|u-v|\bigr|=2\min(|u|,|v|)$.
Our solution (not official)
Answer: $\frac\pi4\le x\le\frac{7\pi}4$
Key idea: the middle expression is $2\min(|\sin x|,|\cos x|)$.
Since $1\pm\sin2x=(\sin x\pm\cos x)^2$, the roots are $|\sin x+\cos x|$ and $|\sin x-\cos x|$. For real $u,v$, the larger of $|u+v|$ and $|u-v|$ is $|u|+|v|$ and the smaller is $\bigl||u|-|v|\bigr|$. So $$\bigl||u+v|-|u-v|\bigr|=2\min(|u|,|v|).$$ Hence the middle expression is $m(x)=2\min(|\sin x|,|\cos x|)$.
Right inequality. $\min(|\sin x|,|\cos x|)\le\frac{\sqrt2}2$, because the two squares add up to $1$. So $m(x)\le\sqrt2$ always holds.
Left inequality. $2\cos x\le m(x)$.
- If $\cos x\le0$, it holds, since $m\ge0$. On $[0,2\pi]$ this is $\frac\pi2\le x\le\frac{3\pi}2$.
- If $\cos x>0$, it says $\cos x\le|\sin x|$ (the condition $\cos x\le|\cos x|$ is automatic). With $\cos x>0$, this means $x\in[\frac\pi4,\frac\pi2)$ or $x\in(\frac{3\pi}2,\frac{7\pi}4]$.
Together: $\dfrac\pi4\le x\le\dfrac{7\pi}4$.
How we checked it: script: both inequalities tested at 200,001 points of [0, 2Ο]
P2 Homogeneous system with dominant positive diagonal
- Hint 1. Suppose a non-zero solution exists. Look at the equation of an unknown with the largest absolute value.
- Hint 2. You may assume that unknown is positive; multiply the solution by $-1$ if needed.
Our solution (not official)
Key idea: look at the equation of the largest unknown. Its diagonal term wins against everything else.
Suppose $(x_1,x_2,x_3)\ne(0,0,0)$ is a solution. Choose $i$ with $|x_i|$ largest. Replacing the solution by its negative if needed, assume $x_i>0$, so $x_j\le x_i$ for every $j$.
In the $i$-th equation, every off-diagonal coefficient $a_{ij}$ is negative and $x_j\le x_i$, so $a_{ij}x_j\ge a_{ij}x_i$. Therefore $$0=a_{ii}x_i+\sum_{j\ne i}a_{ij}x_j\ \ge\ \Bigl(a_{ii}+\sum_{j\ne i}a_{ij}\Bigr)x_i>0,$$ because the coefficients of each equation have a positive sum and $x_i>0$. This is a contradiction. So the only solution is $x_1=x_2=x_3=0$.
(The same proof works for any number of unknowns.)
P3 Slicing a tetrahedron parallel to two opposite edges
- Hint 1. Every point of the tetrahedron lies on a segment from a point of $AB$ to a point of $CD$.
- Hint 2. The section at fraction $t$ of the way from $AB$ to $CD$ is a parallelogram with sides $(1-t)a$ and $tb$ meeting at the angle $\omega$.
Our solution (not official)
Answer: $\dfrac{k^2(k+3)}{3k+1}$ (the part containing $AB$ to the part containing $CD$)
Key idea: the cross-sections parallel to both edges are parallelograms whose area is a quadratic in the position. Integrating it gives both pieces.
