IMO 1989
Read the problems at the official link. The 1989 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1989 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 Splitting 1 to 1989 into equal-sum groups of 17
- Hint 1. Think of a table with $117$ rows (the groups) and $17$ columns. Two columns can be filled with constant row sums by running one up and one down.
- Hint 2. That handles $14$ columns. For the last $3$ columns you need $117$ triples with equal sums from $1,\dots,351$.
Our solution (not official)
Key idea: build a $117\times17$ table whose rows are the groups. Seven pairs of “up and down” columns have constant row sums. The remaining three columns form triples of equal sum by a shifted pairing.
Note $1989=117\cdot17$. The rows are the sets $A_1,\dots,A_{117}$.
Three columns from $1,\dots,351$. For row $i$ ($1\le i\le117$) let $$j_i=\begin{cases}i+58,& i\le59\\ i-59,& i\ge60\end{cases}$$ which runs over $1,\dots,117$ once. Put $i$, $117+j_i$ and $528-i-(117+j_i)$ in row $i$. The values $i+j_i$ are $2i+58$ for $i\le59$ (the even numbers $60,\dots,176$) and $2i-59$ for $i\ge60$ (the odd numbers $61,\dots,175$). So $i+j_i$ runs over $60,\dots,176$ once each, and the third entries $411-(i+j_i)$ run over $235,\dots,351$ once each. The three columns use $1,\dots,351$ exactly once, and each row sums to $528$.
Fourteen columns from $352,\dots,1989$. These $1638=7\cdot234$ numbers form seven blocks of $234$ consecutive numbers. For the block starting at $s$, put $s+i-1$ in one column and $s+233-(i-1)$ in the next, so row $i$ gets the pair sum $2s+233$, the same for every row. Each block is used exactly once.
Every row then has $17$ numbers, the rows use $1,\dots,1989$ once each, and all row sums are equal.
How we checked it: script: the table uses 1..1989 once, has rows of 17, and one common row sum
P2 Excentral triangle versus hexagon versus triangle areas
- Hint 1. Let $I$ be the incenter. The point $A_1$ is equally far from $B$, $C$ and $I$. Why is it the midpoint of $IA_0$?
- Hint 2. For (ii), compare triangle $BA_1C$ with triangle $BHC$, where $H$ is the orthocenter, using heights over $BC$.
Our solution (not official)
Key idea: $A_1$ is the midpoint of $IA_0$, so cutting both figures into six triangles around $I$ gives (i). For (ii), each cap triangle on a side beats the matching piece cut out by the orthocenter.
Let $I$ be the incenter and $R$ the circumradius. $A_0,B_0,C_0$ are the excenters, and $B$ lies on segment $A_0C_0$ (the external bisector at $B$), and so on.
$A_1$ is the midpoint of $IA_0$. Angle chasing gives $\angle A_1BI=\frac B2+\frac A2=\angle A_1IB$, so $A_1B=A_1I$, and likewise $A_1C=A_1I$. The internal and external bisectors at $B$ are perpendicular, so $\angle IBA_0=90^\circ$. The point of segment $IA_0$ equidistant from $I$ and $B$ is the midpoint of the hypotenuse of right triangle $IBA_0$. So $A_1$ is that midpoint.
(i) The triangle $A_0B_0C_0$ splits around $I$ into the six triangles $IBA_0$, $IA_0C$, $ICB_0$, $IB_0A$, $IAC_0$, $IC_0B$. The hexagon splits into $IBA_1$, $IA_1C$, $ICB_1$, $IB_1A$, $IAC_1$, $IC_1B$. Triangles $IBA_1$ and $IBA_0$ share the side $IB$, and $A_1$, the midpoint of $IA_0$, is half as far from line $IB$ as $A_0$. So $[IBA_1]=\frac12[IBA_0]$, and the same holds for the other five pairs. Hence $[A_0B_0C_0]=2\,[\text{hexagon}]$.
