IMO 1994
Read the problems at the official link. The 1994 IMO problems are Β© the International Mathematical Olympiad, so we don't reproduce them here. Open the 1994 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 Sets closed under small sums have a large average
- Hint 1. Sort the numbers in decreasing order. Try to show that the $i$-th largest plus the $i$-th smallest is more than $n$.
- Hint 2. If not, add the $i$-th largest to each of the $i$ smallest numbers.
Our solution (not official)
Key idea: pair the $i$-th largest with the $i$-th smallest. If such a pair summed to at most $n$, we would get $i$ different sums in the set, all larger than the $i$-th largest, but only $i-1$ elements are larger.
Sort the numbers as $a_1>a_2>\dots>a_m$. We claim $a_i+a_{m+1-i}\ge n+1$ for every $i$.
Suppose $a_i+a_{m+1-i}\le n$ for some $i$. Then for each $j$ from $m+1-i$ to $m$, $a_i+a_j\le a_i+a_{m+1-i}\le n$, so $a_i+a_j$ is in the set. These are $i$ different numbers (the $a_j$ differ), and each is larger than $a_i$. But only $a_1,\dots,a_{i-1}$, that is $i-1$ numbers, are larger than $a_i$. Contradiction.
Adding the claim over $i=1,\dots,m$ counts each element twice: $$2(a_1+\dots+a_m)\ge m(n+1),$$ which is the inequality.
How we checked it: script: brute force over all such sets for n β€ 12
P2 Isosceles triangle: perpendicular from a center versus midpoint
- Hint 1. If $OQ\perp EF$, then $Q$ and $B$ both see $OE$ at a right angle, and $Q$ and $C$ both see $OF$ at a right angle.
- Hint 2. For the converse, the line through $Q$ cut in half by $Q$ between lines $AB$ and $AC$ is unique.
Our solution (not official)
Key idea: if $OQ\perp EF$, two circles with diameters $OE$ and $OF$ make triangle $OEF$ isosceles, so $Q$ is the midpoint. The converse follows because only one segment through $Q$ is bisected by $Q$.
By symmetry in line $AM$, $OC\perp AC$ and $\angle OBC=\angle OCB$. We use directed angles mod $180^\circ$.
($\Rightarrow$) Let $OQ\perp EF$. Then $\angle OQE=\angle OBE=90^\circ$, so $O,Q,B,E$ lie on the circle with diameter $OE$. Likewise $O,Q,C,F$ lie on the circle with diameter $OF$. Hence $$\angle(EO,EF)=\angle(EO,EQ)=\angle(BO,BQ)=\angle(BO,BC),\qquad \angle(FE,FO)=\angle(FQ,FO)=\angle(CQ,CO)=\angle(CB,CO).$$ The reflection in $AM$ fixes $O$ and swaps $B,C$, so $\angle(BO,BC)=\angle(CB,CO)$. Thus the angles of triangle $OEF$ at $E$ and $F$ agree, so $OE=OF$. The foot of the perpendicular from $O$ to $EF$, which is $Q$, is then the midpoint of $EF$.
($\Leftarrow$) Let $QE=QF$. Let $E'F'$ be the segment through $Q$ perpendicular to $OQ$ with ends on lines $AB$, $AC$. (It exists: if the perpendicular were parallel to $AB$, then $OQ\perp AB$ would put $Q$ on line $OB$, so $Q=B$; likewise for $AC$.) By the first part $Q$ is its midpoint. The half-turn about $Q$ maps line $AB$ to a parallel line, which meets line $AC$ in one point (since $AB\nparallel AC$). Both $F$ and $F'$ are that point, so $F=F'$, $E=E'$, and $OQ\perp EF$.
How we checked it: numeric: 500 random configurations give QE = QF when OQ β₯ EF
P3 Counting numbers with three binary ones in a doubling range
- Hint 1. Compare $f(k+1)$ with $f(k)$. The range gains $2k+1$ and $2k+2$ and loses $k+1$.
- Hint 2. $f$ goes up by $1$ exactly when $k$ has two binary ones. Also $f(2^j)=\binom j2$.
Our solution (not official)
Answer: (b) Exactly the numbers $m=\binom j2+1$ with $j\ge2$: $2,4,7,11,16,\dots$
Key idea: $f$ steps up by $0$ or $1$, and steps up exactly when $k$ has two binary ones. So $f$ hits every value, and a value is hit once only when two step-ups happen in a row.
Let $w(x)=1$ if $x$ has exactly three binary ones, else $0$.
Steps. Going from $k$ to $k+1$, the range gains $2k+1$ and $2k+2$ and loses $k+1$. Since $2k+2$ is $k+1$ shifted left, $w(2k+2)=w(k+1)$, so $$f(k+1)-f(k)=w(2k+1),$$ which is $1$ exactly when $k$ has two binary ones (as $2k+1$ is $k$ with a $1$ appended).
(a) $f(1)=0$, and $f(2^j)=\binom j2$: the numbers from $2^j+1$ to $2^{j+1}-1$ with three ones choose two more ones among $j$ lower places, and $2^{j+1}$ has one. So $f$ is unbounded and rises in steps of $0$ or $1$; it takes every value $m\ge0$.
(b) As $f$ never decreases, $f(k)=m$ has exactly one solution iff the steps into and out of $k$ are both $1$: both $k-1$ and $k$ have two binary ones. If $k-1$ were even, $k$ would have three ones. So $k-1=2^j+1$ and $k=2^j+2$, which has two ones when $j\ge2$. Then $$m=f(2^j+2)=f(2^j)+w(2^{j+1}+1)+w(2^{j+1}+3)=\binom j2+0+1 .$$ So the values with a unique $k$ are $\binom j2+1$ for $j\ge2$.