Every point of the tetrahedron can be written as $(1-t)X+tY$ with $X$ on $AB$, $Y$ on $CD$ and $0\le t\le1$. The planes parallel to $AB$ and $CD$ are exactly the sets of fixed $t$, and the one with parameter $t$ is at distance $td$ from $AB$. Its section is $$\{(1-t)X+tY\}=(1-t)\cdot AB+t\cdot CD,$$ a parallelogram with sides $(1-t)a$ and $tb$ at angle $\omega$. Its area is $t(1-t)\,ab\sin\omega$. Integrating over the distance $d\,dt$: $$\text{vol}=abd\sin\omega\int_0^1t(1-t)\,dt=\frac{abd\sin\omega}{6},$$ which is the familiar formula. The plane $\varepsilon$ has distances in the ratio $k$, so it sits at $t=\lambda=\frac k{k+1}$. The piece next to $AB$ has volume $$abd\sin\omega\int_0^\lambda t(1-t)\,dt=abd\sin\omega\Bigl(\frac{\lambda^2}2-\frac{\lambda^3}3\Bigr)=\frac{abd\sin\omega}{6}\,\lambda^2(3-2\lambda).$$ With $\lambda=\frac k{k+1}$: $$\lambda^2(3-2\lambda)=\frac{k^2(k+3)}{(k+1)^3},\qquad 1-\lambda^2(3-2\lambda)=\frac{(k+1)^3-k^3-3k^2}{(k+1)^3}=\frac{3k+1}{(k+1)^3}.$$ So the two pieces are in the ratio $$\frac{k^2(k+3)}{3k+1},$$ independent of $a,b,d,\omega$. For $k=1$ the plane is halfway and the ratio is $1$, as symmetry demands.
How we checked it: script: Monte Carlo on random tetrahedra for k = 0.5, 1, 2, 3 agrees with the formula to within sampling error
P4 Four numbers: each plus the others’ product is 2
- Hint 1. Let $P$ be the product of all four. If $x_i\ne0$, then $x_i+P/x_i=2$.
- Hint 2. So every $x_i$ is a root of the same quadratic $t^2-2t+P=0$: there are at most two different values.
Our solution (not official)
Answer: $(1,1,1,1)$ and the four arrangements of $(3,-1,-1,-1)$
Key idea: all four numbers are roots of one quadratic, so they take at most two values. A short case check finishes.
No $x_i$ is $0$: if $x_1=0$, the other equations give $x_2=x_3=x_4=2$, and then the first equation says $0+8=2$. Let $P=x_1x_2x_3x_4\ne0$. The $i$-th equation is $x_i+\frac P{x_i}=2$, that is $$x_i^2-2x_i+P=0 .$$ So each $x_i$ is a root $r$ or $s$ of $t^2-2t+P$, where $r+s=2$ and $rs=P$.
- All equal, $x_i=r$: then $r+r^3=2$, so $(r-1)(r^2+r+2)=0$ and $r=1$. This gives $(1,1,1,1)$.
- Three equal to $r$, one equal to $s\ne r$: $P=r^3s$ and $P=rs$, so $r^2=1$. $r=1$ forces $s=1=r$, so $r=-1$ and $s=3$. Check: $3+(-1)^3=2$ and $-1+(-1)(-1)\cdot3=2$. This gives $(3,-1,-1,-1)$ in any order.
- Two and two: $P=r^2s^2=rs$, so $rs=1$ with $r+s=2$, which forces $r=s=1$, a contradiction.
The solutions are $(1,1,1,1)$ and the four arrangements of $(3,-1,-1,-1)$.
How we checked it: sympy: solving the system returns exactly these real solutions (plus non-real ones)
P5 Orthocenter of a pedal triangle moves linearly
- Hint 1. $O,P,Q,M$ lie on the circle with diameter $OM$. Its center is the midpoint of $OM$.
- Hint 2. For a triangle with circumcenter $K$, the orthocenter is $X+Y+Z-2K$ (as vectors). With $O$ at the origin, $H=P+Q-M$.
Our solution (not official)
Answer: (a) The segment joining the feet of the perpendiculars from $A$ to $OB$ and from $B$ to $OA$; (b) the interior of the triangle these feet form with $O$
Key idea: $H$ is a linear function of $M$: it is $P+Q-M$ with $O$ as origin. A linear map sends segments to segments and triangles to triangles.
$H$ in vectors. Put $O$ at the origin. The right angles at $P$ and $Q$ put $O,P,Q,M$ on the circle with diameter $OM$, so the triangle $OPQ$ has circumcenter $\frac12M$. For any triangle $XYZ$ with circumcenter $K$, the orthocenter is $X+Y+Z-2K$. (Check: $(X+Y+Z-2K)-X=(Y-K)+(Z-K)$, and this is perpendicular to $Y-Z=(Y-K)-(Z-K)$ because $|Y-K|=|Z-K|$.) Hence $$H=O+P+Q-M=P+Q-M .$$ $P$ and $Q$ are the projections of $M$ onto the lines $OA$ and $OB$, which are linear in $M$. So $H=L(M)$ for a linear map $L=\pi_A+\pi_B-\mathrm{id}$, where $\pi_A,\pi_B$ are these projections.