(ii) The hexagon is triangle $ABC$ plus the three caps $BA_1C$, $CB_1A$, $AC_1B$. The distance from $A_1$ (midpoint of arc $BC$) to $BC$ is $R(1-\cos A)$. The orthocenter $H$ is inside (acute triangle) at distance $2R\cos B\cos C$ from $BC$. Now $$1-\cos A=1+\cos(B+C)=1+\cos B\cos C-\sin B\sin C,$$ $$1-\cos A-2\cos B\cos C=1-\cos(B-C)\ge0 .$$ So $[BA_1C]\ge[BHC]$, and similarly for the other caps. Adding, the caps total at least $[BHC]+[CHA]+[AHB]=[ABC]$. So the hexagon has area at least twice $[ABC]$, and by (i) $[A_0B_0C_0]\ge4[ABC]$.
How we checked it: numeric: 1,000 random acute triangles confirm both parts
P3 Many equidistant neighbors force few of them
- Hint 1. Count pairs $\{Q,R\}$ together with a point $P$ equally far from both.
- Hint 2. A point equally far from $Q$ and $R$ lies on their perpendicular bisector. How many points of the set can that line hold?
Our solution (not official)
Key idea: double count triples $(P,\{Q,R\})$ with $PQ=PR$. Each point $P$ gives many such pairs, but each pair has at most two such $P$.
For each $P$ there is a circle centered at $P$ through at least $k$ points of $S$, which gives at least $\binom k2$ pairs $\{Q,R\}$ with $PQ=PR$. So there are at least $n\binom k2$ triples.
For a fixed pair $\{Q,R\}$, every such $P$ lies on the perpendicular bisector of $QR$. No three points of $S$ are collinear, so there are at most $2$ such $P$. So there are at most $2\binom n2$ triples. Hence $$n\,\frac{k(k-1)}2\le n(n-1)\quad\Longrightarrow\quad k^2-k-2(n-1)\le0 .$$ So $k\le\frac12+\sqrt{2n-\frac74}<\frac12+\sqrt{2n}$.
How we checked it: reasoning (double counting)
P4 Three touching circles and a tangent line inequality
- Hint 1. Draw the circle about $A$ of radius $AD$, the circle about $B$ of radius $BC$, and the circle about $P$ of radius $h$. How are they placed?
- Hint 2. Look at the angles at $P$ between $PA$, $PB$ and the perpendicular from $P$ to line $CD$. Show first that this perpendicular, extended away from $CD$, crosses side $AB$.
Our solution (not official)
Write $a=AD$, $b=BC$, and $\ell$ for the line $CD$.
Key idea: since $AB=a+b$, $AP=a+h$ and $BP=b+h$, the circles about $A$, $B$, $P$ with radii $a$, $b$, $h$ touch each other from outside. The line $\ell$ touches the circle about $P$ and comes within $a$ of $A$ and within $b$ of $B$. Turning this into angles at $P$ gives $\angle APB\ge\alpha+\beta$ for two explicit angles, and the half-angle formula finishes.
Step 1: the perpendicular from $P$ crosses $AB$. Let $u$ be the unit vector at $P$ perpendicular to $\ell$ and pointing away from $\ell$. The ray from $P$ in direction $u$ leaves the convex quadrilateral at a boundary point $E$, at distance $e=h+PE$ from $\ell$. It is not on $CD$, which lies on $\ell$. Suppose $E$ were on side $BC$. As $C$ lies on $\ell$, $EC\ge e$, so $EB=b-EC\le b-e$ and $$PB\le PE+EB\le(e-h)+(b-e)=b-h<b+h=PB,$$ which is impossible. In the same way $E$ is not on $DA$. So $E$ is on $AB$, and $u$ points into the angle $APB$: $$\angle(u,PA)+\angle(u,PB)=\angle APB.$$
Step 2: two angle bounds. $A$ is on $P$’s side of $\ell$, at distance at most $AD=a$ from it, and that distance equals $h+(a+h)\cos\angle(u,PA)$. Hence $$\cos\angle(u,PA)\le\frac{a-h}{a+h}=\cos\alpha,$$ where $\alpha$ is the angle in $(0^\circ,180^\circ)$ with this cosine. Since $\tan^2\frac\alpha2=\frac{1-\cos\alpha}{1+\cos\alpha}=\frac ha$, we have $\tan\frac\alpha2=\sqrt{h/a}$. So $\angle(u,PA)\ge\alpha$. Likewise $\angle(u,PB)\ge\beta$ with $\tan\frac\beta2=\sqrt{h/b}$. With Step 1, $\gamma=\angle APB\ge\alpha+\beta$.