How we checked it: script: computed f for k < 5000; the unique values are 2, 4, 7, 11, 16, β¦
P4 When mn minus one divides n cubed plus one
- Hint 1. Show the condition is symmetric: $mn-1\mid n^3+1$ exactly when $mn-1\mid m^3+1$.
- Hint 2. Write the quotient as $an-1$. Then $(a,n)$ is also a solution, with $a$ small when $m>n$.
Our solution (not official)
Answer: $(1,2),(2,1),(1,3),(3,1),(2,2),(2,5),(5,2),(3,5),(5,3)$
Key idea: the quotient has the form $an-1$, and then $(a,n)$ is a smaller solution. When $m>n$ this forces $a=1$, and a short divisor check finishes.
Symmetry. $m^3(n^3+1)=(mn)^3+m^3\equiv1+m^3\pmod{mn-1}$, and $\gcd(m,mn-1)=1$. So $mn-1\mid n^3+1$ exactly when $mn-1\mid m^3+1$. Assume $m\ge n$. (The pair $(1,1)$ is excluded since $mn-1=0$.)
$n=1$: $m-1\mid2$, so $m=2,3$.
$m=n\ge2$: $n^2-1\mid n^3+1$ means $(n-1)\mid n^2-n+1$ after dividing by $n+1$, so $n-1\mid1$ and $n=2$.
$m>n\ge2$: let $k=\frac{n^3+1}{mn-1}$. Reducing $k(mn-1)=n^3+1$ mod $n$ gives $k\equiv-1$, so $k=an-1$ with $a\ge1$. Then $(an-1)(mn-1)=n^3+1$, and since $m\ge n+1$, $$an-1\le\frac{n^3+1}{n^2+n-1}<n$$ (as $n^3+n^2-n>n^3+1$ for $n\ge2$). So $a=1$ and $mn-1=\frac{n^3+1}{n-1}=n^2+n+1+\frac2{n-1}$. Hence $n-1\mid2$: $n=2$ gives $mn-1=9$, $m=5$; $n=3$ gives $mn-1=14$, $m=5$.
With symmetry, the solutions are $(2,1),(3,1),(2,2),(5,2),(5,3)$ and their swaps.
How we checked it: script: search with m, n < 400 gives exactly these nine pairs
P5 Functional equation under a shifted product operation
- Hint 1. Write $x\circ y=x+y+xy=(1+x)(1+y)-1$. Then the equation is $f(x\circ f(y))=y\circ f(x)$.
- Hint 2. Put $x=y$: the number $x\circ f(x)$ is a fixed point of $f$. How many fixed points can there be?
Our solution (not official)
Answer: $f(x)=-\dfrac{x}{1+x}$
Key idea: with $x\circ y=(1+x)(1+y)-1$, the number $x\circ f(x)$ is always a fixed point. Fixed points come in chains $a,\ a\circ a,\dots$, which the monotonicity condition forbids, so the only fixed point is $0$.
Let $x\circ y=(1+x)(1+y)-1$; it maps $S\times S$ to $S$. The equation reads $f(x\circ f(y))=y\circ f(x)$.
Fixed points. Put $y=x$: $f(x\circ f(x))=x\circ f(x)$, so $a=x\circ f(x)$ is a fixed point. If $f(a)=a$, then putting $x=y=a$ gives $f(a\circ a)=a\circ a$: so $a\circ a$ is fixed too.
A fixed point $a\ne0$ has $\frac{f(a)}a=1$. Since $\frac{f(x)}x$ is strictly increasing on $(-1,0)$ and on $(0,\infty)$, each interval contains at most one fixed point.
- If $a>0$ is fixed, then $a\circ a=a^2+2a>a$ is another fixed point in $(0,\infty)$.
- If $-1<a<0$ is fixed, then $a\circ a=(1+a)^2-1$ is in $(-1,0)$ and differs from $a$ (equality would need $1+a\in\{0,1\}$).
Both are impossible, so $0$ is the only fixed point, and $x\circ f(x)=0$ for all $x$. That is $(1+x)(1+f(x))=1$, so $f(x)=\frac1{1+x}-1=-\frac x{1+x}$.
Check. Here $1+f(x)=\frac1{1+x}$, so $f(x)\in S$, and $x\circ f(y)=\frac{1+x}{1+y}-1$. Then $f(x\circ f(y))=\frac{1+y}{1+x}-1$, while $y\circ f(x)=\frac{1+y}{1+x}-1$. Also $\frac{f(x)}x=-\frac1{1+x}$ increases on each interval.
How we checked it: algebra: the solution satisfies both conditions
P6 A set that every infinite set of primes splits
- Hint 1. Let membership depend on the number of prime factors and on the smallest prime factor.
Our solution (not official)
Key idea: put a squarefree number in $A$ when its number of prime factors equals its smallest prime factor. Any infinite set of primes can build a product of each kind.
Let $A$ be the set of numbers $p_1p_2\cdots p_k$ with primes $p_1<p_2<\dots<p_k$ and $k=p_1$.
Let $S$ be an infinite set of primes and $p$ its smallest element; note $p\ge2$.
- In $A$: $m=p\,q_2\cdots q_p$ with $q_2,\dots,q_p$ other elements of $S$. It has $p$ distinct prime factors and smallest factor $p$, so $m\in A$.
- Not in $A$: $n=q_1q_2\cdots q_p$ with $p$ elements of $S$ other than $p$. Its smallest factor exceeds $p$, but it has $p$ factors, so $n\notin A$.
Both are products of $k=p\ge2$ distinct elements of $S$.
How we checked it: reasoning