Images of $A$ and $B$. $L(A)=A+\pi_B(A)-A=\pi_B(A)=:A_1$, the foot of the perpendicular from $A$ to $OB$. Likewise $L(B)=B_1$, the foot from $B$ to $OA$. Since the angle at $O$ is acute and not right, $A_1$ is a non-zero point of ray $OB$ and $B_1$ a non-zero point of ray $OA$. They are linearly independent, so $L$ is invertible.
(a) $M=(1-u)A+uB$ with $0\le u\le1$ gives $H=(1-u)A_1+uB_1$. The locus is the segment $A_1B_1$.
(b) $L$ maps the interior of triangle $OAB$ onto the interior of triangle $OA_1B_1$. (For every $M\ne O$ in the closed triangle the projections $P,Q$ are non-zero, because the angle is acute, so the triangle $OPQ$ is not degenerate.)
How we checked it: script: 200 random configurations; the orthocenter computed directly equals the predicted point on A1B1
P6 A point set has at most n diameters
- Hint 1. Two diameters that share no endpoint must cross. Otherwise some pair of their four endpoints would be farther apart than $d$.
- Hint 2. If a point is the end of three diameters, the middle one of the three far ends belongs to no other diameter. Then remove it and use induction.
Our solution (not official)
Key idea: diameters pairwise meet. So a point with three diameters has a middle neighbour that can have no other diameter, and removing it allows induction.
Call the diameters edges. The degree of a point is the number of edges at it.
Lemma 1. Two edges with four different endpoints cross. Suppose edges $XY$ and $ZW$ do not meet.
- If the four points are in convex position, then $XY$ and $ZW$ are not the two diagonals (diagonals of a convex quadrilateral cross). So they are opposite sides. The two diagonals cross at a point, and the triangle inequality at that point gives: sum of diagonals $>$ sum of the opposite sides $=2d$. So some diagonal is longer than $d$. Contradiction.
- Otherwise one point, say $W$, lies in the triangle formed by the other three (the four points cannot be collinear, since on a line only one segment can have the maximum length). The distance from $Z$ to a point of a triangle with vertex $Z$ is largest only at the other two vertices, so $ZW<\max(ZX,ZY)\le d$. Contradiction. The same argument works whichever point lies inside.
Lemma 2. If $P$ has three edges $PA,PB,PC$, then one of $A,B,C$ has degree $1$. $A,B,C$ lie on the circle of radius $d$ about $P$, and their pairwise distances are at most $d$. So they lie within an arc of at most $60^\circ$. Name them so that $B$ is between $A$ and $C$, strictly inside the angle $APC$. Then $BA<d$ and $BC<d$. Suppose $BQ$ is an edge with $Q\ne P$. Then $Q\ne A,C$, so by Lemma 1, $BQ$ meets both segments $PA$ and $PC$. It cannot pass through $P$ (that would make $BQ=BP+PQ>d$). But segment $BQ$ starts inside the convex angle $APC$. Once it leaves the angle through one side, it cannot come back to cross the other side. Contradiction. So $B$ has degree $1$.
Induction on $n$. For $n=3$ there are at most $3$ segments in all. Let $n\ge4$.
- If some point $V$ has degree at most $1$, remove it. If the remaining set still has the largest distance $d$, its edges are the old edges not at $V$, and by induction there are at most $n-1$ of them, so there were at most $n$. If the largest distance drops, every old edge used $V$, so there was at most $1$.
- If every degree is at least $2$ and some point has degree at least $3$, Lemma 2 gives a point of degree $1$, a contradiction. So every degree is exactly $2$, and the number of edges is $\frac{2n}2=n$.
Either way there are at most $n$ diameters. (A regular polygon with an odd number of vertices has exactly $n$.)