Step 3: the triangle $APB$. Its sides are $a+h$, $b+h$, $a+b$ and its half-perimeter is $s=a+b+h$. The half-angle formula gives $$\tan\frac\gamma2=\sqrt{\frac{(s-PA)(s-PB)}{s\,(s-AB)}}=\sqrt{\frac{ab}{h\,(a+b+h)}}.$$ As $\gamma<180^\circ$, also $\alpha+\beta<180^\circ$, and tangent increases on $(0^\circ,90^\circ)$, so $$\tan\frac\gamma2\ \ge\ \tan\frac{\alpha+\beta}2=\frac{\sqrt{h/a}+\sqrt{h/b}}{1-h/\sqrt{ab}}.$$ Put $x=\sqrt h$, $p=\sqrt a$, $q=\sqrt b$ (the denominator above is positive, so $x^2<pq$). The inequality says $\Phi(x)\ge0$, where $$\Phi(x)=\frac{pq}{x\sqrt{p^2+q^2+x^2}}-\frac{x(p+q)}{pq-x^2}.$$ The first term decreases as $x$ grows and the second increases, so $\Phi$ is decreasing. At $x_0=\frac{pq}{p+q}$ both terms equal $\frac{(p+q)^2}{p^2+pq+q^2}$; this uses the identity $(p^2+q^2)(p+q)^2+p^2q^2=(p^2+pq+q^2)^2$. So $\Phi(x_0)=0$, and $\Phi(x)\ge0$ forces $x\le x_0$, that is, $$\frac1{\sqrt h}\ \ge\ \frac{p+q}{pq}=\frac1{\sqrt{AD}}+\frac1{\sqrt{BC}}.$$
How we checked it: numeric: 13,000 random configurations meeting all the conditions; the inequality always holds and the perpendicular from P always meets AB as in Step 1
P5 Long runs of integers that are not prime powers
- Hint 1. A number divisible by two different primes is not a prime power.
- Hint 2. Choose $2n$ different primes and use the Chinese remainder theorem.
Our solution (not official)
Key idea: force each number of the run to be divisible by two different primes.
Take $2n$ distinct primes $p_1,q_1,\dots,p_n,q_n$. By the Chinese remainder theorem there is a positive integer $x$ with $$x\equiv-i\pmod{p_iq_i}\qquad(i=1,\dots,n).$$ Then $x+i$ is divisible by the two different primes $p_i$ and $q_i$, so it is not a power of a prime. The numbers $x+1,\dots,x+n$ are $n$ consecutive positive integers, none a prime power.
How we checked it: reasoning
P6 Permutations with a partner pair side by side
- Hint 1. Call $i$ and $i+n$ partners. Find a one-to-one map from permutations with no adjacent partners into those with exactly one adjacent pair.
- Hint 2. Move the first entry so that it lands just before its partner.
Our solution (not official)
Key idea: move the first entry next to its partner. This maps the permutations without the property one-to-one into those with exactly one adjacent partner pair, and some permutations with the property are missed.
Call $i$ and $i+n$ partners. Let $\pi=(x_1,\dots,x_{2n})$ have no adjacent partners. The partner of $x_1$ sits at some position $k\ge3$. Define $$\Phi(\pi)=(x_2,\dots,x_{k-1},\,x_1,\,x_k,\dots,x_{2n}).$$ In $\Phi(\pi)$ the entries $x_1,x_k$ are adjacent partners. The other new neighbor of $x_1$ is $x_{k-1}$, not its partner, and every other adjacent pair was already adjacent in $\pi$. So $\Phi(\pi)$ has exactly one adjacent partner pair, with $x_1$ as its first entry.
One-to-one. From $\Phi(\pi)$ we recover $\pi$: find the unique adjacent partner pair and move its first entry to the front. So $\Phi$ is injective, and there are at least as many permutations with the property as without.
Strictly more. For $n\ge2$, the permutation $(1,n+1,2,n+2,3,\dots)$ has at least two adjacent partner pairs, so it is not in the image. For $n=1$ both permutations of $\{1,2\}$ have the property. So the inequality is strict.
How we checked it: script: for n β€ 4 the counts are 2β0, 16β8, 480β240, 26496β13824